A historical database for use in rainfall-runoff modeling of the Okavango River Basin in Southwest Africa is presented. The work has relevance for similar data-sparse regions. The parameters of main concern are rainfall and catchment water balance which are key variables for subsequent studies of the hydrological impacts of development and climate change. Rainfall estimates are based on a combination of in-situ gauges and satellite sources. Rain gauge measurements are most extensive from 1955 to 1972, after which they are drastically reduced due to the Angolan civil war. The sensitivity of the rainfall fields to spatial interpolation techniques and the density of gauges was evaluated. Satellite based rainfall estimates for the basin are developed for the period from 1991 onwards, based on the Tropical Rainfall Measuring Mission (TRMM) and Special Sensor Microwave Imager (SSM/I) data sets. The consistency between the gauges and satellite estimates was considered. A methodology was developed to allow calibration of the rainfall-runoff hydrological model against rain gauge data from 1960-1972, with the prerequisite that the model should be driven by satellite derived rainfall products for the 1990s onwards. With the rain gauge data, addition of a single rainfall station (Longa) in regions where stations earlier were lacking was more important than the chosen interpolation method. Comparison of satellite and gauge rainfall outside the basin indicated that the satellite overestimates rainfall by 20%. A non-linear correction was derived used by fitting the rainfall frequency characteristics to those of the historical rainfall data. This satellite rainfall dataset was found satisfactory when using the Pitman rainfall-runoff model (Hughes et al., this issue). Intensive monitoring in the region is recommended to increase accuracy of the comprehensive satellite rainfall estimate calibration procedure