Simulation Of Daily Flow Data Using A Stochastic Nonparametric Model (K-Nearest Neighbor)
Szafranski, Bill 1
1 University of Colorado, Boulder
Daily flow data for the Usk River in Wales was simulated using a nonparametric bootstrap procedure that captures autocorrelation called k-nearest neighbor (k-NN) (Lall and Sharma, 1996). The objective of the modeling approach was to obtain a lengthy and robust daily flow record, on the order of 10,000 years. This data was used to validate a reservoir guide curve that was optimized using a genetic algorithm. Each resampled flow sequence was generated for a complete 39 year period, which was identical in length to the original flow data. The statistical comparison between modeled and observed data was facilitated by maintaining the identical period of record for the 500 model simulations. The simulated flow data captured the statistics of the observed data well when compared at the daily, monthly, and annual interval, while providing valuable variability. The daily lag-1 correlations for simulated and observed data is shown in Figure 1, the monthly total flow volumes for simulated and observed data are shown in Figure 2, and the annual flow duration curves are provided in Figure 3.
Lall, U., Sharma, A., 1996. A Nearest Neighbor Bootstrap For Resampling Hydrologic Time Series. Water Resour. Res. 32, 679?693. doi:10.1029/95WR02966.