Disaggregation models of seasonal streamflow forecasts

TitleDisaggregation models of seasonal streamflow forecasts
Publication TypeConference Proceedings
Year of Conference1993
AuthorsGrygier, J., Stedinger J. R., Hongbing Yin, and Freeman G. J.
Conference Name61st Annual Western Snow Conference
Series TitleProceedings of the 61st Annual Western Snow Conference
Date PublishedJune 1993
PublisherWestern Snow Conference
Conference LocationQuebec City, Quebec
KeywordsForecasts, Model, Statistical methods

Pacific Gas and Electric Company (PG&E) has developed a regression-based procedure for disaggregating a seasonal runoff forecast into monthly flow forecasts. These forecasts can be used to estimate the most likely or mean value for each month’s runoff, or to estimate runoff for particular meteorological scenarios, given current conditions. Four basic seasonal-to-monthly disaggregation model families were studied: (1) linear, (2) polynomial, (3) exponential, and (4) logit. The logit model, which ensures that monthly forecasts are always non-zero and less than the forecasted seasonal volume, had larger or comparable R² than the polynomial and exponential models, so the later two were dropped from further analysis. A program was developed which considers the linear and logit models with up to three explanatory variables: the total seasonal forecast, previous month’s flow, and future monthly precipitation. It can select, using standard statistical criteria, a model reasonable for each site and season.