Nonlinear Forecasting of River Flows in Colombia Based Upon ENSO and Its Associated Economic Value for Hydropower Generation

Abstract

There is ample evidence of hydrologie variability at annual and interannual time scales over the northern regions of tropical South America. Hydroelectric power provides a cheap regional energy source, yielding over seventy percent of Colombia’s national energy annually. The El Peñol scheme on the Nare River is the country’s largest providing roughly 14% of national production. The region is particularly susceptible to droughts during warm phases of ENSO, however stream flow inputs are also subject to a variety of regional and local factors other than ENSO, which may makes forecasting difficult. The identification of a model which permits the reliable incorporation of readily available ocean-atmosphere variables, and variables derived from standard forecasts, to a potentially non-linear prediction of monthly stream flows is crucial to the optimal operation of the reservoir. The MARS (Multiple Adaptive Regression Splines) model is calibrated to provide forecasts of monthly stream flows over the period 1956–86. The applicability of the forecast technique is discussed by reference to comparisons between observed and forecasted flows in a separate model validation series (1987–92). Reliability of the stream flow forecasting methodology is investigated over “forecast horizons” ranging from 3 to 12 months, and the potential economic value of incorporating the methodology into the operation of the national power generating system is illustrated.