ESA POLinSAR 2015 - 1st BIOMASS Science Workshop > Session details
The Effect of Wave Polarization on Object Texture Characteristics in Radar Images
Show abstractThe effect of wave polarization on object texture features in polarimetric SIR/C-X-SAR images is examined through the calculation of surface fractal dimension and Haralick’s second- order statistics. The performed analysis of variances of fractal dimension mean values of natural objects 'forest' and 'water' didn't reveal polarization influence on mean values before speckle filtering, but revealed statistically reliable differences in mean values for some pairs of samples after speckle filtering. The incidental nature of observed differences in mean values of fractal dimension is shown for co- polarization VV-HH and cross –polarization HV-VH pairs of samples. This conclusion is true also for Haralick's textural features 'contrast', ‘inverse moment’ and 'entropy' for the object 'water'. Statistically valid differences by wave polarization are revealed in mean values of Haralick's textural features for objects 'city' and 'water' both until and after speckle filtering. Analysis of variances showed that differences in mean values of textural features for object 'forest' are statistically doubtful at change of polarization. Calculations were made in 28*28 pixels window series sized N=20 for every image object (‘forest’, ‘town’, ‘water’) in four polarizations (HH,VV,HV,VH), two wave lengths (L- and C-) SIR/C-X-SAR images
A tool for monitoring woody biomass (change) in woodland ecosystems
Vaccari, Simone (1); Ryan, Casey (1); Lincoln, Philippa (2); Abbot, Paddy (2); Gou, Yaqing (1) 1: University of Edinburgh, United Kingdom; 2: LTS International LTD
Show abstractAbstract Changes in forest area, forest biomass and related carbon emissions are altering the global carbon cycle. Knowledge of the magnitude, location and causes of forest carbon stock changes is crucial to achieving sustainable forest management. In this paper we present the Forest Radar Assessment Tool (FoRAsT), which was first developed over a woodland forest area of 1,160 km2 located in central Mozambique, where an inventory of ninety-six plots was used to calibrate L-band high resolution radar data (ALOS PALSAR). Biomass maps were generated and carbon stock quantified with per-pixel random errors of 10 MgC ha-1 and scene-wide bias of 1.6 MgC ha-1 (95% confidence intervals) over a three-year period between 2007 and 2010. To investigate the tool’s versatility in assessing woody carbon stocks and their change in similar African woodland ecosystems, two case studies were investigated, in collaboration with LTS International Ltd. The first project involved assessing the effects of degradation in the middle Shire river basin in Southern Malawi. Degradation accounted for more than 36% of the total biomass loss, illustrating the important role of degradation alongside deforestation in woodland landscapes. The objective of the second project was to test Forast on different forest ecosystems, which confirmed the importance of site-specific parameterisation of the tool. Introduction Changes in forest area, forest biomass and related carbon emissions are altering the global carbon cycle and the global climate. The most common causes of forest area and biomass loss in African woodlands are small-scale cultivation and farming, timber and charcoal production, and urbanization expansion. The conversion of forest into a non-forest land is termed deforestation, whereas the reduction of forest biomass stock within a forested area is termed forest degradation. Deforestation is often detected and quantified from optical/infra red remote sensing data. Forest degradation is a more complex process as it affects the forest structure and density at smaller scale, which limits its identification from optical data. To account for degradation effects, the use of RAdio Detection And Ranging (RADAR) systems is suggested and presented in this paper. Beyond the advantage of being able to operate at any atmospheric conditions and equally either at day- and night-time, long wavelength RADAR systems measure the backscatter signal, which is often related to aboveground biomass and thus forest carbon stock. For these reasons RADAR is considered a more suitable approach, compared to optical systems, to quantify forest biomass changes due to deforestation and degradation in tropical wet forests and African woodlands. The biomass monitoring tool proposed in this paper makes use of the Phased Array L-band Synthetic Aperture RADAR (PALSAR) instrument, which is part of Japan’s Advanced Land Observing Satellite (ALOS). To test the applicability of ALOS PALSAR data, and to account for forest differences, two case studies have been investigated in collaboration with LTS International Ltd. The first project, at sub national scale, focused on Southern Malawi and aimed to quantifying the magnitude of deforestation and forest degradation in the middle Shire river basin. The objective of the second project was to test the use of the Forast in a higher biomass forest/savanna mosaic landscape. For both projects, yearly biomass maps were generated (2007-10) and compared to obtain a biomass change map, which was classified in order to distinguish between deforestation, forest degradation and biomass increase. This paper summarizes the methodical approach of a forest biomass monitoring tool, developed and tested in Mozambique and used in two commercial project frameworks by LTS International. The tool is being developed for use with data from the upcoming second mission of ALOS PALSAR in 2014, which launch is planned on the 24th of May, 2014, and might potentially be compatible with the ESA BIOMASS mission (P-band radar), planned for 2020. The applicability of the tool and its use at different scales and in different ecosystems is evaluated and potential futures areas of work are outlined.
SAR Tomography Polarimetric Parameters Performance Enhancement by Compressed Sensing Via Atomic Decomposition
Biondi, Filippo Italian Ministry of Defence, Italy
Show abstractSAR Tomography is the extension of the conventional interferometric radar signal processing, where the radar image formation is extended in the height dimension. In order to improve the vertical resolution with respect to the classical Fourier methods, high resolution approaches, based on the Convex Optimization, has been implemented. This methods recast in the Compressed Sensing framework that optimize tomographic smooth profiles via atomic decomposition, in order to retrieve the sparsity condition. The optimum solution has been estimated by Interior Point Methods. The problem for such kind of signal processing is the loss of the tomographic phase information, so only the optimized energy information is available. In this paper a method to estimate an optimized spectra and phase information, is proposed, projecting each vector components of each tomographic resolution cell spanned on the real and the imaginary component. The tomographic solutions has been estimated by processing multi-baseline SAR datasets, in a full polarimetric mode, acquired by the TropiSAR campaign in 2009 by ONERA. The performance enhancement of the principal decomposition parameters for the tomographic vertical planes analysis has been demonstrated.
Ship detection using RADARSAT-2 MSSR OSVN data
Liu, Chen; Vachon, Paris W.; Sandirasegaram, Nicholas Defence R&D Canada - Ottawa Research Centre, Canada
Show abstractRADARSAT-2 (R-2) provides various multi-polarization, wide swath beam modes that have been widely used for maritime surveillance. Two new ScanSAR modes, referred to as Maritime Satellite Surveillance Radar (MSSR), provide improved ship detection over wide swaths, compared to R-2 ScanSAR Narrow B (SCNB) or ScanSAR Wide A (SCWA). The Detection of Vessels, Wide swath, Far incidence angle (DVWF) MSSR mode is designed for detection of vessels with HH polarization, large incidence angles and 450 km swath, while the Ocean Surveillance, Very wide swath, Near incidence angle (OSVN) mode is designed for more general ocean surveillance with HH and HV dual polarization and 530 km swath. This paper focuses on the ocean characteristics of R-2 OSVN MSSR mode imaging and the detection of ship targets. The OSVN data is processed using 4 looks in the azimuth direction and 1 look in the range direction. The six images that were used in this study were acquired over the Canary Islands, from 16 December 2012 to 14 January 2013. Each image contains various types of known ships that were operating with various velocities. Shore-based Automatic Identification System (AIS) data were acquired along with the RADARSAT-2 data and were used to identify the ship targets considered in this study. A novel dual-polarization ship detection approach is presented that is based on a combination of the Likelihood Ratio Test (LRT) with K-distribution constant false alarm rate (KCFAR) detection. In this method, the decision variable is derived using the LRT. For a dual polarization (i.e. HH and HV) SAR system, the inter-channel correlation of the ocean is low. In this case, it can be shown that a good decision variable is the sum of the intensities of the channels, each normalized by the corresponding average ocean intensity. This is referred to as the normalized intensity sum (NIS). This decision variable also applies to a multi-look system, with the individual looks being assumed to be independent. In the actual data, the looks are not independent and the equivalent number of independent looks (ENIL) is used in the derivation of the decision variable. However, the number of looks does not affect the form of the equation for the decision variable. Ocean clutter is often modeled using a K-distribution probability density function (PDF). In this study, it was found empirically that the individual HH and HV channels as well as the NIS decision variable derived from the LRT are well-modelled by a K-distribution PDF. The ENIL and order parameter define the shape of the K-distribution. These parameters were estimated from the data using the method of moments. The detection performance was then evaluated using a two-set Venn diagram: candidate targets versus detected targets, where the candidate targets were based on the AIS reports and a visual assessment using validation critetia developed at Defence R&D Canada (DRDC). The results demonstrate that the probability density function of the ocean decision variable for dual-channel, as well as for the individual HH and HV channels, follows a K-distribution in most ocean areas. The results demonstrate also that a dual-polarization system achieved better detection performance than an HH system. The dual-polarization system declared more ships, but in general, it declared more confusers (i.e. false alarms) compared to an HV system. These confusers often appeared in inhomogeous ocean areas and were visible in the HH channel and not the HV channel. Some confusers near shore had very large spatial scale. At the present time, the algorithms to recognize and reject such confusers are under development.
RETRIEVAL OF SOIL MOISTURE IN PLATEAU PASTURE USING RADARSAT-2 IMAGERY
Chai, Xun; Shao, Yun; Zhang, Tingting; Gong, Huaze; Liu, Long; Xie, Chou; Xie, Kaixin Chinese Academy of Sciences, China, People's Republic of
Show abstractWith the development of microwave remote sensing, especially Synthetic Aperture Radar (SAR), it has been demonstrated the potential of deriving accurate soil moisture content (SMC) distribution on large scale in high resolution with SAR data, which satisfies the needs of environmental studies. Most of previously mentioned research focused on the retrieval of SMC for cropland, retrieval of soil moisture in mountain and plateau areas are rarely reported in literature. Accurate soil moisture retrieval in large scale with high resolution is significant for plateau pasture. The object of this paper is to investigate the estimation of volumetric soil moisture in vegetated areas of plateau pasture using fully polarimetric RADARSAT-2 SAR images at C-band. The study area was a plateau pasture region located in northeastern Tibetan Plateau, China. Two fine quad polarization RADARSAT-2 SAR images and two optical images comprising study area were acquired in September 2012 and May 2013. Experimental campaigns were conducted in study area with simultaneously satellite overpass. Water-cloud model, Chen model and Dubois model were applied in this study. Water-cloud model was widely used to separate significant impact on the radar backscattering coefficient owing to vegetation. Chen model and Dubois model are all soil moisture retrieval model for bare soil. Based on water-cloud model, Chen model and Dubois model, two algorithms for soil moisture retrieval in vegetated soil surface were proposed. Then the performance of models were tested and validated using simulated and experimental data. Root Mean Square Error (RMSE), correlation coefficients square (R2) and the ratio of (standard error of) prediction to standard deviation (RPD) were used to evaluate these two models. Compared with the developed Chen model, predicted soil moisture by the developed Dubois model agreed better with field measurements in terms of accuracy and stability. RMSE, R2 and RPD value of developed Dubois model are (5.4, 0.8, 1.6) and (3.05, 0.78, 1.74) for two experiments respectively. Elimination of vegetation cover effect for backscattering was achieved using water-cloud model in the developed model, and solving the Dubois equation was used to minimize the effect of soil surface roughness. Validation results indicated that the developed Dubois model can satisfied the requirement of soil moisture inversion in study region.
Polarimetric SAR Tomography Via Compressed Sensing and Atomic Decomposition by Linearly Frequency Modulated – Continuous Wave Radar
Biondi, Filippo Italian Ministry of Defence, Italy
Show abstractThe Work described in this paper was done together to the Ingegneria Dei Sistemi (IDS) S.p.A. Via Enrica Calabresi 24, 56121 PISA (ITALY). The main activity was done in order to demonstrate two important subjects. The first one is the feasibility to perform Synthetic Aperture Radar (SAR) tomography by Linearly Frequency Modulated Continuous Wave (LFM-CW) radars and the second one is the feasibility to perform polarimetric SAR classification of scattering events occurred in the third dimension after optimization by Compressed Sensing (CS) and Atomic Decomposition. This work started in order to contribute for the future to solve a problem: the performing of single-pass SAR tomography by Unmanned Aerial Vehicles (UAVs). This work was conducted in laboratory, using a CW radar, in a full polarimetric configuration, working on a slide plotter, in order to simulate the single-pass condition. Three kinds of different environments has been designed. The first is a point scatterer environment, the second is a distributed scatterers environment and the last one represents an hybrid environment constituted by point and distributed scatterers. The reason for this choice was to include all the circumstances that may be encountered when performing SAR tomography. The used optimization method was based on the CS, focusing our expectations on a particular and very powerful solver based on Interior Point Methods (IPM). The results are very promising because they provide to appreciate detailed properties of the principal polarimetric electromagnetic parameters extended along the vertical direction.
Relating Rice Ear Biomass to SAR Data
Liu, Long (1,2); Shao, Yun (1); Li, Kun (1); Yang, Zhi (1,2); Gong, Huaze (1); Xie, Chou (1); Chai, Xun (1,2) 1: The Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences,China; 2: University of Chinese Academy of Sciences, China
Show abstractEar biomass is the most relevant portion to grain yield in the whole above ground biomass of rice.A novel method which aimed at estimating ear biomass from SAR data was proposed. A simplified Monte-Carlo scattering model of ear layer related ear geometrical and physical parameters to SAR observations directly and a complete inversion procedure of ear biomass inversion was presented afterwards. Promising result was acquired by conducting the inversion procedure in ground measurements.
ROLE OF VEGETATION PHENOLOGY (LEAF-ON, LEAF-OFF) IN THE ACCURACY OF FOREST HEIGHT MAPS DERIVED FROM TANDEM-X-HH INTERFEROGRAMS
Sadeghi, Yaser (1); St-Onge, Benoît (1); Leblon, Brigitte (2); Simard, Marc (3); Papathanassiou, Kostas (4) 1: University of Quebec, Canada; 2: University of New Brunswick, Canada; 3: JPL, USA; 4: DLR, Germany
Show abstractThis study explores the sensitivity of TanDEM-X interferograms to tree phenology in the context of global mapping of forest height or biomass of the boreal and temperate forests. At such scales, the available images for a given site may not always be optimal in terms of the conditions of their acquisition. For example, deciduous trees may be without leafs, partially leaved in spring, or senescent in fall (in the northern hemisphere) depending on the season of acquisition. Moreover, in winter, conifer crowns can be covered by snow. Because SAR penetration may be affected by these phenological variations, a greater uncertainty in the height or biomass estimates could occur. Our goal was therefore to assess the influence of these conditions on the InSAR TANDEM-X coherence image, and on the accuracy of the digital surface models (DSMs) obtained from the interferogram. Two dual polarimetric (HH and VV) datasets, one for the leaf-on and one for the leaf-off period, were used to analyze the effect of leaf conditions on the TanDEM-X coherence and the error in the resulting TanDEM-X DSMs. The InSAR coherence image was first computed from each interferogram. Because the coherence images for both polarizations at each season was found to be almost the same, polarization effects were not assessed, and only the HH interferograms were used in the study. Moreover, we have masked pixels with coherence values less than 0.6 because they produced errors due to low coherence because of phase unwrapping blunders during the extraction of the InSAR DSMs. The resulting InSAR DSM was used to compute a canopy height model (CHM) by subtraction of a lidar DTM. These hybrid CHMs were generated at a 5x5 m cell size and compared to the corresponding lidar-only CHM acquired in leaf-on conditions. The accuracy of the hybrid TanDEM-X CHMs was quantified by calculating the bias, the root mean square error (RMSE), and the coefficient of determination r2 relative to the corresponding lidar-only CHM. For leaf-on conditions, a bias of 1.4 m, a RMSE of 1.9 m, and a r2 of 0.75 were obtained, while for leaf-off condition, these values were 1.7 m, 2.4 m and 0.70, respectively. Since the study region is mostly covered by coniferous trees, the greater penetration in leafless deciduous trees has a limited overall impact, but is more accentuated in areas with higher deciduous component. The coherence values in leaf-off condition were also found to be lower than those of leaf-on conditions. For deciduous or mixed forests with a large deciduous component, the lower coherence in leaf-off condition is related to the increase of volume decorrelation, because of the deep penetration of the SAR signal into barren crowns compared to leaf-on crowns. We thereby conclude that the TanDEM-X coherence image and the related computed DSM are influenced by seasonal forest phenology or seasonal variations. This effect will have to be taken into account when producing global maps of height or biomass of biomes in which deciduous trees are present if only leaf-off acquisitions are available.
Forest biomass estimation by the use of RADARSAT-2 data in Ocnele Mari mining area
Poenaru, Violeta Domnica; Badea, Alexandru; Cimpeanu, Sorin Mihai
Show abstractThe paper investigates the potential of C-band dual polarization RADARSAT-2 imagery for forest biomass estimation in a hilly region covered with vegetation and affected by land degradation. The Ocnele Mari area poses scattering problems: due to layover effects the RADARSAT-2 data could contain the corrupted height information which can be adjusted by a DEM or using different combinations of raw backscattering, polarization texture parameters and texture polarization ratios. A set of 4 data images acquired in the joint ESA-CSA SOAR Europe-16605 scientific proposal framework covering July to September 2014 period is analyzed. The main problem is to preserve the coherence between RADARSAT-2 satellite overpasses. A both coherent and non-coherent multi-temporal analysis will be carried out for forest assessment. The results will confirm the ESA’s BIOMASS sensor data capabilities to estimate more accurately the trees height and forest biomass.
Towards operational near real time oil spill detection service using polarimetric TerraSAR-X images
Singha, Suman; Velotto, Domenico; Lehner, Susanne German Aerospace Center (DLR), Germany
Show abstractAbstract: Exploitation of polarimetric features for oil spill detection is relatively new and until recently those properties have not been used for operational services. European flagship operational Near Real Time (NRT) oil spill detection service, ‘CleanSeaNet’ (hosted by European Maritime Safety Agency, EMSA) shows regular occurrence of accidental and deliberate oil spill over European waters especially over offshore installations. ‘CleanSeaNet’ currently uses manual image interpretation technique on single pol (mainly VV polarized) multi looked ground range ‘ScanSAR’ images in order to report oil spill to responsible authorities in member states. In last decade, a number of semi-automatic and automatic techniques have been proposed in order to differentiate oil spill and look alike spots based on single pol SAR data, however these techniques suffers from high miss classification rate which is completely undesirable for operational services like ‘CleanSeaNet’. In addition to that small operational spillages from offshore platforms are often ignored as it appears insignificant on traditional ‘ScanSAR’ images which affect the decision making process of manual classification. In order to mitigate this situation and anticipating regular and large amount of data from ESA’s Sentinal-1 mission (under Copernicus Service), a major focus of research in this area is the development of automated algorithms based on polarimetric images to distinguish oil spills from ‘look-alikes’ complementing the visual analysis carried out by current operational services. This paper describes the development of a semi-automatic approach for oil spill detection from dual-polarimetric TerraSAR-X images exploiting polarimetric features (e.g. Co-Polarised Phase Difference, Entropy etc). Numbers of feature parameters were extracted; include both single polarimetric and dual polarimetric based features from each segmented dark spot which are then used for training and validation of the decision tree classifier. Initial performance estimation was carried out for the proposed methodology in order to evaluate its suitability for operational service. Index Terms — Polarimetric Oil spill detection, Co-Polarized Phase Difference, Near Real Time Services,
Contribution to aboveground biomass estimations in tropical forests: Linking LiDAR data and forest simulations
Fischer, Rico; Huth, Andreas Helmholtz Centre for Environmental Research - UFZ Leipzig, Germany
Show abstractDisturbances and the natural spatial heterogeneity of tropical forests make it difficult to determine aboveground biomass with an adequate accuracy. LiDAR is currently the most accurate remote sensing method to estimate forest aboveground biomass, e.g. for REDD+ projects. Recent studies have used forest inventory data to evaluate several metrics to estimate forest biomass from LiDAR canopy height profiles. These validations are restricted to selected forest plots, especially in tropical forests. In this study, different metrics to estimate aboveground biomass were evaluated and compared using virtual forest stands generated by a well-tested individual-based forest model (FORMIND). Therefore we present a new approach for modeling discrete LiDAR returns for forest stands. The simulated LiDAR data will also be validated with real airborne LiDAR data from a larger tropical moist forest (50 ha plot of Barro Colorado Island). Additionally the influence of different disturbance regimes on biomass estimations of forests was investigated. It turned out that simple regression models can provide good predictions of aboveground biomass of tropical forests on hectare scale over a wide range of successional stages and disturbance intensities. In future investigations forest models could be used for ground truthing for remote sensing in forestry and play an important role for balancing the global carbon budget in higher accuracy than today.
Time-frequency polarimetric analyses of maritime environments in high resolution SAR data
Banda, Francesco (1); Ferro-Famil, Laurent (2); Tebaldini, Stefano (1) 1: Politecnico di Milano, Italy; 2: Université de Rennes 1, France
Show abstractSynthetic Aperture Radar (SAR) systems represent a major tool for maritime applications nowadays . Moreover, especially recent research has proved that both polarimetry and high resolution bring improvements in terms of ship detection, identification of the ambiguities and imaging [2,3]. Due to the coherent nature of SAR, the estimation of the velocity of ships is also possible(e.g. ). Very often the analysis of SAR data is based on the assumption that targets are observed from the look direction of the antenna and at the fixed frequency of the signal carrier. Complex targets like man-made objects may exhibit a variable behavior though, both with respect to azimuth angle due to target anisotropy and motion within the integration time and within the illuminating bandwidth of frequencies [5,6]. The assumption of isotropic scattering often made in SAR data analysis may thus lead to erroneous interpretations over such targets. In this work preliminary results from a phenomenological analysis carried out on high resolution Synthetic Aperture Radar data of maritime environments is presented. Time-frequency analysis tools [5,7,8] have been applied in order to detect and characterize the different scattering mechanisms contributing to the received signal. The impact and correction of data perturbations due to the motion of vessels are also discussed. References  P. Vachon, “Ship detection in Synthetic Aperture Radar Imagery,” in OceanSAR, 2006.  C. Brekke and S.N. Anfinsen, “Ship Detection in Ice- Infested Waters Based on Dual-Polarization SAR Imagery,” Geoscience and Remote Sensing Letters, IEEE, vol. 8, no. 3, pp. 391–395, 2011.  M. Martorella, F. Berizzi, D. Pastina, and P. Lombardo, “Exploitation of Cosmo SkyMed SAR images for maritime traffic surveillance,” in Radar Conference (RADAR), 2011 IEEE, 2011, pp. 113–117.  S. Brusch, S. Lehner, T. Fritz, M. Soccorsi, A. Soloviev, and B. van Schie, “Ship Surveillance With TerraSAR-X,” Geoscience and Remote Sensing, IEEE Transactions on, vol. 49, no. 3, pp. 1092–1103, 2011.  L. Ferro-Famil and E. Pottier, “Urban area remote sensing from L-band PolSAR data using Time-Frequency techniques,” in Urban Remote Sensing Joint Event, 2007, 2007, pp. 1–6.  S. Hinz, F. Meyer, A. Laika, and R. Bamler, “Spaceborne Traffic Monitoring with Dual Channel Synthetic Aperture Radar Theory and Experiments,” in Computer Vision and Pattern Recognition - Workshops, 2005. CVPR Workshops. IEEE Computer Society Conference on, 2005, pp. 7–7.  L. Ferro-Famil, A. Reigber, and E. Pottier, “Nonstationary natural media analysis from polarimetric SAR data using a two-dimensional time-frequency decomposition approach,” Canadian Journal of Remote Sensing, vol. 31, no. 1, pp. 21–29, 2005.  Hu, Canbin; Ferro-Famil, Laurent; Kuang, Gangyao. 2013. "Ship Discrimination Using Polarimetric SAR Data and Coherent Time-Frequency Analysis." Remote Sens. 5, no. 12: 6899-6920.
Classification of Low Backscatter Ocean Regions Using Log-cumulants
Skrunes, Stine; Brekke, Camilla University of Tromsø - the Arctic University of Norway, Norway
Show abstractOil is released into the world’s oceans on a regular basis. Large scale accidents that take place during oil production and transportation receive much attention from the public and the media. However, large quantities of oil are also released during smaller but frequent operational discharges from ships, that are intentional and often illegal. Satellite synthetic aperture radar (SAR) is an effective tool for detection of oil spills. A recent review on oil spill observation by SAR can be found in . One of the main challenges for oil spill detection by SAR is natural phenomena that produce similar SAR signatures as oil spills. These are called look-alikes and include low wind regions, grease ice, rain cells and biogenic surface slicks. Traditionally, features related to region geometry, characteristics of the backscatter levels, contextual information and spatial texture have been used to classify possible oil spills . Over the last decade, a potential for using multipolarization techniques for discrimination between oil spills and other phenomena has been demonstrated, see e.g., [3, 4, 5, 6, 7]. In  and , log-cumulants (defined in [10, translated in ] and ) were investigated for the purpose of oil versus look-alike discrimination. First results presented in  investigated the second and third order log-cumulants calculated from single-look VV intensity data. Promising results for discriminating mineral oil from simulated biogenic slick and a natural phenomenon, particularly using the second order log-cumulant, were found for both Radarsat-2 and TerraSAR-X data. The study was expanded to the multipolarization case in . In that paper, log-cumulants of first and second order were extracted from multilooked dual-copolarization data (HH/VV), and a separation between mineral oil slicks and other low backscatter phenomena was found in the log-cumulant space. The log-cumulants were normalized with respect to water to account for variation between scenes, e.g., due to varying incidence angle and wind conditions. The first and second order log-cumulants represent the mean and variance in the log-domain, respectively. The results in  indicated that mineral oils are characterized by low values of the first order log-cumulant and large values of the second order log-cumulant, compared to clean sea and other low backscatter regions. The former reflects the reduced backscatter from these areas compared to clean sea due to wave damping (and possibly a reduction in the dielectric constant). The large values of the second order log-cumulant may indicate a larger internal variation (more radar texture) in mineral oils compared to clean sea and look-alikes. Inhomogeneous distribution of oil may cause this behavior. In , a potential future application of the method for classification of low backscatter regions of unknown origin was suggested. However, the need for further investigations on a larger data set was emphasized. The objective of this paper is to further evaluate the potential for classification of low backscatter ocean regions by log-cumulants of first and second order. A larger data set will be analyzed and compared to the results from . In [8, 9], Radarsat-2 and TerraSAR-X data from oil-on-water exercises in the North Sea, containing experimental oil spills, simulated biogenic slicks, and natural low backscatter phenomena, were investigated. For this paper, a larger data set, also including actual oil spills, e.g., from the Deepwater Horizon accident in 2010, and natural seeps, will be analyzed. The effects of varying wind conditions and observation geometries (incidence angles) will also be addressed. References:  A. H. S. Solberg, “Remote sensing of ocean oil-spill pollution,” Proc. IEEE, vol. 100, no. 10, pp. 2931–2945, Oct. 2012.  C. Brekke and A. H. S. Solberg, “Oil spill detection by satellite remote sensing,” Remote Sens. Environ., vol. 95, pp. 1–13, 2005.  F. Nunziata, A. Gambardella, and M. Migliaccio, “On the Mueller scattering matrix for SAR sea oil slick observation,” IEEE Geosci. Remote Sens. Lett., vol. 5, no. 4, pp. 691–695, Oct. 2008.  M. Migliaccio, F. Nunziata, and A. Gambardella, “On the co-polarized phase difference for oil spill observation,” Int. J. Remote Sens., vol. 30, no. 6, pp. 1587–1602, Mar. 2009.  D. Velotto, M. Migliaccio, F. Nunziata, and S. Lehner, “Dual-polarized TerraSAR-X data for oil-spill observation,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 12, pp. 4751–4762, Dec. 2011.  F. Nunziata, M. Migliaccio, and A. Gambardella, “Pedestal height for sea oil slick observation,” IET Radar, Sonar Navig., vol. 5, no. 2, pp. 103 –110, Feb. 2011.  S. Skrunes, C. Brekke, and T. Eltoft, “Characterization of marine surface slicks by Radarsat-2 multipolarization features,” IEEE Trans. Geosci. Remote Sens., vol. 52, no. 9, pp. 5302–5319, Sept. 2014.  S. Skrunes, C. Brekke, T. Eltoft, and V. Kudryavtsev, “Comparing near coincident C- and X-band SAR acquisitions of marine oil spills,” IEEE Trans. Geosci. Remote Sens., Accepted for publication, 2014.  S. Skrunes, C. Brekke, and A. P. Doulgeris, “Characterization of low backscatter ocean features in dual-copolarization SAR using log-cumulants,” IEEE Geosci. Remote Sens. Lett., Accepted for publication, 2014.  J.-M. Nicolas, “Introduction aux statistiques de deuxie`me espe`ce: applications des logs-moments et des logs-cumulants a`l’analyse des lois d’images radar,” Traitement du Signal, vol. 19, no. 3, pp. 139–167, 2002.  J.-M. Nicolas and S. N. Anfinsen, “Introduction to second kind statistics: Application of log-moments and log- cumulants to the analysis of radar image distributions,” Technical note, translation from French of . URL: http://eo.uit.no/publications/JMN-TRANS-12.pdf, Visited 30. Jun. 2014, May 2012.  S.N.AnfinsenandT.Eltoft,“Applicationofthematrix-variateMellintransformtoanalysisofpolarimetricradarimages,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 6, pp. 2281–2295, Jun. 2011.
