Modeling Rainfall Variability in Somalia using Spatial Econometrics: Evidence from the Spatial Autoregressive Model

This study employs spatial econometric techniques, specifically the Spatial Autoregressive Model (SAR), to analyze rainfall variability in Somalia over a 120-year period (1901-2021) using high-resolution gridded precipitation data from the Climate Research Unit Time-Series (CRU TS) dataset. The analysis reveals significant temporal and spatial patterns, with annual rainfall ranging from 150mm to 400mm and no clear linear trend but marked interannual variability. Annual rainfall in Somalia exhibits pronounced interannual variability with no statistically significant monotonic trend over the full 1901–2021 period (Mann–Kendall p = 0.085). However, sub-period analysis reveals a significant decline during 1901–1950 and a significant increase during 1951–2021, highlighting non-stationary rainfall behavior. Decadal analysis highlights extreme fluctuations, including the lowest mean rainfall of 213.22mm in the 1940s and the highest of 315.51mm in the 1960s, followed by a trend toward more stable patterns in recent decades, with variance decreasing to 193.80 in the 2000s and 411.37 in the 2020s. Spatial analysis demonstrates strong spatial autocorrelation, with Moran's I values ranging from 0.29 to 0.35 across different spatial weight specifications, indicating significant clustering of rainfall patterns. The SAR model outperformed traditional OLS regression, with distance-based spatial weights achieving the highest explanatory power (R² = 0.45), followed by K-Nearest Neighbors (R² = 0.42) and contiguity-based weights (R² = 0.40). These findings underscore the importance of spatial dependencies in understanding rainfall variability and highlight the utility of spatial econometric methods for climatological studies. The results have critical implications for climate adaptation, agricultural planning, and water resource management in Somalia, emphasizing the need for regional coordination and flexible strategies to address the interconnected nature of precipitation patterns. This study advances the methodological framework for analyzing rainfall variability and provides actionable insights for policymakers and practitioners working to enhance climate resilience in vulnerable regions.

Actions (login required)

View Item