Abstract
Short-term hydrological forecasts provide valuable information to improve the management of a wide range of human-related activities such as early flood warnings, hydropower generation scheduling and water supply, among the others. Despite the increasing accuracy and reliability of streamflow forecasts there is still the need of improving the quality of such predictions, especially considering mountain areas where the complex topography and the extreme spatial variability of natural phenomena introduce significant uncertainty either in the weather forecasts and in the hydrological response. In particular, hydrological forecasting aimed to support hydropower production needs to be continuous in time and able to predict different aspect of the hydrological cycle, ranging from low flows to flood peaks.
In this work, the set-up of a streamflow forecasting system aimed to support hydropower production scheduling is presented as well as an analysis about the coupling between hydrological models and satellite- derived snow cover area maps in order to reduce uncertainty affecting snow related variable and, in turn, improve streamflow estimates. Indeed, snow cover area maps are available at suitable time and spatial scales and therefore represent a valuable source of information to be used for conditioning hydrological models. Firstly, a validation of satellite-derived snow cover products over South Tyrol region is carried out, formulating a criterion to identify areas where the snow cover maps are unreliable because of the combined effect of the forest canopy and weak solar illumination. After proving the general reliability of satellite-derived snow cover maps, a methodology was then proposed using such snow cover information to reduce the predicted streamflow parametric uncertainty conditioning the model with both streamflow and snow cover observations. Such methodology allowed a reduction of the predicted discharge uncertainty in multiple sites across the studied catchment, also providing a spatial validation of the novel approach proposed.
Such findings are used to implement a potentially real-time streamflow forecasting system, where a Data Assimilation tool is added in order to reduce the uncertainty related to the initial condition, i.e. the estimation of the catchment state before running the forecast. This approach added significant benefits to the streamflow accuracy as proven by a hindcast experiment carried out over the spring-summer 2019.
Finally, along with the investigation on hydrological modelling an optimization tool was implemented combining streamflow and energy prices information, in order to optimize the hydropower production scheduling over a period of 24 hours, i.e. to support energy trading on the day ahead electricity market. The impact of different econometric models on optimization results was evaluated, showing that an econometric model that account for the renewable energy production leads to better price forecasts and the optimization driven by such predictions is more reliable than the one carried out with price predictions formulated neglecting the role played by renewable energy in the price realization.
In conclusion, this work provides the basis for further research ranging from the assimilation of snow- related variables (e.g. snow depth records) to the evaluation of how the different sources of hydrological uncertainty affect the optimization of hydropower production.