A multi-model approach to operational NWP bias correction

Systematic Numerical Weather Prediction model (NWP) bias errors are well known and it is common practice to calibrate the forecast outputs using observations as the baseline. There are numerous techniques for tuning the forecast outputs in order to automatically suppress bias errors; in our implementation we are automatically combining information from the various NWP model outputs and their bias-corrected derivatives using multiple bias correction methods and automatic assessment of the model performance in order to drive an optimum blended solution.

We will describe the method that we employ as well as a flavor of the products that we provide to operational meteorologists engaged in the quality control of forecast data and general (public) weather forecast consumers of our content. Potential future directions will be discussed, including the provision of forecast uncertainty information generated by the blending technique and the application of GIS mapping techniques to support quality control of the overall system.