Predictions of streamflow a month or a season ahead provide valuable information for water resource managers for subsequent planning. This is particularly the case in catchments with a highly variable flow regime, as the streamflow is difficult to predict, and hence, manage. For the purposes of predicting streamflow, it is desirable to not only provide a value for the prediction, but also quantify the uncertainty associated with the prediction to provide an indication of the potential range of outcomes.
This project presented a range of approaches tested to provide such predictions for a study site in the South East of South Australia. Due to the flat topography and drainage infrastructure, decisions on where and when to divert flow around the landscape can be made, often with the aim to improve environmental outcomes. One of these decision points is along Drain M, which terminates at a wetland of high importance, Lake George, but water can also be diverted out of the drain to be used to support other watercourses and wetlands. With the many competing demands on this water resource, it is desirable to forecast future flows at key locations along Drain M in order to maximise the outcomes achieved from the water available.
Some form of hydrological model is required to provide predictions of upcoming streamflow in Drain M. A hydrological model captures the relationships between initial conditions, climate forcings (typically rainfall and potential evapotranspiration) and streamflow. In prediction mode, a hydrologic model calibrated with historical data is run forward in time, with input data representing forecast climate forcings, to predict the streamflow in an upcoming period. Three major factors control forecasting accuracy: (1) the ability of the hydrologic model to predict streamflow with actual forcings; (2) the accuracy of the initial conditions adopted (e.g., soil moisture and groundwater stores); and (3) the accuracy of the forecasts of the climate inputs. While the third factor is obviously important to forecast accuracy, the objective of this project was to investigate approaches to improve model predictions due to the first two factors, and adopting recently developed approaches to provide the climate forecast inputs.
The models developed through this project are intended to provide another source of information to assist the decision making process surrounding diversions from Drain M. With forecasted volumes for the upcoming month, it is possible that the decision to divert flow can be made earlier in the season, while still ensuring that the downstream requirements of Lake George can be maintained. In turn, this results in improved use of the freshwater resource available in the region, balancing the competing environmental outcomes across the landscape.