A hybrid approach for exploring the spatiotemporal patterns of precipitation in Sudan: Insights from neural network clustering and Fourier-wavelet transform analysis

Understanding the spatiotemporal variation of precipitation is critical for climate modeling, resource management, and agricultural planning. This study employs self-organizing maps (SOM), discrete Fourier transform (DFT), and wavelet transform (WT) to analyze monthly precipitation data from 80 monitoring stations across Sudan from 2010 to 2019. Initially, cluster analysis identified six distinct precipitation clusters that align with the north-south precipitation gradient, ranging from arid conditions in the northern desert to tropical rainfall in the south. SOM analysis revealed distinct spatial and temporal precipitation patterns, with notable variations between nodes representing climatic zones, such as arid deserts, semi-arid areas, and tropical rainfall zones. This analysis confirmed that central parts of Sudan serve as a transition between northern and southern climatic regimes. The spectral analysis using DFT identified dominant precipitation cycles across measurement points, including annual and semi-annual cycles, detected at all sites with the highest amplitudes. Cycles of 4 months, 3 months, and 2.5 months were also observed, with varying relative amplitudes across regions. Notably, a 60-month (5-year) cycle appeared at specific locations, potentially linked to the Quasi-Biennial Oscillation (QBO) and El Niño Southern Oscillation (ENSO). Complementary wavelet transformation results confirmed the dominance of annual and semi-annual cycles, although shorter cycles such as 2.5 months and 24 months were detected. The outcomes of this research provided a robust framework for understanding rainfall variability, enabling sustainable water resource management.