Development of a national 7-day ensemble streamflow forecasting service for Australia

[摘要] Reliable streamflow forecasts with associated uncertainty estimates areessential to manage and make better use of Australia's scarce surface waterresources. Here we present the development of an operational 7 d ensemblestreamflow forecasting service for Australia to meet the growing needs ofusers, primarily water and river managers, for probabilistic forecasts tosupport their decision making. We test the modelling methodology for 100catchments to learn the characteristics of different rainfall forecasts fromNumerical Weather Prediction (NWP) models, the effect of statisticalprocessing on streamflow forecasts, the optimal ensemble size, andparameters of a bootstrapping technique for calculating forecast skill. Aconceptual rainfall–runoff model, GR4H (hourly), and lag and route channelrouting model that are in-built in the Short-term Water InformationForecasting Tools (SWIFT) hydrologic modelling package are used to simulatestreamflow from input rainfall and potential evaporation. The statisticalcatchment hydrologic pre-processor (CHyPP) is used for calibrating rainfallforecasts, and the error reduction and representation in stages (ERRIS)model is used to reduce hydrological errors and quantify hydrologicaluncertainty. Calibrating raw forecast rainfall with CHyPP is an efficientmethod to significantly reduce bias and improve reliability for up to 7lead days. We demonstrate that ERRIS significantly improves forecast skillup to 7 lead days. Forecast skills are highest in temperate perenniallyflowing rivers, while it is lowest in intermittently flowing rivers. Asensitivity analysis for optimising the number of streamflow ensemblemembers for the operational service shows that more than 200 members areneeded to represent the forecast uncertainty. We show that the bootstrappingblock size is sensitive to the forecast skill calculation. A bootstrappingblock size of 1 month is recommended to capture maximum possibleuncertainty. We present benchmark criteria for accepting forecast locationsfor the public service. Based on the criteria, 209 forecast locations out ofa possible 283 are selected in different hydro-climatic regions acrossAustralia for the public service. The service, which has been operationalsince 2019, provides daily updates of graphical and tabular products ofensemble streamflow forecasts along with performance information, for up to7 lead days.

[学科分类] 妇产科学