Non-stationarity of extreme rainfalls and its impact assessment in the northeastern part of India

ABSTRACT This study investigates the variability in extreme rainfall in northeast India using stationary and non-stationary frequency analyses. The study identified non-stationarity in the annual maximum daily rainfall time series in northeast India, and it was shown that the non-stationary models exhibited more accurate results compared to modelling the series on the assumption of stationarity. Genetic algorithm (GA) is employed as the optimization algorithm for parameter estimation in the method of maximum likelihood estimation of parameters, hereby referred to as maximum likelihood estimate (MLE)-GA. The results demonstrate better non-stationary parameter fitting using the MLE-GA approach for generalized extreme value (GEV) and Pearson Type III (PE3) distributions. The results obtained found that all the 12 rainfall stations considered in the region show significant non-stationarity with values ranging from a negative impact of −31.97% to a positive impact of 38.6% for a return period of 100 years. Overall, the research findings indicate non-stationarity in the frequency analysis of extreme rainfall series to have a significant impact on the northeastern region as a whole, and non-stationary modelling is more beneficial and reliable for any water management studies under climate change in the region.