Merging satellite rainfall estimates and daily rain gauge observations for improved flood simulation in MelkaKuntire catchment, upper Awash Basin, Ethiopia

Acquiring high-quality rainfall data is very important for studying catchment hydrology. This can be obtained by merging Satellite Rainfall Estimates (SREs) and gauge observations as highlighted several researchers. However, the application of these merged products hydrological modeling has rarely been explored in a complex topographic region like Melkakuntire catchment, Ethiopia. Therefore, this study aims to assess suitability unmerged SREs simulating three flow conditions (daily streamflow, seasonal high flows) using HBV-light model. A total nine which from (raw) (IMERG v06B, TRMM 3B42v7, TAMSAT v3), downscaled-merged (IMERG-D + Gauge, TRMM-D TAMSAT-D Gauge), undownscaled-merged Gauge) were used period 2003–2015. We Random Forest (RF) algorithm merge with gauges, without downscaling spatial resolution selected SREs. For RF algorithm, gridded elevation values independent variables while daily considered dependent variable. Results show that outperform both all when model calibrated each input independently. Similarly, hydrologic forced produces better performances running baseline parameter derived gauges well. Moreover, more improved if than observations. Overall, produce best result closing water budget but further studies could needed validate products, particularly performing IMERG-D Gauge product, other data-scare regions.