It is Tails

The standard evaluation criteria for temperature forecasting accuracy is overall error minimization measured by standard error metrics. Normal metrics in use are mean absolute error (MAE) and root mean squared error (RMSE). However, most economic value for temperature end-users in energy lies in the tails of the historical distribution. In this presentation, we present methodology concepts and statistical evaluation criteria for improving short term temperature forecast in the tails of the distribution. A case study will be presented to show the application of the concepts.