Dans le cadre de sa thèse sur la modélisation hydrologique du delta du Mékong au Cambodge, Christina Orieschnig présentera une méthode utilisant la plateforme cloud-computing de Google Earth Engine pour traiter des images satellitaires et données du Synthetic Aperture Radar dans le but d’estimer l’étendue d’inondations.
Résumé des travaux :
The annual Monsoon inundation of the Cambodian part of the Mekong Delta is fundamental to agricultural productivity. Despite the importance of this seasonal cycle, hydrological data on it is sparse, as there are few measuring stations. In turn, this hinders full understanding of the area’s hydrological dynamics. I am currently working on overcoming part of this obstacle using remote sensing data. The method I am working on combines optical imaging with Synthetic Aperture Radar data, and uses the online cloud computing platform Google Earth Engine for processing. Google Earth Engine offers optical imaging data from the MODIS, Landsat and Sentinel missions, as well as the Synthetic Aperture Radar data from Sentinel used in this analysis. Optical images undergo cloud masking, followed by categorisation into open water surface and flooded vegetation and and dry land. Said categorisation is based on the Normalised Difference Water Index and the Normalised Difference Moisture Index. Since cloud cover is very frequent in the region, especially during Monsoon season, Synthetic Aperture Radar data supplements information that cannot be extracted from optical data even with post-processing. Synthetic Aperture Radar images are classified into the above categories using two-step thresholding. The time series of flood extents resulting from the combination of these two steps is then analysed using the parameters of inundation frequency and variation rate estimation. Validation is undertaken using a manually classified image and, later on, will also include ground-truthing points collected during field missions.
Christina Orieschnig fera une présentation de ses travaux le 21 juin à 11:00 à la salle Mosson.