Modeling Rainfall Intensity-Duration-Frequency (IDF) and Establishing Climate Change Existence in Abakaliki-Nigeria Using a Non-Stationary Approach

Rainfall Intensity Duration Frequency (IDF) models are essential tools for obtaining the rainfall intensity necessary for designing hydraulic structures. Underestimation of rainfall intensities from stationary rainfall models can lead to inadequate designs. Stationary models fail to account for the changing variations in climatic parameters. This study aims to develop non- stationary IDF curves for Abakaliki, Nigeria, using a 31- year rainfall record (1992-2022) obtained from the Nigerian Meteorological Agency (NIMET). The 24- hour rainfall data from NIMET were downscaled to shorter durations using the Indian Meteorological Department (IMD) formula. Non- stationarity in the rainfall data was identified using the Mann-Kendall test, a non-parametric method. Abrupt changes in the rainfall data were detected using two change point tests: Distribution- free CUSUM and Sequential Mann- Kendall. The General Extreme Value distribution was employed to develop the non-stationary IDF rainfall models. Three distinct General Extreme Value (GEV) distribution models were assessed to determine the best-fitting non-stationary model according to the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). For the trend change, results from the Mann- Kendall test provided sufficient evidence that rainfall in Abakaliki shows an increasing trend (p-value = 0.0059). The findings suggest that the statistical parameters are not constant over time and that non-stationary approaches are required for IDF modelling in Abakaliki. The change point test identified 2010 and 2012 as probable points of change in the rainfall trend. Out of the three models assessed, GEVt- III displayed the best performance across most duration intervals (10-1440 minutes) indicated by its lowest AIC values ranging from 218.897 to 328.818. For the 5- minute duration, GEVt- I was the best non-stationary model with the lowest AIC of 207.946. A generalised non-stationary IDF model was created, demonstrating exceptional predictive ability (R² = 0.996, MSE = 37.00). These results emphasise the necessity of incorporating non-stationary methods in infrastructure design in Abakaliki, as conventional stationary approaches may significantly underestimate rainfall intensities in the era of climate change.