Operational use of real-tima flood forecasting: The case of Lake Como
Ezio Todini1, R. Codeluppi2
Lake Como is a natural reservoir in Northern Italy, which is operated by means of a downstream gate in order to provide water for irrigation and energy production.
Unfortunately the relatively small operational volume of the lake and the reduced downstream outflow when compared to the inflow discharges, has been the cause of several episodes of flooding caused by the extremely high levels reached in the lake.
In order to optimise the Lake Como operations aimed at reducing the risk of flooding, a real-time forecasting system of inflows to the lake, together with an estimate of their uncertainty, has been developed.
The system provides the decision maker with two forecasts: the first one based upon a Nearest Neighbour technique produces daily forecasts and a measure of their uncertainty up to 10 days in advance; the second one, based upon a real-time rainfall- runoff forecasting system produces hourly forecasts and a measure of their uncertainty up to 24 hours in advance.
The forecasts and their uncertainty measures are then used in real time as part of a stochastic optimisation algorithm in order to preserve the expected benefits deriving from the use of waters (irrigation and hydro-power) while reducing, at the same time, the expected damages due to the flooding of Como main square.
The system, which is operational since October 1997, has already proved to be a valuable tool for the Lake Como operational manager who reported four cases in which the suggestions provided by the system allowed him to improve his standard operational practice.