Discrete k-nearest neighbor resampling for simulating multisite precipitation occurrence and model adaption to climate change

Abstract

Stochastic weather simulation models are commonly employed in water resources management, agricultural applications, forest management, transportation management, and recreational activities. Stochastic simulation of multisite precipitation occurrence is a challenge because of its intermittent characteristics as well as spatial and temporal cross-correlation. This study proposes a novel simulation method for multisite precipitation occurrence employing a nonparametric technique, the discrete version of the k-nearest neighbor resampling (KNNR), and couples it with a genetic algorithm (GA). Its modification for the study of climatic change adaptation is also tested. The datasets simulated from both the discrete KNNR (DKNNR) model and an existing traditional model were evaluated using a number of statistics, such as occurrence and transition probabilities, as well as temporal and spatial cross-correlations. Results showed that the proposed DKNNR model with GA-simulated multisite precipitation occurrence preserved the lagged cross-correlation between sites, while the existing conventional model was not able to reproduce lagged cross-correlation between stations, so long stochastic simulation was required. Also, the GA mixing process provided a number of new patterns that were different from observations, which was not feasible with the sole DKNNR model. When climate change was considered, the model performed satisfactorily, but further improvement is required to more accurately simulate specific variations of the occurrence probability.