Mission Exploitation

TanDEM-X (TerraSAR-X add-on for Digital Elevation Measurements) is an interferometric SAR mission flying two radar satellites in close orbit formation. Its primary objective is the production of a homogeneous global digital elevation model (DEM) of unprecedented accuracy. Since 2010 all land surfaces have been mapped at least twice and difficult terrain even up to four times or more. The interferomeric data acquisition for the global DEM has been concluded in September 2014, and it is expected to complete the processing of the full DEM in early 2016.

The first part of this paper gives a description and summary of the DEM acquisition planning and presents the current status of the continuous quality monitoring that accompanies and supports the DEM generation process. The acquisition of the first global coverage without Antarctica has been completed in January 2012. The second global coverage was finished end of March 2013 and was followed by a short recovery phase used for gap filling. Antarctica has also been acquired twice in separate phases during the local winter in 2013 and 2014. Interferometric performance parameters like coherence of the two radar images of an acquired scene and estimated height errors have been monitored and evaluated throughout the mission. The quality analysis is based on almost 500,000 single DEM scenes and 5000 final DEM tiles.

In October 2014 the mission entered a new phase dedicated to science activities, which shall be addressed in the second part of this paper. This Science Phase is planned for 15 months (till end of 2015) and is dedicated to experimental and scientific applications. . It offers the opportunity to set special baselines and allows the commanding of new interferometric imaging modes in larger quantities for the fulfillment of the secondary mission objective of TanDEM-X, the demonstration of new SAR techniques and applications. While science acquisitions will have priority during the 15-month Science Phase period, it is nevertheless planned to acquire additional data to fill gaps and further improve the quality of the TanDEM-X DEM and to acquire data for high-resolution DEMs.

Interferometric data acquisition with the TanDEM-X satellite formation can be achieved in three interferometric modes: Bistatic, Pursuit Monostatic and Alternating Bistatic. The three cooperative modes may further be combined with different SAR imaging modes like Stripmap, ScanSAR, and Spotlight, the last mode being in sliding spotlight acquisition geometry. Also the new imaging modes like Staring Spotlight and wide-beam ScanSAR mode will be available during the Science Phase. Furthermore, the full polarimetric capability and short along-track baseline interferometry using the Dual Receive Antenna Mode can be explored for the first time during the TanDEM-X mission.

Pursuit monostatic acquisitions are available from October 2014 till February 2015 and are characterized by a set of drifting across-track baselines at all latitudes and within short time periods ranging between 0 to 750 m. Beginning in December 2014 the Dual Receive Antenna mode is switched on and will be operated until the end of Science Phase. The variety of baselines is ideal for SAR tomography studies and for polar regions applications. The along-track separation between the satellites will be large enabling the observation of small velocities, as they can be detected from ships or drifting sea ice.

The following bistatic acquisition phase with a duration of 10 months is characterised by three elements, the operation of the Dual Receive Antenna mode, the stable huge (3-4 km) and the short (0-250 m) across-track interferometric baselines. This phase will be suitable for fully polarimetric and 4 phase center bistatic experiments.

The flexible adjustment of the interferometric baseline and instrument mode variety is a unique feature of the TanDEM-X mission. This paper will give an overview on the technical capabilities during the science phase and how specific interferometric modes can be achieved in diverse imaging and polarization modes.

The Italian Space Agency (ASI) promotes Earth Observation applications related to themes such as the prediction, monitoring, management and mitigation of natural and anthropogenic hazards: floods, landslides, fires, seismic risks, volcanic hazards, air quality, oil pollution on sea, coastal management, etc. The approach generally followed is the development and demonstration of prototype services, using currently available data from space missions, in particular the COSMO-SkyMed mission, which represents the largest Italian investment in Space System for Earth Observation and thanks to which Italy plays a key role worldwide.

Projects funded by ASI provide the convergence of various national industry expertise, research and institutional reference users. In this context, a significant example is provided by the  ASI’s Pilot Projects such as OPERA (Floods), MORFEO (Landslides), SIGRIS (Seismic Risk), SIGRI (Fires), PROSA (Now-casting/Meteorologocal Alert), SRV (Volcanoes), PRIMI (Oil Spill), SISMA (Seismic Risk), QUITSAT (Air Quality/Environment Monitoring), recently concluded.

