In a context of increasing pressure on water resources, the search for better agricultural productivity of irrigation water leads to optimize watering schedules according to soil water conditions and crop development stages. Spatial remote sensing can now provide spatialized information in near-real time on soil and vegetation characteristics. In particular, radar data have shown great potential for estimating soil moisture. Similarly, optical data have been used for a long time to estimate vegetation parameters (leaf area index, biomass...). This information can be integrated in crop models to simulate in real time the evolution of the yield. The general objective of the thesis is to show how information from high spatial and temporal resolution remote sensing can be used to retrieve the water and vegetation dynamics of an irrigated area. The approach is based on experiments carried out on an irrigated cereal (maize) plot system, with spatial and ground observations with high temporal repetitivity, and the use of a crop model.
The first part of the thesis aims to evaluate the characterization of soil parameters (roughness and moisture) and vegetation by coupling radar remote sensing data in C and L bands (respectively Sentinel-1 and ALOS-2/PALSAR) and optical (mainly Sentinel-2). The coupled use of C- and L-band data will allow a better estimation of soil moisture due to a better penetration of the L-band radar wave in dense agricultural canopies. In addition, we should be able to jointly estimate soil moisture and soil roughness. The current availability of L-band data (ALOS2/PALSAR) and the planned launch of new L-band sensors (SAOCOM-1A and NISAR) gives this objective a strong scientific legitimacy. Vegetation parameters such as LAI will be computed from Sentinel-2 optical data.
The second component, conducted in parallel, will consist in the realization of a remote sensing module for the Optirrig crop model, developed at the UMR G-Eau (Montpellier) and involved in many academic and operational partnerships. The challenge of this component will be at least to carry out the forcing of observations obtained by remote sensing and if possible to go towards data assimilation in the mathematical sense (i.e. allowing a recalibration of the model parameters).
Key words : remote sensing, Optirrig, irrigation, Sentinel-2, vegetation, LAI, soil moisture
Figure 1: Illustration of the difference (E) between the LAI without irrigation (LAI0i) and the noisy one (vLAI) representing the LAI derived from remote sensing that allows us to identify evidence of irrigation between the Sentinel-2 images (ti)
