Context and objectives
Up-to-date information is crucial for the monitoring of environment-related processes. Unfortunately, despite the variety of satellite imagery, up-to-date detailed information is rarely available. Coarse and medium resolution provide cheap and frequent updates but the image is often not detailed enough. High resolution images are increasingly available but on small surfaces and to a prohibitive cost particularly for malaria control. Detailed information from land covers is generally available but often out of date due to the long process implied in developing such dataset for a whole country. The malaria control workers have thus to choose between cheap up-to-date data from coarse or medium spatial resolution, which is often not sufficient for the purpose, or detailed dataset produced by other services but that are often three to five years old and thus not relevant either for the purpose. Innovative methods are thus needed to provide more frequent updates and increase the opportunity of dynamic predictive mapping. Multi-sensor data fusion and downscaling techniques could offer an alternative. This research has thus four objectives: (1) develop a Bayesian Data Fusion method to merge data from various source and spatial resolution referring to the same date (2) adapt this approach to temporally non overlapping dataset (3) use the approach to map land surface descriptors of interest for malaria vector An. dirus s.l. (land cover, vegetation water content) (4) Delineate An. dirus s.l. distribution and dry season habitat using those descriptors.