ABSTRACT
The K-nearest neighbor resampling (KNNR) method was developed by U. Lall and A. Sharma for generating hydrologic time series. This chapter suggests several approaches for statistical temporal downscaling of precipitation time series. Among others, Taesam Lee and C. Jeong presents a theoretical framework for a nonparametric temporal downscaling model that considers the diurnal cycle and the specific details of key hourly statistics using KNNR and the Genetic Algorithm mixing process. The background of this approach lies in a K-nearest neighbor density estimator that employs the Euclidean or Mahalanobis distance to the Kth nearest data point and its volume containing K-data points. The temporal downscaling of the precipitation output of climate models is critical in hydrological assessment of climate change, especially for small or urban watersheds. Nonparametric-based modeling of temporal downscaling successfully produces hourly precipitation conditioned on daily precipitation values.
