Remote sensing reveals spatially uneven vegetation trends and seasonal rainfall associations in Somaliland from 2000 to 2022

Abstract Dryland vegetation in Somaliland is highly sensitive to rainfall variability, warming, and recurrent drought, yet Somaliland-wide evidence on MODIS-era vegetation–climate relationships remain limited. This study assessed vegetation dynamics and climatic variability across Somaliland during 2000–2022 using MODIS NDVI, CHIRPS rainfall, TerraClimate temperature, Mann–Kendall trend testing, Sen’s slope estimation, autocorrelation and modified Mann–Kendall sensitivity checks, correlation analysis, multiple linear regression, seasonal lag-response testing, pixel-wise NDVI trend mapping, and exploratory Random Forest modelling. Seasonal analyses were conducted separately for the MAM/Gu and SON/Deyr analytical windows, with 23 annual observations per season. NDVI showed substantial interannual variability and weak positive tendencies during both MAM/Gu (Kendall’s τ = 0.091, p = 0.561; Sen’s slope = 0.000237 NDVI units year⁻¹) and SON/Deyr (τ = 0.281, p = 0.064; Sen’s slope = 0.000778 NDVI units year⁻¹), but neither trend was statistically significant at p < 0.05. Autocorrelation and modified trend sensitivity results supported a conservative interpretation of national seasonal NDVI trends as weak tendencies rather than robust country-wide greening. Rainfall varied strongly between years but showed no significant seasonal trend, whereas temperature increased significantly during SON/Deyr (τ = 0.573, p < 0.001; Sen’s slope = 0.038 °C year⁻¹). Rainfall showed the strongest and most consistent association with NDVI among the climatic variables evaluated, with significant positive correlations during MAM/Gu (Pearson r = 0.713, p < 0.001) and SON/Deyr ( r = 0.716, p < 0.001). Seasonal lag analysis showed that same-season rainfall had the clearest relationship with same-season NDVI. Random Forest modelling, evaluated using leave-one-year-out cross-validation, provided exploratory support for the higher importance of rainfall relative to temperature but showed only modest predictive performance. Pixel-wise annual NDVI trend mapping indicated significant greening over 27.07% of the mapped MODIS analytical area, significant browning over 1.15%, and non-significant or stable conditions over 71.77%. Overall, the study provides a Somaliland-specific remote-sensing framework showing that vegetation greenness is strongly associated with seasonal rainfall variability, while spatial NDVI change remains heterogeneous. The findings may support drought monitoring, rangeland assessment, and future climate-risk studies in data-scarce dryland environments.