Comparing object-based landslide detection methods based on polarimetric SAR and optical satellite imagery – a case study in Taiwan
Plank, Simon (1); Eisank, Clemens (2); Hölbling, Daniel (2); Friedl, Barbara (2); Martinis, Sandro (1); Twele, André (1) 1: German Aerospace Center (DLR), Germany; 2: Interfaculty Department of Geoinformatics - Z_GIS, University of Salzburg, Austria
Show abstractApplied to a test site located in southern Taiwan, this study compares two object-based image analysis (OBIA) methods for post-failure landslide detection based on (I) polarimetric synthetic aperture radar (PolSAR) and (II) optical satellite imagery. With its day-and-night availability and almost complete weather independency, SAR has several advantages compared to optical imagery. Consequently, in most cases, SAR imagery for a dedicated area of interest is earlier available than the first cloud-free optical data. However, the high spatial and spectral resolution of multispectral optical Earth observation data may enable a more detailed and accurate landslide detection. A comparison of both methods is feasible, as the polarimetric SAR image (dual-polarimetric (HH/HV) TerraSAR-X, StripMap) and the very high spatial resolution optical imagery (QuickBird) were acquired at a temporal baseline of only 20 days. Thus, SAR and optical data show the same state of the environment. After speckle filtering of the PolSAR data using the refined Lee filter and radiometric calibration, the intensity information of both polarization channels (HH and HV) was geocoded. Next, the features of interest (i.e. landslides, debris flows and riverbed) are derived from the PolSAR imagery using a newly developed OBIA procedure, which makes use of the different backscattering behavior of forest and other vegetated areas as compared to bare soil, which was assumed to be an indication for the occurrence of mass-movements and debris/sediment transport and deposition areas. Using the normalized difference standard deviation of the calibrated intensities of both polarimetric channels, HH and HV, the OBIA procedure considers (I) the higher variation of the backscattering intensities in forest areas and (II) the relatively higher backscattering of vegetated areas in the cross-polarized channel compared to bare soil areas. The latter is characterized by a more dominant backscattering in the co-polarized channel. For the object-based detection based on the QuickBird image the Normalized Difference Vegetation Index (NDVI) was applied to detect the unvegetated areas. Based on the NDVI layer an automated threshold was computed, which divides the image into two subsets, i.e. vegetated and non-vegetated areas. To produce suitable image objects for the classification, the areas potentially affected by mass-movements were re-segmented using the multiresolution segmentation algorithm implemented in the eCognition (Trimble) software. Additionally to the QuickBird image a digital elevation model (DEM) with 5 m spatial resolution was used to support the differentiation of classes. As the spectral information alone was not sufficient for class separation, the distinction into landslides, debris flows and riverbed was mainly based on slope values. Finally, a few rules considering spatial and contextual properties were introduced to refine the classification and to remove false positives (e.g. built-up areas and fields). The reference dataset for validation includes vector data of landslides, debris flows and the river bed and was produced through manual digitization, performed by a local expert. Both results are compared to the reference data set and the pros and cons of the imagery utilized for landslide detection are evaluated. The result of the novel object-based method based on PolSAR data reveals a certain potential for landslide detection, especially for rapid assessment of affected areas after landslide triggering events.
Quantitative environmental assessment of oil spill pollution around Oil Rocks Settlement, Chilov and Pirallahi Islands in the Caspian sea using multi-temporal ENVISAT SAR images 2006-2010
Bayramov, Emil Dresden University of Technology, Germany
Show abstractThe main goal of this research was to detect oil spills, to determine the oil spill frequencies and to approximate oil leak sources around the Oil Rocks Settlement, the Chilov and Pirallahi Islands in the Caspian Sea using 134 multi-temporal ENVISAT Advanced Synthetic Aperture Radar Wide Swath Medium Resolution Images acquired during 2006-2010. The following oil spill frequencies were observed around the Oil Rocks Settlement, the Chilov and Pirallahi Islands: 2-10 (3471.04 sq. km.), 11-20 (971.66 sq. km.), 21-50 (692.44 sq. km.), 51-128 (191.38 sq. km.). The most critical oil leak sources with the frequency range of 41-128 were observed at the Oil Rocks Settlement. The exponential regression analysis between wind speeds and oil slick areas detected from 134 multi-temporal ENVISAT images revealed the regression coefficient equal to 63%. The regression model showed that larger oil spill areas were observed with decreasing wind speeds. The spatiotemporal patterns of currents in the Caspian Sea explained the multi-directional spatial distribution of oil spills around Oil Rocks Settlement, the Chilov and Pirallahi Islands. The linear regression analysis between detected oil spill frequencies and predicted oil contamination probability by the stochastic model showed the positive trend with the regression coefficient of 30%. ACKNOWLEDGMENT I would like to acknowledge European Space Agency (ESA) for the provision of access to ENVISAT images under the project ID: 15837.
Modelling carbon fluxes across global forest eddy covariance observations by a new individual-based carbon budget model FORCCHN
Ma, Jianyong; Yan, Xiaodong State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University, Beijing, 100875，China
Show abstractCarbon budget of forest ecosystems, as important components of the terrestrial carbon cycle, always be accurately quantified by ecological models for detecting the mitigation of climate change globally. In this study, based on a new individual-based carbon budget model FORCCHN, we evaluated the modelling of carbon fluxes against observations measured by 37 global forest eddy covariance sites covering 5 continents: Europe, Asia, North America, South America and Oceania for the first time through using the means, the linear regression parameters slope and coefficient of determination, the Pearson correlation coefficient, the root mean square error and the D-index. The results showed that our model successfully captured the seasonal variation as well as the magnitude of gross primary productivity (GPP), ecosystem respiration(ER) and net ecosystem productivity (NEP) across all eddy covariance samples with exception of several Mediterranean sites. Under Mediterranean climate, part of the variability in carbon fluxes that could not described by our model may be attributed to a series of factors, including inappropriate physiological and ecological parameters for representing the phenology in natural-forest and forest management practice for alleviating the water stress in artificial-forest. In summary, it is reasonable to apply FORCCHN to estimate carbon budget of forest ecosystems on large scale, particularly at continental or global levels. Keywords Individual-based model; FORCCHN; Gross primary productivity; Ecosystem respiration; Net ecosystem productivity; Forest ecosystems; Eddy covariance
Building Area Extraction From Polarimetric SAR Data via Stationarity Detection and Circular-Pol Correlation Coefficient
Xiang, Deliang; Su, Yi; Ban, Yifang
Show abstractAlong with the polarimetric Synthetic Aperture Radar (PolSAR) data acquired with airborne and spaceborne sensors such as Pi-SAR, ESAR, RADARSAT-2, etc., building area extraction from PolSAR images has been one of the significant applications in urban remote sensing. Since the buildings have complex geometries and may be misclassified as forests or mountains with volume scattering due to the significant cross-pol backscatter and lack reflection symmetry, especially the slant-oriented buildings, building area extraction is a challenging problem. In this study, the time-frequency decomposition technique is adopted to acquire subaperture images, which correspond to the same scene responses under different azimuthal look angles. Stationarity detection approach with polarimetric distribution is proposed to extract ortho-oriented buildings and the circular polarization correlation coefficient is optimal in characterizing slant-oriented buildings. We test the aforementioned method on two PolSAR datasets, i.e. ESAR image with L-band and RADARSAT-2 image with C-band. The detection results are compared with SVM supervised classifier and conventional entropy/alpha-Wishart classifier, as well as the high resolution optical imageries. The results demonstrate that the proposed method can effectively extract both ortho-oriented and slant-oriented buildings and the overall detection accuracy as well as kappa value is 10%-20% higher than the compared methods.
Multi-polarization C- and X-band SAR data for coastline extraction
Nunziata, Ferdinando; Buono, Andrea; Migliaccio, Maurizio Università di Napoli Parthenope, Italy
Show abstractThe Italian peninsula with Sicily and Sardinia and its many minor islands, is one of the most important European coastal countries with a coastline of 7375 km and it is the sixth European region in littoral extension. Nowadays, the Italian coastal areas suffer the negative consequences of a poor coastal management in which private and personal business has prevailed on the common good. Highly urbanized and infrastructured Italian coastal areas are strongly affected by the effects induced by phenomena related to climate change (e.g. sea level rise, coastal retreat and flooding) and natural and anthropogenic land subsidence (e.g. Venice and Po delta among the others). Among them, coastal erosion and retreat is a key problem, which is accelerated by large storms, flooding, strong wave action, sea level rise, and human activities (e.g. inappropriate land use, alterations, and shore protection structures). Within this context, Monasterace marina, a municipality in the Province of Reggio Calabria in the Italian region Calabria located about 50 km south of Catanzaro and about 90 km northeast of Reggio Calabria along the Ionian Sea, is strongly affected by coastal erosion and retreat. This area is o paramount importance even from a cultural viewpoint. In fact, the ruins of the ancient Greek city of Kaulonia are located a short distance north, along the coast. Kaulonia is an ancient Greek settlement (~700 to 389 B.C.), positioned in a tectonically mobile setting. Extensive shoreline retreats have occurred along this coast since the mid-Holocene. Rapid modifications of the shoreline are occurring since 1896 with an abrupt acceleration since the late 1970’s. The erosion process, which results from tectonics and hydrogeological changes, is significantly increased by human activity mainly in the urban area, where an average shoreline loss of 6m per year has occurred since 1994. In this study, co-located multi-polarization Synthetic Aperture Radar (SAR) measurements collected in the frame of the SOAR-ASI and SOAR-EI projects over the area of Monasterace are analysed for coastline extraction purposes. The data set includes a fine quad-polarization C-band RadarSAT-2 scene and a dual-polarization PingPong (HH-HV) Cosmo-SkyMed SAR scene. In addition, ancillary GPS and drone-base surveys have been collected over the imaged area. The proposed rationale aims at comparing the performance of multi-polarization algorithms for coastline extraction purposes. Single-pol: The performance of like-polarized channels (HH or VV) will be contrasted with cross-polarized channel. This comparison will be undertaken both at C- and X-band. Dual-pol: The performance of the HH-HV/VV-VH correlation, measured from C-band SAR data, will be contrasted with the HH-HV correlation, measured from X-band incoherent PingPong CSK SAR data. Note that the CSK SAR, when acting through the PingPong mode, resembles an along-track interferometer (ATI); hence, different correlation is expected for sea and land areas. Quad-pol: Conventional quad-pol features, e.g. polarimetric entropy, will be contrasted with compact-polarimetric ones in view of the recently lunched (RISAT-1/ALOS-2) and the forthcoming (RCM) SAR missions.
Comparative Study of Speckle Filtering Methods in PolSAR Radar Images
Boutarfa, Souhila; BOUCHEMAKH, Lynda; SMARA, Youcef
Show abstractImages acquired by polarimetric SAR (PolSAR) radar systems are characterized by the presence of a noise called speckle. This noise has a multiplicative nature, corrupts both the amplitude and phase images, which complicates data interpretation, degrades segmentation performance and reduces the detectability of targets. Hence, the need to preprocess the images by adapted filtering methods before analysis. In this paper, we present a comparative study of implemented methods for reducing speckle in PolSAR images. These developed filters are: refined Lee filter based on the estimation of the minimum mean square error MMSE, improved Sigma filter with detection of strong scatterers based on the calculation of the coherency matrix to detect the different scatterers in order to preserve the polarization signature and maintain structures that are necessary for image interpretation, filtering by stationary wavelet transform SWT using multi-scale edge detection and the technique for improving the wavelet coefficients called SSC (sum of squared coefficients), and Turbo filter which is a combination between two complementary filters the refined Lee filter and the wavelet transform SWT. One filter can boost up the results of the other. The originality of our work is based on the application of these methods to several types of images: amplitude, intensity and complex, from a satellite or an airborne radar, and on the optimization of wavelet filtering by adding a parameter in the calculation of the threshold. This parameter will control the filtering effect and get a good compromise between smoothing homogeneous areas and preserving linear structures. The methods are applied to the fully polarimetric RADARSAT-2 images (HH, HV, VH, VV) acquired on Algiers, Algeria, in C-band and to the three polarimetric E-SAR images (HH, HV, VV) acquired on Oberpfaffenhofen area located in Munich, Germany, in P-band. To evaluate the performance of each filter, we used the following criteria: smoothing homogeneous areas, preserving edges and polarimetric information. Experimental results are included to illustrate the different implemented methods. Keywords: PolSAR Image, Speckle, Lee filter, Turbo filter, Sigma filter, SWT.
A Ku band polarimetric-interferometric ground based real aperture radar: calibration and first observations
Baffelli, Simone; Marino, Armando; Othmar, Frey; Charles, Werner; Irena, Hajnsek
Show abstractIn this paper, we present the pol-GBRAR a new fully-polarimetric and interferometric ground based real aperture radar which is based on Gamma Remote Sensing's GPRI-II (Gamma Portable Radar Interferometer). The radar is designed to operate in the Ku band, with a central frequency of 17.2 GHz. The system supports fully polarimetric imaging at each end of a interferometric baseline: this has been obtained by the addition of horizontally polarized antennas and switching hardware in the RF frontend of the original GPRI-II system. The radar is based on a dechirp-on-receive FMCW architecture with 200 MHz of bandwidth, giving a ranging resolution of 0.75 m. The cross-range resolution is obtained with a large antenna with an azimuth beamwidth of approx. 0.4 degrees. Due to the mechanical construction of the system, there is a small bistatic angle between the transmission and the reception antennas and also between the H- and the V- polarized antennas. The polarimetric calibration is complicated by the presence of this angle, as the effective antenna phase centers have a different location for each polarimetric channel. The displaced phase center behave as an interferometer, adding a topographic contribution to the polarimetric phases. To correct this effect, a method making use of the interferometric capabilites of the GPRI-II has been devised. This method could also be adapted to the use of an external DEM. After the removal of the topographic phase, the radar was calibrated for polarimetry using trihedal reflectors and natural targets according to the method developed by Van Zyl. The copolarized channels phase and amplitude imbalance were estimated over a trihedral reflector, while the crosspolarized channel imbalance was reduced by using a large number of distributed targets under the assumption of reciprocity. For the calibration procedure, the polarimetric channel crosstalk could be neglected because of the very low level that was measured in the receiving circuitry (approx -40 dB) and because the polarimetric acquisition is performed using temporal multiplexing, in which only one combination of TX- and RX- is active at the same time, reducing the crosstalk at the antennas. Potential uses for this systems are the monitoring of unstable slopes by zero-baseline differential interferometry, the observation of snow and ice and the extraction of soil and vegetation parameters for agricultural applications. Due to its portability and flexibility, the radar allows experimenting with a wide variety of targets, acquisitions modalities and time interval, making it a precious insturment for the development of multitemporal, polarimetric and interferometric processing techniques. As an example of potential applications, some preliminary polarimetric analysis over a test area have been performed and will be presented, showing the polarimetric response of agricultural fields and of built up areas at Ku band. As second dataset, acquired on the Bisgletscher, a fast moving hanging glacier in the Swiss canton of Wallis, will be presented as well, showing the evolution of the coherency for different polarizations over the course of one day of acquisitions.
Testing different modules combinations of a classifier based on the geometrical perturbation filter.
Marino, Armando; Hajnsek, Irena
Show abstractLand classification is an important subject in remote sensing. This is particularly true in the recent years due to the current availability of long time series of SAR data. Polarimetry has the potentials to improve significantly the detection capability due to the possibility to discriminate between different polarimetric targets. This paper is focused on a methodology proposed by the authors  and named geometrical perturbation filter (GPF). One advantage of such methodology is that the GPF distance is independent of the overall intensity (OI) of the polarimetric acquisition (i.e. Trace of the covariance matrix). This is particularly important when different targets can present the same OI depending on the geometry of observation (e.g. presence of slopes). On the other hand, it can be observed that in some situations the OI is only linked to physical characteristic of the target (e.g. in very flat areas). In such situations, the OI would be expected to improve the separation between targets. Additionally, two different completely depolarised targets, cannot be discriminated by the polarimetric information since both can be expressed as an identity covariance matrix multiplied by a factor (i.e. the OI). Two examples of such targets are a cloud of small spheres and thermal noise. In order to include the information kept in the OI, the authors proposed a new version of the GPF where the final classifier is built with two modules: one module is independent of the OI and the other is exclusively dependent on the OI . This paper deals with the test and comparison of several methodologies to combine the two modules. This is performed following two approaches. In the first, Monte Carlo simulations are considered in order to test the algorithms in an ideal scenario. The second exploits real data to verify the results obtained with the simulations. E-SAR AGRISAR2006 and ICESAR2007 L-band quad polarimetric data are employed in this analysis. References:  A. Marino, S. Cloude, and I. H. Woodhouse, “Detecting depolarizing targets using a new geometrical perturbation filter,” IEEE Transactions on Geosciance and Remote Sensing, vol. 50(10), pp. 3787–3799, 2012.  A. Marino, I. Hajnsek, “Including overall amplitude in the Geometrical Perturbation Filters classifier,” Proceedings on EUSAR2014.
Uncertainties in ground-based estimates of biomass carbon
Paul, Keryn; Roxburgh, Stephen CSIRO, Australia
Show abstractMeasurements for estimation of biomass carbon in forests vary greatly, and many of these have high uncertainty. It is important to understand these uncertainties and, if possible, address these as further investment in forestry carbon projects may be limited without building confidence in biomass estimates. We undertook an assessment of key factors leading to the propagation of errors in biomass carbon estimates in a range of planted forests, including mixed-species and monoculture plantings. To provide comprehensive information on spatial biomass variability we mapped and measured all individuals across a range of contrasting case study forest sites, and within these sites entire plots were fully harvested in order to obtain a ‘true’ measure of biomass carbon above- and below-ground at those locations. A range of different indirect methods were then tested to estimate biomass carbon; (i) various types of spatial sampling designs, (ii) application of various types of generality in above- and below-ground allometric equations, and (iii) application of expansion factors. For these various indirect methods at each case study site, a Monte Carlo analysis was used to assess the uncertainty in biomass carbon estimates through the propagation of key measurement errors. A sensitivity analysis showed the relative importance of these errors on site-based biomass carbon estimates. Those assessed included; (i) sampling design for forest inventories, (ii) sampling design for biomass harvesting, (iii) sampling design for excavation of roots, (iv) stem measurements (i.e. stem diameters or height), (v) moisture content corrections, and (vi) application of allometric equations or expansion factors to inventory datasets. Identifying the relative sensitivity of biomass carbon estimates to these various steps in measurements and calculations facilitates better utilization of resources to ensure the highest degree of accuracy in these estimates possible. Here we found that the sampling design for forest inventories was the main factor contributing to uncertainty in biomass carbon estimates, with application of allometric equations or expansion factors contributing much less to uncertainty. We conclude that confidence in many forest biomass carbon estimates could be greatly improved through the application of more efficient sampling designs which are targeted for the type of planting in question.
Multipolarization and Multifrequency SAR data to observe agricultural crops
Mascolo, Lucio; Lopez-Sanchez, Juan Manuel; D'Urso, Guido; Nunziata, Ferdinando; Migliaccio, Maurizio; Mazzarella, Giuseppe
Show abstractThe capabilities of Synthetic Aperture Radar (SAR) to provide fine spatial resolution and a wide area coverage, both in day and night time and almost under all weather conditions, make it a key tool for agricultural applications. In the special case of SAR polarimetry, such applications are based on the known sensitivity of polarization of microwaves to crop structure (size, shape, and orientation of leaves, stalks, and fruits), dielectric properties of the canopy (related to the water content), and the physical properties of the underlying soil (roughness and moisture). Crop structure and plant water content vary as a function of crop type, growth stage and crop condition. In this study, X-band COSMO-SkyMed (CSK) incoherent dual-polarimetric SAR data and C-band RadarSAT-2 (RS2) polarimetric SAR (PolSAR) measurements are used to monitor agricultural crops. This allows to exploit the synergy between CSK and RS2 polarimetric modes by taking full benefit of their different characteristics in terms of operating frequency and polarization capabilities. Regarding to the incoherent dual-pol analysis, a time series of CSK Level 1 A Single Look Complex (SLC) PingPong mode VV/VH data is analyzed. In particular the behaviours of the backscattering coefficients and the polarization ratio as a function of the crop types, growth stages of crop and crop condition are investigated. With respect to the coherent analysis, a time series of SLC RS2 Fine quad-pol SAR data is used. In this case, the evolution of polarimetric observables (backscatter coefficients and ratios, correlations and phase differences, decomposition outputs, etc.) is analyzed. Within this framework, recent studies - have shown the potential of retrieving crop phenology by means PolSAR methodologies. Both CSK and RS2 data are collected over agricultural sites present in the Piana del Sele zone, which is located in the province of Salerno, South Italy, where ground truth information regarding the agricultural crops are provided.  J. M. Lopez-Sanchez, J. D. Ballester. “Potentials of Polarimetric SAR Interferometry for Agriculture Monitoring”, Radio Science, Vol. 44, RS2010, March 2009.  J. M. Lopez-Sanchez, S. R. Cloude, J. D. Ballester. “Rice Phenology Monitoring by Means of SAR Polarimetry at X-Band”, IEEE Trans. Geosci. Remote Sens., Vol. 50, No. 7, pp. 2695-2709, 2012.  J. M. Lopez-Sanchez, F. Vicente-Guijalba, J. D. Ballester, S. R. Cloude. “Polarimetric Response of Rice Fields at C-Band: Analysis and Phenology Retrieval”, IEEE Trans. Geosci. Remote Sens., Vol. 50, No. 5, pp. 2977-2993, 2014.
Use of pan-tropical biomass maps and other forest structural characteristics to derive biomass emission estimates for the Amazon biome
Langner, Andreas (1); Shimabukuro, Yosio (1,2); Achard, Frédéric (1); Mitchard, Edward (3) 1: European Commission, Joint Research Centre, Institute for Environment and Sustainability, Via E. Fermi, 2749, I-21027 Ispra (VA), Italy; 2: Instituto Nacional de Pesquisas Espaciais (INPE), São José dos Campos (SP), Brazil; 3: School of GeoSciences, University of Edinburgh, Crew Building, Edinburgh, EH9 3JN, UK
Show abstractA recent study by Langner et al (submitted) shows that the pan-tropical biomass maps of Saatchi et al. (2011) and Baccini et al. (2012) can be used to derive more realistic Tier 1 above-ground biomass (AGB) values with lower uncertainties than the hitherto existing default values from the IPCC guidelines (IPCC 2006). For that sake a composite map was created to produce average values per ecological zone and continent. Building on these results, this study investigates the impact of deforestation and biomass spatial datasets on estimates of carbon emission for the Brazilian Amazon. We use a spatially explicit approach, which is more robust than a Tier 1 approach as it takes better into account the geospatial pattern of AGB in natural tropical forests. For deforestation datasets (‘activity data’) we consider three spatially explicit different sources: (1) Achard et al. (2014) for period 2000 to 2010, (2) Hansen et al. (2013) for period 2000 to 2012 and (3) INPE (2014) that provides annual forest losses between 2000 and 2013. For biomass maps (used to derive ‘emission factors’) we use (i) Saatchi map, (ii) Baccini map, (iii) carbon map of Brazil (MCT 2010), (iv) average composite of Saatchi and Baccini maps, and (v) R2-weighted composite of Saatchi and Baccini maps. While the average composite assumes the same accuracy of Saatchi and Baccini maps, the R2-weighted composite analyzes the correlation coefficients between these two AGB maps and other maps of forest structural characteristics, which also relate to AGB: tree height map (Simard et al. 2011) and wood-density and basal area maps (Mitchard et al. 2014). Based on the relative value of the resulting pixel-based correlation coefficients with the Saatchi or Baccini maps a R2-weighted composite is produced, possibly leading to lower uncertainties. A comparison of the five biomass maps with field measurements from the literature (Mitchard et al. 2014) provides information on their uncertainties. The three deforestation datasets are finally combined with the five AGB maps for the Brazilian Amazon biome. The derived carbon emission estimates are compared to the forest reference emission level for Brazilian Amazonia submitted by Brazil to UNFCCC (Brazil 2014) and to other recent estimates (Ometto et al. 2014). We conclude by showing the potential impact of the use of such alternative AGB maps on the forest reference emission level (FREL) of Brazil. References: Achard F et al., 2014. Determination of tropical deforestation rates and related carbon losses from 1990 to 2010. Global Change Biology, 20, 2540–2554 Baccini A et al., 2012. Estimated carbon dioxide emissions from tropical deforestation improved by carbon-density maps. Nat. Climate Change, 2, 182-185 Brazil, 2014. Brazil’s submission of a forest reference emission level for deforestation in the Amazonia biome for results-based payments for REDD+ under the UNFCCC – Report. http://unfccc.int/methods/redd/items/8414.php (accessed 09.2014) Hansen M C et al., 2013. High-Resolution Global Maps of 21st-Century Forest Cover Change. Science, 342, 850-853 INPE, 2014. http://www.obt.inpe.br/prodesdigital/cadastro.php (accessed 09.2014) IPCC (Intergovernmental Panel on Climate Change) 2006, 2006. IPCC Guidelines for National Greenhouse Gas Inventories, Volume 4 , Agriculture, Forestry and Other Land Use IPCC National Greenhouse Gas Inventories Programme, Technical Support Unit, ISBN 4-88788-032-4 Langner A, Achard F and Grassi G. Can recent pan-tropical biomass maps be used to derive alternative Tier 1 values for reporting REDD+ activities under UNFCCC? Environmental Research Letters, submitted at Environmental Research Letters MCT (Ministry of Science and Technology), 2010. Second National Communication of Brazil to the United Nations Framework Convention on Climate Change. Brasília 2 v. : il. col., map.; 30 cm. + 1 CD-ROM (4 3/4 in.) Mitchard E T A et al., 2014. Markedly divergent estimates of Amazon forest carbon density from ground plots and satellites. Global Ecology and Biogeography, DOI 10.1111/geb.12168 Ometto J P et al., 2014. Amazon forest biomass density maps: tackling the uncertainty in carbon emission estimates. Climatic Change, 124, 545–560, DOI 10.1007/s10584-014-1058-7 Simard M, Naiara P, Fisher J B and Baccini A, 2011. Mapping forest canopy height globally with spaceborne lidar. J. Geophys. Res., 116, G04021, DOI 10.1029/2011JG001708
Towards a Global Biomass Data Repository
Schepaschenko, Dmitry; Fritz, Steffen; See, Linda; Perger, Christoph International Institute for Applied Systems Analysis, Laxenburg, Austria
Show abstractTerrestrial biomass is considered an essential indicator for the monitoring of the Earth’s ecosystems and their dynamics. In recent years, many regional and zonal biomass datasets have been produced. These were obtained using a wide range of methods which combine optical, radar and LiDAR remote sensing (RS) data with field measurements. However, existing biomass products quite often contradict one another when compared. To address this problem, the Biomass Geo-Wiki application was developed (Schepaschenko et al., 2012), which is one branch within the family of Geo-Wiki.org tools (Fritz et al., 2009, 2012). Biomass Geo-Wiki was created in order to bring different biomass datasets together in a single place so that they can be viewed and compared with high resolution imagery on Google Earth. Biomass Geo-Wiki provides users with an instant global overview of the available datasets in standard units. The datasets are grouped into 3 blocks: above ground live biomass, forest woody biomass and in-situ forest biomass measurements. They span multiple spatial scales from global to national to regional as well as down to the fine local scale in the form of plot level measurements. An example of a global above ground live biomass product displayed in Biomass Geo-Wiki is the product developed by Kindermann et al. (2008) in which the aggregated country level FAO FRA2010 statistics were downscaled to a 1 km2 spatially explicit global dataset containing forest growing stock; above/below ground biomass, dead wood and total forest biomass; and above ground, below ground, dead wood, litter and soil carbon. This product represents one of the first attempts to produce a globally consistent spatial database at a 1 km2 resolution containing forest growing stock, biomass and carbon stock values. Other above ground live biomass products are based on space band radar, e.g. the pan-boreal dataset by Thurner (2014); LiDAR, e.g. the pan-tropical datasets produced by Saatchi et al. (2011) and Baccini et al. (2012); and spectrometer imagery, e.g. the dataset for the USA produced in combination with ground data at the Woods Hole Research Center (Kellndorfer et al., 2012). The second category of datasets is forest woody biomass, where forest woody biomass (in Mg dry matter) or growing stock volume (GSV in m3) is estimated based on calibration of remote sensing data (Spectroradiometric, radar or LiDAR) with forest inventory data. An example of this kind of product is the recent pan-boreal estimate of forest GSV (Santoro et al., 2011). It is representative for the year 2010, has a spatial resolution of 0.01 degree and covers the boreal zone in North America, Europe and Asia. This product is based on Envisat ASAR ScanSAR backscatter measurements. The GSV has been retrieved using the BIOMASAR algorithm. A number of other products in this category can be viewed on Biomass Geo-Wiki. In addition to products derived from remote sensing, in-situ measurements are also available. This kind of data is very valuable because it represents ground truth. However, the data also have a number of associated weaknesses such as the wide variety of research teams and methods used to collect the data (from simple measurements of tree diameter with the application of allometric equations to full destructive methods of biomass measurement); the small size of typical sample plots compared to the size of a remotely sensed pixel; spatial and temporal inconsistencies; substantial gaps in representativeness of vast territories by ground measurements and regional biomass models; and problems with the lack of data sharing amongst those collecting the data. Biomass Geo-Wiki offers a sample plot collection for Northern Eurasia with around 3,500 measurements but another motivation behind Biomass Geo-Wiki is to act as a global repository for in-situ biomass measurements for use by the entire community, e.g. in calibration and validation of future products. It is hoped that the community will open up in the future and agree to share the information more freely. Moreover, smartphone applications are starting to appear that will allow ordinary citizens to make biomass measurements with their phones, especially with coming 3-D functions. Plans are underway to link data from these apps directly to Biomass Geo-Wiki, which creates new possibilities in gathering much larger quantities of in-situ data on biomass in the future. In addition to visualization, Biomass Geo-Wiki also allows users to compare the different datasets with one another and with imagery from Google Earth. A user can query different datasets at a single location, view the different biomass values at that location and then validate the measurements using the underlying Google Earth satellite imagery (where possible) or use any expert or local knowledge. Finally, Biomass Geo-Wiki was designed to be part of a harmonization process in which the different estimates of biomass would be integrated into a single, global hybrid dataset that would be more accurate than the individual products. These products can be integrated using any number of methods, e.g. geographically weighted regression, which has been used to create a hybrid forest product using crowdsourced data from Geo-Wiki (Schepaschenko et al., 2013). However, the method relies on sufficient quantities of in-situ data for training and validation. This will only be possible once the community begins to more openly share their in-situ data and/or when citizen involvement through crowdsourcing of biomass data is realized in the future. For now, Biomass Geo-wiki remains a valuable resource for the community. The ultimate goal is the harmonization of existing biomass datasets with land cover datasets and robust in situ data collection to produce a global hybrid biomass map.
