Effect of rainfall variability and gauge representativeness on satellite rainfall accuracy in a small upland watershed in southern Ethiopia

Abstract

The actual accuracy of satellite rainfall products is often unknown due to the limitation of raingauge networks. We evaluated the effect of gauge representativeness error on evaluation of rainfall estimates from the CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data) rainfall product. The reference data were collected using an experimental raingauge network within a small watershed of 1690 ha, which is comparable to the CHIRPS resolution. The study applied a total bias approach, decomposed into hit, missed and false biases, and an error-variance separation method to evaluate gauge representativeness error at the scale of CHIRPS pixel size, as well as modeled the spatial correlation field of daily rainfall with a three-parametric exponential model. The results indicate that the gauge representativeness error is still too large to ignore in evaluating satellite rainfall. However, it is significantly affected by sample size and caution should be exercised when the rainfall data has a small sample size.