Soil Moisture Retrieval using Sentinel-1 data: a Bayesian Multitemporal Approach
Pierdicca, Nazzareno1; Pulvirenti, Luca2
1Sapienza University of Rome, ITALY; 2CIMA Research Foundation, ITALY
Soil moisture content (mv) is a key parameter that influences both global water and energy budgets because it controls the redistribution of rainfall among infiltration, runoff, percolation in soil, and evapotranspiration. Its knowledge is therefore essential for many applications, such as drought and flood predictions, weather forecasts, as well as climatology and agronomy. The data provided by Synthetic Aperture Radar (SAR) systems are very useful to map mv over wide areas because of the sensitivity of the electromagnetic surface scattering to the water content and the transparency of the atmosphere in the microwave range of the electromagnetic spectrum, even in cloudy conditions. Moreover, SAR-derived mv maps may have spatial resolution in the order of hundreds of meters, as opposed to those derived from other microwave instruments such as radiometers and scatterometers whose spatial resolution is in the order of tens of kilometres. SAR measurements are sensitive not only to soil moisture, but also to surface roughness and, in the presence of vegetation, to biomass and canopy structure. This implies that mv retrieval is generally an ill-posed problem, especially if the observations from a single configuration (i.e. single frequency, polarization and incidence angle) system are used. The forthcoming European Space Agency (ESA) Sentinel-1 (S-1) mission, which is an element of the European Global Monitoring for Environment and Security (GMES) program, will provide C-band radar data characterized by short revisit time (the two-satellite constellation will offer six days exact repeat). Such a short revisit time will allow for the generation of frequent soil moisture maps, which are expected to fulfil the temporal resolution requirement of a soil moisture product, which is in the order of five days. Moreover, recent works have proved that the ill-posedness of the retrieval problem can be tackled if a time-series of SAR observations is available. In these works, both change detection techniques (e.g., ) as well as Bayesian approaches  have been used. This work presents a multitemporal algorithm conceived to be operationally used to map mv from S-1 data; it has been designed within the framework of the ''GMES Sentinel-1 Soil Moisture Algorithm Development'' project funded by ESA. The algorithm takes advantage of the short revisit time of S-1 data by assuming that, considering a time interval in which a temporal series of S-1 images is available, the average characteristics of surface roughness do not change, as opposed to soil moisture, whose temporal scale of variation is shorter than that of roughness. The temporal series of radar data, possibly corrected for the vegetation effects, is integrated within the retrieval algorithm that is based on the Bayesian maximum posterior probability (MAP) statistical criterion. The MAP estimator is applied by inverting a forward soil backscattering model relating the backscattering coefficient to the bare soil parameters (i.e., not only mv, but also soil roughness). The current version of the algorithm uses the semiempirical forward model proposed by Oh et al. , but a possible change of the selected model can be easily accomplished because the algorithm specification and its numerical implementation are completely independent of the forward model complexity, being based on look up tables. A combination of on-line and off-line processing has been designed in order to decrease the time necessary to produce a soil moisture map, which may be a critical aspect of multitemporal approaches. As for the correction of the vegetation effects, which represents a critical aspect of the retrieval procedure, the well-established water cloud model , useful at least when the vegetation is at a low/moderate stage of growth is used. To deal with dense vegetation, an incremental approach that considers as a reference the images acquired over bare soil, developed in , is currently tested. The results of an operational test of the algorithm, performed by using ERS-1/2 data as well as data of the AGRISAR 2006 campaign, will be presented at the conference. Moreover, the issues related to its implementation into a software prototype conceived to be used in a processor hosted by the S-1 ground segment will be discussed.
 Pathe, C.; Wagner, W.; Sabel, D.; Doubkova, M.; Basara J. B.: Using ENVISAT ASAR Global Mo-de Data for Surface Soil Moisture Retrieval O-ver Oklahoma, USA. IEEE Transactions on Geoscience and Remote Sensing. Vol. 47, No. 2, pp. 468-480, 2009.
 Pierdicca, N.; Pulvirenti, L.; Bignami, C.: Soil moisture estimation over vegetated terrains u-sing multitemporal remote sensing data. Remote Sensing of Environment. Vol. 114, pp. 440-448, 2010.
 Oh, Y.; Sarabandi, K.; Ulaby, F. T.: Semi-Empirical Model of the Ensemble-Averaged Differential Mueller Matrix for Microwave Backscattering From Bare Soil Surfaces. IEEE Transactions on Geoscience and Remote Sensing. Vol. 40, No. 6, pp. 1348-1355, 2002.
 Attema, E. P. W.; Ulaby, F. T.: Vegetation modeled as a water cloud. Radio Science. Vol. 13, pp. 357–364, 1978.