Non-stationary drought patterns in hyper-arid regions: Spatiotemporal and multi-timescale drought analysis.

Droughts rank among the most devastating natural disasters, particularly in arid regions such as Oman. However, traditional drought assessment based on stationarity may not be applicable under climate change. Moreover, most previous studies have been point-based, relying on station observations without capturing the spatial variability of drought. In contrast, this study applies both stationary and non-stationary models on a pixel-by-pixel basis, allowing for detailed spatial and temporal assessment of drought dynamics. To address this, the present study evaluates the trends in both stationary and non-stationary droughts in Oman using the Standardized Precipitation Evapotranspiration Index (SPEI) at 3-, 6-, and 12-month time scales derived from high-resolution ERA5-Land data (1950-2024). We compare stationary and non-stationary log-logistic SPEI parameter models for each pixel, selecting optimal models using the Akaike Information Criterion (AIC), and subsequently, analyze drought duration, frequency, and severity. Results indicate that non-stationarity models, particularly those with location and scale parameters, yield better results than stationary models, particularly for long-term (SPEI-12) droughts. Non-stationarity location shifts best describe short-term droughts (SPEI-3), whereas SPEI-12 requires dynamic mean and variability changes. Spatially, increasing trends of wetness are observed in northern and central Oman, whereas eastern coastal regions exhibit a trend toward increasing dryness. Large timescales are associated with large severity and duration of drought, and SPEI-12 reflects interior long-term deficiencies. The study identifies synchronized drought regimes and demonstrates the value of pixel-wise stationary and non-stationary modeling for drought monitoring in arid regions, offering important insights for climate-resilient water policy and early warning systems.