Simulation of Extreme Precipitation Events Using an Improved K-Nearest Neighbor Model

A major impact of climate change is on the occurrence of extreme precipitation events leading to disastrous consequences for both riverine and urban flooding. Therefore, mitigation and adaptation to extreme events have received considerable attention from researchers and policy makers around the world. The present research describes the simulation of extreme precipitation events at a rainfall station in New Delhi – the capital city of India using an improved K-Nearest Neighbor model. The intent is to assess the vulnerability of the city to future extreme precipitation events under changing climatic conditions. A distinguishing feature of the K-Nearest Neighbor model used herein is that it can simulate, through perturbation of the historical data, extreme events that are significantly more severe than observed in the historical record. A 800-year simulation was carried out to simulate three extreme events, namely, total precipitation in extreme events, duration of wet spells, and duration of dry spells. Analysis of the simulation results clearly indicated that several extreme events not seen in historical record were simulated by the model. A precipitation event with a total precipitation of 715 mm was simulated by the model compared to a historical precipitation event with a total precipitation of 544 mm. A dry spell of 201 days was simulated by the model compared to a corresponding historical value of 175 days. The model simulated an extreme wet spell of 56 days compared to an extreme wet spell of 47 days in the historical record. The K-NN model employed herein could can be potentially utilized for the design of urban drainage system to cope with extreme precipitation events under changing climatic conditions.