Utilizing Multi-Sensor Data Products and high-resolution flood model in Analyzing North African Hydrological Processes

North Africa is an arid region characterized by isolated extreme events such as floods and droughts. Our present understanding of hydrological processes over North Africa is limited due to low rainfall, mixed response of evaporation to temperature and soil moisture gradients, and lack of high-resolution ground measurements. Remote sensing is an excellent way to obtain near real- time data of high spatial and temporal resolution. Satellite estimates of rainfall and evapotranspiration (ET) have uncertainties due to topography, land-sea contrast, complex weather, and climate variability for high-elevated regions. Generally for arid regions, the satellite precipitation instruments are sensitive to soil moisture and land surface geometry. This study analyzes different components of hydrological processes over North Africa based on remote sensing data such as precipitation (NASA-TMPA, CMORPH and PERSIANN), evaporation (ALEXI and MODIS), and elevation (SRTM) along with ground measurements and model simulations. Here we use the Coupled Routing and Excess STorage (CREST) hydrological model-version 2.0, which was originally developed by NASA-GSFC and the University of Oklahoma [Wang J et al., 2011]. The model is driven by real time TMPA and climatological PET, interpolated to model grids. The flexible simulation and calibration enables the model to provide high-resolution runoff and water depth at each time step. Our study mainly focuses on two major basins such as Medjerda over Tunisia and the Sebou basin of Morocco. Case studies of flood events over North Africa were analyzed based on CREST model simulations with respect to ground measurements. The floods are mainly modulated by rainfall associated with synoptic frontal and tropical plumes and orographic mesoscale systems. Occurrences of peak floods simulated by CREST are comparable with diagnostics such as vertically integrated moisture convergence, stratiform and convective precipitation from ECMWF reanalysis. These were also checked with AMSRE-River Watch and MODIS flood maps and uncertainties among different tools will be discussed. The performance of TMPA and ET is also evaluated based on model-simulated runoff and actual ET in comparison with their respective ground measurements. In general, the model is able to simulate the streamflow and flood fairly well with acceptable statistical efficiency limits for some gauge stations, though it has shown considerable sensitivity to fine resolution elevation data of less than 1km and TMPA of different versions. TMPA rainfall and the simulated runoff have shown large bias and inconsistent trends in some months with respect to gauge measurements; however, better correlation were noted for monthly and yearly aggregation with their respective gauges. Our analysis recommends the following steps to improve the model performance. (i) correcting TMPA against gauge observations (ii) using high resolution ET from satellite based ALEXI instead of using climatological mean (iii) obtaining more accurate DEMs and (iv) perform model calibration with long term gauge data to ensure that the model is independent of grid scaling.