Precipitation and temperature ensemble forecasts from single-value forecasts

摘要: Abstract. A procedure is presented to construct ensemble forecasts from single-value of precipitation and temperature. This involves dividing the spatial forecast domain total period into a number parts that are treated as separate events. The divided hydrologic sub-basins. time periods, one for each model step. For event archived values corresponding observations used joint distribution observations. conditional given represent probability events may occur forecast. subsequently create members vary in space using "Schaake Shuffle" (Clark et al, 2004). resulting have same space-time patterns historical so relationships between significant effect on hydrological response tend be preserved. Forecast uncertainty time-scale dependent. lead beginning valid an event, depends length area which applies. Although procedure, when applied time-series single value forecasts, preserve some this scale dependency, it not sufficient without additional constraint. To account more fully time-dependent structure uncertainty, "aggregate" periods defined accumulations different "base" periods. generated can ingested by Ensemble Streamflow Prediction system produce streamflow other variables reflect meteorological uncertainty. methodology illustrated application generate temperature American River California. Parameter estimation dependent validation results based operational archives short-range Center (RFC) medium-range mean National Weather Service (NWS) Global System (GFS).