Potential of rainfall data hybridization in a data-scarce region

Abstract Many impact assessment studies have been using reanalysis precipitation data as an alternative to observed rainfall data even where few rainfall records are available. For that reason, this study investigated fit-to-observations and fit-for-purpose of hybrid rainfall data using four strategically located rainfall stations in the Wami-Ruvu basin, Tanzania. Hybrid rainfall data were generated by randomly filling 15% to 70% hypothetical missing data in the observed records at the rainfall stations with reanalysis precipitation data from nearby grids. Fit-to-observations was used to evaluate the performance of the whole time series of hybrid rainfall data in mimicking that of observed rainfall data at the stations. Fit-for-purpose was used to evaluate the performance of trend and seasonal components of the whole time series of hybrid rainfall data in mimicking those of the whole time series of observed rainfall data at the stations. The findings showed that hybrid rainfall data are superior to reanalysis precipitation data in mimicking observed rainfall data. The findings also revealed that in most cases, trend and seasonal components have greater performances than the whole time series at high degrees of hybridization. Therefore, hybridization of rainfall data is highly encouraged especially in data-scarce regions.