Non-Stationary Frequency Analysis of Annual Rainfall

In northwestern Mexico, where water stress is high, strategies to address periods of water scarcity could fail if availability is evaluated based on mean runoff or precipitation values without considering changes in the behavior of these variables. The objective of this work was to estimate accumulated annual rainfall events in northwestern Mexico based on different return periods and future scenarios. Three non-stationary models were proposed for seven probability functions related to minimum values. The models were validated for series with statistical characteristics that change over time. A total of 221 accumulated annual rainfall series were evaluated, of which 36 (16%) were determined to be non-stationary. For these series, non-stationary models better represented the variability of the data than conventional models. A decrease in annual mean rainfall under 20% is predicted by the year 2044 at stations located on the California peninsula (Sonora and Sinaloa). The results suggest that these areas will be more vulnerable to future droughts.