In this paper a special focus will be addressed to the volcanic risk and to the contribution provided in this field by COSMO-SkyMed data during the last years, starting from the  ASI-SRV (Sistema Rischio Vulcanico) project, coordinated by the INGV - Istituto Nazionale di Geofisica e Vulcanologia, which is responsible at national level for the volcanic monitoring. The SRV project started at the beginning of  2007 with the main objective to develop a pre-operative system based on Earth Observation data and ground measurements integration to support the volcanic risk monitoring of the Italian Civil Protection Department. This project demonstrated that, in order to study the volcanic processes and to produce applications for the volcanic risk management, it is necessary to have temporal series of data acquired in interferometric configuration and always updated over the volcanic areas.

On the basis of this considerations, a wide COSMO-SkyMed images archive has been populated allowing the national and international scientific community, as well as companies developing commercial services and applications, to utilize these data for volcanic applications.

Basically, after defining a list of volcanoes to be monitored worldwide, an acquisition plan has started as part of the global COSMO-SkyMed Background Mission. This is a low priority acquisition plan defined on the basis of both institutional and commercial needs, with the main goal to create an always updated archive for interferometric applications, capitalizing the previous ENVISAT experience. The acquisitions populating this handbook are periodic, with a pre-defined geometry and scheduled over a long period of interest. For what concern the volcanic risk monitoring, the radar images have to be acquired with the same geometry in order not to lose the coherence between consecutive acquisitions. Furthermore, in order to separate the deformation’ components it is necessary to have acquisitions on both the orbit directions. In particular, the extended volcanic areas are acquired in Stripmap mode while the smaller in the Spotlight mode.

This paper gives an overview of the various national and international projects using the data of the COSMO-SkyMed satellite constellation for the volcanic risk mitigation, and also gives a shortly account of the obtained results, highlighting the Italian contribution provided worldwide in the framework of this important topic.

Introduction

The SAOCOM mission, being developed by CONAE from Argentina (Comision Nacional de Actividades Espaciales) is an earth observation system composed of two identical satellites (SAOCOM 1A and SAOCOM 1B) flying in constellation, carrying each one a fully polarimetric Synthetic Aperture Radar (SAR) operating in L-band.

The two SAOCOM satellites will fly in a Low-Earth Orbit at 620 km altitude, providing 16-day revisit-time for single-pass acquisitions and 8-day revisit-time with both SAOCOM satellites. They will also be in constellation with 4 Italian COSMO Sky-Med satellites.

It is in the interest of CONAE to provide interferometry products to final users, by exploiting SAOCOM Stripmap acquisitions. Hence, a mathematical model was developed to assess the expected accuracy of interferometry products, considering multiple error sources (orbit knowledge error, decorrelation noise, DEM errors, etc.)

In this work, we address the suitability of the SAOCOM system to estimate surface deformation mapping. We present the error model developed for estimating the LOS displacement variance given the SAOCOM Stripmap acquisition modes characteristics, and discuss some of the obtained results.

Error model for surface deformation mapping

In absence of other phase components, the relationship between interferometric phase and Line-of-Sight displacement is given by [1]:

d_LOS=λ/4π Δϕ^DInSAR      (1)

In a real case, the interferometric phase contains information related with the acquisition geometry  (Δϕ^orbit) , the topography of the illuminated area (Δϕ^topo), soil displacement produced between S1 and S2 acquisition times (Δϕ^defo), differences in the atmospheric and/or ionospheric state at the time of S1 and S2 acquisitions (Δϕ^defo) , and decorrelation noise (Δϕ^noise ). The differential phase at a point of coordinates (x, R) in range and azimuth can be expressed, after topographic, orbital and atmospheric phase compensation, as:

Δϕ^DInSAR=E{Δϕ^deformation }+Δϕ^{res orbit}+Δϕ^{res topo}+Δϕ^{res atm}+Δϕ^noise     (2)

where residual terms account for uncertainty in estimating the true orbital, topographic an atmospheric phase contributions present in the original interferogram.

Furthermore, in a practical case LOS deformation observations are obtained by integrating spatial gradients between resolution cells. In other words, deformation at one generic point p of coordinates (x_p,R_p) is relative to another point q of coordinates (x_q,R_q) located inside the processed scene. As a consequence, the estimated relative LOS displacement between p and q as a function of the relative interferometric phase change is:

d_pq^LOS=λ/4π∙ΔΦ_pq^DInSAR=λ/4π∙[E{Δϕ_pq^defo }+Δϕ_pq^res_orbit+Δϕ_pq^res_topo+Δϕ_pq^atm+Δϕ_pq^noise ]    (3)

where Δϕ_pq is short notation for Δϕ_p-Δϕ_q.