A STUDY ON PoLinSAR COHERENCE BASED REGRESSION ANALYSIS OF FOREST BIOMASS (BARKOT RESERVE FOREST INDIA), USING RADARSAT-2 DATASETS.
Singh, Jenia; Kumar, Shashi; Kushwaha, SPS Indian Institute of Remote Sensing, ISRO Dehradun 248001, India
Show abstractForests cover 30% of the world’s land surface, and are home to around 90% of the world’s flora and fauna. They serve as one of the world’s largest carbon sinks, absorbing 2.4 million tonnes of CO2 each year and storing billions more in form of biomass. Around 6 million hectares of forest is lost or changed each year and as much as a fifth of global emissions are estimated to come from deforestation. Hence accurate estimation of forest biophysical variables is necessary as it is a key parameter in determination of forest inventories, vegetation modelling and global carbon cycle. Remote sensing techniques are capable of providing accurate and reliable information about forest parameters. Various Remote sensing approaches have been exploited to extract information about diverse bio-physical parameters associated with the forests. But these approaches suffer from multiple limitations like inability to penetrate through clouds, lack of information retrieval about surface roughness, dielectric properties and moisture. All these limitations can be overcome with the employment of Microwave Remote Sensing technique utilizing a suitable wavelength. This research work aims at exploiting the potential of C-band Radarsat-2 Polarimetric Interferometric Synthetic Aperture Radar (PolInSAR) technique for developing a relationship between complex coherence and field estimated forest aboveground biomass(t/ha). In order to attain our objective Radarsat-2 satellite interferometric pair of 4th March 2013(master image) and 28th March 2013(slave image) were acquired. The present study takes into account Barkot Reserve Forest(78.16°-78.28°E and 30.06°-30.17°N),a moist deciduous forest located in Doon valley in Uttrakhand state of India as the study area. The study area comprises of variable forest density classes with mixed Sal (Shorea robusta) and Teak (Tectona grandis) plantations which highly contribute in volume scattering. Other scattering mechanisms like surface scattering is yielded by the bare agricultural fields present around the forest range. Field inventory was done for 30 plots (31.62m x 31.62m) and tree height and stem diameter were procured for each plot. Work emphasises on the application of PolInSAR coherence instead of using SAR backscatter which suffers from a constraint that after a certain volume of biomass the SAR backscatter technique reaches a saturation point after which it is unable to procure more information about the biomass of the forest. Also backscatter information is dependent on environmental factors like stem and branch moisture etc. While coherence remains unaffected by such factors. Complex coherence values for different polarisation channels were computed with the help of polarimetric interferometric coherence matrix. Retrieved complex coherences were investigated individually and then regression analysis was carried with the field estimated aboveground biomass. Results stated that a negative linear relationship exists between complex coherences and aboveground biomass. Reduction in coherence values was observed due to the unstable scatterers; hence as the aboveground biomass increased, coherence values of different polarisation channels suffered a decrease. Coefficient of determination (R2) value of 0.204. 0.126 and 0.71 was obtained respectively for coherence contributed by surface scatterers (HH+VV),double bounce scatterers(HH-VV)and volume scatterers(HV+VH).R2 value of HV+VH complex coherence component was found to be relatively higher than other polarization channel components. Results conclude that the PolInSAR coherence techniques along with various modelling approaches can be utilized for estimating forest biophysical parameters.
FOUR-COMPONENT DECOMPOSITION WITH JOINT ANGLE COMPENSATION FOR ELLIPTICITY ANGLE AND ORIENTATION ANGLE
chen, jiehong; Zhang, Hong; wang, chao; liu, meng; Xie, lei
Show abstractThe objective of target decomposition is to express the scattering matrix, the coherency matrix or the covariance matrix as the sum of independent elements to relate physical scattering mechanisms to the decomposed components. Target decomposition plays an important role in polarimetric classification, change detection, and other applications of POLarimetric SAR (POLSAR) data. The Yamaguchi decomposition is a model-based decomposition based on Freeman and Durden decomposition . It is based on simple physical scattering mechanisms (surface scattering, double-bounce scattering, volume scattering and helix scattering) and is widely used for its simplicity and clear physical meanings. Three shortcomings of the Yamaguchi decomposition are observed in both theory and the experimental results of real POLSAR data . Only six elements of the coherency matrix are used in the decomposition, and the other elements are unused. In addition, a few scattering phenomena are misinterpreted in built-up areas. Negative power pixels also emerge in a few pixels after decomposition, particularly for high-frequency POLSAR data. The second and third shortcomings are due to volume scattering overestimation because the assumed surface, double-bounce and volume scattering models are always applicable and their coherency matrices satisfy scattering reflection symmetry. Great efforts have been made to improve the Yamaguchi decomposition. A rotation has been applied to the coherency matrix before the Yamaguchi decomposition. Lee et al. proposed Orientation Angle Compensation (OAC) based on circular polarization. Xu and Jin developed an algorithm by minimizing the power of the cross-polarization channel. An et al. presented another procedure by maximizing the co-polarization power using Huynen parameters. All three techniques, which are essentially identical, can produce the same results. They can increase the number of elements of the coherency matrix used in the decomposition to seven. Different volume scattering models have also been proposed to decrease the volume scattering. Hajnsek et al. presented asymmetrical volume scattering matrices using narrowing of orientation angle distribution. An et al. and Shan et al. proposed unitary volume scattering model. Sato et al. developed a new volume scattering model using distributed dihedral corner reflectors for double-bounce structures. All these new volume models are effective in decreasing the volume scattering and increasing surface and double-bounce scatterings and are under the condition of scattering reflection symmetry. We propose an image processing technique to alleviate the three shortcomings. As discussed above, only six elements are used in the Yamaguchi decomposition under the assumption of scattering reflection symmetry. We believe that the assumption of scattering reflection symmetry is the main reason for the volume scattering overestimation, causing the misinterpretation of scattering mechanisms and the emergence of negative power pixels. In fact, most measured coherency matrices of targets do not satisfy this assumption. Therefore, the JAC is first introduced to rotate the coherency matrix into a form more consistent with the scattering reflection symmetry. The JAC algorithm includes the EAC and the OAC algorithms. An improved four-component scattering decomposition, i.e., the Yamaguchi decomposition with JAC algorithm (JAC+Yamaguchi), is proposed in this paper. Compared with the Yamaguchi decomposition and the Yamaguchi decomposition with OAC algorithm (OAC+ Yamaguchi), the improved algorithm can increase the number of elements of the coherency matrix used in the decomposition to eight, correctly interpret the scattering mechanisms in built-up areas, and decrease the number of negative power pixels. The idea of the JAC algorithm, i.e., minimization of the cross-polarization, is similar to that of Cameron decomposition , i.e., maximization of the symmetric scattering component, and the deorientation algorithm , i.e., minimization of the cross-polarization. The difference between the JAC and the other two algorithms is that there are two factors, i.e., the ellipticity angle and the orientation angle, used in the JAC to minimize the cross-polarization, whereas in the other two algorithms, only orientation angle is used to achieve the goal. Using the ellipticity angle is beneficial to minimize the cross-polarization because the ellipticity angle relates to the degree of the asymmetry of the targets and causes the cross-polarization. Joint angle compensation combines ellipticity angle compensation and orientation angle compensation. Joint angle compensation rotates the measured coherency matrix into a form proximately satisfying scattering reflection symmetry using ellipticity angle and orientation angle. After joint angle compensation, the cross-polarization is minimized. Therefore, in the proposed decomposition, i.e., the JAC+Yamaguchi decomposition, the number of elements in the coherency matrix used in the decomposition is increased to eight, the interpretation of scattering phenomena is more correct, and the experimental results demonstrate that the number of negative power pixels is also reduced drastically.
Biomass Mapping for Shrublands by Polarimetric Radar Vegetation Index
chang, Jisung Geba; shoshany, Maxim Technion, Israel
Show abstractInformation on shrublands biomass mapping plays important role in better understanding Mediterranean and semi-arid ecosystem changes. Technical problems concerning the cutting and measuring of shrubs’ biomass in the field, and restrictions regarding the application of such methods in natural regions severely limit the amount of the available data over large regions. Remote sensing provide potential solution for this purpose. However, existing methods in the optical and radar spectral regions are more suitable for homogenous forest regions rather than to the high geodiversity of Mediterranean and semi-arid environments. The low-frequency (L-, P-band) radar sensor has good performs because the electromagnetic waves penetrate deeper into the canopies than in the case of an optical sensor. However, the radar signal is seriously affected by the soil roughness and rocks at low and moderate vegetation densities. Therefore, more research is required to find suitable algorithms for biomass estimation for Mediterranean and semi-arid regions with high heterogeneity of surface cover. The goal of the proposed research is to develop a quantitative mapping methodology for biomass of semi-arid shrublands. Using the polarimetric Synthetic Aperture Radar (SAR), a new Polarimetric Radar Vegetation Index (PRVI), based on the degree of polarization (DOP) which facilitate decomposition of the signal into shrub and non-vegetation components. For validation, field estimates of shrubs biomass would be facilitated using allometric methods based on their size data (radius and height) derived from airphotographs and field work. The study area consists of the climate transition zones of Israel between the Judean Mountains and the Judean Desert in the east to the Negev desert in the south. Polarimetric L-band PALSAR data were acquired for this region together with multi-spectral, rainfall, and ASTER digital elevation maps (DEM). A wide coverage of ortho-rectified airphotograph with 0.21 meter resolution was acquired for 2010 for almost all of this study area. Using the PALSAR data and existing biomass information for the climatic gradient area in Central Israel, the proposed PRVI model showed good performance and improving accuracy of estimation (cross-pol regression fit has R-square 0.41, PRVI model has R-square-0.74).
A Unified Crop Classification Framework for FP, DP and CP Data at X-band in Southern China
Xie, lei (1); Zhang, Hong (1); Li, Hongzhong (2); Wang, Chao (1); Liu, meng (1) 1: Institute of Remote Sensing and Digital Earth, CAS, China, People's Republic of; 2: Shenzhen Key Laboratory of Spatial Information Smart Sensing and Services & Key Laboratory for Geo-Environment Monitoring of Coastal Zone of the National Administration of Surveying, Mapping and GeoInformation, Shenzhen University, Shenzhen 518060, China.
Show abstractRemote sensing techniques provide frequent, large-scale observation of the Earth. Both optical and radar remote sensing techniques have been widely used in various fields. However, in some tropical and subtropical regions, optical sensors cannot work well due to the rainy weather and frequent cloud cover. In contrast, synthetic aperture radar (SAR) is a suitable tool in photogrammetry and remote sensing due to its unlimited capabilities in terms of weather and time of day . Agriculture is the foundation of local economy in some cities in Southern China, and crops classification plays an important role in agriculture management. Recently, polarimetric SAR (PolSAR) data in various imaging modes have been successfully introduced in crop classification. Previous investigations shown that performances with fully polarimetric (FP) SAR data were better than that using dual-pol (DP) and single-pol (SP) data . However, some radar systems, such as Envisat/ASAR, do not provide FP data, so the capability for crop classification using DP and SP data is also worthy of study. During the last decade, the compact polarimetric (CP) radar systems have also been proposed and developed, CP data provide a large swath widths than the FP data at the cost of information loss . Anisworth et al.  compared the land cover classification results between CP and FP data, and the overall accuracy using CP data was less than that of FP data. Currently, it is widely accepted that CP SAR cannot replace FP SAR, but it performs better than the classical DP mode in various applications. Although various investigations on land cover classification have been carried out using PolSAR data in different imaging modes, there is still a lack of common approaches. Therefore, the main objective of this paper is to introduce an unified framework for crop classification in Southern China using data in FP, coherent DP and CP modes. Data acquired by TerraSAR-X over Leizhou Peninsula, Guangdong Province, China, are used in our experiments. There are four main crops (rice, banana trees, sugarcane, eucalyptus), along with built-up areas and water, in the experiment site. In situ measurements are used for investigations and validation. We explore the similarities between data in FP, DP and CP modes, and find that the surface scattering component (SC=SHH+SVV) is the common feature. Based on this property, the alternative (q= SC 2/SPAN) to average scattering angle α is also shared by these data, SPAN in CP mode is half of that in FP mode, so q in CP mode is quite similar to that in FP mode. SPAN in coherent DP mode is less than FP mode due to the loss of cross-polarization, so q in DP mode is larger than FP mode. The q-SC space is divided into four zones, each zones represents an land cover type. Based on q-SC space, these land cover types show great separability. However, rice and eucalyptus still confuse in q-SC space. As shown in previous publications , the HH/VV ratio is feasible for rice mapping due to its vertical structure. Superficially, the HH/VV ratio in the CTLR mode cannot be obtained. However, The 2*(RH intensity) and the 2*(RV intensity) are similar to the HH intensity and VV intensity under the reflection symmetry assumption, resulting in the similarity between the HH/VV ratio and RH/RV ratio. From the above analyses, the HH/VV ratio is also a common feature shared by data in the FP, coherent HH/VV and CTLR modes. Thus, rice fields can be separated from eucalyptus. The experimental data also show that double-bounce scattering component (DBC=SHH-SVV) of built-up areas, banana trees and rice fields is quite larger than other areas, and the intensity in the built-up areas is 6.07 dB and 6.47 dB larger than that in the rice fields and banana trees, so the built-up areas can be easily extracted. Similarly, in CP mode, the RH intensity behaves similarly to DBC, so RH intensity can be applied in CP mode for built-up areas extraction. Based on the similar properties, the unified classification framework can be put forward and tested with experimental data. According to the samples of validation group, the overall accuracy of each data type reaches 94.54%, 91.91% and 91.12% using FP, coherent DP and CP data respectively, which demonstrates the effectiveness of the proposed classification framework. From another perspective, data in the coherent HH/VV and CTLR modes can be used as an alternative to FP data when FP data are not available. The overall accuracies for the data in the FP, HH/VV and CTLR modes using the Wishart ML classifier are 88.63%, 86.14% and 86.31%, respectively. The proposed method achieves a higher classification accuracy because more polarimetric features that describe the scattering mechanisms and structures of the land cover types are added to the classification scheme. References  J. S. Lee and E. Pottier, Polarimetric Radar Imaging: From Basics to Applications. Boca Raton, FL, USA: CRC Press, 2009.  R. K. Raney, Hybrid-polarity SAR architecture, IEEE Transactions on Geoscience and Remote Sensing, 2007, 45, pp. 3397-3404.  T. Ainsworth, J. Kelly, and J. S. Lee, Classification comparisons between dual-pol, compact polarimetric and quad-pol SAR imagery, ISPRS Journal of Photogrammetry Remote Sensing, 2009, 64, pp. 464-471.  Fan Wu, Chao Wang, Hong Zhong, Bo Zhang, Yixian Tang, Rice Crop Monitoring in South China With RADARSAT-2 Quad-Polarization SAR Data, IEEE Geoscience and Remote Sensing Letters, 2011, 8, pp. 196-200.
Merchant Ships Classification in Polarimetric SAR Image
Zhang, Hong; wang, ziwei; wang, chao; liu, meng; Xie, Lei
Show abstractShip classification is one of the most important applications for maritime remote sensing. PolSAR (polarimetric synthetic aperture radar), having high resolution and polarimetric information, is playing a more significant role in this field. Recently, many researchers studied the PolSAR ship classification. Wu  investigated the polarimetric backscattering signals of bulk carrier, container ship and oil tanker using Pauli, SDH coherent decompositions and gave a negative conclusion on using mean value and peak value for PolSAR ship classification. Margarit [2-4] published a serious of paper on PolSAR ship classification using simulation data, and concluded that the classifier would work in PolSAR image with suitable resolution which could separate independent scatters. Paladini  proposed a nonhomogeneous man-made object classifier using the H/α decomposition to search for the rotation invariant. But up till now, not even a mature PolSAR ship classifier is available using real PolSAR image owning to varieties of reasons. In this paper, we propose a ship classifier in PolSAR images for three main categories of merchant ships: bulk carrier, container ship and oil tanker. Three Radarsat-2 Fine Quad images and corresponding AIS (automatic identification system) data are used to validate our method. Feature vector is widely used for feature description in target classification. We use three features in the proposed classifier including length width ratio (f1), peak distribution along the ship length (f2), and ship ROI (region of interest) backscattering value (f3). Polarization acts the role to extract the feature more accurately: the pol-filter (polarimetric filter) and pol-decomposition (polarimetric decomposition) are used. Pol-filter gives an optimal way to combine multiple polarimetric channels, with which the structure detail of f1 is well displayed. In this paper, after a comparison of effect of the pol-filter on ship-sea contract and ship moving phenomenon, the SPAN pol-filter is used to get the minimum enclosing rectangle so that to measure the ship length and width. The pol-decomposition holds the ability on decomposing the backscattering signal into different components according to its physical scattering mechanism. So for f2 and f3, which represents the function of the ship deck by indicating different cargos or equipments, the SDH polarimetric decomposition is used. The sphere component (Ks) is used for counting the strong peaks along the ship length after analyzing the structure difference of deck. The ROI backscattering value is described by polarimetric feature space by the three components (Kd,Kh,Ks) of SDH decomposition. Then, we get a flow chart of the polarimetric ship classifier as (1) Ship detection using CFAR. (2) Get the minimum enclosing rectangle and ship ROI in SPAN image, then the ship length and width can be calculated. (3) Process SDH decomposition only in the ship ROI achieved in step (2) (4) Use Kd component for the peak distribution along the ship length direction (5) Use the Kd, Kh, Ks feature spaces to get the f3. In a preliminary experiment to calculate the overall classification rate, the most positive one can reach about 70%. Comparing to using the single-pol image of the same image resolution, the polarimetric classifier also holds more robustness. In summary, with polarimetric information of merchant ship, the classifier will work more efficient. Reference  F. Wu, C. Wang, H. Zhang, B. Zhang, “Analysis of polarimetric ship signatures with Radarsat-2 quad-pol imagery,” IGARSS 2012.  G. Margarit, J. J. Mallorqui, “Analysis of polarimetric ship signatures with Radarsat-2 quad-pol imagery,” IEEE Trans. On Geosci. Remote Sens. vol. 47, no. 4, pp. 1224-1235, 2009.  G. Margarit, “Marine applications of SAR polarimetry,” Ph. D. dissertation, Departament de Teoria, Universitat Politecnica de catalunya, Barcelona, April 2007.  G. Magarit, X. Fabregas, J. J. Mallorqui, A. Broquetas, “Analysis of the limitations of coherent polarimetric decompositions on vessel classification using simulated images,” IGARSS 2004.  R. Paladini, M. Martorella, F. Berizzi, “Classification of Man-Made Targets via Invariant Coherency-Matrix Eigenvector Decomposition of Polarimetric SAR/ISAR Images,” IEEE Trans. On Geosci. Remote Sens. vol. 49, no. 8, pp. 3022-3034, 2011.
WATER EXCTRACTION OF POLARIMETRIC SAR DATA USING MODIFIED STATISTICAL REGION
Zhang, Hong; Deng, ying; wang, chao; xie, lei; liu, meng
Show abstractWith the new generation of airborne and spaceborne SAR sensors，most of the recently launched Polarimetric SAR (PolSAR) systems have the capability of polarimetric imaging. PolSAR systems transmit and receive electromagnetic wave in two different channels (e.g. vertical and horizontal) respectively, that preserve the scattering properties of ground targets from random waves, provide abundant feature information. Numerous studies indicate that PolSAR data can be used in classification, target identification and forest parameter estimation and so on. Water detection methods for single-polarimetric SAR data are relatively mature. However water parts appear to dark pixels, very much similar to shadow, bare area or any target with high reflection, that features contained in one single channel are insufficient to distinguish these dark pixels . Existing water detection methods for PolSAR data can be divided into two broad categories: pixel-based method and object- oriented method. The former algorithms calculate based on pixels, extract interest features according to specific rules, while correlations between center pixels and their neighboring ones depend on window size of current algorithm. Those methods are simple to operate but affect greatly by speckle, yet fail to use spatial and texture feature as well. The latter methods provides a new idea for high resolution remote sensing image analysis, which is becoming one of the hot spots of remote sensing, photogrammetry and GIS. Object- Oriented detection method uses multi-scale segmentation for scaling that breaks the shackles of the natural scale of the image, analyzing the land covers in homogeneous adjacent pixels, applying classifications through geometry, semantics, texture and topology between each neighbor unites, with anti-spackle property, high computational efficiency and applicability. The statistical region merging (SRM) algorithm is a member of the Object-Oriented technology proposed by Nock and Nielsen (2004) , which exhibits efficient performance in solving significant noise corruption and does not depend on the data distribution. However the original SRM algorithm was design for optical data and gray images characterized by additive noise while SAR images characterized by multiplicative noise. This method was later generalized by Lang et al. (2014) , so that it is suitable for images with larger range and multiplicative noise. Although the generalized SRM (GSRM) has fair performance on all kinds of SAR data, but still fail to use the greatest strength of full-polarization specifically. This paper aim at explore the potential of full-polarized information and achieve fast and high accuracy water extraction. The proposed method modified the merging criterion of GSRM, design an algorithm that applies on full-polarimetric data only, can distinguish dark pixel on the merging section, and does not need a pose-classification. As these pixels of different land covers with similar intensities can be distinguished by scattering characteristics, which express by heterogeneity  , here use heterogeneous clutter models to described PolSAR data, and build heterogeneity measurements. First, calculate the gradient of each pair of adjacent pixels according to the 8-neighborhood templates using the polarization coherent matrix，arrange these pairs by the gradients in ascending order. Then, a merging criterion is deduced from the heterogeneity measurements that derived from the scattering properties of different land covers. Traverse all gradients in turn, merge the corresponding pixels (regions) if meet the criterion. Finally, a simple threshold segmentation is applied to achieve precise water extraction. The method is tested on Rasarsat-2 data of Suzhou, China, acquired on August 29, 2014. The original SRM algorithm using single-polarimetric data and the region-based Mean Shift method  are also applied as comparisons. Experimental results show that the proposed method can acquire more accurate water area while maintain the outline. REFERENCES B. Brisco, R. Touzi, J. J. van der Sanden, F. Charbonneau, T. Pultz, and M. D'Iorio, "Water resource applications with RADARSAT-2–a preview," International Journal of Digital Earth, vol. 1, pp. 130-147, 2008. R. Nock and F. Nielsen, "Statistical region merging," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 26, pp. 1452-1458, 2004. F. Lang, J. Yang, D. Li, L. Zhao, and L. Shi, "Polarimetric SAR Image Segmentation Using Statistical Region Merging," Geoscience and Remote Sensing Letters, IEEE, vol. 11, pp. 509-513, 2014. L. Bombrun, G. Vasile, M. Gay, and F. Totir, "Hierarchical Segmentation of Polarimetric SAR Images Using Heterogeneous Clutter Models," IEEE Transactions on Geoscience and Remote Sensing, vol. 49, pp. 726-737, Feb 2011. M. Liu, H. Zhang, C. Wang, and F. Wu, "Change Detection of Multilook Polarimetric SAR Images Using Heterogeneous Clutter Models," IEEE Transactions on Geoscience and Remote Sensing, vol. 52, pp. 7483-7494, Dec 2014. B. Zhang, G. Ma, Z. Zhang, and Q. Qin, "Region-based classification by combining MS segmentation and MRF for POLSAR images," Systems Engineering and Electronics, Journal of, vol. 24, pp. 400-409, 2013.
Supporting forest inventory in Durango, Mexico by combining in-situ data from cellular phones with satellite image mapping
Toivanen, Timo (1); Molinier, Matthieu (1); Häme, Tuomas (1); Corral, Javier (2); Lopez, Carloz (2); Vega, Daniel (2); Rauste, Yrjö (1) 1: VTT Technical Research Centre of Finland, Finland; 2: Juarez University of Durango State, Mexixo
Show abstractIn this paper, we present a concept how volunteers will gather forest inventory data to be utilized as in-situ data for forest biomass computation from satellite imagery. A VTT-developed forest measurement tool called ‘Relasphone’ will be used to collect in-situ forest inventory data. We will demonstrate the concept in the state of Durango in Mexico that contains natural and managed forest area. We will concentrate on measurements of Cooper’s Pine (Pinus cooperi) trees. The study area includes over 390 permanent sampling plots that will provide reference data for concept validation and verification. Landsat-8 imagery will be used as a proxy for Sentinel-2 data. Forest biomass is widely monitored using a variety of remote sensing data and techniques (Gleason and Im 2011). Remote sensing offers benefits compared to traditional in-situ forest inventories made by experts. One of the main benefits is that the number of ground reference plots can be significantly reduced. Remote sensing of forest biomass can provide reduced costs and time requirement compared to full forest inventories. The amount of biomass in the forests is crucial information for forest owners and managers to understand the value and the use of the forest. The future Biomass mission by ESA, to be launched 2020, aims to take measurements of forest biomass to assess terrestrial carbon stocks and fluxes. The information is needed to understand how forests affect the carbon cycle and climate change. The availability of ground reference data has been a bottleneck in remote sensing analysis over wide forested areas, as the acquisition of this data is an expensive and long process. A report by the Global Forest Observation Initiative (GFOI 2013) identified that many countries suffer from a lack of in-situ calibration and validation data for biomass estimation. Presently the ground data for management plans are collected by forestry professionals, but it can be foreseen that by combining ground based data, UAV (Unmanned Aerial Vehicle) and satellite data, principal material for forest management plans of acceptable quality can be collected without professional foresters’ work on the field. Such opportunity could significantly increase the area coverage of forest management plans, particularly in countries where funds for forest management planning are restricted. Recent advances in information and communication technology (ICT) and increased awareness of the status of the environment, in particular global climate change, have activated people to participate in monitoring (Burke et al. 2006). Nowadays mobile devices are present wherever there are people, which enable new possibilities in building a platform for utilizing citizen observers as a spatially and temporally dense observation network. This opportunity is strengthened by the growing common concern about the state and future of the environment. People are increasingly willing to contribute their time and resources to the analysis and processing of scientifically collected data. This citizen engagement in scientific work – Citizen Science – is becoming more efficient, pervasive and useful with the help of the latest ICT technology. Citizen observations are one promising way to increase both temporal and spatial coverage of collected in-situ forest inventory data. The Relasphone is a forest measurement tool on a mobile phone. The Relasphone concept has been developed in Finland to help forest owners and timber brokers make forest measurements. With the Relasphone application, users gather in-situ forest inventory data including estimated biomass, tree species, tree height and age. The biomass estimation is based on the relascope functionality (Bitterlich 1984). The collected in-situ forest measurements are utilized as a ground reference material for spaceborne or airborne remote sensing data analysis. The measurements are put into a planner engine that outputs forest resources information for every forest stand and feeds them into a GIS. The GPS based location information with measured forest data makes it possible to introduce measurements easily as in-situ reference data. Once the concept is successfully transferred from boreal to tropical forests, it can dramatically change the way forest databases and management plans are produced globally and may lead to major cost reductions. A low-cost solution can increase the coverage of forest management plans particularly in regions where possibilities to invest on expensive planning work are limited. The concept can be easily applied for other purposes that require similar information. The users of this category include the Non-Governmental Organizations that are concerned about nature conservation and illegal cuttings, the European Commission that will fund EU-wide land cover mapping, and the FAO that conducts global survey of forest resources, for instance. One potential commercial use of the tools is the REDD (Reduction of Emissions from avoided Deforestation and Degradation of forests) process that will require inventory of forest resources over wide areas in the tropical zone. For example, in the developing countries inexpensive smart phones and GPS could be given to people free of charge on the condition that they collect data and images from forests to support satellite image mapping that further serves the economy of the country through sustainable use of forest resources. REFERENCES Bitterlich, W. (1984), The Relascope Idea: Relative Measurements in Forestry, Commonwealth Agricultural Bureaux, Farnham Royal. Burke J, Estrin D, Hansen M, Parker A, Ramanathan N, Reddy S, Srivastava M B (2006) Participatory Sensing. In: Proceedings of the International Workshop on World-Sensor-Web (WSW'2006), ACM, Boulder, CO, USA, 31 Oct 2006. Gleason, C. J., Im, J. (2011) A Review of Remote Sensing of Forest Biomass and Biofuel: Options for Small-Area Applications, GIScience & Remote Sensing, 48:2, 141-170. GFOI (2013): Review of Priority Research & Development Topics: R&D Related to the use of Remote Sensing in National Forest Monitoring. Pub. GEO, Switzerland, 2013 ISBN 978-92-990047-5-3.
