Utilisation de la teledetection pour l'estimation de la reserve hydrique au bassin du Mackenzie au nord ouest Canadien

This work elaborates a method to estimate soil moisture and open water extent using remote sensing data. Specifically, Microwave data are used in this work in combination with other sources of information. This study is applied to Mackenzie River Basin as it was conducted in the framework of the MAGS project (Mackenzie GEWEX Study). The methodology developed in this work is based on the concept of the wetness index which is a surrogate of the measured soil moisture. Firstly, it was proposed to used wetness index derived from a linear combination of brightness temperatures measured by the SSM/I sensor. Beside the correlation of this index with the observed discharge at some observing points of the Mackenzie River basin, it was noticed that the empirical parameter of the index were affected by the variation in space and time of the surface conditions. The sensitivity of the wetness index was improved by considering this effect. Secondly, the potential of combination of passive microwave data and discharge observation for flood forecasting was investigated. The methodology was based on the concept of the rating curve model. The Kaman filter was used to update the parameters of the rating model at the reception of each new satellite image. The estimated water surface fractions using the discharge observations showed a reasonable concordance with those derived from visible images (NOAA-AVHRR). The publication of the AMSR-E data has enabled us to improve the spatial resolution of the passive microwave data. Thus, a new formulation of a wetness index was proposed in the third part of this work. The proposed Basin Wetness Index is based on the difference of the sensitivity of passive microwave and visible images to soil moisture. Microwave data are sensitive to the soil water content and open water extent. However, visible images can mirror only the water extent. Information about soil moisture can be derived from the difference of these sensitivities. This index was able to qualitatively describe the temporal evolution of the wetness over the Mackenzie River Basin. It is worth mentioning that vegetation distribution was considered in the formulation of the proposed index by the segregation between bare and vegetated soil soils by the mosaic approach. The BWI was finally downscaled using Digital Elevation Model data. The sensitivity of the topography attributes to the spatial distribution of soil moisture was assessed. A classical formulation of topographical index was modified in this study to take into account the difference of sensitivity of topographical attributes over bare and covered soils. The combination of the modified topographical index and the proposed BWI enabled us to downscale the soil water content estimation and provide an estimate of soil moisture at the scale of the Digital elevation Model. Overall, the downscaling approach provided satisfactory resultants when compared to precipitation and temperature variation. Further works are needed to assess the reliability of the proposed approaches using results of distributed hydrological model outputs.