Statistical variability of precipitation and the detection approaches

Precipitation variation driven by climate change adversely affects agriculture and the hydraulic infrastructures. The two ends of precipitation extremes, flood and drought, have become highly frequent, compromising the design of hydraulic structures. This necessitates novel detection approaches in terms of statistical modeling. A robust statistical model that can accurately capture the precipitation variability requires a deep understanding of the precipitation trends and the decadal shifts at various spatiotemporal scales. Identifying the baseline for comparison is equally pertinent. The global precipitation data reveal rising trends in both extremes with inconsistency at the spatial scales (regional disparity in floods and droughts). This requires a model that can accurately decipher the trend at different spatiotemporal scales and model the future values. Nonlinear dynamics govern precipitation, and models like ANN, autoregressive, and wavelet-based models perform well at different scales. The nonlinear wavelet model has been the most robust in unraveling the spatiotemporal variation and its linkages with the predictor variables. AI-based robust prediction model tested by the ground feedback mechanism (though user apps) holds the key to capturing precipitation's statistical variability. This study is significant in highlighting all the detecting approaches associated with the changing patterns in precipitation.