By James R. Prairie and Balaji Rajagopalan, Journal of Hydrology , 344 (1-2):, 43-54, 2007.
Abstract: In the western United States many rivers experience high salinity resulting from natural and anthropogenic sources. This impacts the water quality and hence, is closely monitored. The salinity is closely linked with streamflow quantity in that, a higher flow brings with it more salt but also provides substantial dilution to reduce the salt concentration and vice-versa during low flow regimes. Decision makers typically plan strategies for salinity mitigation and evaluate impacts of water management policy options on salinity in the basin using decision support models. These models require statistically consistent basin wide scenarios of streamflow and salinity. Recognizing this need, we develop a basin wide stochastic generator of salinity that is consistent with the flows. The framework integrates a nonparametric space–time disaggregation model, applied to both flow and salt, and a nonparametric regression model, which simulates natural salt magnitude for a given natural flow. The nonparametric techniques do not require prior assumptions about the underlying functional relationship unlike traditional parametric methods, are data driven, and robust at capturing any functional form present in the data. They are also easily portable across sites. Within this integrated framework we propose two representative approaches, from several possible variations, for generating salt scenarios, each capturing different aspects of the flow and salt statistics. We demonstrate the performance of these approaches by applying them to flow and salt data from four gauges in the Upper Colorado River Basin. The spatial and temporal statistics of salt and their relationship with flows are well captured. This framework has utility in a variety of applications such as, short term salinity forecast conditioned on streamflow forecast, long term salinity scenarios conditioned on climate change projections, and also in reconstructing paleo salinity from paleo streamflows.