Exploitation of agricultural Polarimetric SAR time series with Binary Partition Trees
Alonso-Gonzalez, Alberto; Jagdhuber, Thomas; Hajnsek, Irena
Show abstractDuring the last years, the presence of space-borne SAR systems has empowered the construction of time series datasets, containing SAR images of the same scene at different time instants. These datasets are relevant as they contain information not only about the scene itself but also about its temporal evolution. Moreover, this tendency is expected to upscale in the near future, since most of the currently planned SAR missions are focused on the construction of time series, as the ESA Sentinel-1 or the DLR TerraSAR-X, TanDEM-X and the future TanDEM-L. In this work, the Binary Partition Tree (BPT)  is proposed in order to analyse PolSAR time series. The BPT may be considered as a hierarchical region-based and multi-scale data representation. Recently, this technique has been applied to process PolSAR  and hyperspectral data , demonstrating its ability to detect the homogeneous regions of the scene while also preserving contours and small details of the data. The BPT may also be extended to process SAR time series, as described in . This representation has proven to be attractive for different applications as, for instance, speckle filtering, polarimetric estimation, classification, segmentation and change detection. The use of the BPT is particularly interesting for agricultural scenes, as large homogeneous regions may be found on these areas, allowing a strong level of speckle filtering over these areas while also maintaining the spatial resolution. This allows a precise investigation and characterization of the polarimetric signature of these areas and also their temporal evolution. Furthermore, the monitoring of these areas may be crucial for planning and optimizing the use of the limited resources and ensuring no disruptions in the food supply. In this work, we propose to perform a more detailed analysis of these datasets, with special attention to the characterization of the temporal evolution of the scene. The technique will be applied to the AgriSAR 2006 campaign, consisting of a set of E-SAR acquisitions over agricultural fields and an extensive collection of in-situ ground measurements. This ground truth information will be used in order to analyse and verify the obtained results. Particular attention will be taken to the final land applications that may take profit of the BPT region-based processing as, for instance, crop identification and monitoring, change detection and characterization or bio/geophysical parameter retrieval.  Salembier, P.; Garrido, L., “Binary partition tree as an efficient representation for image processing, segmentation, and information retrieval,” IEEE TIP, vol. 9, no. 4, pp. 561–576, 2000.  Alonso-Gonzalez, A.; Lopez-Martinez, C.; Salembier, P., “Filtering and segmentation of polarimetric SAR images with Binary Partition Trees,” in Proc. IEEE IGARSS, 2010, pp. 4043–4046.  Alonso-Gonzalez, A.; Lopez-Martinez, C.; Salembier, P., “Filtering and segmentation of Polarimetric SAR data based on Binary Partition Trees,” IEEE TGRS, vol. 50, no. 2, pp. 593 –605, 2012.  Alonso-Gonzalez, A.; Valero, S.; Chanussot, J.; Lopez-Martinez, C.; Salembier, P., “Processing Multidimensional SAR and Hyperspectral Images With Binary Partition Tree,” Proceedings of the IEEE, vol.101, no.3, pp.723,747, 2013.  Alonso-Gonzalez, A.; Lopez-Martinez, C.; Salembier, P., “PolSAR Time Series Processing With Binary Partition Trees,” IEEE TGRS, vol.52, no.6, pp.3553,3567, 2014.
The Tropical managed Forests Observatory: a research network addressing the resilience and Carbon cycle of tropical logged forests
Sist, Plinio (1); Rutishauser, Ervan (2) 1: Cirad, France; 2: CarboForExpert
Show abstractTropical forests represent a major reservoir of global carbon, accounting for up to half of the estimated 652 Pg of carbon stored in vegetation. While deforestation in the tropics remains a major environmental issue to be tackled, forest degradation deserves more attention from a broad range of stakeholders concerned about social and ecological well-being. Degraded forests are a major component of tropical landscapes; over half of all tropical forests have been cleared or logged, and almost half of standing primary tropical forests, up to 400 million ha, are designated by national forest services for timber production. The portion of tropical forests managed for timber extraction, hereafter referred to as “managed forests”, will therefore play key roles in the trade-off between provision of goods and maintenance of carbon stocks, biodiversity, and other services. However, so far, most of our understanding of tropical forest yields from plot networks located in old-growth undisturbed forests or in secondary forests, while the dynamics of managed forests at the regional and continental scale remains poorly studied. Monitoring of managed forests is important for myriad reasons including the need to understand their roles in the global carbon cycle and the trade-offs between environmental impacts and human benefits. In regards to these trade-offs, the results of monitoring can be used to design forest management practices that mitigate any deleterious impacts of forest use and enhance the resilience of forest subjected to unavoidable impacts so as to maximize the conservation values of those forests. Moreover, given that forest management practices, forest structure, and dynamics differ widely among tropical countries and regions, assessments of the impacts of different practices are needed at regional and continental scales to inform policy. To address the challenge posed by the resilience and management of tropical forests, 20 research institutes are now collaborating on studies on the effects of logging and post-logging silviculture on forest structure, productivity, biodiversity and carbon fluxes at large spatial and temporal scales. These efforts resulted in the creation of the Tropical managed Forests Observatory (TmFO), an international network that merges information from experimental sites across South America, Africa and South East Asia. TmFO aims to address the data and knowledge gaps by promoting collaboration among institutions that monitor permanent plots in managed tropical forests, some of which were established by foresters many decades ago. These institutions house irreplaceable and unique data on logged forests from before CO2, temperature and precipitation regimes reached contemporary levels. TmFO’s guiding questions include: (1) how resilient are tropical forest structure, function and composition to logging; (2) how do forest responses to logging and other silvicultural treatments vary across regions and continents; (3) what are the trade-offs between financial viability and environmental sustainability of commercial logging; and (4) what is the role of silviculture in forest conservation? Here we will present the main characteristics of the network as well as preliminary results for the Amazon Basin. TmFO currently comprises 24 experimental sites located in 9 countries across three main regions: Amazon Basin (11 sites, 5 countries); Congo Basin (6 sites, 2 countries); and South East Asia (7 sites, 2 countries). A total of 921 ha of tropical managed forests spread over 490 permanent sample plots (PSP) are inventoried. Among them, 59 control plots (12 %) are in undisturbed forests. TmFO spans a large gradient of logging intensities (5-60% of above-ground biomass removed) and silviculture practices (reduced-impact logging, conventional logging, and post-logging silvicultural treatments). These features offer unique opportunities to investigate post-logging dynamics over a wide range of conditions, up to those that were heavily degraded by high-intensity unplanned logging. In the Amazon basin, a standardized protocol to estimate above-ground biomass (AGB) stocks before and after logging was carried out over 81 permanent forest plots which span the breadth of climate, soil and logging intensity gradients occurring in the Amazon region. We computed the time to recover initial AGB stock (year) for each plot. The main biophysical predictors driving biomass recovery time were identified using multiple General Linear Mixed Models (GLMM). Our results show that despite marked geographical pattern in natural forest structure and dynamics, we observed a markedly consistent biomass recovery across the Amazon region, with time to recover initial aboveground biomass highly correlated with the proportion of initial biomass removed by logging activities. This general biomass recovery response to logging will lead to improved modeling exercises predicting responses of tropical managed forest carbon cycling to land use and climate change.
Scattering Function Optimization for Growth Stage Determination of Rice Fields: Preliminary Results
Yuzugullu, Onur (1); Erten, Esra (2); Hajnsek, Irena (1,3) 1: Institute of Environmental Engineering, ETH Zurich, CH-8093 Zurich, Switzerland; 2: Faculty of Civil Engineering, Istanbul Technical University, TR-34469 Istanbul, Turkey; 3: Institute of Remote Sensing Technology, German Aerospace Center, D-82234 Oberpfaffenhofen, Germany
Show abstractRice is one of the major crops that is highly consumed worldwide. Thus, there is a requirement for frequent monitoring. This study aims to provide an approach to determine the growth stage of rice fields by means of polarimetric Synthetic Aperture Radar (PolSAR) images. To estimate the growth stage of rice fields, a morphology based forward-scattering model and solution-space search algorithm is used at X-band. Firstly, the forward-scattering model is used for simulating the back-scattering intensity in different polarimetric channels to create the solution-space. Subsequently, solution-spaces are constrained based on the intensity values from the SAR image. Next, the resulting spaces are compared in order to check the matching morphologies and their physical feasibilities with respect to growth constraints. Finally, probability of different growth stages are calculated and translated to BBCH scale. The proposed method is tested over co-polar TerraSAR-X (TSX) data by exploiting a single acquisition and a priori information.
Potential of various polarimetric signatures for retrieval of pine forest parameters
Ziolkowski, Dariusz; Lewandowska, Aneta
Show abstractFull polarimetric ALOS-2 radar images collected in May and November 2007 were used in the presented work. High-quality Digital Forest Map was applied as a reference; it served for deriving information on stand age, height, DBH, density and stem volume. 150 polygons (2 ha) were selected for research works; they were located in the stands from 5 to 150 years old, with stem volume from 25 to 390 m3/ha. Homogeneity of particular polygons was verified using high-resolution orthophotomap. Radar images were processed using various polarimetric decompositions, e.g. Freeman-Duren decomposition, Yamaguchi, H/A/Alpha in POLSARPRO software; the images were next rectified and further analysis was conducted within ENVI SARSCAPE software. Two different approaches were applied in the course of the research works. At first approach each polygon was treated as a whole; various statistical parameters were calculated for every polygon for particular decomposition products and for their combinations. The obtained results were next correlated with forest stand parameters derived from Digital Forest Map (age, DBH, height, density and stem volume). At the second phase of the works the detailed analysis of polarimetric signatures for particular pixels located within polygons was performed. They were next classified due to contribution of various scattering mechanisms (surface, double bounce, volume scattering). At the next stage relations between information gained from radar images and stand parameters were searched for each class separately and next combined in order to obtain one value for the whole forest unit. The obtained results reveal high potential of the method for estimating biomass and other parameters of forest stands.
First results of AfriScat, a tower-based radar experiment in African forest
KOLECK, Thierry (1); ALBINET, Clément (2); BORDERIES, Pierre (4); VAGLIO LAURIN, Gaia (3); NICOLINI, Giacomo (3); LE TOAN, Thuy (2); HO TONG MINH, Dinh (2); VILLARD, Ludovic (2); HAMADI, Alia (4) 1: CNES, France; 2: CESBIO, France; 3: CMCC, Italy; 4: ONERA, Toulouse
Show abstractIn the framework of the EE7 BIOMASS mission, AfriScat is the second experiment based on long-term radar backscattering measurement in P-band. It provides feedback on two key issues linked to mission operations and Level-2 product quality: - the coherence of P-band radar as a function of time of day, time intervals and environmental conditions which directly impacts the quality of the Level-2 products - the scattering characteristics of P-band radar within the tropical forest to support the development of robust widely-applicable retrieval algorithms and the development of algorithm theoretical baseline documents. The AfriScat campaign is built on the results from the previous TropiScat campaign which took place in French Guyana from 2011 to 2013, extending the results to the African tropical rain forest whose structure and environmental conditions are different. The Ankasa Conservation Area Tower in Ghana, operative as part of the CarbonAfrica eddy covariance network, has been selected for the AfriScat experiment. Even if this site is also located in a tropical forest, some important characteristics are different from the TropiScat experiment. As the Ankasa tower is located on a plateau, the ground topography is flat, contrary to French Guiana, with a slopping ground. Because of the flat configuration, double bounce effect should be more pronounced and can impact the temporal decorelation behaviour. So, in addition to the TropiScat experiment in amazonian forest, the AfriScat campaign will provide feedback on: - the long-term statistical characterisation of temporal coherence at P-band over tropical forests (diurnal, weekly, monthly, yearly); - the identification and characterisation of the main sources of temporal decorrelation (environmental, instrument, geometrical, structural); - the 3-dimensional distribution of radar scatterers and their backscattering characteristics; - the impact of double bounce effects in the radar backscattering of tropical forests; - the validity of simplified physical and statistical scattering models (intensity, polarimetry, PolInSAR) used as basis for geophysical information retrieval. The AfriScat instrument is a scatterometer located on a 60 meters high tower. As it is associated with an array of wide-band antennas, it provides backscattering measurement in full polarimetric and tomographic configuration. The measurement configuration is identical to the TropiScat one. In addition to these P-band measurements, in-situ data will be simultaneously collected. The experiment will be deployed on site on November/December 2014. A feedback of the setup and tests, as well as first scientific results will be presented to the workshop.
Terrestrial Laser Scanning for measuring forest biomass change
Lau Sarmiento, Alvaro Iván (1); Calders, Kim (1); Herold, Martin (1); Avitabile, Valerio (1); Raumonen, Pasi (2); Gonzalez de Tanago, Jose (1); Bartholomeus, Harm (1); Joseph, Shijo (3); Martius, Christopher (3); Goodman, Rosa (4) 1: Wageningen University and Research Centre, the Netherlands; 2: Tampere University of Technology, Finland; 3: CIFOR - Center for International Forestry Research; 4: School of Geography, University of Leeds, UK.
Show abstractEstimating tropical forest biomass change and its related carbon emissions due to anthropogenic activities is crucial for the establishment of measurements, reporting and verification (MRV) systems for a consistent framework in which estimates could be measured. For this, remote sensing techniques are considered as an essential REDD+ tool; and in combination with ground measurements provide an objective, functional and cost-effective approach to assess the impact on forest carbon in REDD+ projects. Improving the tropical forest emission estimates requires accurate biomass measurements before and after the impact events at local level, and the effective use and integration with remote sensing techniques to monitor impacts over larger areas. Traditional methods involve ground-based measurements, which are often constrained by physical accessibility, time consuming, labour-intensive and expensive to implement at large scale. However, more recent technologies such as terrestrial laser scanning (TLS) offers a detailed 3D measurements of tree and canopy structure, and allometry in a rapid and non-destructive approach. Four major field campaigns (Gabon, Peru, Indonesia and Guyana) have been successfully completed and tropical forest datasets were acquired using TLS (RIEGL VZ-400 laser scanner), field inventory and data from destructive harvesting. We also developed and tested methods for acquiring a high accurate pointcloud at plot level scale. The Peruvian and the Gabon campaign scanned 1.0 ha plots of intact tropical forest, whereas the Indonesian and Guyana campaigns focused on 0.12 ha partial harvesting plots, scanning more than 15 pre- and post-harvest plots. Moreover, a previous Australian campaign in native Eucalypt forest demonstrated that LiDAR scanning is more accurate than traditional methods for assessing biomass, providing a more robust and less biased estimates. The TLS data processing uses available techniques for processing point clouds and employs tree modelling algorithms, such as Quantitative Structure Model (QSM) to estimate volume directly and related changes in canopy structure. Our preliminary results evidenced that QSM tree modelling algorithm is able to reconstruct branches with a RMSE less than 5 cm deviation. This TLS analysis is expected to result in direct biomass change measurements that will be validated and compared to those from traditional forest methods for inventories and harvest estimates.
Multiple data sources for analyzing, integrating and validating biomass maps
Avitabile, Valerio (1); Herold, Martin (1); Lewis, S.L. (2); Phillips, O.L. (2); Asner, G. P. (3); Bayol, N. (4); de Jong, B. (5); Kearsley, E. (6); Lin, X. (7); Lindsell, J. (8); Lopez-Gonzalez, G. (3); Lucas, R. (9); Malhi, Y. (10); Morel, A. (10); Mitchard, E. (11); Piao, S. (12); Ryan, C. (11); Vaglio Laurin, G. (13); Valentini, R. (14); Verbeeck, H. (6); Wijaya, A. (15); Willcock, S. (16) 1: Wageningen University, Netherlands, The; 2: University of Leeds, UK; 3: Carnegie Institution for Science, USA; 4: Foret Ressources Management, France; 5: Ecosur, Mexico; 6: Ghent University, Belgium; 7: Laboratoire des Sciences du Climat et l'Environnement, France; 8: A Rocha International, UK; 9: Aberystwyth University, Australia; 10: University of Oxford, UK; 11: University of Edinburgh, UK; 12: Peking University, China; 13: Centro Euro-Mediterraneo sui Cambiamenti Climatici, Italy; 14: Tuscia University, Italy; 15: Center for International Forestry Research, Indonesia; 16: University of Southampton, UK
Show abstractDuring the last decade large efforts have been conducted to better quantify amounts and spatial distribution of aboveground biomass, a key parameter for estimating carbon emissions and removals due to land-use change, and related impacts on climate. Several biomass datasets have been recently produced at different scales, using a variety of modelling approaches based on remote sensing data. Particular attention has been given to the tropical region where uncertainties are higher. Wall-to-wall biomass maps have been produced for the tropical belt by integrating different satellite and field observations using empirical modelling approaches calibrated by biomass field observations. Such empirical approaches rely on field observations for training and testing the biomass models but measuring biomass in the field is feasible only for a limited number of small sample areas. In addition, field observations are not error-free since they often employ generalized allometric relationships to estimate biomass. As a result, proper validation of regional products remains problematic, given their large area coverage and large mapping unit. The present research compiled a biomass reference dataset for the tropics with the aim of better understand the spatial accuracy of the existing regional maps, quantify their regional biases and develop an optimal data integration method. A variety of quality high-resolution biomass data were screened and harmonized, developing specific procedures to upscale the field observations to the map resolution and to combine datasets with different characteristics. The reference dataset includes research field observations, forestry inventory plots and high-resolution biomass maps underpinned by substantial field measurements. The reference data were used to identify errors and biases of the regional biomass maps and eventually integrate them with a data fusion approach, performed independently for each vegetation stratum to account for different error characteristics. The output of this analysis, performed in the context of the EU GEOCARBON project, was an improved pan-tropical forest biomass map at 1 km resolution.
Combining TanDEM-X and Landsat 8 data for biomass retrieval of boreal forests
Antropov, Oleg (1,2); Rauste, Yrjö (1); Praks, Jaan (2); Häme, Tuomas (1) 1: VTT Technical Research Centre of Finland, Finland; 2: Aalto University, Finland
Show abstractA solution to improve accuracies in forest biomass mapping has been searched via combination of data from different data sources or attracting additional reference data into biomass models. One critically important variable capable of improving biomass estimation is forest tree height. In this study, we assess the potential of combining forest tree height derived from interferometric SAR data with satellite optical data for improving accuracy of forest stem volume mapping. Study site was located near the Hyytiälä forestry station in central Finland, with terrain representative of the boreal coniferous forest. The ground reference data were information on stand level from forest management plans. A total of 104 stands with a dynamic range of 0-390 m3/ha were used in the study area. Average forest stand size was 4 ha, mean and median stem volume were 135 and 140 m3/ha. Forest stands of size smaller than 2 ha were not considered. The whole reference data was split into two equal groups: training stands and validation stands. Forest tree height is derived using interferometric coregistered single look complex (CoSSC) data of the TanDEM-X mission in presence of auxiliary ground DEM. TanDEM-X data used in the study spans several seasons. More particularly, a set of 12 dual-polarization (HH and VV) interferometric CoSSC pairs acquired during 2013 and 2014 were investigated. Also, a set of three single-polarization (HH) interferometric CoSSC pairs acquired during 2012-2013 in standard DEM production mode were studied separately. Multivariate regression is used to optimally combine results in stand-wise forest tree height estimation. Then, a combination of forest tree height information and optical spectral information is studied in order to obtain improved estimate of forest biomass. Landsat 8 images are used as satellite optical data in the study. Results provided by the combined model are compared to use of either optical data or InSAR-derived forest tree height separately. Quantitative results are reported in detail at the workshop.
EVOLUTION OF PLANT COVER AND CARBON STOCK IN THE FUELWOOD SUPPLY BASIN OF KINSHASA (D. R of CONGO)
GOND, Valéry; DUBIEZ, Emilien; BOULOGNE, Marine; PENNEC, Alexandre; PELTIER, Régis CIRAD, France
Show abstractContext: The European Makala project has map the trees and forests resources of the fuelwood supply basin of Kinshasa, in D. R. of Congo and predict its evolution. The analysis of past land cover change, used satellite imaging techniques to understand and document the spatial organization, mechanisms of forest degradation and the mechanisms for the recovery of forest stands. Method: Through high resolution LANDSAT satellite imagery, we mapped the supply area of the city of Kinshasa, from 1984 to 2012. Subsequently we applied for each class obtained an estimate of biomass measured in the field and finally we have applied volume tables to obtain the potential energy wood. Estimating aboveground biomass was conducted in 2012 as a result of forest inventories in the field. The 317 plots (4337 trees of 44 species) across the four types of plant cover were used to quantify the aboveground biomass. These measurements were performed over the area of Batéké plateau, in the valley of the river Lufimi. These initial data, combined with satellite data allowed the first comprehensive assessment of the aboveground biomass in the study area. Results: The average volume of wood available per hectare between 2000 and 2012 fell by more than 50% in 10 years. Through the combined use of field measurements and remote sensing, each finage of the supply basin can be analyzed individually. The Kinduala village, in Lower Congo, near Kisantu is under great anthropogenic pressure. The analysis shows a strong deforestation until 2005 to leave room in 2012 as an open landscape where juxtaposed agricultural parcels and savannah. The disappearance of woodland is responsible for the observed decrease in carbon stocks (less than 75% in 28 years). Conclusion: We have available for the first time qualitative and quantitative estimates of the dynamics of ecosystem degradation (forest and non-forest) in the supply area of Kinshasa. Drastic reduction of forest cover, significant decline in fallow periods and savannah, declining stocks of biomass and carbon constitute particularly strong signals. However, these initial estimates are derived from a small sample was extrapolated to the entire supply basin. Should be in the months and years ahead increased sampling to approach fairer and concrete values. Only a conscious policy and a sustainable community land management, with a very high dynamic of trees reintroduction in agricultural land, can reverse the curve.
Active-Passive Microwave Polarimetry for Soil Moisture Estimation
Jagdhuber, Thomas; Hajnsek, Irena; Montzka, Carsten; Papathanassiou, Konstantinos P.
Show abstractThe estimation of surface soil moisture with active or passive microwave data is a research topic since decades. As electro-magnetic waves in the microwave regime show sensitivity to dielectric particles (water) in porous media (soil), the link between soil moisture content and microwave signals was mostly conducted exclusively either in the active or in the passive domain and rather seldom jointly in both domains. However, both domains are linked by Kirchhoff’s law and provide a physical basis for a combined soil moisture estimation. In addition, both microwave techniques feature polarimetric acquisition capabilities, which add another information space for a joint, polarimetric retrieval of geo-physical parameters, like soil moisture. Triggered by these opportunities and by the upcoming Soil Moisture Active Passive (SMAP) mission in November 2014, the German Aerospace Center (DLR) and the Research Centre Jülich (FZJ) conducted an airborne active-passive microwave campaign in April and May 2013 within the TERENO Eifel observatory, near Jülich, in Western Germany. The fully polarimetric F-SAR sensor was operated together with the Polarimetric L-band Multi-beam Radiometer (PLMR2) in a multi-sensor configuration on DLR’s Do-228 platform to acquire an active-passive microwave data set at L-band. The PLMR2 sensor features medium-resolution mapping of H- and V-polarized brightness temperatures using multiple beams (+/-8°, +/-22°, +/-38° incidence) in across-track direction . The F-SAR sensor obtains fully polarimetric data in high spatial resolution (range: 4m, azimuth: 0.6m) in slant range configuration. The scattering/emission mechanisms and the respective scatterers/emitters, which can be also identified from in situ measurements of soil and vegetation provided by the Research Centre Jülich, are analyzed to understand their polarimetric signatures in active and passive microwave acquisitions . In this way, communalities and differences in polarimetric scattering/emission can be identified and understood for different land cover types. In a second step, methods to upscale the high-resolution SAR or to downscale the coarse resolution radiometer information to a common reference frame will be investigated to enable an active/passive fusion of the polarimetric observation spaces and to allow a direct comparison of signals. The polarimetric entropy can be used, for instance, to identify regions with homogeneous scatterers, which fit into the coarser radiometer footprints . In a third step, active and passive polarimetric descriptors are combined for a joint analysis of the fused observation space. This observation space will then be exploited for a polarimetric, active/passive inversion of soil moisture, in order to merge the benefits of both acquisition techniques.  Jagdhuber, T., Hajnsek, I. & Horn, R.: Polarimetric SAR Soil Moisture Processing Report: SOIMEX 2013 L-band Active / Passive Microwave Sensing. Processing Report, Project HA 310 - Helmholtz-Alliance: Remote Sensing and Earth System Dynamics, Oberpfaffenhofen, Deutschland, 2014.  Sorrentino, A., Jagdhuber, T., Hajnsek, I., Montzka, C., Papathanassiou, K.P.: Investigation of Polarimetric Active/Passive Microwave Scattering towards Soil Moisture Estimation. Proc. of EUSAR, June 3-5, Berlin, Germany, 2014.
Mapping tropical forest disturbance due to selective logging using Radarsat-2 data
Antropov, Oleg (1); Rauste, Yrjö (1); Häme, Tuomas (1); Frank Martin, Seifert (2) 1: VTT Technical Research Centre of Finland, Finland; 2: European Space Agency
Show abstractGenerally, detecting forest disturbance due to selective logging, resulting in temporal change of forest biomass, is quite a challenging task. Use of optical satellite data for this purpose is problematic due to near-permanent cloud coverage in the area of interest. Natural solution is to use SAR capability, especially relying on longer and denser time series. One particular focus is on ESA Sentinel-1 offering such an opportunity for frequent revisit. The problem is complicated by C-band being suboptimal for biomass state assessment, sometimes demonstrating a relatively week contrast between forested and non-forested areas, especially in rainy conditions. While potential of Sentinel-1 data itself is still yet to be evaluated, here we used a historical set of Radarsat-2 to develop mapping approach for detecting selective logging at C-band. A technique was developed for the mapping of selective logging in tropical forest using time series of Radarsat-2 images acquired both before and after the logging operations. The technique developed utilizes automatic detection of new forest roads, multi-temporal aggregation of ortho-rectified SAR imagery acquired before and after forest disturbance, followed by the analysis of textural features of SAR backscatter temporal log-ratio image. The study site was located in the northern part of the Republic of the Congo. The area is mostly covered with rain forest, some of which is on swamp where the forest canopy is lower. Few rivers and agricultural areas are also present in the study site, covering relatively small fraction of the study site. A time series of twelve Wide Multi-look Fine (MF3W) Radarsat-2 images acquired between November 2012 and December 2013 were studied in order to capture signs of forest disturbance. SRTM DEM was used in the ortho-rectification process. A radiometric normalization was performed to eliminate topography-induced radiometric variation. The power of each pixel was normalized with respect to the projected area of the scattering element. Time series or stack of co-registered images in the “gamma-nought” format was formed. Visual inspection of SAR scenes, as well as of individual temporal ratios and log-ratios between them has revealed only land-cover dependence or seasonal/weather artifacts. Due to limited sensitivity to biomass variation at C-band, the approach developed had to rely strongly on capturing spatial texture of backscatter log-ratio image. Multitemporal aggregation of SAR scenes allowed to decrease effects of speckle and made extracted textural features more prominent. The method performance was promising and convincing when cross-validation was done via visual interpretation of optical satellite imagery. The only Landsat 8 scene suitable for visual interpretation in 2013 was acquired just few months after the selective logging operation was performed.
Sensitivity of Dual Polarimetric X-band SAR Parameters to Grass Cutting Practices
Voormansik, Kaupo (1); Zalite, Karlis (1); Jagdhuber, Thomas (2); Hajnsek, Irena (2,3) 1: Tartu Observatory, Estonia; 2: German Aerospace Center (DLR), Germany; 3: ETH Zürich, Switzerland
Show abstractDuring the vegetative season of 2013 Estonian grasslands in Rannu were monitored intensively with SAR and in situ measurements in order to find a systematic behaviour of SAR parameters with regards to grass cutting practices. TerraSAR-X dual polarimetric HH/VV spotlight mode images were recorded every 11 days from the beginning of May until the end of July. Simultaneously with the satellite data takes, grass height, cutting status, wet and dry biomass and soil moisture were recorded in 11 different test fields of the Rannu area covering a wide range of different grass types in different growth stages. The dual polarimetric entropy/alpha decomposition (H2α) was applied and the polarimetric coherences were calculated from the TerraSAR-X data takes. Out of the many parameters investigated, a systematic behaviour was observed in H2α entropy and HH/VV polarimetric coherence concerning grass cutting practices. After grass cutting the entropy increased by up to 0.21 units and the HH/VV coherence decreased by down to 0.27 units. The H2α entropy parameter in the fields ranged from 0.6 to 0.9 whereas HH/VV coherences stretched from 0.3 to 0.7 in the different fields including all acquisition dates. The grass growing phase was generally associated to systematic entropy decrease and HH/VV polarimetric coherence increase. There as only an exception in the very early stage, from completely bare field to emerging vegetation, which was connected with entropy increase and HH/VV coherence decrease. In the study, data sets of 40.0° incidence angle ascending orbit (evening pass) as well as of 44.5° incidence angle descending orbit (morning pass) were incorporated to form a dense time series. Their derived SAR parameters showed no clear differences in behaviour due to the look angle variation. These results suggest that the observed grasslands are look direction invariant. In addition, no visible difference within the derived SAR parameters was observed when comparing morning and evening conditions within the grassland areas. In order to explain the behaviour of the observed SAR parameters, the particle cloud model for vegetation scattering, developed by S. R. Cloude, was applied. In this model the main vegetation constituents are synthesized with spheroidal particles having varying shapes and orientation randomness. As noted in the field surveys, the growing grass plants could be for instance modelled with primarily vertical dipoles whereas cutting changed the orientation of the plants to a more horizontal or more random layer. The modelling results are in line with the observed behaviour from the SAR data. According to the model horizontal dipoles (cut grass) correspond to lower HH/VV polarimetric coherence and higher H2α entropy values than vertical dipoles (growing grass). The increase of particle orientation randomness (after cut) corresponds to an increase in entropy and a decrease in HH/VV polarimetric coherence.
