In this paper kernel estimates of the joint and conditional probabilit
y density functions are used to generate synthetic streamflow sequence
s. Streamflow is assumed to be a Markov process with time dependence c
haracterized by a multivariate probability density function. Kernel me
thods are used to estimate this multivariate density function. Simulat
ion proceeds by sequentially resampling from the conditional density f
unction derived from the kernel estimate of the underlying multivariat
e probability density function. This is a nonparametric method for the
synthesis of streamflow that is data-driven and avoids prior assumpti
ons as to the form of dependence (e.g., linear or nonlinear) and the f
orm of the probability density functions (e.g., Gaussian). We show, us
ing synthetic examples with known underlying models, that the nonparam
etric method presented is more flexible than the conventional models u
sed in stochastic hydrology and is capable of reproducing both linear
and nonlinear dependence. The effectiveness of this model is illustrat
ed through its application to simulation of monthly streamflow from th
e Beaver River in Utah.