Performance comparison of Kriging models used for estimation of rainfall variability in the northern coast of Tanzania

Accurate rainfall estimation underpins effective disaster preparedness and sustainable socioeconomic planning. The performance of Ordinary Kriging (OK), Universal Kriging (UK), and Simple Kriging (SK) in estimation of rainfall variability were assessed across eight stations in the northern coast of Tanzania for the period from 1960 to 2020. Using ArcGIS 10.3, each method was evaluated and cross-validated using six performance metrics, namely; Mean Error (ME), Mean Square Error (MSE), Root Mean Square Error (RMSE), Root Mean Standardized Square Error (RMSSE), Average Standardized Error (ASE), and the Coefficient of Determination (R²). Performance comparison results indicate OK to have achieved the lowest bias, best overall accuracy, and the highest R². UK underperformed, while SK performed comparably, but slightly lower than OK. These results demonstrate that OK provides the most reliable and precise spatial rainfall predictions. While performance metrics have indicated OK to be more accurate, the standard deviation also shows that OK delivers more consistent performance across spatial scales. It is the best linear unbiased prediction which minimizes estimation variance, although suffer from non-linear spatial patterns. This study recommends that Kriging models be adopted and integrated into TMA’s forecasting and early warning system to enhance spatial precision in rainfall estimation.