Simulations of Tropical Cyclone Interannual Variability Using a High-Resolution Tropical Channel Model

Tropical cyclones (TCs) are among the most destructive of natural disasters, motivating further research to understand the physical drivers of TC variability and to improve TC prediction on seasonal and longer timescales. Here we present a tropical channel model (TCM) configuration tailored to address this need. The TCM is based on the Weather Research and Forecasting model (WRF), and is configured with a model domain that covers the entire tropics and part of the subtropics from 30°S to 50°N, with a TC-permitting horizontal resolution of 27km. The TCM physical parametrizations are carefully tuned to achieve more realistic TC simulations in the global tropics. We performed a large set of retrospective TC simulations from 1982 to 2016 during the northern hemisphere TC season, with a 10-member ensemble for each season, using boundary conditions derived from the 6-hour National Centers for Environmental Prediction – Climate Forecast System (NCEP-CFS) dataset.

TCM shows high skill in simulating observed TC characteristics, including the climatology of TC numbers and tracks in each TC basin, and TC physical structures, the latter of which is more directly relevant to the potential threat that TCs pose. Interannual variations of TCs over the western North Pacific (WNP), eastern North Pacific (ENP), and North Atlantic (NA) are also captured well, with correlation coefficients between simulated and observed accumulated cyclone energy (ACE) of 0.81, 0.59, and 0.60, respectively. To investigate the impact of the lateral boundary condition (LBC) on the model simulation skill, sensitivity experiments were carried out by replacing the interannually varying LBC with a perpetual 1996 LBC, which was characterized by a neutral Atlantic Multidecadal Oscillation (AMO), while keeping everything else unchanged. Preliminary results indicate that the interannual TC variability is relatively insensitive to LBCs, suggesting that SST variability is the primary factor determining seasonal TC predictability. On the basis of these skillful hindcast simulation results, we build a statistical-dynamical hybrid TC forecasting model and assess its prediction skill.