Step Ahead State Water Requirement Forecasting Using Streamflow Observations: Po River Watershed, Italy.

摘要

A simple method of forecasting the future monthly degree of satisfaction of an assigned water demand for irrigation is presented. The supply series /Q(t)/ is transformed into a state series /S(t)/, each state being defined by the ordinal number j=l,J of the interval containing the value of supply, the intervals being defined by assigned J-l water demand levels. Small collections of c=1,2,3 successive states are considered to form representational seasonal objects Ok(t)=(S(t-c+l); ...; S(T) which define local terminal month k shapes within the state series. For every month k, the frequency of occurence of each object in the historical series is used to select dominant objects, which provide the decision criterium for the future state value in time t, when the c-l states S(t-c+l),...,S(t-l) are known. Monthly data from the Po river watershed were used to simulate step ahead state forecasting with various values of J and c. The comparison of actual and forecasted states shows the existence of an optimal value of c for J=2,3,...,5, which maximizes the total number of right state forecasts. The application of the method, for both constant and seasonal water demand levels, to two streamflow series with different characteristics of persistence, shows the convenience of considering objects with c=2 consecutive states. The increase in the number of right state forecasts from the c=l to the c=2 model is about 5% for constant demand and 10%-40% for seasonal demand. (ERA citation 14:004254)