EVALUATION OF GREAT-LAKES NET BASIN SUPPLY FORECASTS

Evaluation of the Great Lakes Environmental Research Laboratory's (GLE RL's) physically-based monthly net basin supply forecast method reveal s component errors and the effects of model improvements for use on th e Laurentian Great Lakes. While designed for probabilistic outlooks, i t is assessed for giving deterministic outlooks along with other net b asin supply forecast methods of the U.S. Army Corps of Engineers and E nvironment Canada, and with a stochastic approach commissioned by the Corps. The methods are compared to a simple climatological forecast an d to actual time series of net basin supplies. Actual net basin suppli es are currently determined by estimating all components directly, ins tead of as water-balance residuals. This is judged more accurate and a ppropriate for both forecasting and simulation. GLERL's physically-bas ed method forecasts component supplies while the other methods are bas ed on residual supplies. These other methods should be rederived to be based on component supplies. For each of these other methods, differe nces between their outlooks and residual supplies are used as error es timates for the rederived methods and component supplies. The evaluati ons are made over a recent period of record high levels followed by a record drought. Net basin supply outlooks are better than climatology, and GLERL's physically-based method performs best with regard to eith er component or residual net basin supplies. Until advances are made i n long-range climate outlooks, deterministic supply outlooks cannot be improved significantly.