IMPROVED RIFT VALLEY FEVER VECTORS AGGRESSIVENESS PREDICTION FROM HETEROGENEOUS RAIN FIELDS USING HIGH-RESOLUTION SATELLITE DATA

Rainfall is a key factor for vector-borne diseases such as Malaria, Rift Valley Fever (RVF), and Dengue Fever among others. What are the consequences on epidemics from spatial rainfall heterogeneity at local scale (i.e., 1 – 10 km)? A model for vector risk prediction of RVF in the Ferlo (Senegal) is being proposed. It is based on three steps: 1) rainfall estimation; 2) a simple model for pond dynamic (primary larvae breeding grounds); and 3) mosquito life cycle model and associated epidemiological risks. Initially the rainfall input for the prediction model was provided by a single rain gauge from the Senegalese weather service. As gauge-based estimation is affected by spatial sampling errors, one can legitimately question the reliability of a single gauge estimation in a region (Sahel) where convective rainfalls are highly heterogeneous. Additionally a sole rain gauge cannot provide information about the rainfall spatial structure. Because of non-linearity in the causality chain from rain to vectors (mainly when runoff is concerned) rainfall heterogeneity strongly impacts the level of risk and thus can not be ignored.

Here the analysis includes the sensitivity of the prediction to the spatial variability of rainfall inside the studied domain (45 km x 45 km) using high resolution satellite rainfall. Several state-of-the-art high resolution satellite rainfall products were tested: TRMM-3B42V6, TRMM-3B42RT, GSMAP-NRT, GSMAP-MKV, RFE2, CMORPH and PERSIANN. The satellite products were first evaluated by comparison with a dense rain gauge network in Niger and then corrected using a probability matching method. They were subsequently used as forcing for the prediction. The dynamic, number and size of the predicted ponds could be thus compared with that from satellite monitoring (2003 to 2010 period). The results show that high resolution satellite rainfall products improves prediction of ponds dynamic in the Sahel and provides a more realistic spatial distribution of the RVF risks of epidemics.