Session: Land III

Fire is a key component of the carbon cycle, related to greenhouse gases and aerosol emissions to the atmosphere. Furthermore, there is also increasing evidence on the interaction between climate and fire. Expected temperature and rainfall changes are likely to increase wildfire frequency in some regions, while others may experience decrease in fire activity either by higher precipitation or lower biomass production. The growing interaction between fire and climate highlights the relevance of accessing global burned area (BA) information, particularly for climate and dynamic vegetation models. The relevance of fire in the global system justifies the identification of fire disturbance as one of the Essential Climate Variables (ECV) by the Global Climate Observing System (GCOS). The European Space Agency´s Climate Change Initiative (ESA-CCI) selected fire as one of the 13 target ECVs to develop in the first phase of the programme. The Fire_CCI project aims at obtaining long-term and consistent time series of BA information, with proper error and uncertainty characterization, and to improve the use of this information by climate modelers.

This paper presents the MERIS BA algorithm, the results obtained and the most recent developments. The algorithm is based on MERIS reflectance bands, spectral indices and both post-fire and multi-temporal analysis controlled by MODIS hotspots locations. Auxiliary information (confidence level, days between valid images, valid images…) is also provided in order to better characterize the burned area product. In the first phase of the fire_CCI project, global burned area maps for years 2006 to 2008 were produced, as well as validation results. BA estimates for these three target years range from 3.6 to 3.8 million km². Results from global validation datasets provided an average overall accuracy higher than 0.95. Omission and commission errors (OE,CE) were lower for areas with higher proportion of burned area (OE = 0.51, CE  = 0.52) than for areas with very low BA proportion (OE=0.74, CE = 0.60). In terms of total BA estimation, errors were generally well balanced, with a tendency towards underestimation (34%). MERIS errors were significantly lower than the ones obtained with other European BA products.  A comparison with external fire perimeters in four larger sites was performed, showing overall agreements between 0.8 and 0.99, with higher omission than commission rates. Intercomparison with other existing BA datasets pointed out similar spatial and temporal trends, with high correlation with GFED v4 BA estimations for the three years (r²>0.974).

The MERIS sensor has a follow up version on the OLCI, and since the SLSTR will include active fire detection capabilities, the launch of Sentinel 3 will ensure the continuity of the BA collection obtained with this algorithm.

Land surface temperature (LST) will be one of the main products to be generated by the Sea and Land Surface Temperature Radiometer (SLSTR) instrument on board of the upcoming Sentinel-3 mission. LST is an important parameter for a wide variety of earth surface processes and in particular for evapotranspiration. The ESA-funded project WACMOS-ET aims at advancing the development of evapotranspiration estimates at global and regional scales using various earth observations products. As part of this project, LST is computed globally using a consistent retrieval algorithm for satellite data from both low-earth orbit and geostationary instruments. These instruments include the Advanced Along-Track Scanning Radiometer (AATSR), the Spinning Enhanced Visible and Infrared Imager (SEVIRI), the Geostationary Operational Environmental Satellite (GOES), and the Multi-functional Transport Satellites (MTSAT).

In order to evaluate the quality of the resulting LST products, a comprehensive global validation study was carried out. The validation was performed by comparing satellite-derived LST against a) in situ observations acquired at stations located in various land cover types and b) the independent observations of the well-validated MOD11 LST product, which is generated from data acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite instrument onboard of the Terra platform.

A direct comparison of AATSR-derived LST against in situ observations indicated a mean nighttime bias of 0.3 °C and a mean daytime bias of 1.4 °C. The standard deviations were found to be 1.3 °C and 2.5 °C, respectively. The root mean squared error (RMSE) as a measure of overall product accuracy was found to be 1.4 °C and 3.2 °C for nighttime and daytime data, respectively. LST derived from AATSR was found to be negatively affected by insufficient cloud masking during nighttime observations. However, the WACMOS-ET AATSR product was found to provide slightly more accurate retrievals than those of the GlobTemperature AATSR product when the same cloud mask is used. No suitable in situ sites were available for validating MTSAT LST but inter-comparisons with MODIS and AATSR LST showed a good correspondence. LST retrievals from GOES-E showed a good agreement with the ground-based in situ observations. The average bias over seven stations was found to be 0.37 °C for daytime data and as low as 0.21 °C for nighttime data. The standard deviations were found to be 2.9 °C and 2.4 °C, respectively. The RMSE for daytime data was 3.2 °C and for nighttime data 2.6 °C.

Overall, the LST retrievals evaluated here show good to very good performance within the limits of what is currently achievable for LST products. The study indicates that applying a consistent retrieval algorithm for multiple instruments is feasible and provides promising results. The results from this study can be seen as an important first step towards producing merged LST products with high spatial and temporal resolution by combining data from both geostationary and low-earth orbit instruments such as SLSTR.

