A Novel k -Nearest Neighbors Approach for Forecasting Sub-Seasonal Precipitation at Weather Observing Stations

This study introduces a novel k -nearest neighbors ( k NN) method of forecasting precipitation at weather-observing stations. The method identifies numerous monthly temporal patterns to produce precipitation forecasts for a specific month. Compared to climatological forecasts, which average the observed precipitation over the prior thirty years, and other existing contemporary iterations of k NN, the proposed novel k NN method produces more accurate forecasts on a consistent basis. Specifically, the novel k NN method produces improved root mean square errors (RMSE), mean relative errors, and Nash–Sutcliffe coefficients when compared to climatological and other k NN forecasts at five weather stations in Oklahoma. Rather than looking at the daily data for feature vectors, this novel k NN method takes so many days and evenly groups them, using the resulting average as one feature each. All methods tested were lacking in the ability to forecast wet extremes; however, the novel k NN method produced more frequent higher precipitation forecasts compared to climatology and the two other k NN methods tested.