From previous equation, and assuming that all terms are uncorrelated, LOS displacement variance is:

σ²_dpq^LOS=λ/4π∙[σ²_{Δϕpq^{res orbit}} +σ²_{Δϕpq^{res topo}} +σ²_Δϕpq^atm+σ²_Δϕpq^noise ]    (4)

Residual orbital variance (σ²_{Δϕpq^{res orbit}}) can be estimated by considering that the perpendicular baseline is known except for an error ΔB_⊥. Under this assumption, residual orbital variance between two points located at ranges p and q can be expressed as a function of perpendicular baseline uncertainty as :

σ²Δϕ_pq^{res orbit} =[4π/λ· ∫1/(r∙sin ϑ₀) (cos ϑ₀-r/(R_E+H) )∙dr]²∙σ²_Bperp    (5)

Topographic residual term, (σ²_{Δϕpq^{res topo}}), is estimated by considering the expression derived in a previous work [2]. Uncertainty sources are the baseline error (ΔB_⊥) and the DEM error (Δz). Residual topography variance is expressed as a function of the perpendicular baseline uncertainty and DEM errors at point p and q:

σ²_{Δϕpq^{res topo}}=[(4πB_perp/λ · 1/(r_p sin ϑ_p)]² ∙σ²_zp+[(4πB_perp/λ · 1/(r_q sin ϑ_q)]² ∙σ²_zq-[4πB_perp)/λ]² · 2/(r_p r_q sin ϑ_p sin ϑ_q)∙σ²_{zp zq}+[4πz_q/(λ r_q sin ϑ_q)]²·σ²_Bperp    (6)

Decorrelation variance have been estimated by numerically integrating the interferometric phase probability density function for distributed targets [3]. In order to relate the SAOCOM operating parameters with coherence we decomposed the last in the following terms:

γ=γ_thermal∙γ_geom∙γ_doppler∙γ_proc∙γ_temporal    (7)

Thermal decorrelation term (γ_thermal) have been estimated as γ_thermal=1/(1+SNR^-1 ), following Curlander and McDonough [4]. Geometric term _(γgeom) is derived from the linear model presented by Zebker and Villasenor [5]. Doppler term (γ_doppler) was computed following the linear model presented in Hanssen [6]. Processing term was computed from values provided by CONAE. Temporal decorrelation is like atmospheric errors: do not depend on the system settings, so we considered a fixed value.

Results

We implemented the described error model in a Python programmed interface. A plain text parameter file provides an easy way for setting acquisition geometry parameters like SAOCOM Stripmap mode, perpendicular baseline, perpendicular baseline error and parallel baseline rate error; terrain parameters like temporal coherence, slope, height change; and processing parameters like filtering, DEM error and multilook pixel dimensions. Satellite operating parameters are taken from a database which contains all the relevant information like carrier frequency, bandwidths, PRF, sampling frequency, look angle, etc.

With this tool we estimated the expected LOS displacement errors along a range line for SAOCOM Stripmap acquisition modes. We present here the results for acquisition mode DP5, which is representative of the performance achievable with the rest of the SAOCOM Stripmap acquisition modes. The main characteristics of DP5 mode are summarized in Table 1.

Figure 1 shows the results. Green, blue, solid and dashed lines represent different acquisition / processing conditions. Common parameters used for computing them are the following: mode S5 DP, perpendicular baseline error σ_Bperp=0.1 m, multilook pixel is a 25m square, processing includes spectral filtering, DEM error Δz=3m, flat terrain (slope = 0, height change = 0m), coherence due to processing errors is 0.95

The horizontal axis shows the characteristic spatial width of the imaging system, and ranges between 25m (black solid line) and 50km (black dashed line) that is the swath width for the investigated acquisition mode. Vertical axis show the measurable LOS deformation. Solid lines at the bottom represent the LOS deformation standard deviation computed by considering temporal coherence of 0.99, null perpendicular baseline (green) and 2000m perpendicular baseline (blue). Dashed lines are obtained by setting temporal coherence to 0.6, null perpendicular baseline (green) and 2000m perpendicular baseline (blue). All those lines represent the system’s precision under different circumstances. Red lines at the figure’s top represent the maximum LOS change measurable without phase aliasing. Dashed line is computed considering full-resolution processing, i.e. unwrapping the full resolution interferogram, whereas solid line is computed by processing at the selected multilooking factors.

The space defined by the lines represents the deformation pattern that S5 DP SAOCOM mode is capable of characterizing by DInSAR techniques. We also included typical LOS deformations for volcanoes (pre-eruptive, inter-eruptive and eruptive cycles), subsidence and glaciers after 8 days (purple) and 1 year (green) taken from the literature.