Motion Compensation Strategy for Airborne Differential Interferometric SAR data
Show abstractOver the last decades, Differential Interformetry (D-InSAR) using spaceborne SAR systems has become a powerful and established tool for detecting earth’s surface deformation and landslides on a sub-centimetre scale. Although the stable trajectory of a SAR satellite effectively eliminates artefacts related to imprecise knowledge of the acquisition geometry, the long data acquisition span can often represent a bottleneck, especially in terms of low coherence. On the other hand, airborne sensors can overcome these limitations by providing more flexibility in terms of short time intervals, spatial resolution and wavelengths used. Nevertheless, the airborne case introduces a further challenge since it is subject to the limitations of the adopted Motion Compensation (MoCo). The fact that the platform is not following an ideal rectilinear trajectory introduces several errors that must be compensated in order to achieve the accuracy required for D-InSAR (e.g. the phase must be better than 3° to reach millimetre accuracy in L band) . This study aims to describe the strategy and the algorithm used for processing a set of images acquired by DLR’s F-SAR sensor over an agricultural area near Wallerfing in south Germany. A major limitation introduced by the test site taken into account is the critical loss of coherence especially in the middle of the vegetation growth period. Even though the multi-squint (MS) approach has proven to be a reliable method for the estimation of residual motion errors, its performance is not optimal in the case of highly decorrelated scenes as frequently occurs in the case of vegetated areas and long temporal baselines . The adopted motion compensation strategy is summarized as follows: in order to cope with the temporal decorrelation, the short term SAR image pairs forming the stack are interferometrically processed using the DLR’s STEP processor, which features an interferometric processing chain including the conventional MS algorithm for residual motion estimates. Successively, the entire acquisition stack is modelled as a directed graph whose nodes contain the baseline errors estimated in the pair wise processing, each with respect to the predecessor node. Consequently, the baseline errors for long term acquisition pairs are computed by algebraically adding the baseline errors falling along the path from the first acquisition to the second one. Afterwards, all the slave tracks are updated with the correction so computed by explicitly correcting the underlying navigation data at the SAR raw data processing level. Successively, the complete stack is reprocessed, this time without using MS. The approach has been tested with F-SAR data acquired by DLR in the frame of the 14CROPEX campaign. The data set comprises 15 acquisitions in L- and C- band over the agricultural test site of Wallerfing (Germany), spanning three months and half from May 12th to August 4th, 2014. This time span covers a large part of the crops’ growth cycle, making this data set suitable for vegetation modeling and the study of agricultural parameter retrieval applications. REFERENCES  P. Prats, A. Reigber, J. J. Mallorquì, R. Scheiber, and A. Moreira, “ Advanced D-InSAR Techniques Applied to a time Series of Airborne SAR Data”, In: Proceedings of the IEEE Geoscience and Remote Sensing Symposium ) IGARSS, Barcelona, Spain, 2007.  A. Reigber, R. Scheiber, M. Jäger, P. Prats, I. Hajnsek, T. Jagdhuber, K.P. Papathanassiou, M. Nannini, E. Aguilera, S. Baumgartner, R. Horn, A. Nottensteiner, A. Moreira, “Very-High-Resolution Airborne Synthetic Aperture Radar Imaging: Signal Processing and Applications”, In: Proceedings of the IEEE, 101 (3), pp. 759-783, 2013.
On the retrieval of phenological stages of agricultural crops by means of C-band polarimetric SAR data in Barrax, Spain
Mascolo, Lucio (1); Lopez-Sanchez, Juan Manuel (2); Vicente Guijalba, Fernando (2); Ballester Berman, Josep David (2); Nunziata, Ferdinando (3); Migliaccio, Maurizio (3); Mazzarella, Giuseppe (1) 1: Università di Cagliari, Italy; 2: University of Alicante, Spain; 3: Università di Napoli "Parthenope", Italy
Show abstractThe monitoring of crop phenology is fundamental for the assessment of crop conditions, crop classification and crop yield, and it is also important for the management of land and water resources. Due to the sensitivity of polarization of microwaves to crop structure and dielectric properties of the canopy, which in turn depend on the crop type, retrieval of phenology of agricultural crops by means of polarimetric SAR (PolSAR) measurements is a promising application of this technology , especially after the launch of a number of polarimetric satellite sensors. As a matter of fact, different crops types, or the same type at different growth stages, produce different polarimetric signatures, which can be identified in the acquired images and used to establish classification algorithms aimed at the retrieval of the phenological stages. In this study, a time series of C-band Single Look Complex (SLC) RadarSAT-2 (RS2) fine quad-pol data collected over the Barrax region, Spain, from April to September 2009, is exploited to estimate the phenological stages of cereals fields. Ground truth information about crops growth stages are provided by the ESA-funded Agricultural bio/geophysical retrieval from frequent repeat pass SAR and optical imaging (AgriSAR) field campaign conducted in 2009. The RS2 data were previously used in , where the sensitivity of the backscattering coefficients to crops growth and soil condition was analyzed for corn, barley, wheat, onions and alfalfa fields. In this study, the estimation of phenology of cereals fields from the same images is carried out by first analyzing the evolution of a wide number of polarimetric observables (backscatter coefficients and ratios, correlations and phase differences, decomposition outputs, etc.) as a function of phenology. This analysis provides information about the most sensitive parameters and the scattering mechanisms that take part at each stage of the growth cycle. Based on this analysis, a reduced set is selected, from which a retrieval algorithm is constructed. This methodology has proven successful for rice monitoring [3,4] and, recently, for cereals and rape in another test site . References:  J. M. Lopez-Sanchez, J. D. Ballester. “Potentials of Polarimetric SAR Interferometry for Agriculture Monitoring”, Radio Science, Vol. 44, RS2010, March 2009.  M. S. Moran, L. Alonso, J. F. Moreno, M. P. Cendrero Mateo, D. F. De la Cruz, and A. Montoro. “A RADRASAT-2 Quad-Polarized Time Series for Monitoring Crop and Soil Conditions in Barrax, Spain”, IEEE Trans. Geosci. Remote Sens., Vol. 50, No. 4, pp. 1057-1070, 2012.  J. M. Lopez-Sanchez, S. R. Cloude, J. D. Ballester. “Rice Phenology Monitoring by Means of SAR Polarimetry at X-Band”, IEEE Trans. Geosci. Remote Sens., Vol. 50, No. 7, pp. 2695-2709, 2012.  J. M. Lopez-Sanchez, F. Vicente-Guijalba, J. D. Ballester, S. R. Cloude. “Polarimetric Response of Rice Fields at C-Band: Analysis and Phenology Retrieval”, IEEE Trans. Geosci. Remote Sens., Vol. 50, No. 5, pp. 2977-2993, 2014.  J. M. Lopez-Sanchez et al., “Estimating Phenology of Agricultural Crops from Space”, Proceedings of ESA Living Planet Symposium, ESA SP-722, Edinburgh, UK, Sept. 2013.
Toward aboveground biomass estimation with RADAR, LiDAR and optical remote sensing data in southern Mexico
Urbazaev, Mikhail; Thiel, Christian; Migliavacca, Mirco; Schmullius, Christiane
Show abstractInformation on the spatial distribution of aboveground biomass (AGB) over large areas is needed (1) for understanding and managing the processes involved in the carbon cycle, and (2) supporting international policies for climate change mitigation and adaption. Using remote sensing techniques it is possible to provide spatially explicit information of AGB from local to global scales. In this work we present the first results on the use of multi-sensor remote sensing data to estimate AGB over three test sites in southern Mexico. In order to develop a set of AGB retrieval algorithms, we firstly compared different SAR parameters (e.g. multi-polarized backscatter intensities) obtained from ALOS PALSAR sensor and Landsat imagery with field-based AGB estimates using empirical regressions and analyzed the relationships between them. The next steps of the work will be development of a two stage up scaling approach: firstly, to enlarge the cal/val data, we propose to estimate AGB along airborne LiDAR (from G-LiHT sensor) transects using field-based AGB and LiDAR metrics. With LiDAR-based AGB we will then calibrate SAR parameters in a non parametric model (e.g., randomForest) to create AGB maps over the study areas. An overall aim of the study is the analysis of capabilities and limitations of SAR data for AGB mapping and the investigation of the potential synergistic use of SAR, LiDAR and optical systems. The proposed monitoring tool will facilitate quantitative estimations in loss of carbon storage and support the selection of terrestrial (e.g. tropical dry forests, shrublands) sites for conservation priorities with high value for the national carbon budget.
Quantifying the causes of deforestation and degradation and creating transparent REDD+ baselines: a method and case study from central Mozambique
Ryan, Casey M (1); Berry, Nicholas J (1); Joshi, Neha (2) 1: University of Edinburgh, United Kingdom; 2: University of Copenhagen, Denmark
Show abstractReductions in deforestation and forest degradation are advocated as a means to mitigate climate change. The formulation and implementation of policies to achieve such reductions requires an understanding of current and historic land-use change and associated greenhouse gas emissions. In addition, it is often proposed that any reduction in emissions be measured against a reference scenario that describes future land-use in the absence of intervention. However, the information needed to progress this agenda is rarely available, as robust data on the extent and causes of land-use change, and the associated changes in carbon stocks, are sparse, particularly in African woodlands. Here we present a novel method for obtaining such information by combining data from radar remote sensing and ground surveys with a simple aspatial model. Using this approach we quantify changes in woody biomass and investigate the land-use activities that caused these changes in a 7500 km2 area of Manica province, Mozambique. We use the data to construct a model linking the activities causing biomass loss to hypothesised drivers, allowing the definition of future scenarios. Within the study area, biomass was lost at a rate of 2.8±1.9% per year, from 19.4±0.9 TgC in 2007 to 17.6±0.9 TgC in 2010. Small-scale agriculture was the direct cause of 46±17% of the total biomass loss, followed in magnitude by construction and miscellaneous activities (24±11%), charcoal production (18±9%), logging (9±5%) and large-scale agriculture (3±2%). Uncertainties remain on the biomass accumulated by regrowing vegetation. Extrapolating into the future, a scenario that includes projected population growth shows 41% of forest biomass being lost from 2010-2020 (a loss of 7.2 TgC). A scenario of intensive policy interventions gives reduced losses of 3.8 TgC by factoring in improvements in crop yields, charcoal production efficiency, and sustainable timber harvesting. Our case study demonstrates the importance of low intensity losses of biomass in African woodlands, and highlights the broad range of activities that will need to be addressed to develop locally appropriate mitigation actions. The simple modelling framework allows for the transparent creation of scenarios in data sparse areas, which could be used as local or national reference emissions levels under REDD+.
FOREST ABOVE GROUND BIOMASS EXTRACTION FROM AIRBORNE P-BAND POLSAR DATA
Feng, Qi; Chen, Erxue; Li, Zengyuan; Zhao, Lei; Li, Lan
Show abstractThe performance of one forest above ground biomass (AGB) estimation model using P-band polarimetric synthetic aperture radar (PolSAR) data was evaluated over one forested area with complex topography and forest structure in the Northeast of China. The P-band PolSAR data used were acquired by one China airborne system in 2013. Airborne LiDAR and ground forest plots data were acquired in both year 2012 and 2013, and the AGB extracted from LiDAR and plots data will be used as ground truths to evaluate the model. We will introduce the key SAR data processing steps, such as, SAR image geocoding terrain correction (GTC), forest AGB estimation from LiDAR and ground plots data. Then, we will investigate the relationship between SAR backscatter coefficients and LiDAR derived AGB. Furthermore, one empirical inversion model, involving the local incidence angle derived from LiDAR DEM through SAR image simulation, was used to estimate the AGB, and the performance of this model will be compared with the other models. Preliminary results show that the model has good performance, the R2 is 0.6 and AGB estimation RMSE is 15.98 ton/ha with thousand LiDAR derived AGB plots as reference data.
Suitability of different observables and polarizations for DInSAR-based soil moisture estimation
Zwieback, Simon (1); Hajnsek, Irena (1,2); Hensley, Scott (3) 1: ETH Zurich, Institute of Environmental Engineering, Zurich, Switzerland; 2: German Aerospace Center (DLR), Microwaves and Radar Institute, Wessling, Germany; 3: Jet Propulsion Laboratory, Radar Science & Engineering, Pasadena, USA
Show abstractDInSAR measurements can be affected by soil moisture changes: these do not only influence the phase (leading to spurious deformation estimates) but also the coherence and the phase triplets. Here we propose to analyse based on a first-order solution of Maxwell's equations whether these impacts on the DInSAR data allow in turn the retrieval of soil moisture from such data. We focus particularly on the suitability of different polarizations and of the different observables. All three of these polarimetric observables are found to be insensitive to certain patterns of the soil moisture time series (such as an offset). These properties of the soil moisture time series thus cannot be reliably estimated using the particular observable. By contrast, the impact of the polarization is found to be considerably smaller.
On the Potentials of PolInSAR Inversion for Monitoring Long-Term Forest Height Dynamics at L-band
Bertoluzza, Manuel; Pardini, Matteo; Kim, Jun Su; Bruzzone, Lorenzo; Papathanassiou, Kostas
Show abstractForest height is one of the most important parameters for (forest) stand characterization, and it is closely related to biomass. It provides information on stand condition and site index, and it allows characterizing the successional state of the forest. At the same time, the distribution of forest heights within a stand may be used to assess the disturbance regime and to detect logging activities. In the characterization of dynamic forest processes, the knowledge of forest height changes is even more important than absolute height measurements itself. Indeed, forest height changes can be directly used to characterize forest growth, mortality and deforestation, and to conclude about the associated carbon fluxes independently from the successional status of the forest. Forest denormetric variables are usually measured by airborne laser scanning by using LiDAR instruments that are very precise in characterizing the tree heights and the main related parameters. However, LiDAR technology is expensive and in many cases not affordable, especially in a multitemporal monitoring scenario. Low frequency synthetic aperture radar (SAR) sensors in a polarimetric-interferometric configuration represent a very promising alternative tool to LiDAR for estimating forest height and its dynamics as they can provide large continuous coverage and high spatial and temporal resolution with a limited cost. From the methodological point of view, polarimetric SAR interferometry (PolInSAR) combines the inherent sensitivity of the interferometric coherence to the vertical structure of volume scatterers combined with the potential of SAR polarimetry to interpret and characterise the individual scattering processes. Exploiting this combination, forest height estimations obtained in the last years from a pre-operational to an operational PolInSAR product that have been validated in the frame of several campaigns over a wide range of forests, terrains and environmental conditions. The overall obtained estimation error is in the order of 10% or smaller. The objective of this work is to assess qualitatively and quantitatively the capability of monitoring forest dynamics by using PolInSAR height maps. To this purpose, a multibaseline-multitemporal PolInSAR L-band airborne dataset over the temperate forest site of Traunstein (South of Germany) has been used. Data have been acquired since 2003 with the DLR’s E- and F-SAR radar platforms covering more than a decade, with temporal baselines ranging from hours to years. For each acquisition date, a multibaseline PolInSAR inversion with short temporal baseline (in the order of one hour) has been carried out. Accordingly, a time series of accurately estimated forest heights has been obtained without disturbances due to (long-term) temporal decorrelation. Depending on the time lag between two different height maps, several comparisons could be carried out in order to characterize height differences induced by scattering changes within the same stand and/or variations of the baseline distribution. Such scattering changes can in general be caused by different sources, like the water content of the vegetation (e.g. before and after a rain event), the presence of wind, the seasonal cycles, disturbances and management activities. These components could be observed in the data and their effects characterized. Extensive results of this analysis will be presented, and the capability of distinguishing between different sources of PolInSAR height changes will be discussed together with their potentials for identifying and monitoring ecosystem changes.
The TEAM initiative
Sheil, Douglas 1: Norwegian University of Life Sciences, Norway; 2: CIFOR, Indonesia
Show abstractThe poster will outline the tropic wide network of permanent plots maintained by the TEAM network.
The concept and results of new polarimetric signature-based decomposition
Porzycka-Strzelczyk, Stanisława; Strzelczyk, Jacek
Show abstractIn this work the concept of new, model-based decomposition of quad-polarimetric SAR data will be presented. This method is based on the analysis of polarimetric signatures of canonical objects (trihedral, dihedral, helix) and models representing volume scattering. Using these signatures, that are treated as patterns, the single-bounce, double-bounce and helix and volume scattering mechanisms are identified. In proposed method the recognition of the amount of particular scattering mechanism within one pixels of SAR images is preceded by division of pixels into classes. Two main of them are: canonical coherent scatterers and incoherent scatterers. The latter is further divided into 4 following classes: built-up areas with reflection symmetry, built-up areas without reflection symmetry, vegetation and surface. After that, for each class the appropriate set of polarimetric signatures of canonical objects is used in the pattern recognition-based decomposition procedure. Such approach help to overcome the problem of distinction between volume scattering and scattering form rotated man-made objects. The validation of the proposed decomposition method was done for TerraSAR-X quad-polarimetric data of Mahlow region that is located on the South of Berlin. The decomposition results have been compared with results of Yamaguchi four-components decomposition as well as with optical images of the studied region. Based on this comparison the pros and cones of proposed decomposition method have been defined and described in detail.
SAR image coregistration procedure using dual-polarimetric signatures
Strzelczyk, Jacek; Porzycka-Strzelczyk, Stanisława
Show abstractIn this work the concept and preliminary results of SAR image coregistration procedure based on dual-polarimetric signatures will be presented. In the proposed method, both coarse and fine SAR image coregistration procedures have been proposed without taking into account the coherence parameter. In this method the coregistration is performed using canonical coherent scatters (CCS) that are recognized using dual-polarimetric signatures calculated for trihedral, dihedral and helix. The CCS are identified in the master and slave SAR images. For each CCS some parameters like total power, pedestal and type of scattering mechanism (SB, DB, HX) are determined. After that, the set of CCSs in slave images is fitted to the set of CCS in the master image. Like in the classical approach, in the proposed procedure the spatial window is used. However, instead of using coherence parameter the quality of the fit is assessed using differences between parameters that characterized each CCS.
3D Forest Structure Estimation from SAR Tomography by means of a Full Rank Polarimetric inversion based on Compressive Sensing
Cazcarra Bes, Victor; Tello Alonso, Marivi; Papathanassiou, Kostas
Show abstractIn the context of forest applications, remote sensing observations have been mainly focused up to now on the evaluation of 2D spatial distributions of image patterns. However, in the last years, attention is shifting more and more from 2D to 3D information by adding vertical structure observables, since it has been demonstrated that the horizontal distribution is not sufficient for an appropriate assessment of many physical parameters of interest. In front of this need, SAR tomography (TomoSAR) techniques allow a direct 3D imaging of the forest body by exploiting angular diversity with different passes of the sensor. One of the main drawbacks of SAR tomography is that the estimation of the vertical reflectivity profile has to be performed through a limited set of multibaseline acquisitions, which requires solving a highly underdetermined system of equations. In order to do so, in the TomoSAR literature, the most widely employed inversion techniques are up to now the Capon and the Beam Forming spectral estimators . As an alternative, the application of Compressive Sensing (CS) techniques to the estimation of forest profiles has been recently introduced . In this paper, a different algorithm based on CS is proposed. It performs a full rank polarimetric inversion, allowing thus an estimation of 3D coherency matrices. Essentially, the theory of CS assumes that in a highly underdetermined system of equations, the unknown signal can be recovered from a few measurements with a high probability by solving a minimization problem, provided that it is sparse or compressible in a certain projection space and that the sensing matrix satisfies the Restricted Isometry Property . Compared to the full rank polarimetric inversion with the Capon spectral estimator , the main difference introduced by the CS algorithm proposed is that the inversion is not performed independently for every height, but on the contrary, for each spatial resolution cell in range and azimuth, the whole set of coherency matrices along the vertical dimension is retrieved at once, assuming some degree of correlation between them. Once the coherency matrices are estimated, 3D classical polarimetric parameters, such as the entropy, the anisotropy or the alpha angle, among others, can be evaluated. In order to study the goodness of the full rank polarimetric TomoSAR inversion approach proposed, it is applied to a temporal series of airborne data, acquired by the E-SAR system of the Microwave and Radar Institute of the DLR. The multibaseline acquisitions are fully polarimetric at L-band and they cover the area of Traunstein, in the Southeast of Germany. The scenario monitored is constituted by a highly heterogeneous managed forest in a temperate climate, with coniferous, deciduous and mixed stands at different growth stages. The results of the 3D polarimetric inversion will be contrasted to in situ measurements at two different spatial scales. In a first step, inventory plots provided by the Forest Growth and Yield Department of the Technical University in Munich (TUM) in the framework of the HGF Alliance will be considered. Each inventory plot covers an area of around 20x20 meters and it gives the spatial distribution of the different trees within this area, together with some of their morphological characteristics, such as height, diameter of the crown, diameter of the stem… Therefore, the comparison of polarimetric TomoSAR products and in situ data at inventory plot level will allow a first assess of the consistency of the results of the inversion technique proposed, especially in terms of the quality of the polarimetric parameters derived. In a second step, homogeneous stands will be considered. The objective is to analyze if distinct 3D polarimetric signatures are obtained for the different types of stands. References  L. Ferro-Famil, Y. Huang, and A. Reigber. High-resolution SAR tomography using full rank polarimetric spectral estimators. Proc. IGARSS, July 2012.  E. Aguilera, M. Nannini, A. Reigber. Wavelet-Based Compressed Sensing for SAR Tomography of Forested Areas. IEEE Trans. Geosci. Remote Sens., pp(99): 1, 2013  R. Baraniuk. “Compressive Sensing”. IEEE Signal Processing Magazine, 24(4), pp. 118-121, 2007.  E. Candes and M. Wakin, “An Introduction To Compressive Sampling” IEEE Signal Processing Magazine, vol. 25, no. 2, pp. 21–30, Mar.2008.
Derivation of Soil Moisture Patterns on Agricultural Fields from SAR-data in the Rur Catchment
Esch, Sabrina; Korres, Wolfgang; Reichenau, Tim; Schneider, Karl
Show abstractSoil moisture is one of the main variables in hydrology, meteorology and agriculture, as it determines the partitioning of both incoming solar radiation into latent and sensible heat and precipitation into surface runoff and infiltration. Nevertheless, its spatio-temporal patterns in agriculturally used landscapes, that are affected by multiple natural (rainfall, soil, topography etc.) and agronomic (fertilisation, soil management etc.) factors, are often not well known. The aim of this study is to derive a long term time series of surface soil moisture patterns from C-band SAR data in order to create a soil moisture climatology, to observe long term surface soil moisture pattern development and to identify the main drivers for these patterns in a mesoscale catchment (2354 km^2). A time series of 124 single-channel C-band ERS SAR scenes with a spatial resolution of 20 m from 1992-2003 was used to create a soil moisture climatology. Our research area, the Rur catchment, is located in Western Germany and is part of SFB/TR32 “Patterns in soil, vegetation, atmosphere systems: monitoring, modeling and data assimilation”. It consists of two major landscape units, with the northern part belonging to the mainly agriculturally used fertile loess plain and the southern part belonging to the low mountain range of Eifel, dominated by forests and pasture. Surface soil moisture maps of agricultural areas and pasture were calculated, using a semi-quantitative soil moisture index concept, grouping the backscatter values from the SAR scenes into different classes of soil moisture content. This approach was used because no ground truth data was available for that period, so that calibration/validation of a quantitative semi-empirical inversion scheme was not possible. The radar backscattering coefficient (sigma nought) is sensitive to vegetation, surface roughness and soil moisture content. To eliminate the influence of vegetation and surface roughness and to derive soil moisture information from single channel SAR, the data was sorted into classes before the analysis as follows. Using land use maps for each year (derived from optical remote sensing) the data was grouped by agricultural land use class (cereals, sugar beet, maize, bare soil, pasture) and stage of the vegetation development. Within each class the influence of biomass and surface roughness is assumed to be constant. Thus, changes in backscatter intensity can be accounted to changes in surface soil moisture. Within each group, we assume that pixels exist which represent a wet and a dry soil moisture state. An index was spanned between high and low backscatter values, thus identifying wet and dry areas. By using soil parameters (porosity, field capacity and wilting point) according to the soil texture of the given location, the qualitative index, indicating high and low surface soil moistures, can be converted into volumetric soil moisture.The approach was tested i) using the full resolution 20 m pixels for the computing of the histograms and ii) using medians of the agricultural fields for the histograms. In order to validate the approach, resulting soil moisture maps were compared to modelled soil moisture patterns. The surface soil moisture maps were then used to create a soil moisture climatology, which depicts the typical, average state of soil moisture for every month. The climatology can be used as a reference for comparison to upcoming soil moisture studies in the Rur catchment. From the vegetation period of 2015 on, dual-and full-polarimetric SAR data at C- and L-band from ALOS-2, Radarsat-2 and Sentinel-1 will be combined to enhance knowledge of surface soil moisture pattern evolution in the Rur catchment. Based on previous work in our research group on ALOS-1 data, polarimetric decomposition techniques will be used in order to increase the amount of SAR observables and to develop new polarimetric retrieval algorithms, which then can be used to estimate soil moisture content, biomass and surface roughness independently.
GlobBiomass - Estimates of Biomass on a Global Scale
Schmullius, CC (1); Thiel, C. (1); Pathe, C. (1); Quegan, S. (2); Carreiras, J. (2); Wegmüller, U. (3); Santoro, M. (3); Balzter, H. (4); Tansey, K. (4); Le Toan, T. (5); Mermoz, S. (5); Bouvet, A. (5); Dabrowska-Zielinska, K. (6); Hoscilo, A. (6); Siegert, F. (7); Englhardt, S. (7); Fransson, J. (8); Herold, M. (9); Avitabile, V. (9); Häme, T. (10); Rauste, Y. (10); Shvidenko, A. (11); Fritz, S. (11); Schepaschenko, D. (11); Sterenczak, K. (12); Eriksson, L. (13); Ulander, L. (13); Reichstein, M. (14); Carvalhais, N. (14); Rodriguez-Veiga, P. (4); Ziolkowski, D. (6); Matejka, E (1); Seifert, FM (15) 1: University Jena, Germany; 2: University Sheffield, UK; 3: Gamma Remote Sensing, Switzerland; 4: University Leicester, UK; 5: CESBIO, France; 6: IGIK, Poland; 7: RSS, Germany; 8: SLU, Sweden; 9: WUR, Netherlands; 10: VTT, Finland; 11: IIASA, Austria; 12: FRI, Poland; 13: Chalmers, Sweden; 14: MPI-BGC, Germany; 15: ESA ESRIN, Italy
Show abstractFor the upcoming three years, a new Data User Element (DUE) will focus on creating improved knowledge about the Essential Climate Variable Biomass. The main purpose of the DUE GlobBiomass project is to better characterise and to reduce uncertainties of AGB estimates by developing an innovative synergistic mapping approach in five regional sites for the epochs 2005, 2010 and 2015 and for one global map for the year 2010. The project team includes leading Earth Observation experts of Europe and is linked through Partnership Agreements with further national bodies from Brazil, Canada, China, Russia and South Africa. Progress will be closely monitored by an Advisory Board and a User Group and further intergovernmental organisations included through Letters of Interest. GlobBiomass will demonstrate how EO observation data can be integrated with in situ measurements and ecological understanding to provide improved biomass estimates that can be effectively exploited by users. The target users will mainly be drawn from the climate and carbon cycle modelling communities, but will include users concerned with carbon emissions and uptake due to biomass changes within initiatives such as REDD+. It will provide a harmonised structure that can be immediately exploited to address user needs for biomass information, but will be capable of being progressively refined as new data and methods become available. Its detailed objectives are: 1. provision of improved quantitative biomass maps at regional and global scale by making best use of satellite and in situ data, coupled with knowledge of ecological and social constraints; in addition to provide quantified estimates of biomass changes within the limitations imposed by data; 2. provision of associated uncertainty maps; 3. validation of the biomass maps by working with major holders of biomass measurements and user organisations, and working with them to establish common practices and standards; 4. contribution of new scientific results on biomass stock and change estimation; 5. with users, assessing the impact of these new products; 6. identification of the limitations of current data and methods to estimate biomass, and how these may be reduced by new methods and new missions, including the Sentinels, BIOMASS, SAOCOM and ALOS-2. The DUE GlobBiomass project involves several categories of external project members of either users or close associates to user organisations: 1) Regional Forestry Representatives from the five selected regions for regional biomass estimation in Sweden, Poland, Cameroon, Indonesia and Mexico, 2) the GlobBiomass User Group dedicated to the project through Commitment Letters 3) dedicated Partnership Agreements with representatives from biomass monitoring networks in Australia, Brazil, Canada, China and Russia, 4) Endorsement Letters from the following relevant international programmes CIFOR, GCP, GEOBON, GEOCARBON, GFOI, ICOS and TERN, and last-but-not-least, 5) an Advisory Board consisting of international EO-experts for biomass monitoring who themselves closely cooperate with user organisations. The global biomass estimates and related uncertainties have two major innovative features: 1. They will be derived from a combination of SAR, LiDAR and optical datasets further supported by auxiliary EO-derived datasets (land cover, land surface temperature etc.) and in situ. We intend to maximize the information available from spaceborne EO by exploiting the most beneficial aspects of each in terms of biomass estimation. Furthermore, the biomass estimation will rely as much as possible on physically-based methods and adapt to regional forest and environmental conditions. Existing algorithms will also be expanded in their ability to adapt to local forest and environmental conditions. The inclusion of land-cover, soil moisture, land surface temperature, etc. in the process of model parameterization is a new and important development. 2. The mapping shall be pursued at an unprecedented resolution (< 100 m). It is intended to start from the native resolution of the input EO data and derive biomass estimates at that scale. While it is known that individual estimates are of poor quality, data combination as well as spatial aggregation is then foreseen to improve the results. Such activity implies a detailed analysis of scaling issues, which is unprecedented. The comparison between global and regional products is likely to lead to considerable insight into flaws in the global products and what may be necessary to mitigate them. This should give greater understanding of how to optimise both data (in situ and satellite) and methods to give the best possible biomass maps, and the implications for how GlobBiomass should develop beyond this project. This presentation will give an overview of the technical prerequisites and challenges for the first year of the DUE GlobBiomass project.
