Simulation of Weather Data in Brahmaputra Basin Using K-Nearest Neighbour Model

This paper describes the application of a K-nearest neighbor weather-generating model that allows resampling with perturbation of the observed data to simulate (1) duration of extreme wet spells, and (2) duration of extreme dry spells in the Brahmaputra River Basin. The results of the simulation of extreme events carried out by the K-NN model clearly indicated that the model generated unprecedented extreme events that were not seen in the observed record. Several extreme wet spells were simulated by the KNN model, which are a critical input to flood management models. The model generated several extreme dry spells that are important for evaluation of effective drought management policies for the basin. The analysis conducted herein has the potential for providing valuable aid in developing efficient flood and drought management strategies for the Brahmaputra basin because of the ability of the model to simulate extreme dry and wet spells. It may be concluded that the utility of flood prediction models in estimating the probability of extreme events may be greatly enhanced if their performance is evaluated based on synthetic sequences generated in the present research.