A NEAREST-NEIGHBOR BOOTSTRAP FOR RESAMPLING HYDROLOGIC TIME-SERIES

A nonparametric method for resampling scalar or vector-valued time ser ies is introduced. Multivariate nearest neighbor probability density e stimation provides the basis for the resampling scheme developed. The motivation for this work comes from a desire to preserve the dependenc e structure of the time series while bootstrapping (resampling it with replacement). The method is data driven and is preferred where the in vestigator is uncomfortable with prior assumptions as to the form (e.g ., linear or nonlinear) of dependence and the form of the probability density function (e.g., Gaussian). Such prior assumptions are often ma de in an ad hoc manner for analyzing hydrologic data. Connections of t he nearest neighbor bootstrap to Markov processes as well as its utili ty in a general Monte Carlo setting are discussed. Applications to res ampling monthly streamflow and some synthetic data are presented. The method is shown to be effective with time series generated by linear a nd nonlinear autoregressive models. The utility of the method for resa mpling monthly streamflow sequences with asymmetric and bimodal margin al probability densities is also demonstrated.