TRIHEDRAL CORNER REFLECTOR FOR POLARIMETRIC CALIBRATION OF BIOMASS
Show abstractTrihedral radar reflectors are passive and low-cost alternatives for polarimetric calibration of the spaceborne P-band SAR on ESA’s future BIOMASS mission. One of the main challenges with BIOMASS will be to mitigate Faraday rotation induced by the Earth’s ionosphere. A polarimetric calibration technique based on trihedrals is described in [1,2] which includes estimation of Faraday rotation using the method by Bickle and Bates . The calibration technique is based on the assumption that the polarization cross-talk is sufficiently small to be neglected, which is also expected to be the case for the BIOMASS antenna. In such a case, the two channel imbalances can be estimated from the average power over distributed (assuming reciprocity) targets in the two cross-pol channels together with the trihedral co-pol responses. The technique has been tested and validated by simulations and measurements using data from PALSAR-1 on ALOS-1 at L-band . In this past work, however, it was assumed that the trihedrals were optimally pointed towards the satellite orbit to give the maximum co-pol RCS and that the cross-pol RCS could be neglected. Hence, the trihedrals must be repointed for each different orbit. In the present paper, we investigate the possibility of mounting the trihedrals in a fix position and use them for polarimetric calibration across multiple offset satellite orbits. This is a relevant scenario for BIOMASS which will have a drifting orbit during its mission. The expected calibration performance depends on two main characteristics, i.e. 1) large and predictable RCS for both HH- and VV-polarization and 2) a small cross-pol RCS compared to the co-pol RCS. The first condition implies that signal-to-clutter and signal-to-noise is sufficient to fulfil RCS calibration uncertainty requirements  as well as to determine the channel imbalances. The second condition suggests that the Bickle and Bates method will perform well in estimating Faraday rotation. The trihedrals developed for PALSAR-1 have a (short) side length of 5.1 m which implies that the RCS is 47.2 dBm2 based on geometrical optics . The corresponding RCS for the BIOMASS center frequency is 9.3 dB lower or 37.9 dBm2, again assuming geometrical optics. The resulting signal-to-clutter and signal-to-noise will depend on the equivalent backscattering coefficient for clutter-and-noise multiplied with the equivalent single-look image resolution determined by the processed pulse and Doppler bandwidths. The equivalent single-look resolution cell of BIOMASS will be about four times larger compared to PALSAR-1. However, the clutter-and-noise equivalent backscattering coefficient is typically lower over open fields at P-band where the trihedrals have been deployed. The smaller trihedral RCS and larger resolution area for BIOMASS implies a 15 dB disadvantage compared to PALSAR-1. However, this value is partly compensated by the lower clutter backscattering coefficient at P-band. A 10 dB lower backscattering coefficient is not unrealistic and would thus offset the disadvantage to 5 dB. Further work is required to quantify the backscattering coefficients of open fields and thus define the trihedral size requirement. The polarimetric scattering matrix for a trihedral with a side length of 5 m has been computed using a commercial software package (SPECRAY) based on physical-optics. The complex scattering amplitudes have been determined for a set of elevation and azimuth angles as well as frequencies. The data set forms the basis for investigating the co- and cross-pol signatures of the trihedral. In this paper, we restrict the analysis to close to the symmetry axis in azimuth and +/- 10 deg from the maximum RCS in elevation angle which encompasses the typical range for satellite SARs. SPECRAY results indicate that the cross-pol RCS will be significantly less than the co-pol RCS, i.e. sufficiently small to be neglected. We conclude that large passive triehdrals could be useful for polarimetric calibration of BIOMASS provided that the antenna cross-talk can be neglected.  G. Sandberg, L.E.B. Eriksson, and L.M.H. Ulander, Measurements of Faraday Rotation Using Polarimetric PALSAR Images, IEEE Geoscience Remote Sensing Letters, 6 (1), 142-146, 2009  A. Freeman, Calibration of linearly polarized polarimetric SAR data subject to Faraday rotation, IEEE Trans. Geosci. Remote Sens., 42 (8), 1617–1624, 2004  S. H. Bickle and R. T. Bates, Effects of magneto-ionic propagation on the polarization scattering matrix, Proc. IEEE, 53 (8), 1089–1091, 1965  L.M.H. Ulander, Accuracy of using Point Targets for SAR Calibration, IEEE Trans. Aerospace Electron Syst., 27 (1), 139-148, 1991
Demonstrating the Potential of ALOS PALSAR Backscatter and INSAR Coherence for Forest Growing Stock Volume Estimation in Central Siberia
Thiel, Christian; Christiane, Schmullius
Show abstractDuring its lifetime, ALOS PALSAR acquired a valuable L-band SAR data set of to date unprecedented quality and quantity. Thanks to the sophisticated observation strategy and the high acquisition success rate, a large amount of PALSAR data is globally available. So far, the potential of ALOS PALSAR L-band coherence for the estimation of forest GSV (growing stock volume) in the boreal forest has hardly been investigated. Moreover, ALOS PALSAR backscatter and coherence has not yet been jointly used to delineate GSV in Siberia. This work investigates the capability of L-band backscatter and coherence data for the estimation of GSV in Siberia, Russia. Altogether, 87 PALSAR images are used and eleven forest inventory sites are investigated. It is observed that coherence acquired at frozen conditions offers the largest potential for GSV estimation. For single coherence images saturation occurs in average at 230 m³/ha, the average correlation (R²) between coherence and GSV is 0.58. PALSAR backscatter is also sensitive for GSV. However, saturation occurs at lower GSV levels (75 - 100 m³/ha). Moreover, the average R² is lower (0.42 - 0.48). HV backscatter offers a slightly greater potential then HH backscatter. For one exemplary forest inventory site a GSV map was generated and validated. For the delineation of the map three coherence images acquired at frozen conditions and six backscatter images acquired at unfrozen conditions are used. The total area covered by the map is approximately 4,900 km² and the pixel spacing is 25 m x25 m. The delineation is based on a rather simple and straight forward approach that could be operationally applied over large areas. The generated GSV map features a relative RMSE of 26%. Thus, the PALSAR based map has an accuracy which is close to the accuracy of the forest inventory data.
Temporal decorrelation affect on tomographic SAR for hemi-boreal forest aboveground biomass estimation
Li, Wen Mei (1); Chen, Erxue (2); Li, Zengyuan (2); Feng, Qi (2) 1: Nanjing University of Post and Telecommunications, China, People's Republic of; 2: Institute of forest resources information technique, Chinese Academy of Forestry, Beijing, China
Show abstractForest aboveground biomass (AGB) study is a crucial field of research and has received increasingly attention recent years. Remote sensing techniques based on airborne or spaceborne platforms have provided useful information for forest AGB and forest height estimation. Among these techniques, tomographic synthetic aperture radar (Tomo-SAR) is an effective tool for forest vertical structure retrieval. The rationale of Tomo-SAR is to employ multiple flight tracks, which are usually parallel to each other and characterized by small differences in the incidence angle. The parallel flight lines allows us to form a synthetic aperture not only along each track but also in elevation. It follows that the signal could be integrated over a 2-D synthetic aperture, and leads to 3-D resolution capabilities. The 3-D resolution cell is determined by the pulse bandwidth in the slant range direction and by the length of the synthetic aperture in the azimuth and the elevation direction. Pol-InSAR data collected by repeat-path airborne or spaceborne platforms usually has time interval, from dozens of minutes to days or months. Therefore, temporal decorrelation is not disregarded and has to be considered, becoming an important challenge for the airborne Tomo-SAR. The research of Dinh et al. shows that the signal coming from upper vegetation layers exhibits a much more stable connection to the forest AGB, it is mean that Tomo-SAR could be used for the tropical forest AGB estimation. The finding was applied to define a simple tropical forest AGB estimator based on the backscatter associated with upper layers only, and finally resulting in accuracy with respect to in-situ data of about 10% at tropical forest test site. There are few research on hemi-boreal forest AGB estimation with Tomo-SAR. The first objective is to assess the affect of temporal decorrelation on Tomo-SAR, and the second objective is to construct a hemi-boreal forest AGB estimator with Tomo-SAR technique. Repeat-path multi-baseline P-band Pol-InSAR data were collected during March and May in 2007 at Remningstorp test site. The largest time interval is about 60 days and the smallest time interval is about 20 minutes, which are key information for temporal decorrelation research.
Polarization Orientation Angle Compensation of PolInSAR Data for building model reconstruction
Wang, Changcheng; Wang, Ping; Hellwich, Olaf
Show abstractThe polarization orientation angle (POA) is related with the azimuth slope of the surface, the range slope and radar look angle[1,2]. For the urban area, the special structure of buildings will cause the shift of polarization orientation angle for Polarimetric SAR data. For example, the tilted roof or the vertical wall which is not perpendicular to the radar incident plane. This type of polarization orientation angle shift will affect the accuracy of building model reconstruction by using polarimetric interferometric SAR (PolInSAR) data. It will cause the interferometric phase errors. Generally, the phase errors caused by the sloped terrain are ignored in traditional single-channel InSAR data processing. This type of phase errors are more obvious for urban building model reconstruction than that of other applications. This paper investigates interferometric phase correction for polarimetric interferometric SAR data by using polarization orientation angle compensation. We select a pair of high-resolution full polarization TerraSAR-X data for analyzing the errors caused by POA shift. Experimental results indicate that the interferometric phase with POA compensation is more reasonable. Therefore, the polarization orientation angle compensation is necessary for high resolution polarimetric interferometric SAR data processing. Key words- synthetic aperture radar (SAR), polarization orientation angle (POA), interferometric phase, polarimetric interferometric SAR (PolInSAR). References  J. S. Lee, D. L. Schuler and T. L. Ainsworth, “Polarimetric SAR data compensation for terrain azimuth slope variation,” IEEE Trans. Geosci. Remote Sens., vol. 38, no. 5, pp. 2153-2163, Sep. 2000.  J. S. Lee, D. L. Schuler, T. L. Ainsworth, E. Krogager, D. Kasilingam and W. M. Boerner, “On the estimation of ratar polarization orientation shifts induced by terrain slopes,” IEEE Trans. Geosci.Remote Sens., vol. 40, no. 1, pp. 30-41, Jan. 2002.  K. Iribe and M. Sato, “Analysis of polarization orientation angle shifts by artificial structures,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 11, Nov. 2007.
ESTIMATION OF ABOVEGROUND BIOMASS IN LARGE AREAS COMBINING LIDAR DATA AND FOREST IN SITU NETWORKS: LESSONS LEARNED IN SPAIN AND COSTA RICA
FERNÁNDEZ, ALFREDO; MARCHAMALO, MIGUEL; ALGEET, NUR; FERNANDEZ-MOYA, JESUS; NAVARRO, JOSE ANTONIO; EZQUERRO, PABLO; REJAS, JUAN GREGORIO; BONATTI, JAVIER; MARTINEZ, RUBEN; TOME, JOSE LUIS
Show abstractThe estimation of aboveground biomass is a key parameter for the assessment of ecological and economical goods, services and trade-offs in forests nowadays. Traditional forest inventories, aimed at estimating timber existences for a given purpose, are giving path to large-scale biomass inventories in the framework of REDD and REDD+ initiatives. This work presents the experience of biomass assessment in two contrasting countries: Costa Rica and Spain, as well as the outlines of future activities to improve these estimates. The "Comunidad, Agua y Bosque-CAB" Program, comprises a set of actions developed by the Technical University of Madrid and the University of Costa Rica in partnership with other institutions, as Agresta S.Coop., aiming at comprehensively address the issue of sustainable management of forests, water and soil resources in Central America under a community perspective. The Program has developed several demonstrating case studies since 2007, using geomatic techniques, intense field work and coordination with focus social communities. Agresta S. Coop. leads applied research on Geomatics applied forest and biomass inventories in Spain since 2001. Agresta has collaborated with CAB Program applying novel geomatic approaches to biomass estimation in tropical forest ecosystems. Costa Rica is a pioneer tropical country in the implementation of conservation-oriented policies, as the Payment for Environmental Services Program and the ban of land use change on forest areas. The country is pioneer also in establishing solid baseline for REDD and REDD+ Programs. Costa Rica is considered a "tropical laboratory" because of the high level of their academic institutions and the availability of good quality temporal series. In the last decade, large areas have been acquired with LiDAR sensors, mainly high priority basins for hydropower production and areas under risk of seismic-induced landslides. These data, combined with Forest In Situ networks, provide the basis for a more accurate biomass estimation. This work presents learned lessons from the Spanish and Costa Rican experiences in forest biomass assessment. Spanish experience is presented in detail, from first LiDAR supported local forest inventories to nowadays large scale monitoring experiences in biomass estimation. First works applied LiDAR data to extend plot values to forest mass through multiple regression models. These works were carried out in singular forests selected by the Administration for pilot studies. In a second phase, mass variable estimations were extended to second level European NUTS as La Rioja, Murcia and Madrid provinces, providing continuous maps of variables such as tree height, basal area or volume. The extension of forest statistics to large areas was possible in Spain as the National Program of Aerial Orthophotography provided increasing LiDAR coverage since 2009 as well as updated orthophotos for the whole territory. This coverage provides a LiDAR point cloud at 0,5 point/sq.m that has been successfully validated for timber and biomass studies, compared to more dense clouds at 2, 4 and 6 points/sq.m. Besides, the National Forest Inventory, provides updated information of 98,000 permanent plots every ten years. The combination of both sources of information allows strong statistical assessment of biomass and forest resources nationwide. Further research in Spain has been conducted in order to extract individual trees and their attributes from LiDAR point clouds. This research led to the presentation of forestUP (Agresta & Cesefor, 2013), a Google Earth application that allows viewing and simulating 3D forests based on object-oriented databases. The first demo of forestUp was been developed with a database of 600,000 parameterized and georeferenced trees from LiDAR information. Simultaneously, the University consortium and Agresta began the exploitation of LiDAR data in Costa Rica to improve forest biomass estimations, as REDD + program requires reliable mechanisms for Monitoring, Reporting and Verification (MRV). Airborne LiDAR technology was validated as a practical solution for the evaluation of biomass and carbon for the Poas Volcano National Park. A total aboveground biomass (AGB) model was calibrated, depending on top of canopy heights (TCH) and regional approaches of basal area and wood density. The methodology was validated for Poas National Park, that hosts forests corresponding to Holdridge life zones of montane rainforest (bp-M), lower montane (bp-MB), montane transition to lower montane (bp-M6), montane (bp-P) and lower montane wet forest (bmh-MB), with altitude ranging from 1,099 to 2,713 m.a.s.l. Nowadays the consortium is working on the estimation of biomass with radar data in two approaches: InSAR dynamic updating of canopy MDS and the analysis of radar backscatter from tree canopies. ESA Biomass mission, designed specifically for forest resources evaluation, will trigger these research lines, improving the accuracy of forest biomass estimations in the near future.
Quad-Polarimetric SAR for Roughness and Deformation Characterization of Sea Ice at Hopen
Brekke, Camilla; Grahn, Jakob; Eltoft, Torbjørn; Doulgeris, Anthony
Show abstractIn the transition between April and May 2014, University of Tromsø – The Arctic University of Norway, Norwegian Polar Institute and the Norwegian Meteorological Institute’s manned weather station at Hopen joined forces and conducted a combined satellite and on-ground sea ice campaign. The aim of this campaign was to collect high-resolution quad-polarimetric (QP) synthetic aperture radar (SAR) data together with near-coincident high-resolution multispectral optical data and in-situ measurements of thin sea ice classes in the Arctic. Under near clear sky weather conditions, the following dataset was collected: three Fine QP RADARSAT-2 SAR scenes (5.2 m × 7.6 m) acquired on the 1, 3 and 5 May, one multispectral WorldView-II (WV-II) image with 8 spectral and 1 panchromatic channels (0.5 m and 1 m) recorded 4 May, and finally, in-situ data consisting of photos, snow and ice measurements and sample ice cores. The in-situ measurements were collected on 30 April, 1,3, 4 and 6 May. In the highly detailed WV-II measurements, a distinct land fast sea ice class (approximately 20 cm thick) with a smooth surface is easily discriminated visually from an adjacent ice class with a rougher and more deformed appearance. The first objective of this study is to investigate the characterization capacity of QP SAR measurements with respect to discrimination between visually appearing smooth and more deformed and rougher sea ice types. As a second objective, we will also look at the temporal discrimination consistency with respect to the varying imaging geometry of the QP SAR data set. Differences in scattering mechanisms will be investigated by exploiting polarimetric decompositions and descriptors, aiming at identifying suitable parameters for a feature based classification scheme.
Glacier zones detection through C-band quad-polarimetric SAR data
Callegari, Mattia; Carturan, Luca; Notarnicola, Claudia; Rastner, Philipp; Seppi, Roberto; Zucca, Francesco
Show abstractGlaciers have a significant impact on stream-flow runoff. During summer months, glacial meltwater may provide the only source of fresh water in alpine catchments. Future climate change will reduce the glacierization and thus also alter the runoff characteristics. Glacier monitoring with different methods is thus of highest relevance. Remote sensing is a valuable tool to fulfil this scope thanks to the capability of monitoring wide and inaccessible areas. In particular, the identification of glacial surface zones from remote sensing imagery can be exploited to extract the so-called snow line, which is defined as the boundary between snow and ice or firn. At the end of the ablation period, the late summer snow line (LSSL) can be interpreted as an approximation of the equilibrium line (Huang et al., 2011), which is the line that separates accumulation and ablation areas of the glacier. It is an important parameter in glaciology as it plays a significant role in the energy balance of a glacier. It controls the quantity of melt and thus the availability of water in a glacier catchment (Adam et al., 1997). At present, the identification of glacial surface zones through satellite imagery is mainly done exploiting optical sensors, such as Landsat imagery. One of the advantage of using SAR data is the possibility to acquire information in cloud-covered sky conditions, thus mitigating the issue of having lack of data in cloud-covered areas, which is typical of optical sensors. Moreover, multi-polarimetric C-band SAR data can be sensitive to different glacial surface conditions and can be used for discriminating snow from ice (Li et al., 2012; Huang et al., 2011; Huang et al., 2012). For instance, during the summer, the accumulation area is generally covered by wet snow and therefore the backscattering is attenuated due to the water available. Moreover, as the snow surface is usually smooth, the backscattering in this area is typically low in this period, unless the incidence angle is also low. The ablation area is also controlled by surface scattering but as the surface roughness is typically higher with respect to the snow covered area (e.g. for the presence of crevasses) the backscattering can be higher (Rees, 2006). In this framework, we tested the potential of C-band quad-polarimetric SAR data for discriminating different zones of the glaciers. This problem has been addressed by applying the H/A/α technique, which is a feature extraction procedure developed by Cloude and Pottier (1997), designed for polarimetric data. This technique is based on the eigen-decomposition of the polarimetric coherency matrix and allows one to extract three parameters: the alpha angle α, the scattering entropy H and the scattering anisotropy A. These parameters express the characteristics of the scattering mechanisms and the degree of randomness of the scattering process. We are also testing the possibility to exploit the capability of the C-band signal to penetrate the dry snow in order to recognize the different zones of the glacier under the winter snow. In this case, the backscattering is dominated by the scattering coming from the interface between snow and ice and, in particular, from ice lenses and pipes that can act as efficient scatterers. Typically, in this period the accumulation area may exhibit a higher value of backscattering with respect to the ablation area for the more abundant presence of ice (Rees, 2006). The identification of different glacial zones during the winter season may overcome the problem of the non-perfect coincidence between the acquisition time and the actual end of the ablation season. Thanks to the opportunity given by the ESA/CSA SOAR initiative, we planned 10 RADARSAT-2 acquisitions in fine quad-pol mode over the Ortles-Cevedale massif in the Eastern Italian Alps. This area includes the Careser glacier, which is one of the most studied glaciers of the Alps, having a 48-year continuous time series of direct mass balance measurements (Carturan et al., 2013). The neighbouring La Mare glacier is also under systematic investigation since 2003, monitoring the snow- and firn-cover distribution in the ablation season, the mass balance and the supraglacial meteorology (Carturan et al., 2012). Two quad-pol RADARSAT2 scenes covering the Ortles-Cevedale massif were acquired at the end of the summer 2014 (i.e. 7th of August 2014 and 24th of September 2014) and have been combined with field observations. One cloud-free Landsat scene and terrestrial photographs are also available in this period, and has been used for cross comparison and validation of the glacier zones mapped using SAR data. The data collected in summer will be also used as basis for detecting the different signature of ice, firn and snow. This analysis will be useful for identifying the glacial zones underneath the dry snow in the winter scenes, whose acquisition is planned in December 2014 and January 2015. References: Adam, S., Pietroniro, A., & Brugman, M. M. (1997). Glacier snow line mapping using ERS-1 SAR imagery. Remote sensing of environment, 61(1), 46-54. Carturan, L., Cazorzi, F., and Dalla Fontana, G. 2012. Distributed mass-balance modelling on two neighbouring glaciers in Ortles-Cevedale, Italy, from 2004 to 2009. Journal of Glaciology 58, 467-486. Carturan L., Baroni C., Becker M., Bellin A., Cainelli O., Carton A., Casarotto C., Dalla Fontana G., Godio A., Martinelli T., Salvatore M. C., and Seppi, R. 2013. Decay of a long-term monitored glacier: the Careser glacier (Ortles-Cevedale, European Alps). The Cryosphere Discussion, 7: 3293-3335. DOI: 10.5194/tcd-7-3293-2013. Cloude, S.R. and R. Pottier, (1997) “An entropy based classification scheme for land applications of polarimetric SAR” IEEE Transaction on Geoscience and Remote Sensing, vol. 35, pp. 68-78. Huang, L., Li, Z., Tian, B. S., Chen, Q., Liu, J. L., & Zhang, R. (2011). Classification and snow line detection for glacial areas using the polarimetric SAR image. Remote Sensing of Environment, 115(7), 1721-1732. Li, Z., Huang, L., Chen, Q., & Tian, B. S. (2012). Glacier Snow Line Detection on a Polarimetric SAR Image. Geoscience and Remote Sensing Letters, IEEE, 9(4), 584-588. Rees, W. G. (2006). Remote sensing of snow and ice. CRC Press.
Assessing the use of ALOS PALSAR data to estimate and map above-ground woody biomass in a partially degraded East African mountain forest
Sassen, Marieke (1); Sheil, Douglas (2); Mitchard, Edward (3) 1: UNEP-WCMC, United Kingdom; 2: Norwegian University of Life Sciences; 3: University of Edinburgh
Show abstractThe distribution of carbon in forests is affected by disturbance, including the history of deforestation, degradation and regeneration. Degradation is often overlooked because it is more difficult to detect using remote sensing than deforestation. Nevertheless, many tropical forests are degraded to some extent and are composed of many intermediate vegetation cover classes that can store significant amounts of carbon. These degraded and secondary forests are often also important for local livelihoods and conservation. Understanding the impact of disturbance on carbon stocks, and their rate of recovery following such disturbance, is critical in light of planned implementation of the Reduced Emissions from Deforestation and Forest Degradation (REDD+) policy mechanism under the United Nations Framework Convention on Climate Change (UNFCCC). A relatively new approach to forest conservation that attempts to reconcile local livelihood improvement and increased carbon sequestration is based on payments for Environmental Services (PES) schemes based on REDD+. Such a scheme is being initiated on Mt Elgon, Uganda. These schemes need information on the carbon content of the forest. They would benefit from consistent remote sensing observations and space borne (ALOS PALSAR) biomass estimations to provide carbon levels, detect deforestation and degradation and monitor forest dynamics. In this paper we assess the above ground biomass (AGB) on Mt Elgon. We explore the use of field-derived AGB data in combination with satellite L-band synthetic aperture radar backscatter data (ALOS PALSAR) to monitor the impact of forest degradation on biomass in relation to planned local REDD+ schemes. We also explore the potential of direct basal area estimations as an effective field method to assess carbon stocks. We found that above ground biomass (AGB) is very high in some areas of Mt Elgon, reaching above 800Mg/ ha. However, the biomass map produced from the relationship between plot-AGB and radar backscatter values did not meet our expectations. Limiting factors likely included the sampling strategy and topography.
Proposed program to assess and monitor the impact of tobacco farming on savannah woodlands in post fast track land reform era in Zimbabwe
Paradzayi, Charles Midlands State University, Zimbabwe
Show abstractZimbabwe is ranked as the largest producer of tobacco in Africa and the world’s fourth-largest producer of flue-cured tobacco in the world. Tobacco is the golden leaf of the Zimbabwean economy. Income from tobacco sales has been rising steadily over the past decade as more and more farmers resettled under the fast track land reform program take up tobacco farming. During the 2012/2013 marketing season, 144 million kilograms of tobacco were sold earning the country over half a billion United States Dollars. There are reports that farmers from areas with sensitive environments and not suitable for tobacco farming are being lured into tobacco farming by the apparent quick financial returns on investment. However, the harsh economic environment is forcing resettled tobacco farmers to use firewood to cure their crop. Firewood consumption is being fuelled by the prohibitive cost of coal and electricity, coupled with the high cost of related infrastructure required to use such alternative energy sources. It is estimated that over 300 000 hectares of natural forests are lost annually, with over 7.5 million trees being cut down to accomodate for tobacco farming expansion and to provide fuel wood for curing tobacco. The alarming levels of deforestation have prompted the Zimbabwean President to threaten a ban on tobacco farming unless farmers resorted to using alternative energy sources. There is a paucity of empirical data for quantifying savannah woodlands lost to tobacco production. Zimbabwe does not have ongoing or planned national forest inventory based on remote sensing technologies. This paper proposes a phased approach for assessing the impact of tobacco farming on savannah woodlands using a combination of remote sensing and GIS technologies. A baseline study to assess the state of savannah woodlands prior to the fast track land reform program will be undertaken using archived optical and microwave retrievals. The results from this phase will include comprehensive database of tobacco producing farms and maps showing the spatial temporal distribution and extent of tobacco production. The second phase will develop biomass quantification and woodland monitoring algorithms using retrievals from the BIOMASS Mission for long term observation of savannah woodlands. Development of key human resources will be undertaken as part of postgraduate programs at Midlands State University, in collaboration with a partner university from a developed country with experience in processing and analysing microwave remote sensing retrievals.
Anisotropic diffusion speckle-filter integrating single-target detector for polarimetric SAR data
Sun, Sheng; Liu, Zhusong; Wen, Wen; Tan, Taizhe
Show abstractAn anisotropic diffusion speckle reduction method for polarimetric SAR data has been proved to be effective. While the strong scatters within a resolution cell are not preserved well in this filter. The single-target dectector, called geometrical perturbation filter, has been put forward using the polarimetric fork for identificating single targets (i.e. the strong targets) in the scene. Owe to the simplicity of the single-target dectector, it can be esaily integrated into other processing procedure. We propose a new anisotropic diffusion speckle-filter integrating single-target dectector. The comparative experiments will be carried on the real data. This improved filter will be more effective in preserving of polarimetric properties.
Polarimetry and Shape from Shading
Di Martino, Gerardo; Di Simone, Alessio; Iodice, Antonio; Riccio, Daniele; Ruello, Giuseppe Università degli Studi di Napoli, Federico II, Italy
Show abstractThe recent ESA Sentinel project for Earth observation and monitoring confirms the actual interest from the international research community in exploiting remotely sensed data for the management of natural resources and monitoring of environmental hazards. Synthetic Aperture Radar (SAR) certainly represents a fascinating and irreplaceable tool for these aims; indeed SAR applications very often require some kind of inversion procedure, in order to retrieve the parameter(s) of interest from the data. The inversion of SAR models is almost always a challenging problem for which no closed form solutions exist. The reason can be found in the numerous parameters influencing scattering at microwaves and then Synthetic Aperture Radar (SAR) data formation. The objective of shape from shading (SfS) techniques is to provide an estimate of the topography of the sensed (natural) surface from an individual SAR intensity image. It is noteworthy that a proper model linking the parameters of interest (in this case the range and azimuth slopes of the surface) to the SAR data is needed. The model proposed by the authors  is based on the fractal geometry both for the surface and the scattering phenomena and the developed inversion technique overcomes the ill-posedness issue proposing a linearization of the direct model. In this paper a study of the role of the polarization is conducted and how polarimetry concepts can assist SfS methods is explored. It is clear that if all the phenomena influencing the scattering and image acquisition were taken into account properly, the estimated Digital Elevation Model (DEM) would be the same irrespective of the sent-received polarization. However, assuming calibration constant estimated as suggested in  and the independence of the β_mn reflection coefficient from the incidence angle in the small slope regime, polarization dependency vanishes in the model proposed by the authors. This causes slight differences in the estimated topography. In the examples considered here, better results are obtained in the HH polarization. The reason lies in the behavior of the reflection coefficient, that is smoother in the HH than VV polarization, so that the angle-independence hypothesis about β_mn is a better approximation in the HH case. A possible approach to overcome the differences in the estimated DEMs is to simply combining them, for example through a weighted averaging. A better approach would be thought if polarization issues were taken into account thus considering the variation of the β_mn coefficient over the image. In this way, using both HH and VV SAR image, we have two equations and two unknowns (the range and azimuth slopes) so that linearization with respect the range slope is no more required (at least in principle: ill-conditioning may arise), thus eliminating the small slope regime hypothesis. Another possible improvement of the proposed technique thanks to polarimetry is represented by the extension of the proposed direct model  to other polarization channels data through a polarimetric two-scale Small Perturbation Model (PTSM) , so to extend the applicability of the technique to polarimetric operative modes, also in order to improve the accuracy of the topography estimation. A possible approach in this direction, i.e., if fully-polarimetric data are at hand, is analogous to that proposed by Chen et al., in which shape-from-shading and polarimetric concepts are combined to estimate the range and azimuth slopes of the surface respectively. A deeper analysis and more detailed results will be provided in the full paper. REFERENCES  G. Di Martino, A. Di Simone, A. Iodice, D. Riccio, G. Ruello, “On shape from Shading and SAR Images: an Overview and a New Perspective,” International Geoscience and Remote Sensing Symposium (IGARSS) 2014, Quebec City.  A. Iodice, A. Natale, D. Riccio, “Retrieval of Soil Surface Parameters via a Polarimetric Two-Scale Model,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 7, pp. 2531-2547, Jul. 2011.  X. Chen, C. Wang, H. Zhang, X. Chen, “DEM Generation Combining SAR Polarimetry and Shape-From-Shading Techniques,” IEEE Geoscience and Remote Sensing Letters, vol. 6, no. 1, pp. 28-32, Jan. 2009.
