Exploration of Predictors for Statistical-Dynamical Subseasonal Prediction of Western North-Pacific Tropical Cyclone Activity in Earth System Models

Abstract

Subseasonal prediction of tropical cyclones (TCs) has many potential applications but remains a challenge due to biases in both model-based large-scale conditions and TCs in coupled global models. Model forecasts of environmental parameters can be linked to TC activity and then be used to extend the horizon of useful skill through statistical-dynamical models. The aim of this work is to assess the utility of incorporating model forecasted environmental fields in a statistical model compared with skill coming from model forecasted Madden Julian Oscillation (MJO) state in predicting TC activity over the Western North Pacific (WNP). In this study, we evaluate the European Center for Medium-Range Weather Forecasts (ECMWF) from the Subseasonal-to-Seasonal (S2S) database and the Navy Earth System Prediction Capability (ESPC) as part of the Subseasonal Experiment on their ability to predict WNP TC activity using environmental fields. To isolate the environmental signals associated with subseasonal variability of TC activity, we examine events of anomalous accumulated cyclone energy, genesis, and TC days. These events are used to create composites of ERA5 reanalysis fields of environmental conditions related to WNP TC activity, which are used to select predictors for statistical dynamical hybrid models. The ECMWF statistical-dynamical scheme exhibits an improvement in skill by using a tailored outgoing longwave radiation (OLR) predictor compared with the MJO predictors. The Navy-ESPC generally performs worse than the ECMWF and has OLR biases that impede it from improving skill in the statistical-dynamical schemes. Using shear and humidity fields as predictors did not improve predictability in either model.