Short-term streamflow forecasts are used operationally at Snowy Hydro to support day-to-day water management decisions, such as setting target levels for reservoirs to optimise hydro-electric generation. Two different methodologies are used to perform the forecasts: A semi-objective, daily-resolution reservoir inflow regression based on forecaster-determined maximum temperature and precipitation; and an objective statistical system which uses direct weather model output to deliver a 3-hourly streamflow forecast. However, the reliability of these forecasts is not well understood, and this can lead to reservations about using the output directly, and confusion when the two methodologies provide different results.

This research will provide an intercomparison of the two forecasting methodologies used at Snowy Hydro and an analysis of their performance under different hydrometeorological regimes. The forecasting methodologies are assumed to be sensitive to errors in the meteorological inputs, but this sensitivity will be quantified by repeating the forecasts after calibrating these inputs. The material impact of quantifying the reliability of streamflow forecast on operational decisions will be discussed.