Post-fire regrowth monitoring using COSMO-SkyMed multipolarization products
Avezzano, Ruggero Giuseppe; Vaglio Laurin, Gaia; Bacciu, Valentina; Covello, Fabio; Virelli, Maria; Del Frate, Fabio; Schiavon, Giovanni; Valentini, Riccardo
Show abstractThe COSMO-SkyMed system was designed specifically for operational monitoring purposes. Its unique revisit time, ensured by the constellation of 4 X-band Synthetic Aperture Radars (SAR) allowed the definition of specific programs for disaster management. The exploitation of system capabilities for post-fire management activities was considered in this work. In particular, the multipolarization acquisition mode (PING-PONG HH-HV) was tested over the area of Capo Figari, located in Sardinia region, seriously damaged by a wildfire on the 25th, June 2013. The ecosystem is composed by combinations of evergreen woody shrubs 0.5-5 m tall, with sclerophyllous leaves, with bare soil and rock covers and by single plants or vegetation patches. The heterogeneous morphology of the area, with heights reaching 340m above MSL and slopes between 0° and 35°, required the use of a combination of Ascending and Descending acquisitions, with incidence angles between 43° and 49°. The time-series started on December 2013, and it is still ongoing. All acquired products were coregistered and radiometric and geometric corrections were applied. In order to develop a strategy for regrowth monitoring, the Corine Land Cover 2006 (CLC2006) database, an IKONOS product acquired in 2006, an orthophoto of 2006 and in situ observations were exploited to extract pre-fire conditions. Two areas, which were classified as Mediterranean Maquis, were selected: the first was included in burnt area (derived using a GPS survey of the Italian National Forest Guard) (Burnt Maquis - BM) and the second was not affected by the wildfire, and used as a control area (Unburnt Maquis - UM). Three features were extracted from the PING-PONG images: the HH polarization and HV polarization backscattering values and the ratio HV/HH. The two regions of interest were sampled considering the Local Incidence Angle (LIA) of the two acquisition configurations, using 5° steps: the area were seen with LIAs varying from 40° to 60°. The average value for HH polarization was 4dB higher in the BM area, compared to UM area, @40°, but this difference was reduced to <1dB @55°. A similar difference was noticed for the HV polarization backscattering value, having the BM a value 2.5dB higher @40°, and a <1dB lower value @55°. For the HV/HH parameter the difference between the two areas was almost constant at all incidence angles, with the UM area having 1-1.7dB higher values. The linear trend information was extracted from the time-series for all features and areas, at all incidence angles. The slope of the regression line was considered, as a measure of changes inside the scene. At HH polarization, no significant changes were detected in both areas, while the HV polarization showed positive slopes for the BM area, increasing with incidence angles. Slopes value for UM area lower than the BM: +0.3%/day for BM, -0.01%/day for UM @40°, +0.9%/day for BM and +0.02%/day for UM @55°. The HV/HH parameter showed the most interesting results. Its values were positive and almost constant for both areas, with the BM area showing an increase of +0.8%/day, and the UM area showing an average value of +0.2%/day. In situ measurements showed that, in the period between April 2014 and July 2014, no significant changes happened in the cover fractions of burned areas, while the average vegetation height increased, depending on various geomorphological parameters. This allowed to associate changes in the scene to changes in vegetation biomass, increasing thanks to rapid regrowth phases of Mediterranean ecosystem. The slope of the HV/HH time-series was selected as regrowth monitoring parameter, and a map of the area was derived considering the time series acquired at both ascending and descending modes. The map highlighted the areas inside the burnt perimeter, which resulted in an average value of +0.8%/day, 3 times larger than the unburnt area (+0.25%/day). The PING-PONG products, acquired by the COSMO-SkyMed constellation, were exploited for recovery monitoring of Mediterranean Maquis. Results showed the importance of multiple polarization acquisitions, being the ratio HV/HH a relative parameter, and for this reason less influenced by variations in soil moisture and incidence angle. The combination of Ascending and Descending acquisitions allowed the complete observation of a region with heterogeneous morphology. The Mediterranean ecosystem is mainly composed of evergreen woody shrubs, which are characterized by rapid regrowth phases. The COSMO-SkyMed system was able to track such changes, thanks to its capability of acquisition of dense time-series. This work was a first test for a definition of a monitoring strategy for hazard recovery phase. It can be integrated in existing monitoring programs and further developed including ESA Sentinel-1A products for multifrequency observations. A considerable step ahead will be performed when the COSMO-SkyMed Second Generation will be operative thanks to its fully polarimetric (quad-pol) capability.
Mapping the biomass of African savannas at 50m resolution using ALOS PALSAR mosaics
Bouvet, Alexandre; Mermoz, Stéphane; Le Toan, Thuy CESBIO, France
Show abstractIn this paper, we investigate methods to derive above-ground biomass maps of savannas in Africa based on dual-polarization (HH and HV) PALSAR mosaics developed by JAXA. A regression model between the backscatter and the above-ground biomass (AGB) is fitted by using AGB estimates from a relevant selection of field plots. Yearly biomass maps of the savannas of Eastern and Austral Africa are produced at a 50m resolution from 2007 to 2010. The validation of the products is ongoing but preliminary assessments in South Africa show promising results.
ON POTENTIAL ESTIMATION OF PERMAFROST-VEGETATION IN ARTIC AREA
Sarti, Maurizio; Mascolo, Lucio; Migliaccio, Maurizio; Nunziata, Ferdinando; Brugnoli, Enrico; Calfapietra, Carlo
Show abstractThe project entitled "SAR polarimetry for land monitoring" (PN5195), developed in the framework of the CSA SOAR-EI (Science and operational applications research Education International) aims to contribute to the application development in demonstrating the exploitation of fully polarimetric time-series datasets for monitoring the soil at the permafrost-vegetation interface. The objective of this paper is to address the issue of evaluating fully polarimetric RADARSAT-2 time-series datasets to determine the permafrost-vegetation interface dynamics.
COASTLINE EXTRACTION AND COAST CLASSIFICATION BASED ON HYBRID-POL FEATURES
Paes, Rafael L (1); Nunziata, Ferdinando (2); Migliaccio, Maurizio (2) 1: National Institute for Space Research (INPE), Brazil; 2: University of Naples "Parthenope", Italy
Show abstractRecent studies present some methods to extract the coastline through multi-polarimetric data , but there are still few contributions using SAR Hybrid-Pol (HP), principally to provide polarimetric information about the different types of coast. Present study shows the easiness of using HP features to extract the coastline, and how such features can be used complementarily to provide information about coast polarimetric characteristics, allowing its classification based on polarimetric response. It will help further physical analysis about the evolution of coastal phenomena. So, Six Radarsat-2 Full-Pol (FP) data are used to emulate HP data. Then, HP features are extracted according wave polarimetry concepts, likewise: entropy (H), degree of polarization (p); complex correlation coefficient (μhv); phase difference between h-v components (δhv); and circular polarization ratio (μc). Proposed methodology consists into two main stages: coastline extraction; and coastline classification. First, Principal Component Analysis (PCA)  is used to select the polarimetric feature more representative between the deterministic features, e.g. H, p, and μhv. Then, a low-pass filter is used based on db4 wavelets aim to reduce noise fluctuations. Then, 2-class k-means algorithm is set to put apart sea from land, and Canny filter is achieved to extract the coast border. Alternatively, when the coastline is highly irregular, morphological operators can be applied. Here, any segment can be chosen (sea or land segment) as reference to extract the board line. So, the coastline classification is achieved by means of Object Oriented Approach . Thus, taking the land segment of the initial 2-class segmentation is possible to analyze more details on the coastal polarimetric behaviors. PCA is achieved again to analyze the mutual contribution between p and μc, and between p and δhv. Correlation analysis among the former pair shows that they have mutually high contribution for the first PC, but varying inversely. It does not occur with the latter pair of parameters. It allows the identification of more two segments, named as: object1 (where high p values and low μc values can found); and object2 (where the inverse behavior is found). A new 2-class by k-means is achieved. In the end, inside each object, Linear Discriminant Analysis (LDA) algorithm  is applied to separate odd bouncing from even bouncing information. Thus, final classification is established according four objects, e.g. object1.1 (high p/ low μc/ backscattered even bouncing signal); object1.2 (high p/ low μc/ backscattered odd bouncing signal); object2.1 (low p/ high μc/ backscattered even bouncing signal); and object2.2 (low p/ high μc/ backscattered odd bouncing signal). Results are presented for different types of coast (e.g., Strait of Gibraltar’s coast, Naples’ coast, Sardegna’s coast), and conditions environment (e.g., different sea states).
Comparison of efficiency of optical and radar data for land cover classification
Surek, György; Nádor, Gizella; Friedl, Zoltán; Rotterné Kulcsár, Anikó; Török, Cecilia; Tar-Andrási, Agnes
Show abstractWe have shown in an earlier work, that radar polarimetry is suitable for identification of geometrical structural changes due to damages in agriculture parcels. In this presentation the contribution of radar polarimetry for distinguishing land cover categories was examined, because these categories show unique spectral as well as geometrical structural characteristics (e.g.: forest, pasture, arable land, reed, water bodies) For this purpose a study area was selected in the vicinity of lake Tisza, in Hungary, having complex characteristics and very rich in different type of land cover categories. Comparative analysis was carried out focused on the efficiency in distinguishing land cover categories using different type of optical and radar satellite images for classification. The following optical HR and VHR satellite images were taken into our analysis: Landsat TM8, SPOT5, Pleiades, Worldview2, having nearly the same acquisition date. Polarimetric radar satellite images involved in our analysis are listed below: RADARSAT2 fine with different polarisations such as quad, dual (HH+HV), single (HH), and 2 TERRASAR-X stripmap dual polarisation (HH+HV) satellite images, with different acquisition dates. Different type of spectral indices were derived from optical satellite images. Polarimetric descriptors were generated based on three different decomposition (H/A/Alpha, Yamaguchi, Touzi) of coherency matrix in case of quad-pol images, and H/A/Alpha decomposition of covariance matrix for dual-pol satellite images. The first results are shown in this presentation. Our project demonstrates the potential of integrated usage of polarimetric radar and optical satellite images for identification and classification of land cover types. Keywords: Land cover, Worldview2, Pleiades, Radarsat2, Terrasar-x, Yamaguchi decomposition, H/A/Alpha decomposition, Tuozi decomposition, polarimetric descriptors, spectral indices
P- And L-Band backscatter and coherence temporal variations of a tropical forest - The TropiScat Experiment
ALBINET, Clément; HAMADI, Alia; LE TOAN, Thuy; BORDERIES, Pierre
Show abstractTropical forests contain the major part of the world forest biomass and their changes in biomass by deforestation and/or by forest regeneration affect strongly the terrestrial carbon budget. To measure with accuracy tropical forest biomass and its temporal change is one of the objectives of the Biomass mission , the 7th Earth Explorer Mission of the European Space Agency (ESA). Biomass will be the first spaceborne Synthetic Aperture Radar (SAR) operating at P-band (435 MHz). The retrieval algorithms currently developed for Biomass are based on the use of backscatter measurements derived from intensity, polarimetry and interferometry. For tropical forests with very high biomass density (>300 ton/ha), biomass inversion requires both intensity and PolInSAR measurements. However, the condition to have exploitable interferometric information is to have high temporal coherence over forests for time intervals compatible with the Biomass mission repeat passes. During the airborne TropiSAR experiment , performed in August 2009 in French Guiana, the temporal coherence has been found high (> 0.8) for time interval up to 22 days. In order to complement the airborne datasets, a ground based experiment over tropical forests was necessary to produce a well controlled radar and in situ dataset for various weather conditions and different seasons. In this context and with ESA and CNES (French space agency) support, the TropiScat ground based experiment  was developed in order to investigate the evolution of the backscattering coefficient and the temporal coherence at P-band of tropical forest at different time scales range. The analysis of P-band time series exhibited daily cycles on the backscattering coefficient and on the temporal coherence, with seasonal effects . In addition, temporal coherence was found high (>0.8) for baselines up to a 20 day . In addition to P band data, the TropiScat dataset also contains L band measurements, acquired between 800 MHz and 1 GHz and synchronously to P-band acquisitions. With the consideration of the SAOCOM-CS satellite to be launched with Argentina's L-band SAR satellite SAOCOM , it is interesting to exploit the L band data and to compare the temporal variation of synchronous L-band and P-band backscattering coefficient and temporal coherence. The results exhibited at L-band also a diurnal cycle a stronger and faster temporal decorrelation than at P-band. The paper will discuss about the impact of these observations on L-band spaceborne missions.  “BIOMASS Phase 0 Report for Assessment”, European Space Agency report, ref. SP1313/2, Nov. 2008.  Dubois-Fernandez Pascale and al. "TropiSAR: Exploring the temporal behavior of P-Band SAR data", International Geoscience and Remote Sensing Symposium, pp.1319-1322, Honolulu, Hawaii, July 2010.  C. Albinet, P. Borderies, T. Koleck, F. Rocca, S. Tebaldini, L. Villard, T. Le Toan, A. Hamadi, and D. Ho Tong Minh, “TropiSCAT: A Ground Based Polarimetric Scatterometer Experiment in Tropical Forests”, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 5, no. 3, June 2012.  A. Hamadi, C. Albinet, P. Borderies, T. Koleck, L. Villard, D. Ho Tong Minh, and T. Le Toan, "Temporal Survey of Polarimetric P-Band Scattering of Tropical Forests", IEEE Transactions on Geoscience and Remote Sensing, vol. 52, no. 8, pp. 4539-4547, August 2014.  A. Hamadi, P. Borderies, C. Albinet, T. Koleck, L. Villard, D. Ho Tong Minh, T. Le Toan, and B. Burban, “Temporal Coherence of Tropical Forests at P-Band: Dry and Rainy Seasons”, IEEE Geoscience and Remote Sensing Letters, vol. 12, no. 5, pp. 557-561, March 2015.  Gebert, N.; Carnicero Dominguez, B.; Davidson, M.; Diaz Martin, M. and Silvestrin, P.; “SAOCOM-CS - A passive companion to SAOCOM for single-pass L-band SAR interferometry”, European Conference on Synthetic Aperture Radar, Berlin, Germany, 2014.
Methodological Issues in Extracting Information from ALOS PALSAR imagery in African Miombo Woodlands.
GOU, Yaqing; Ryan, Casey; McNicol, Iain University of Edinburgh, United Kingdom
Show abstractRadar remote sensing is a promising alterative for monitoring regional forest in tropical areas, where the use of optical imagery is constrained by constant cloud cover. Radar remotes sensing imagery provides biomass information that is essential to monitor forest change, understand carbon flux, broader applications in social, political and scientific fields such as forest management, environmental planning, carbon related projects (REDD, REDD+ and PESs), and international conventions (UNFCCC). However, only a limited number of guidelines have proposed standard methodologies and procedures for generating wall-to-wall radar image mosaics (IFCI 2012), and none of them provide further guidelines for processing image mosaics to biomass maps。 Retrieving woody biomass information from radar imagery is not that straight forward, mainly because of the following 2 aspects. Firstly, speckles, radiometric and geometric distortions result from the nature of radar backscatter and side-looking configuration (Woodhouse 2005) might increase the uncertainties of biomass estimation if the imagery is not properly processed. Thus, inputted terrain information and the choice of filtering approaches will greatly affect the accuracy of biomass estimation. Theoretically, terrain products with higher spatial resolution will lead to better terrain correction results and studies have been focus on exploring methodologies of terrain correction to help forest mapping. However, source (optical or radar) of DEM products might greatly affect the accuracy of dataset itself in cloudy tropical areas, which is somehow neglected in previous studies. Especially for 90m STRM (radar) and 30m ASTER (optical), which are the most widely used DEM products for regional studies. While filtering is suggested in radar preprocessing to reduce speckles, there is no suggested filter for African Miombo woodlands. Thus, this paper provides a comparison of accuracies and uncertainties of biomass estimation using these 2 DEM dataset, and also the effects of using enhance lee filter and median filter with different window size. Secondly, large uncertainties might be introduce to regional biomass estimation when generating biomass maps from processed mosaic of radar imagery, either through interpolate ground plots to a biomass-backscatter relationship or using an existing relationship from literature. It remains a question that which plot size is the best to support a radar remote sensing based biomass estimation. Or how many uncertainties will be introduced to regional biomass estimation when no ground plot is available. These two questions will be answered in this paper. In sum, this paper evaluates if 1) different filters used in processing Advanced Land Observing Satellite Phased Array L-band Synthetic Aperture Radar (ALOS PALSAR) imagery 2) different data for preforming terrain correction, and 3) whether the use of different sized plots as ground reference points will yield improved accuracy in regional woody biomass estimation. We also analyzed 4) if the backscatter-biomass relationship is constant in Miombo woodland in 3 different countries: Mozambique, Malawi and Tanzania, and 5) change in biomass estimation uncertainty when field plot is not available and a backscatter-biomass relationship from previous study is applied directly. To assess the accuracy of biomass estimation, we used the calibration R2, validation RMSE, bias and uncertainty as index. Parameters result in the highest R2, lowest RMSE, bias and uncertainty is suggested as the best processing framework for each country. The backscatter-biomass relationship generated from it is then validated by filed data from the other two countries to test if there is a consistent backscatter – biomass relationship in Miombo woodlands. Results showed that plot size will significantly influence the accuracy of biomass estimating, with 1ha plot size provide the highest calibration R2 (0.81) and lowest calibration RMSE (2.29 tC/ha), bias (0.09tC/ha) and uncertainty (1.62%). Applying 90m STRM (Shuttle Radar Topography Mission) DEM for terrain correction and Enhance Lee (with window size 5) or median filter (with window size 3) will improve biomass estimation accuracy. In situations that ground plots are not available and a backscatter-biomass relationship from previous studies is applied directly, an average of 16 tC/ha of RMSE and -7.52 tC/ha bias will be introduced.
Carbon Losses due to Tropical Forest Fragmentation
Huth, Andreas (1); Brinck, Katharina (2); Fischer, Rico (1); Lehmann, Sebastian (1); Groeneveld, Juergen (1); Puetz, Sandro (1) 1: Helmholtz Center for Environmental Research - UFZ, Germany; 2: University of California, Berkeley, US
Show abstractTropical forests play an important role in the global carbon cycle. Thereby, deforestation is not only responsible for direct carbon emissions but also alters the forest structure and extends the forest edge area in which trees suffer increased mortality due to altered microclimatic conditions. Our aim is to quantify the global amount of anthropogenically created forest edge area and the resulting additional CO2-emissions by combining remote sensing data with previous empirical and modelling results. We found that 1,106 million ha and thereby 10% of the global tropical forested area lies within the forest edge area and that 84% of this area is anthropogenically created. From this area, a total amount of 8 Gt C is emitted due to tropical forest fragmentation, which accounts for an annual loss of 0.25 Gt C equalling 17% of the annual carbon losses due to deforestation. Fragmentation in the tropics hence augments carbon loss from deforestation substantially and should be taken into account both when analysing the role of vegetation in the global carbon balance and when adopting new management strategies in tropical forests.
Understanding contrasting estimates of biophysical parameters from dynamic global vegetation models and global datasets across the Amazon Basin
Carreiras, Joao M. B.; Quegan, Shaun; Lomas, Mark
Show abstractTropical forests hold approximately 40% of the terrestrial carbon stocks, and are recognised to play an important role in the global carbon cycle. The Amazon Basin is the largest contiguous tropical forest in the world and home to vast biodiversity. In the past decades large-scale deforestation and other disturbance events (e.g., selective logging) have dramatically changed some areas within the basin, with severe consequences in terms of, e.g., landscape fragmentation and increased fire susceptibility. Therefore it is important to know these dynamics and its influence in the global carbon cycle. In the past decade several global maps of biophysical parameters (i.e., biomass, tree height) were published, mainly obtained from a combination of in situ and remote sensing data. Over some regions of the Amazon Basin there is a relatively high degree of disagreement among those datasets, which was confirmed recently with data collected exclusively from in situ measurements. Furthermore, biomass estimates (and other biophysical parameters, e.g., net primary production, turnover time) from dynamic global vegetation models disagree among themselves and are also not able to replicate the spatial distribution of biomass exclusively obtained from in situ data. A specific and recent disagreement relates to the inability of biomass datasets obtained from a combination of in situ and remote sensing data to properly characterise an observed biomass gradient over old growth forests from the southwest to the northeast of the basin. In this poster we aim to explore and understand the observed differences among different global biomass datasets, either obtained from relationships between in situ and remote sensing data or generated from dynamic global vegetation models.
REGROWTH-BR: The impact of land use history on the biomass accumulation and biodiversity recovery of secondary forests in the Brazilian Amazon
Carreiras, Joao M. B. (1); Jones, Joshua (2); Lucas, Richard M. (3); Melo, Joana B. (4); Arai, Egidio (5); Cassol, Henrique (5); Shimabukuro, Yosio E. (5); Pereira, Virgílio (6) 1: National Centre for Earth Observation (NCEO), Centre for Terrestrial Carbon Dynamics (CTCD), University of Sheffield, United Kingdom; 2: Department of Geography and Earth Sciences, Aberystwyth University, Aberystwyth, Ceredigion, United Kingdom; 3: Centre for Ecosystem Science, School of Biological, Earth and Environmental Sciences, University of New South Wales, Sydney, New South Wales, Australia; 4: Tropical Research Institute (IICT), Lisbon, Portugal; 5: Instituto Nacional de Pesquisas Espaciais (INPE), São José dos Campos, SP, Brazil; 6: WayCarbon, Belo Horizonte, Brazil
Show abstractThroughout the Brazilian Amazon region, the age and composition of forests regenerating on previously deforested land is determined, in part, by the period of active land use prior to abandonment and the frequency of clearance, both of which can be quantified by comparing time-series of remote sensing data. Using these time-series of near annual data from 1973-2011 in Manaus (Amazonas), 1984-2010 in Santarém (Pará) and 1984-2011 in Machadinho d’Oeste (Rondônia), the changes in the area of primary forest, non-forest and secondary forest were documented from which the age of regenerating forests, period of active land use, and the frequency of forest clearance were derived. A field campaign in the Manaus site was carried out in August 2014. A set of plots was established for the collection of biophysical parameters in areas known to be covered by secondary forests. Plots were randomly selected, but stratified by the age of secondary forest, the period of active land use, and the frequency of clearance. Additionally, at each plot site several combinations of plot shape and size (transect: 100x20 m and 100x60 m, circular: 20 m radius), and minimum measured tree diameter (5 cm, 10 cm and 20 cm) were also evaluated. Measurements at the tree level were used in combination with published allometric relationships to generate biomass estimates first at the tree-level and then at the plot-level. Metrics related to biodiversity were also produced, to further evaluate the impacts of deforestation and the regeneration dynamics. In this communication an analysis of the influence of prior land use on the capability of secondary forests to recover biomass and restore biodiversity will be presented. Using dual-polarisation (HH and HV) data from Japan Aerospace Exploration Agency's (JAXA) Advanced Land Observing Satellite (ALOS 1 and 2) Phased Array L-band Synthetic Aperture Radar (PALSAR 1 and 2) this communication will also provide information about the capability of L-band SAR data to distinguish types of tropical secondary forests subjected to different prior land use and to test the ability to estimate their corresponding biomass. Additionally, the influence of plot size and shape, as well as the influence of the minimum measured tree diameter will also be evaluated in combination with the available L-band SAR data. Several generalizations about the type and composition of secondary forest occurring in any given site can be made. However, a correct assessment can only be achieved with the information obtained from the data collected in situ. Therefore, the research presented in this communication will provide the opportunity to assess the influence of previous land use dynamics on the biomass accumulation and biodiversity restoration. In the end, a framework related to the influence of prior land use on the biomass accumulation and biodiversity could be used to inform land management policies in the region.
A Series Expansion of Statistical Distributions for the Polarimetric Covariance Matrix
Anfinsen, Stian Normann University of Tromsø - The Arctic University of Norway, Norway
Show abstractThis paper introduces a series expansion suitable for the statistical distributions used to model the polarimetric covariance or coherency matrix. The derivation is analogous to that of the well-known Edgeworth expansion , but uses the Mellin transform  instead of the Fourier transform. This leads to a series expansion in terms of Laguerre polynomials in matrix argument  with logarithmic cumulants  as coefficients. Whereas the Edgeworth expansion is targeted at Gaussian-like distributions, the proposed series expansion appears a natural tool to approximate matrix-variate probability density functions close to the complex Wishart distribution. The paper discusses the practical values of this result as well as the theoretical insights it provides. Approximations obtained by the truncated series expansion are useful in model-based image analysis algorithms, since they can be evaluated in terms of basic functions and polynomials and avoids special functions of the Bessel and Kummer type. The computational complexity of such algorithms is further reduced since the approximation is parameterised by logarithmic cumulants whose estimates are formed from simple sample moments , and avoids the costly computation of distribution parameter estimates that require numerical methods. The series expansion further enables us to write the K distribution and other product model distributions as members of the exponential family , and to analyse these distributions and their estimators under this theoretical framework. The extension of the method to tail approximations similar to the Cornish-Fisher expansion  is also discussed. The goodness-of-fit of truncated series approximations is evaluated numerically as a function of model parameters and the number of terms in the truncated series.  O.E. Barndorff-Nielsen and D.R. Cox, Asymptotic Techniques for Use in Statistics, Chapman and Hall, 1989.  S.N. Anfinsen and T. Eltoft, Application of the matrix-variate Mellin transform to analysis of polarimetric radar images, IEEE Trans. Geosci. Remote Sens., vol. 49, no. 6, pp. 2281-2295, 2011.  Y. Chikuse, Hermite and Laguerre polynomials with complex matrix arguments, Linear Algebra and its Applications, vol. 388, pp. 91–105, 2004.  O. Barndorff-Nielsen, Information and Exponential Families in Statistical Theory, 2nd ed., Wiley, 2014.
Developing Methods for using L-band Synthetic Aperture Radar (SAR) in Above-Ground Biomass (AGB) Estimation for the Hindu Kush-Himalayan (HKH) Region and Pakistan
Qazi, Waqas Ahmed (1); Baig, Shahbaz (1); Gilani, Hammad (2) 1: Institute of Space Technology, Pakistan; 2: International Centre for Integrated Mountain Development, Nepal
Show abstractFile attached.
Modelling Mediterranean woodland carbon fluxes using airborne remote sensing
Coomes, David A; Simonson, William D
Show abstractRecent analyses indicate that natural disturbances (which kill trees and release there carbon to the atmosphere) are predominantly small-scale in Amazonian rainforests, with intermediate and large-scale disturbances making little contribution to forest carbon fluxes. We set out to explore whether Mediterranean woodlands function similarly, using repeat-measures airborne LiDAR surveying of canopy heights alongside permanent plot and fire frequency datasets. We find that large-scale disturbance by fire is much more important than small-scale disturbance in these seasonally dry systems, and that variation in fire frequency among years creates considerable inter-annual variation in regional carbon fluxes. Our work emphasizes that forest responses to climate change will vary among biomes.
Evaluating Moisture Effects on L-Band SAR Classification Performance over Eastern Amazon
Liesenberg, Veraldo Santa Catarina State University, Brazil
Show abstractMultitemporal single (HH) and dual-polarizations (i.e., HH, HV) for L-band spaceborne synthetic aperture radar (SAR) scenes were evaluated under different moisture conditions caused by precipitation prior to the SAR data acquisition. We evaluated such changes for backscattering intensity, polarimetric decomposition, backscattering mechanism changes, and land use/land cover classification performance. The study area is a shifting-cultivation environment in the eastern Amazon (Brazil). Several data input scenarios were proposed in the classification scheme (i.e., backscattering intensity alone and combined with alpha/entropy decomposition parameters, band ratios, and textural parameters) using the random forest (RF) classifier ensemble. Integration with optical data was also examined. The classification accuracy values were then compared with the accumulated precipitation values. The results showed that the vegetation moisture increases the backscattering intensity for pasture, riparian forest and young regenerated forest by at least 1 dB compared with old successional forest stages due to its more uniform vertical structure and the landscape’s increased dielectric constant. The overall classification accuracy proved low for each SAR acquisition date compared with the performance of the Landsat data. Based on SAR data, misclassification occurs for the young successional forest stages and increases in scenes with higher moisture conditions. The classification performance benefits from data integration. The best classification performance was yielded for SAR scenes that were acquired during the dry season and integrated with optical data. In addition, the results highlight the importance of selecting proper temporal intervals for the different SAR polarization modes of the forthcoming SAR missions.
