A Methodology for Optimizing Numerical Weather Prediction Models

Optimizing the performance of numerical weather prediction models is a very complicated process due to the numerous parameterization choices provided to the user. In addition, improving the predictability of one model’s variable (e.g., temperature) does not necessarily imply the improvement of another (e.g., precipitation). In this work the Technique of Preference by Similarity to the Ideal Solution (TOPSIS) is suggested as a method to optimize the performance of a numerical weather prediction model. TOPSIS provides the ability of using multiple statistical measures as ranking criteria for multiple forecasting variables. The Weather Research and Forecasting model (WRF) is used here for application of TOPSIS in order to optimize the model’s performance by the combined assessment of temperature and precipitation over Europe. Six ensembles optimize model’s physics performance (i.e., microphysics, planetary boundary layer, cumulus scheme, Long–and Short– wave and Land Surface schemes). The best performing option for each ensemble is selected by using multiple statistical criteria as input for the TOPSIS method, based on the integration of entropy weights. The method adopted here illustrates the importance of an integrated evaluation of weather prediction models’ performance and suggests a pathway for its improvement.

Acknowledgments LIFE CLIMATREE project “A novel approach for accounting & monitoring carbon sequestration of tree crops and their potential as carbon sink areas” (LIFE14 CCM/GR/000635).