Quantile Regression and Logistic Regression Combined for Calibration of a Mesoscale Ensemble Prediction System (EPS)

Meteorologists at U.S. Army Dugway Proving Ground, UT use a mesoscale ensemble prediction system (EPS) known as the Ensemble Four-Dimensional Weather System (E-4DWX), which was developed by NCAR. The ensemble’s predictions of selected near-surface variables are dynamically calibrated through algorithms that combine logistic regression and quantile regression, conditioned on the dispersion of the ensemble. Cross-validation is an explicit step in the process. The result is a set of probabilistic predictions that are realistically dispersive and statistically unbiased. The authors will describe the calibration technique and highlight how it is innovative, and will review its performance.