Abstract
There is an urgent need for operational models that can accurately predict soil moisture patterns in space and time. High spatial and temporal variability of soil moisture and its low degree of autocorrelation complicate the modelling with process-based models. The aim of this research was to evaluate different methods to
... read more
estimate top soil moisture patterns in terms of accuracy, coverage and support: field measurements, process-based modelling and remote sensing based modelling. Additionally, a data assimilation approach that combines the advantages of each method for operational soil moisture modelling was developed and evaluated. In this research three study sites, Barrax (Spain), Sehoul (Morocco) and La Peyne (France), were used. The accuracy, support and coverage of field measurements of top soil moisture was evaluated using Time Domain Reflectometry (TDR) measurements in the Sehoul study area. For the evaluation of a process-based model, the Soil Moisture System (SOMS) model has been developed. SOMS estimates the evolution of top soil moisture patterns for agricultural and semi-natural areas after rainfall events. Remote sensing can be used to estimate soil moisture patterns using empirical, non-linear, relationships between surface temperature, vegetation indices and soil moisture content. On the other hand, remote sensing can provide valuable spatial input data for process-based models, e.g. land cover data, albedo and actual evapotranspiration (AET). The calculation of AET using the Surface Energy Balance System (SEBS) model was explained and evaluated with field measurements from the Barrax study area. In addition am upscaling algorithm, DisTrad, that can be used to estimate sub-pixel temperatures from remote sensing images was evaluated. For the integrated approach, which uses field measurements (TDR), a process-based model (SOMS) and remote sensing data from MODIS, a particle filter data assimilation algorithm was implemented. The algorithm integrates AET derived from remote sensing using error propagation scenarios of SEBS with error propagation scenarios of SOMS. Factors influencing the performance of the particle filter algorithm have been identified. It can be concluded that the use of a particular method for the derivation of soil moisture patterns in space and time depends on the specific demands of the application in terms of support and coverage, in relation with the uncertainty of the method. For applications demanding spatially distributed, operational, soil moisture predictions, the assimilation of all available data, either remote sensing data or field measurements, in a process-based model can benefit from the advantages of each method separately. In this case, the particle filter algorithm used in this study can give good results. Further research is needed to confirm this. Because the availability of field measurements of soil moisture and high resolution, cloudless, remote sensing images with a good temporal coverage remains a problem, research should focus on the optimal use of the many datasets that are available in order to provide operational models for the many environmental problems that we are facing. New data assimilation algorithms can play an important role in this research.
show less
