Evaluating Medium-Range Tropical Cyclone Forecasts

A method to evaluate medium-range tropical cyclone (TC) forecasts is presented. Medium range is defined here as forecasts out to 10 days. The method is based on identifying TC tracks as objects and applying matching criteria to forecast and observed tracks. Tracking is done using the GFDL tracker. The method accounts for storms not present at initialization time, and combines intensity and track evaluation in a multi-category contingency table approach. The method is applied to forecasts produced by the Model for Prediction Across Scales (MPAS) and by the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS). Two configurations of MPAS are considered: a uniform 15-km mesh and a variable-resolution mesh transitioning from 60 km globally to 15 km over the eastern Pacific. Heidke skill scores derived from the multi-category framework indicate that the GFS has an overall higher level of skill for TC forecasts, but MPAS has a better ability to discriminate hurricanes from tropical storms. Both MPAS, using a modified Tiedtke cumulus parameterization, and the GFS suffer from biases in genesis at longer time ranges; MPAS produces too many storms whereas the GFS produces too few. These biases are traced to errors in physical parameterization of the coupling of convective precipitation, surface wind and surface moisture fluxes in the ITCZ region of the eastern North Pacific. Reductions of MPAS biases in 2015 due to modifications in the Tiedtke scheme will also be discussed.