Evaluating Classification Performance at L-band SAR using Single, Dual and Quad Polarization Mode Scenes in Eastern Amazon
Show abstractSingle, interferometric dual, and quad-polarization mode data were evaluated for the characterization and classification of seven land use classes in an area with shifting cultivation practices located in the Eastern Amazon (Brazil). The Advanced Land-Observing Satellite (ALOS) Phased Array L-band Synthetic Aperture Radar (PALSAR) data were acquired during a six month interval. A clear-sky Landsat-5/TM image acquired at the same period was used as additional ground reference and as ancillary input data in the classification scheme. We evaluated backscattering intensity, polarimetric features, interferometric coherence and texture parameters for classification purposes using support vector machines (SVM) and feature selection. Results showed that the forest classes were characterized by low temporal backscattering intensity variability, low coherence and high entropy. Quad polarization mode performed better than dual and single polarizations but overall accuracies remain low and were affected by precipitation events on the date and prior SAR date acquisition. Misclassifications were reduced by integrating Landsat data and an overall accuracy of 85% was attained. The integration of Landsat to both quad and dual polarization modes showed similarity at the 5% significance level. SVM was not affected by SAR dimensionality and feature selection technique reveals that co-polarized channels as well as SAR derived parameters such as Alpha-Entropy decomposition were important ranked features after Landsat’ near-infrared and green bands. We show that in absence of Landsat data, polarimetric features extracted from quad-polarization L-band increase classification accuracies when compared to single and dual polarization alone. We argue that the joint analysis of SAR and their derived parameters with optical data performs even better and thus encourage the further development of joint techniques under the Reducing Emissions from Deforestation and Degradation (REDD) mechanism.
Characterizing Tropical Secondary Forest in Eastern Amazon with Multifrequency SAR
Liesenberg, Veraldo Santa Catarina State University, Brazil
Show abstractMultifrequency SAR scenes from ALOS/PALSAR, RADARSAT-2 and TERRASAR-X were evaluated for classification of successional forest stages and land use classes in an area with shifting cultivation practices located in the Tapajos National Forest (Eastern Amazon, Brazil). Results showed that complex secondary forest areas with shifting cultivation practices can be classified with SAR intensity backscattering and their derived parameters with relative success. The addition of SAR-derived parameters with intensity backscattering increases overall accuracy at the 5% significance level when compared with backscattering alone at different frequencies. However, this classification accuracy did not overcome the classification accuracy obtained from the optical data alone. However, such approaches could be an alternative source of data for mapping purposes in tropical regions affected by cloud coverage. In tropical regions where optical data is usually limited, SAR L-band may perform the best classification results as well as for the retrieval of biophysical parameters. Hence, the integration of SAR backscattering, their derived parameters and optical data increase classification accuracy using the random forest (RF) classifier. Significant improvements can be observed for the selected classes and overall accuracy reached up to 90%. Misclassification of forest types reported as the most difficult to be mapped in the Amazon environment could be minimized by using the multi-sensor approach. Joint approaches of dual SAR frequency data are equivalent at the 5% significance level for any classification scheme. The classification results are well in agreement to those reported in the literature employing similar SAR datasets in complex environments worldwide. Precipitation events on the date and prior data acquisition should be taken into account in mapping approaches, since classification accuracy is strongly impacted in all frequencies investigated. Incidence angle is also an important parameter to be considered in classification settings, mainly at higher frequencies (i.e. X- or C-bands). The regression model based on RF was not robust enough to handle large datasets, requiring feature selection or data mining approaches. RF-based classification reveals that cross-polarized channels, texture and alpha-entropy decomposition were important ranked features in both classification and regression procedures.
Full and compact polarimetric SAR data analysis for various land features
Kumar, Vineet; Rao, Yalamanchili S. Indian Institute of Technology Bombay, India
Show abstractFully polarimetric and compact polarimetric SAR systems have their own advantages and disadvantages in mapping various land features. While fully-polarimetric SAR data is available from various airborne and satellite SAR systems, compact polarimetric SAR data is only available from Indian RISAT-1 satellite in right circular transmit and horizontal and vertical receive mode (RH and RV). Recently launched ALOS-2 PALSAR will also provide data in compact polarimetric mode. Future systems are planned to include compact mode in view of its wide swath coverage and less power consumption. Many studies (-) have so far confined to simulated hybrid polarimetric data from fully-polarimetric mode data. We have analyzed actual hybrid polarimetric data from RISAT-1 and compared with fully-polarimetric SAR data from RADARSAT-2 and also simulated hybrid polarimetric SAR data for two test sites of India. RADARSAT-2 and RISAT-1 data sets were acquired over Karjan, Vadodara district of Gujarat state of India on Nov. 03 and 04, 2013 respectively. Similarly, the data were acquired on March 11 and 09, 2014 over Vijayawada, Andhra Pradesh state of India respectively. As the time interval difference between two satellite acquisitions was less, major features will not change within the interval and comparison can be made. Comparison between RARDARSAT-2 and RISAT-1 has been in sigma-0, decomposition analysis in terms of odd, even and volume scattering components and classification. For comparison of sigma-0, RADARSAT-2 full-pol data is converted to right-circular transmit and linear receive and compared with RISAT-1 data bare soil, cotton, sugarcane, banana, rice, scrub, trees and settlement classes. We observed the difference of 3 dB for bare field and rice crop and maximum 10 dB for settlements. RISAT-1 sigma-0 values are always higher than RADARSAT-2. M-chi decomposition has been performed on two data sets and analyzed surface, double and volume scattering components for all classes for two data sets. Due to the limitation of m-chi decomposition, more volume component is observed for all classes. Even surface component is also more in m-chi decomposition for rice crop which gives more double component. When compared with Freeman-Durden decomposition, we found hybrid SAR data is inferior to RADARSAT-2 full-pol data for decomposing the target scattering mechanism into surface, double and volume components. Wishart supervised classification is applied on full and hybrid PolSAR data for classification of various land features and the classification accuracy is shown in Table 1 and 2 for two test areas. Over all accuracy of full-pol data classification is slightly higher than LC hybrid polarization for Karjat test site. All classes except sugarcane give lower than full pol data. For Vijayawada test site, the difference between RISAT-1 and RADARSAT-2 classification is 5%. Simulated RADARSAT-2 hybrid pol data give 15% lower accuracy than full pol data. Bare filed and Urban-1 (orderly constructed buildings) accuracy is same in full-pol and simulated LC pol. Sand in the river bed and Banana is well classified in RISAT-1 than full-pol data. Other hybrid parameters such as degree polarization, phase difference, ellipticity, and circular polarization ratio are also studied for various classes.  Charbonneau, F. J., B. Brisco, R. K. Raney, H. McNairn, C. Liu, P. W. Vachon, J. Shang et al. "Compact polarimetry overview and applications assessment."Canadian Journal of Remote Sensing 36, no. sup2 (2010): S298-S315.  Li, Y., Lin, H., Zhang, Y., and Chen, J., “Analysis of polarimetric features from CTLR compact polarimetric SAR data for discriminating oil slick damping status” IGARSS 2014  Ainsworth, T. L., J. P. Kelly, and J-S. Lee. "Classification comparisons between dual-pol, compact po-larimetric and quad-pol SAR imagery." ISPRS Journal of Photogrammetry and Remote Sensing 64, no. 5 (2009): 464-471. Table 1. Comparison of classification accuracy of RISAT-1 hybrid pol, RADARSAT-2 full-pol, RA-DARSAT-2 simulated Left and Right Circular polarizations data for Karjan test site. Class RISAT-1 RS-2 RC RS-2 LC RS-2 Full Water 100.00 99.02 99.51 100.00 Sand 94.26 77.65 66.89 85.93 Banana 99.66 80.51 90.95 96.52 Rice 91.56 85.03 74.83 92.29 Cotton 45.83 52.54 54.56 61.66 Bare 90.24 76.95 91.92 92.51 Urban-1 87.41 90.82 91.01 92.13 Urban-2 36.53 41.81 34.01 67.25 Trees 58.24 40.71 43.58 89.29 Overall % 81.20 71.95 71.46 86.13 Table 2. Comparison of classification accuracy of RISAT-1 hybrid pol, RADARSAT-2 full-pol, RA-DARSAT-2 simulated Left and Right Circular polarizations data for Vijayawada test site. Class RISAT-1 RS-2 RC RS-2 LC RS-2 Full Water 100.00 99.02 99.51 100.00 Sand 94.26 77.65 66.89 85.93 Banana 99.66 80.51 90.95 96.52 Rice 91.56 85.03 74.83 92.29 Cotton 45.83 52.54 54.56 61.66 Bare 90.24 76.95 91.92 92.51 Urban-1 87.41 90.82 91.01 92.13 Urban-2 36.53 41.81 34.01 67.25 Trees 58.24 40.71 43.58 89.29 Overall% 81.20 71.95 71.46 86.13
USE OF SAR-DERIVED BIOMASS MAP FOR REDD+
Mermoz, Stephane (1); Le Toan, Thuy (1); Fichet, Louis-Vincent (2); Sannier, Christophe (2) 1: CESBIO, France; 2: SIRS, France
Forest Biomass and Carbon Stock Mapping Using Active and Passive Remote Sensing in the Southwest US
Wu, Zhuoting US Geological Survey, United States of America
Show abstractAboveground biomass was estimated from active and passive remote sensing sources, including high point density airborne lidar, high spatial resolution WorldView-2 and moderate resolution Landsat 8 satellites, in a natural pinyon-juniper-oak dominated woodland and ponderosa pine forest area in the Southwestern U.S. Discrete-return airborne lidar provided direct measurement of individual tree height with a slope of 1 and R-square value of 0.98, and allowed estimation of aboveground biomass with a slope of 0.93 and R-square value of 0.55, compared to the field-based measurements. Plot-level aboveground biomass models for all species inclusive and dominate species based were created using lidar, WorldView-2 and Landsat 8. Combined height and intensity metrics derived from lidar performed best among all remote sensing sources for the general model (all species) with R-square value of 0.69 and RMSE of 32 Mg ha-1, whereas WorldView-2 and Landsat 8 based models produced errors greater than 42 Mg ha-1. For all woodland and forest species, lidar-derived height and intensity metrics in combination provided the most robust estimate for aboveground biomass, producing models with R-square values above 0.8 and RMSE less than 14 Mg ha-1. Dominate species based aboveground biomass models yielded errors ranging from 9 to 28 Mg ha-1 using Landsat 8, whereas WorldView-2 based model yielded errors of 17 to 44 Mg ha-1. Structural information (e.g., tree height) extracted from lidar provided accurate estimates for fine-scale aboveground biomass mapping, while spectral information derived from Landsat 8 can be used for large scale biomass mapping beyond lidar spatial coverage.
FOREST BIOMASS AND ENVIRONMENTAL FACTORS AS PREDICTORS FOR SOIL RESPIRATION IN A 30 YEAR DIPTEROCARPUS BAUDII AND VERRUSCOSUS IN THE TROPICAL RAINFOREST, PENINSULAR MALAYSIA
Mande, Hosea Kato; Abdullah, Ahmad Makmom; Aris, Ahmad Zaharin; Nuruddin, Ahmad Ainuddin; Ghadimzadeh, Azadeh Universiti Putra Malaysia, Malaysia
Show abstractThe holistic function of forest biomass, Total Organic Carbon, Soil Organic Carbon, Soil Organic Carbon Stock, soil moisture and environmental factors are likely to have a wide-reaching consequence for soil carbon dynamic as their contribution and alteration in the carbon balance are important. The objective is to determine the effect of forest biomass and environmental factors on soil respiration. The research was conducted in a 30-years old Dipterocarpus Baudii and Verruscosus trees species forest of Sungai Menyala, Peninsular Malaysia. Soil respiration was measured using a continuous open flow chambers technique connected to a multi gas-handling unit and infrared gas analyser. Forest biomass, total above ground biomass (TAGB), below ground biomass (BGB), total forest carbon (SOCs), soil organic carbon stock (SOCstock) and soil properties were quantified based on standard method and analysed using the Kjeldahl method and Walkley-black wet oxidation technique. The associated forest biomass, microclimate condition and soil properties were much related to the forest canopy density and this was explained by determining the leaf area Index (LAI). Forest biomass, soil properties, environmental factors and soil respiration was measured on daily basis between February and June 2013 representing the indistinctive entire seasons of the Peninsular. The average mean soil respiration for the entire measurement period ranges from 99.23 to 520.58 mg m-2 h-1, while forest biomass was found to contribute considerable percentage of TAGB, BGB, SOCs and SOCstock of 1621636.18 kg, 74736.18 kg, 797295.8 kg and 44.59 Mg ha-1 respectively (Fig. 1) and LAI was 2.97. The canopy density development in the study forest compartment is largely controlled by the shade tolerance characteristics of the species, further proven that recovering forests increased their carbon use efficiency, as the increase in the canopy cover. The recorded values of leaf area index in the study area indicated LAI can reaches a maximum at certain age by complete canopy closure alongside the continue canopy stand increase to increase in tree height and thereby increases the forest biomass. The variation pattern of soil respiration coincided with the corresponding trend of forest biomass, soil temperature and moisture dynamic indicating a symmetrical parabola curve and significantly correlated at p<0.01, R2 0.96, proven to be predictor factors responsible for the spatial and temporal variation in soil respiration. Subsequently, in relating forest biomass and it input on soil nutrient, several biophysical and environmental conditions in terms of litter fall C/N concentration were significantly correlated and the percentage contributions of C/N ratio range from 47.99 to 50.89%. This is responsible for the rate of organic matter decomposition for microorganism. The estimated carbon concentration in the soil resulted from forest biomass i.e soil organic carbon stock (SOCstock) at each 10 cm of soil depth of top 30 cm was found to decrease between 0 and 30 cm. The higher percentage of carbon centration occurred at the top 10 cm of soil and decreased with depth (Fig. 2). The estimated total organic carbon (TOC) and soil organic carbon (SOC) was significant (p<0.001) with forest biomass, as they were recorded to occur in considerable quantities with higher percentage between 0 and 30 cm soil depth and decreased between 10 and 100 cm soil depth. Pearson correlation was further confirmed by indicating a significantly positive correlation between forest biomass and TOC and SOC, This suggested that the concentration of the soil properties is being influenced by forest biomass. These showed that an overall input from recovering forest and management practices, could influence the total forest biomass, change the rate of soil properties and soil CO2 efflux. Keywords: Environmental factor, Forest biomass, soil Carbon stock, Soil CO2 efflux, Spatial and Temporal.
OIL SPILL DETECTION IN COMPLEX ARCTIC ENVIRONMENT WITH C- AND X-BANDS POLARIMETRIC SAR DATA
Longépé, Nicolas (1); Vincent, Pauline (1); Husson, Romain (1); Miegebielle, Véronique (2); Dubucq, Dominique (2) 1: CLS, Radar Application Division, France; 2: TOTAL, France
Show abstractThe aim of this study is to further evaluate the feasibility of oil spill detection in complex arctic environment by using polarimetric satellite radar images. Oil spilled in ice is either encapsulated under the icepack, in the icepack, on the ice surface or between ice floes. However, the attenuation of sea ice at X-, C- and L-bands implies that no ElectroMagnetic (EM) return from potential polluted water under standard icepack can be properly measured by current SAR system. In addition, the backscattering of sea ice depends on many parameters, and the difference of scattering properties between oil and ice may not be detectable. As a consequence, this study only focuses on the possibility to discriminate oil spill from look-alike, open water or thin ice in leads/polynias, and low-wind areas. The general idea is to find the most robust polarimetric feature(s) which can intrinsically discriminate the dielectric properties of the oil material, and not its effect on the sea surface roughness or the preponderance of noise effect due to its low EM backscattering. In this study, polarimetric C-band SAR imagery has been assessed using an extensive dataset with hundreds polarimetric Radardat-2 images covering oil spill, look alike, clean open water or sea ice. The classifications result gives an Overall Accuracy ranging from 60% to 90% for a SVM-based classification with 4 classes (oil, clear water, lookalike and lead in ice). An optimal set of about 10 full-polarimetric features could be extracted, but results may vary depending on the classification/ranking scheme. We also analyzed the impact of the noise floor (NESZ) of RS2 sensor for the estimation of full-polarimetric features, as the expected backscattering from polluted water, low wind areas, or leads is supposedly low. In addition, the Polarisation Ratio PR (VV/HH) seems an efficient tool for oil detection in leads: literature and first experiments at C-band show the potential discrimination of oily areas from open water (PRoil < PRwater) and open water from thin ice (PRwater < PRthinice). However, two important assumptions have been further assessed: - Wind fetch in leads should be large enough to produce small waves potentially damped by oil pollution - Thin ice should generate high PR values for a large set of ice/radar configurations. Finally, X-band data is supposedly of higher interest for oil spill detection and characterization. The wavelength at X-band being about 3 cm (about 5 cm for C-band), the sensitivity to Bragg resonance waves in the case of oil presence is likely to be higher for this frequency. In this study, we also evaluated the potential of dual-polarimetric X-band (TSX and CSK).
Evolving biomass monitoring requirements: REDD+ and supporting developing countries
Herold, Martin (1); Avitabile, Valerio (1); Seifert, Frank-Martin (2) 1: Wageningen University, Netherlands, The; 2: ESA-ESRIN
Show abstractIn 2005, official discussions initiated at the United Framework Convention on Climate Change (UNFCCC) established a process to analyze how reducing greenhouse gas emissions from deforestation and degradation (REDD) in developing countries could contribute to the ultimate goal of the Convention. A basic underpinning requirement is the systematic long-‐term observation of forests and other land cover characteristics and changes on national and global levels. In response to this need several international activities have been established like the REDD working group within the “Global Observation of Forest and Land Cover Dynamics” (GOFC-‐GOLD) and the “Global Forest Observation Initiative” (GFOI) of GEO. Both initiatives foster a continuous interaction mechanism with the UNFCCC to provide technical contributions to the negotiations and for national-‐level capacity development for monitoring emphasizing the important role of satellite remote sensing in this context. The presentation will elaborate the state of play of REDD+ in the political context and discuss the approach and needs for biomass monitoring with the REDD political discussions. It addresses the research community with open scientific questions to improve methodologies to support REDD early actions and readiness mechanisms for building national REDD monitoring systems.
Investigation of parameters configurations influence on inversions methodes.
TAHRAOUI, Sofiane; M.OUARZEDINE, Mounira
Show abstractThe wide range of radar systems frequency offers deferent signal scattering comportment. The X-band, presents a great sensitivity to leaf-scale objects. Moreover, it low penetration permit a forest cover cartography. P-band allow a deeper penetration, and ground contribution is present in all polarization states. The high sensitivity of the interaction behavior between the electromagnetic wave and naturel medium to the chosen frequency band. In addition to the high sensitivity of the interferometric coherence behavior to the polarization state. Permit as to consider a new pattern of scattering signature of the mediums by combining these two Features in inversions methods for estimation of mediums parameter. These aforementioned facts, leading to a discussion on the potential of the use of deferent frequency bands in inversion methods, and the comportment of its scattering in deferent polarization state, where the response of every medium type has a specific shape and location in complex plan with a respect to a chosen parameters configurations.
Changes of crop biomass in response to climate change in China
Zhang, Wen (1); Zhang, Qing (1); Huang, Yao (2) 1: LAPC, Institute of Atmospheric Physics, Chinese Academy of Sciences; 2: LVEC, Institute of Botany, Chinese Academy of Sciences
Show abstractCropland is of great importance by providing food and other agricultural supply to human’s life. It may also be a significant sink of the atmospheric carbon dioxide when managed properly. On the other hand, a lot of studies showed the vulnerability of agro-ecosystems to climate changes and anthropogenic activities. Exploring the responses of carbon procedures to climate changes and agronomic managements in croplands is the prerequisite to improve our capabilities of policy making. With the field observations at 16 CERN stations under different crop rotation systems, we conducted a modelling analysis to the impacts of climatic warming to crop biomass in China. Compared with the 1960s, the temperature during wheat–rice, wheat–maize and double rice growing seasons in China has increased. The temperature frequency distributions in the growth season move towards the biologically high–temperature end at all sites. The days of the biologically low–temperature decreased during the wheat, rice and maize seasons in northeast China. And more days with optimum temperature to crop growing achieved. In southern regions, however, the temperature distributions during rice seasons remained unchanged. As a result, the proportion of crop biomass, upon the contribution from fertilizer application etc., increased significantly in north and northeast China, while the temperature-induced rice biomass in south China showed no obvious change. It was found that the warming during the period from trans-/planting to heading account for major proportion of the temperature-induced crop biomass increment, in comparison to the period from heading to harvesting, especially for wheat. But the climatic warming contributed trivially to the increase of rice biomass in south China and maize biomass in north China. The adaption of crops to climate change will help increase the photosynthetic production of crops. Using the Agro-C model, we analyzed the impacts of climate changes, with and without crop adaption, to the crop growing. Without adaptive adjustment to the increasing temperature, the growing duration of crops will shorten significantly and thereafter, a decrease of aboveground biomass of crops will occur, especially in the northeast China. With proper adaption, e.g., by keeping/lengthening their growing duration via variety renewal, the aboveground biomass of crops will probably get increase but with great diversity. The increase of aboveground biomass for winter wheat will the most significant, while the rice in south China will show little increase and even decrease due to the stress of high temperature.
Study of river ice structure with polarimetric SAR data
Los, Helena Warsaw University of Technology
Show abstractRiver ice is a natural element of river regime in Poland as well as other countries with harsh winters such as Canada, the USA or Russia. The presence of ice cover changes work conditions of hydropower. Uncontrolled water lever rise caused by ice-jam may result in flood. These are the main reasons for river ice cover monitoring. For several years satellite data have been involved in this domain. Especially SAR (Synthetic Aperture Radar) systems are highly effective due to their capacity for all weather data acquisition. Although basic information can be obtained from single-pol data there are also limits with some ice types classification and ice structure study. It is expected to overcome them by applying polarimetric data. The aim of the ongoing project is to study thickness and structure of ice cover on Lower Vistula River (Poland) with polarimetric RADARSAT-2 imagery. Satellite data were acquired through the Sciences and Operational Application Research (SOAR2) opportunity. We investigate potential of quad-pol and dual-pol imagery. As reference data ground measurements of ice cover structure collected in selected river profiles are used. Channel filling with ice as well as ice types and density of frazil ice are estimated based on parameters obtained from polarimetric decomposition. It is expected that developed methodology will be able to apply also with dual-pol Sentinel-1 data.
Tropical Forest Network Data for Calibration and Validation of Remotely Sensed Biomass
Lewis, Simon University College London, United Kingdom
Show abstractBIOMASS is a major effort to map the biomass of the world’s forests. Its calibration and validation are going to need a co-ordinated, global approach on the ground, with a special focus on the tropics where most of the biomass is. This talk aims to suggest ways to help maximise the return on investment in remote sensing technology by the European taxpayer, so that the mission can best achieve the ambitious aim to map across the world’s tropical forests.
Assessing the Biomass of Northern Eurasia Forests: Integration of Ground Measurements and Remote Sensing Data
Shvidenko, Anatoly (1,2); Schepaschenko, Dmitry (1,3) 1: International Institute for Applied Systems Analysis, Schlossplatz 1, A-2361 Laxenburg, Austria; 2: V.N. Sukachev Institute of Forest, Siberian Branch of the Russian Academy of Sciences, Akademgorodok, 660036 Krasnoyarsk, Russia; 3: Moscow State Forest University, Mytischi, 141005 Moscow region, Russia
Show abstractThe network of ground measurements over Northern Eurasia (NE) forests has been substantially destroyed after disintegration of the Soviet Union. Particularly damaged were series of long period measurements on permanent sample plots. However, a large amount of ground measurements in forests has been collected, basically during the period of 1960-1990. The most complete database (DB) of field measurements of biomass across the Eurasian continent in temperate and boreal forests accounts for above 8000 sample plots that were established during 1960-2010s. The DB includes biometric characteristics of forest stands (species composition, age, average diameter and height of stands, relative stocking, site index, basal area, growing stock volume) and mass (dry matter) of live biomass by components including stems, branches, foliage, roots, understory (shrubs + undergrowth), green forest floor, as well as dead wood (dry stems, snags, dry branches of living trees, stumps, and logs). While the DB contains a unique collection of empirical data, it has some substantial shortcomings: 1) different approaches were used for assessing LB components (basically a different number of measured sample trees) that resulted in different accuracy of the LB estimates; 2) very rough coordinates of sample plots; 3) omission of measurement of some important indicators (e.g., roots, dead wood, green forest floor) on part of sample plots; and 4) availability of vast forest areas in Northern Eurasia without any measurements of LB. Based on the DB, a spatially distributed system of Biomass Extension Factors (BEF) for assessing LB of NE forests was developed. Spatially and parametrically regionalized by ecoregions, a unified set of BEF for Northern Eurasia (NE) forests was presented in forms of multiple regressions for aggregated data by forest enterprises [BEF = f (TS, A, SI, RS), where TS-tree species, A-age, SI-site index, RS-relative stocking] taking into account that these indicators are presented in forest inventory data for territories of the former Soviet Union. BEF were defined for the above six LB components. Statistical significance and adequacy of the equation were satisfactory. Combining the BEF equations and growth (yield) models, the dynamic spatially distributed tables of biological productivity of NE forests have been developed. These tables contain age dynamics of LB components by region, tree species and site index for fully stocked and modal (currently existing) forest ecosystems. At a country level, the system has been applied to forests of Russia and Ukraine. Accuracy of the assessment depends mostly upon reliability of forest inventory data. Based on data of the Integrated Land Information System, which contained updated forest inventory data, “within method” uncertainties in defining forest LB at the national level were estimated at ±5-7% and dead wood at ±8-12% (CI 0.9). Such an organization of empirical data in multi-dimensional modeling and numerical clusters generates a basis for different promising applications of ground information in multi-sensor remote sensing products aiming at increasing the products’ quality and presenting more detailed information. We discuss some typical examples. (1) Together with well represented for NE forests models of distribution and interconnection of biometric indicators for homogeneous groups of forest ecosystems, the above modeling sets present a comprehensive multi-dimensional description of forest ecosystems. This allows inter alia a) partition in components the above ground LB that is defined, e.g. by remote sensing; b) assessment of “hidden” elements of LB, e.g., below ground or green forest floor LB; c) use of “most appropriate” components of LB for assessing other LB components, e.g. estimation of total forest LB, based on short-wave imagery from SAR (X or C bands) associated with amount of foliage and twigs; however, reliability of such operations requires special experimental examination. (2) Availability of the detailed spatially distributed forest biomass inventory could be used for consideration of some debatable questions of application of radar methods, like how much the share of dry stem biomass impacts RS signals or which errors in defining the above ground LB are generated by omissions of accounting for lower layers of forest ecosystems. (3) Possibility to measure LB dynamics by remote sensing opens new fields for assessing other indicators, e.g. Net Primary Production, based on modeling of productivity of LB by components (Shvidenko et al. 2007). However, there is a need for harmonizing definitions and classification schemes used in forest management with indicators that are delivered by remote sensing. Typical example includes the mismatch between canopy cover, canopy closure and relative stocking; different definitions of growing stock volume; and others. For instance, replacement of relative stocking (used in forest inventory) by canopy closure could generate errors in LB assessment up to 30% for some categories of forests of high latitudes.
Monitoring for REDD+ with InSAR
Solberg, Svein (1,); Næsset, Erik (1); Gobakken, Terje (2); Hofstad, Endre Hofstad (2); Zahabu, Eliakimu (3); Malimbwi, Rogers (3) 1: Norwegian Forest and Landscape Institute, P.O.Box 115, 1431 Ås, Norway; 2: Norwegian University of Life Sciences; 3: Sokoine University of agriculture P.O. Box 3013, Morogoro, Tanzania
Show abstractA method for forest monitoring within REDD is being developed. The basic idea is to derive estimates of forest Carbon stock changes from height changes in InSAR DSMs over time. One study area is a 400 km2 area located in Liwale, Tanzania, - a miombo savannah forest. First, an AGB model was developed, where AGB was found to increase with 13.6 t/ha per m change in InSAR height, i.e. the height above ground of the center of the backscatter. This was done by covering the entire area by airborne laser scanning (ALS) from which we derived a Digital Terrain Model (DTM). A field inventory of forest biomass was carried out on 88 circular 700 m2 geo-referenced plots. Tandem-X data were processed against the DTM providing InSAR height, or canopy height, for each field plot. We then derived height changes from SRTM to Tandem-X DSMs representing an 11-year business-as-usual baseline for changes. Prior to this the C-band SRTM DSM was corrected, i.e. lifted, to simulate a wall-to-wall X-band DSM in 2000. Then we derive height changes from repeated Tandem-X data in 2011 and 2012, as an example of an annual MRV effort. The mean DSM change 2000 – 2011 was a 0.44 m decrease, corresponding to an annual decrease of AGB of 0.54 t/ha. The mean height change from 2011 to 2012 was -0.041 m, corresponding to an AGB decrease of 0.56 t/ha. Hence, in this study area the forest C loss in 20111-2012 was larger than the 11 preceding years, and hence, no C credit was obtained. With the near-global SRTM and Tandem-X data, this could be further developed into an operational MRV effort for REDD+. A further processing taking coherence into account, provided changes in categories, i.e. deforestation, degradation and forest growth (gain).
The International Forest Biomass Network (IFBN) Initiative
Chave, Jerome CNRS, France
JOINT POSTER SESSION
2015-01-29 18:10 - 2015-01-29 19:40