Abhatoo - Centre Nationale de Documentation

Optimizing water management in agriculture is essential over semi-arid areas in order to preserve water resources which are already low and erratic due to human actions and climate change. This thesis aims to use the synergy of multispectral remote sensing observations (radar, optical and thermal data) for high spatio-temporal resolution monitoring of crops water needs. In this context, different approaches using various sensors (Landsat-7/8, Sentinel^-1 and MODIS) have been developed to provide information on the crop Soil Moisture (SM) and water stress at a spatiotemporal scale relevant to irrigation management. This work fits well the REC "Root zone soil moisture Estimates at the daily and agricultural parcel scales for Crop irrigation management and water use impact : a multi-sensor remote sensing approach" (http://rec.isardsat.com/) project objectives, which aim to estimate the Root Zone Soil Moisture (RZSM) for optimizing the management of irrigation water. Innovative and promising approaches are set up to estimate evapotranspiration (ET), RZSM, land surface temperature (LST) and vegetation water stress through SM indices derived from multispectral observations with high spatio-temporal resolution. The proposed methodologies rely on image-based methods, radiative transfer modelling and water and energy balance modelling and are applied in a semi-arid climate region (central Morocco). In the frame of my PhD thesis, three axes have been investigated.
In the first axis, a Landsat LST-derived RZSM index is used to estimate the ET over wheat parcels and bare soil. The ET modelling estimation is explored using a modified Penman-Monteith equation obtained by introducing a simple empirical relationship between surface resistance (r_c) and a RZSM index. The later is estimated from Landsat-derived land surface temperature (LST) combined with the LST endmembers (in wet and dry conditions) simulated by a surface energy balance model driven by meteorological forcing and Landsat-derived fractional vegetation cover. The investigated method is calibrated and validated over two wheat parcels located in the same area near Marrakech City in Morocco.
In the next axis, a method to retrieve near surface (0-5 cm) SM at high spatial and temporal resolution is developed from a synergy between radar (Sentinel^-1) and thermal (Landsat) data and by using a soil energy balance model. The developed approach is validated over bare soil agricultural fields and gives an accurate estimates of near surface SM with a root mean square difference compared to in situ SM equal to 0.03 m³ m^-3.
In the final axis a new method is developed to disaggregate the 1 km resolution MODIS LST at 100 m resolution by integrating the near surface SM derived from Sentinel-1 radar data and the optical-vegetation index derived from Landsat observations. The new algorithm including the S-1 backscatter as input to the disaggregation, produces more stable and robust results during the selected year. Where, 3.35 °C and 0.75 were the lowest RMSE and the highest correlation coefficient assessed using the new algorithm.