Stochastic models are often ®tted to historical data in order to produce stream¯ow scenarios. These scenarios are used as input data for simulation/ optimization models that support operational decisions for water resource systems. The stream¯ow scenarios are sampled from probability distributions conditioned on the available information, such as recent stream¯ow data. In this paper we introduce a procedure for further conditioning the probability distributions by considering the recent measurements of climatic variables, such as sea temperatures, that are used to describe the occurrence of El Niño. We adopt an auto-regressive model and use the``El Niño information'' to re®ne the parameter estimation process for each time step. The corresponding methodology is tested for the monthly energy time series,``in¯owing'' to the power plants of Colombia. This is a linear combination of stream¯ow values for the 18 most important rivers of the country.

No abstract

No abstract