This paper describes the procedure for estimating the uncertainties budget of the fraction of absorbed photosynthetic active radiation (FAPAR). This product is obtained from Top Of Atmosphere (TOA) data acquired by the ENVISAT/MERIS and Sentinel-3/Ocean Land Colour Instrument (OLCI). The FAPAR algorithm itself takes the form of a set of several formulae which transform calibrated spectral bidirectional directional reflectances into a single numerical value. These formulae are designed to extract the direct green FAPAR in the plant canopy and the rectified channels in the red and near-infrared bands. Following the guidelines of QA4EO, the uncertainties estimation uses uncertainties propagation theory taking into account the one associated to the inputs, i.e. the spectral reflectance measurements at the top of the atmosphere. Methodology and results using simulated data and actual MERIS data will be presented. 

We present a generic inversion package (the JRC-TIP) enabling us to optimally retrieve land surface fluxes and associated land parameters based on a 1D two-stream radiation transfer model. This package implements the minimization of a cost function balancing 1) the misfit between observed and modeled remote sensing fluxes and 2) the deviation from prior knowledge on the 1D model parameter values. This procedure delivers a Gaussian approximation of the PDF of the retrieved model parameters which characterize the radiative status of the vegetation-soil system. Propagation of this PDF to the flux space provides a Gaussian approximation of the posterior PDFs of the transmitted, absorbed and scattered fluxes within the vegetation layer, included  those that are not measured, for example, the absorbed fraction by the background below vegetation.
 
The system is highly flexible and can assimilate any combination of narrowband, broadband or hyperspectral  radiation flux observations. Another definite asset of the JRC-TIP lies in its capability to control and  ultimately relax a number of assumptions that are often implicit in traditional approaches. Through a series of selected examples, the inverse procedure implemented in the JRC-TIP is shown to be robust, reliable and compliant with large scale processing requirements. This is demonstrated through application of the JRC-TIP to the MODIS collection 5 broadband albedo product at one km spatial resolution on global scale over a period of 14 years. The detailed analyses of the retrieval uncertainties highlight the central role and contribution of the effective LAI, the main process parameter to interpret radiation transfer observations over vegetated surfaces. As a further demonstration we describe the application of the package to white sky albedos provided by the Globalbedo processing chain over a period of 10 years. Finally, results from a comparison study of the  JRC-TIP products, including the effective single scattering albedo of the canopy, against in-situ information available from a Fluxnet site will also be presented.

The presented abstract relates to the ESA DUE Project DiversityII, specifically the Dryland component of this dual project with a dryland and an Inland Water part. The 10 Years of MERIS data have been exploited and investigated with FR (Full Resolution) and RR (Reduced Resolution) based fAPAR (Fraction of Absorbed Photosynthetically Active Radiation) data to profile and trace the vegetation development in 22 dryland sites all over the globe. Objectives were to map and assess status and trends of vegetation productivity and express the results in a suite of dryland indicators. This work was driven by the information needs of the CBD (Convention on Biological Diversity) and the UNCCD (UN Convention to Combat Desertification).

The fAPAR were generated from the MERIS data as halfmonthly time series data using the fAPAR algorithm of Gobron (2011). They were compared to in situ fAPAR data as well as to modelled fAPAR data using the BETHY model developed by DLR. Further, co-analysed data sets include TRMM (and other EO based) rainfall data and CCI Soil Moisture data. Phenological and productivity parameters were extracted from the MERIS fAPAR time series data using ERDAS based processing modules specifically developed in the project. The resulting processing chain was also applied to NDVI or fAPAR data of other sensors, and can be potentially used for any other vegetation indices or biophysical variables with seasonal and cyclic behaviour through time. It will thus be also applicable to Sentinel 3 data, and to other sensors bridging the time gap between MERIS and Sentinel 3. The latter is an important aspect, as for drylands a continuous monitoring of the highly variable climatic and consequently vegetation conditions is indispensable.

The work resulted in so called first level and second level indicator maps showing average vegetation conditions of the observation period, trends, and epochal changes. The fAPAR data were integrated for the phenological vegetation cycles at every (pixel) location, being comprised of the growing period(s), the dry season(s), and the entire vegetation year. Special attention was given to these seasonal vegetation indices and their relation (status and trends), pointing to potential degradation and/or functional shifts of the vegetation type. The latter were expressed in second order indices, for which several first order indices were combined. First order indicators include seasonal and yearly (vegetation year) primary productivity indices (NPP proxies), as well as seasonally differentiated rain use efficiency indices for average conditions and yearly trends. In addition, “soil moisture use efficiency” indices were generated, where rainfall data have been replaced by soil moisture data.

The results were in some cases contrasted with faunal species abundance data derived from models and partly from in situ investigations. Their spatial distribution patterns and temporal variability were compared to the derived NPP proxies, to check if expected commonalities can be seen. While a generic interpretation of the derived indicators in terms of primary productivity, vegetation degradation and functional vegetation composition/diversity is presented and related to the overall developments in the test sites, the systematic and direct validation with in situ data remains a problem due to the lack of such data. The project outcome nevertheless point to potentially useful dryland indicators, which have been backed by user requirements and user feedback during the project.