<![CDATA[Precipitation variability presents significant challenges for disaster risk reduction and water resource management, particularly in flood and drought-prone regions such as East Kalimantan. This study aims to develop and evaluate two statistical approaches for spatio-temporal precipitation modeling: spatio-temporal kriging (ST-Kriging) and spatio-temporal median polish kriging (ST-MPK). Using monthly precipitation data obtained from seven observation stations provided by BMKG and BPS for the period 2021 to 2023, both models were assessed using performance metrics. ST-Kriging employed a simple sum-metric semivariogram model that combines exponential spatial and Gaussian temporal components. This model achieved an RMSE of 84.05, MAE of 69.95, and MAPE of 52.67%. Meanwhile, ST-MPK model, incorporating robust median polish decomposition and ST-Kriging of residuals, produced a lower MAPE of 44.83% with higher RMSE (122.44) and MAE (91.35). This suggests that while ST-Kriging offers better absolute error performance, ST-MPK provides greater relative accuracy and improved robustness to outliers, critical advantages for modeling precipitation in regions undergoing environmental shifts, where anomalies and extremes are increasingly common. These findings highlight ST-MPK’s potential to produce more reliable forecasts under irregular precipitation conditions, supporting early warning systems and informed water resource planning. Scientifically, this research contributes a robust modeling framework suitable for data-scarce and outlier-prone contexts. Practically, it can aid policymakers in designing adaptive flood mitigation strategies and sustainable water management policies tailored to the evolving climate realities of East Kalimantan.]]>
Copyrights © 2025
