Abstract
In southern Tunisia, not all irrigated oasis-type perimeters have undergone the same development, we observed an evolution of the covered surfaces which more than doubled in the last half-century, while for other regions they have remained practically stable and in some cases, a decrease in these areas. These changes have affected environmental and economic systems. In this context, the evaluation of the state of oases and the development of a typology of oases systems is a key-issue for sustainable agriculture. To know the state of vegetation in these oases, monitoring systems for oasis ecosystems must be informed by data on cultivated areas. These data can be obtained in part by satellite observation systems with high and moderate spatial resolution and high temporal repetitiveness, offer a synoptic vision that makes them a particularly appropriate information source for the estimation of such data. The research work presented here focuses on the exploration of methods developed from two types of time series of Earth observation images: those produced by the SPOT-5 experiment (Take5) and the MOD13Q1 product of the MODIS sensor, at 10m and 250m spatial resolution respectively. These methods and data were tested in the Djerid region with the final aim of setting up an oasis monitoring system based on the analysis of time signatures from Earth observation images made very regularly over time.Two change detection approaches based on NDVI time series. The first consists on temporal variations in vegetation activity over a short period from April to September 2015 through the SPOT-5 time series (Take5): the comparison between oases was made at the scale of the irrigated perimeter (an oasis can be composed of several irrigated perimeters) using agglomerative hierarchical clustering (AHC) method.The second uses a temporal decomposition technique to extract the trend from a multi-year time series at the scale of a geographical point (a 250mx250m pixel) across the MOD13Q1 time series (2000-2016).Results obtained from the processing and analysis of optical time series have shown that it is possible to identify the main types of irrigated perimeters present in the Djerid region, and to retrospectively trace their recent development history. They also highlight the fact that SPOT-5 (Take5) images, which prefigure those currently available with images produced by Sentinel2 satellites, significantly improve the spatio-temporal characterization of oases functioning through their 10m spatial resolution and 5-day temporal repetitiveness.The results of this thesis highlight new possibilities for the combination of remote sensing, field data and statistical analysis, delivering nonstop information in time and space on the characterization of oases systems. Indeed, with a single sensor such as Sentinel2, coupled with the historical data of MOD13Q1, it is now possible to accurately characterize oases on a continuous basis.