Geostatistical Analysis of Rainfall Spatial Variability and Interpolation Uncertainty in the Semi-Arid Triffa Plain (Northeast Morocco)

Rainfall has shown significant variability during recent decades, particularly in semi-arid regions, where its spatial distribution strongly influences water availability and agricultural activities. In North Africa, rainfall remains irregular and unevenly distributed in space, making accurate spatial estimation essential for climate analysis and water resource management. The objective of this study is to assess rainfall spatial variability in the Triffa Plain (northeastern Morocco) and to identify the most reliable interpolation method under conditions of limited observational data. Annual rainfall data from nine stations covering the period 1992–2021 (30 years) were first subjected to a pre-processing phase, including normality, homogeneity, and break detection tests. The results indicate that all rainfall series follow a normal distribution and show no significant breaks at confidence levels ranging from 90% to 99%, confirming the homogeneity and reliability of the dataset. Two commonly used spatial interpolation techniques were then applied and compared: the deterministic Inverse Distance Weighting (IDW) method and the geostatistical Ordinary Kriging approach. Cross-validation was used to evaluate the performance of both methods. The results reveal that mean annual rainfall over the Triffa Plain ranges from 303 to 353 mm, with an overall average of approximately 326 mm and a low coefficient of variation (CV = 0.04), indicating relatively uniform interannual rainfall at the regional scale. Spatial interpolation highlights marked contrasts across the study area. Lower rainfall values are observed in the northwestern communes of Cap d’Eau, Madagh, and Boughriba, particularly around the CMV 105 and CMV 108 stations, where annual rainfall approaches 310–320 mm. In contrast, higher rainfall amounts are recorded near the Berkane commune, where annual rainfall reaches approximately 336 mm, influenced by the surrounding mountainous massifs. Cross-validation results show that Ordinary Kriging outperforms the IDW method, with a lower root mean square error (RMSE = 0.007 mm) and a mean error close to zero (0.000024 mm), compared to the IDW method (RMSE = 0.31 mm). These findings demonstrate that Ordinary Kriging provides more accurate and less biased rainfall estimates than deterministic methods in semi-arid Mediterranean environments characterized by sparse monitoring networks. The study identifies communes exposed to increasing water stress, particularly Cap d’Eau, Madagh, and Boughriba, and highlights the importance of geostatistical approaches for improving rainfall mapping and supporting climate-related water resource management in data-scarce regions..