In a very favorable situation (solid green line) system’s precision is about 2mm. Error increasing at far range is due to the perpendicular baseline error which becomes more significant as distance from the reference point  (near range) increases. However, considering a temporal coherence of 0.6 (solid blue line), a precision of about 1cm is achievable. With increasing baseline, precision degrades to between 1 and 2 cm depending on temporal coherence (dashed lines). Note that for baselines between 0m (uncommon) and 2000m and good coherence the precision always maintains sub-centimeter.

Conclusion

A mathematical model was presented to assess the surface deformation mapping accuracy achievable by processing a single interferometric pair. The developed model is general enough to account for multiple error sources and multiple sensors characteristics. SAOCOM Stripmap acquisition parameters were particularly considered to compute the final errors.

As a conclusion of the analysis, we can say that SAOCOM system, although its main objective is soil moisture characterization, has also good interferometric capabilities. As shown by the figure, it is capable of resolving a good number of real deformation patterns. Taken into account that the shown simulations only consider one image pair, even better error metrics are expected from time-series processing.

References

[1] H. Zebker, P. A. Rosen and S. Hensley, Atmospheric effects in interferometric synthetic aperture radar surface deformation and topographic maps, J. Geophys. Res., 102(B4), 7547–7563, doi:10.1029/96JB03804., 1997

[2] A. Pepe, Advanced differential Interferometric SAR techniques, Chapter 1: SAR Fundamentals, PhD. Thesis, Universita’ degli Studi di Napoli “Federico II”, 155 pp., 2006

[3] R. Bamler and P. Hartl, Synthetic aperture radar interferometry, Inverse Problems, 14, R1-R54, 1998

[4] J. C. Curlander and R. N. McDonough, Synthetic Aperture Radar. Systems and Signal Processing. New York: John Wiley & Sons, 1991.

[5] H. A. Zebker and J. Villasenor, Decorrelation in interferometric radar echoes, IEEE TGARS, DOI - 10.1109/36.175330, vol. 30, no. 5, pp. 950–959, 1992.

[6] R. F. Hanssen, Radar Interferometry - Data Interpretation and Error Analysis, vol. 2. U.S.: Kluwer Academic Publishers, 2001.

Interferometric synthetic aperture RADAR (InSAR) has been used for geophysical modeling of the earth’s surface deformation for decades. The advantages of the interferometric methods over the terrestrial geodetic approaches; such as leveling, GNSS networks, seismic sensors etc., have drawn the attention of the geophysicists to this tool. Although a great number of publications have focused on the application of InSAR in deformation source modeling as well as the development of different algorithms in this regard, little investigation has been dedicated to the sensitivity analysis of the interferometric methods in deformation source modeling. The purpose of the current study is to address this issue by analyzing the reliability of interferometric approaches in modeling the deformation sources due to landslides, seismic and volcanic activities, with special focus on the L band SAR measurements.

The sensitivity analysis is considered for three commonly used deformation models in case of subsidence, seismic and volcanic activities; namely, the Gaussian subsidence bowl, Okada and Mogi point source, respectively. In each of the cases, the InSAR sensitivity is analytically formulated and its performance is investigated using simulated SAR data. The investigations are carried out using stochastic error propagation approaches to infer the precision of the models’ parameters as well as their mutual covariance. The limiting factors in SAR interferometry are categorized in two groups and investigated separately in sensitivity analysis; with the first dealing with the geometrical limits imposed by the side looking geometry of the SAR measurements and the second focusing on the InSAR stochastic characteristics in the L band.

With the deformation pattern mapped on the line of sight direction of the SAR, it is important to quantize the respective geometrical limits imposed on the deformation source modeling. To overcome the geometrical limit, two possibilities have been considered: the fusion of SAR measurements acquired from multi aspects as well as adding the resolution scale differential range measurements (i.e. pixel offset) in azimuth direction to the InSAR measurements. Firstly, each of these possibilities is discussed and their limitations are introduced. Secondly, the accuracy of the aforementioned methods in retrieval of the three dimensional deformation is investigated. Finally and most importantly, the impact of three dimensional motion decomposition on retrieval of the source models’ parameters is investigated. In this regard, four different cases are considered and compared: the single aspect InSAR measurements, fusing multi aspect InSAR measurements with ascending/descending right looking geometries, adding the left looking geometry to the multi aspect fusion scenario and finally adding the azimuth offset measurements to the three considered cases. Based on the achieved results, the relevance of geometrical limit on retrieving each of the introduced models of Gaussian bowl, Okada and Mogi is finally discussed.

In the second part, the error sources affecting the L-band InSAR measurements, such as temporal decorrelation, atmospheric delays and instrumental thermal noises, are described in an appropriate stochastic model. The stochastic model is further used in error propagation from the InSAR measurements to the deformation source parameters. Finally the possibility and effect of advanced stacking techniques such as persistent and distributed scatterer interferometry on improving the accuracy of source modeling is discussed.

The primary objective of the TanDEM-X mission is the generation of a global, high-precision digital elevation model (DEM) by using synthetic aperture radar (SAR) interferometry. The constant monitoring of quality parameters is necessary for optimizing the acquisition strategy and achieving the final mission specification. TanDEM-X offers a vast global data set of multiple InSAR acquisitions, each of them supplemented by quicklook images of different SAR quantities, such as amplitude, coherence, and DEM.

One of the main parameters for asserting the quality of interferometric products is the interferometric coherence between the monostatic and the bistatic images. Global mosaics of such a quantity are operationally used to quickly analyze the mission performance and to derive further key-quality parameters, such as the relative height error. Their analysis represents a powerful mean for optimizing the acquisition planning. For example, large scale mosaics of the relative height error over sandy desert areas has allowed to determine critical regions to be re-acquired using dedicated geometries, in order to improve the final DEM quality. Moreover, large scale mosaics of the interferometric coherence represent a valid starting point for further scientific applications and for the generation of by-pass products from the TanDEM-X global data set. The quality of interferometric data pairs is strongly influenced by the specific land-cover type of the illuminated scene on ground. Over forested regions the presence of multiple scatterers at different heights and within a single resolution cell results in an increase of the interferometric phase uncertainty, whose intensity depends on several factors, such as the acquisition geometry, the sensor parameters, and the canopy density. Hence, from each bistatic TanDEM-X image the volume decorrelation contribution can be isolated from the interferometric coherence and used for the generation of a global mosaic of forest/non-forest. The amount of volume decorrelation is also a good indicator of the vegetation density and sets the basis for the development of a forest-type classifier.

 In the field of glaciology, the classification of snow pack facies is an interesting and open topic. The combination of the information coming from both backscatter intensity and interferometric coherence can be exploited for better classifying different kinds of snow, dominated by different scattering mechanisms. An example is the definition of the snow facies of the Greenland ice-sheet, obtained by properly combining amplitude and coherence mosaics of the area.

The aim of this paper is to present the described mosaics derived from TanDEM-X data, to explain their operational use within the TanDEM-X mission, and to introduce the scientific potential of the global TanDEM-X interferometric data set, not only for topography measurements, but for scientific applications as well. 

Earth Explorers are the backbone of the science and research element of ESA’s Living Planet Programme, providing an important contribution to the understanding of the Earth system.  Following the User Consultation Meeting held in Graz, Austria on 5-6 March 2013, the Earth Science Advisory Committee (ESAC) has recommended implementing Biomass as the 7th Earth Explorer Mission within the frame of the ESA Earth Observation Envelope Programme. This paper will give an overview of the satellite system and its payload. The system technical description presented here is based on the results of the work performed during parallel Phase A system studies by two industrial consortia led by EADS Astrium Ltd. and Thales Alenia Space Italy. Two implementation concepts are described and provide viable options capable of meeting the mission requirements.

The mission will be implemented as fully polarimetric P-band (435 MHz) Synthetic Aperture Radar (SAR) with an orbit that will allow repeat pass interferometry. The primary scientific objectives of the Biomass mission are to determine the distribution of aboveground biomass in the world forests and to measure annual changes in this stock over the period of the mission to greatly enhance our understanding of the land carbon cycle. Secondary mission objectives cover mapping of the topography below forest, desert subsurfaces and ice. 

The Biomass space segment comprises a single low Earth orbit satellite platform carrying the SAR instrument. The SAR antenna is based on a large deployable reflector (12 m circular projected aperture) with an offset feed array and a single-beam. The spacecraft carrying the P-band SAR is planned for launched by Vega in the 2020 timeframe and will operate in a near-polar, Sun-synchronous quasi-circular frozen orbit at an altitude of 637–666 km, depending on the different mission phases. The orbit is designed to enable repeat pass interferometric acquisitions throughout the mission’s life. The baseline observation principle is based on double-baseline interferometric acquisitions, with a repeat cycle (RC) of 4 days. The mission is designed to acquire data at dawn/dusk, i.e. 06:00/18:00 local time (at the equator), to minimise the adverse influence of the ionosphere on the radar signal.