Comparison of North Atlantic Tropical Cyclones in Variable-Resolution Climate Models
Variable-resolution climate models, which have high-resolution over an area of interest and low-resolution over the rest of the global domain, are less computationally expensive to run than global high-resolution models. This study explores how the representation of tropical cyclones and their associated rainfall in the North Atlantic compares three variable-resolution climate models with high-resolution over different extents of the North Atlantic Ocean and one global low-resolution model.
Variable-resolution model configurations are demonstrated to represent observed North Atlantic tropical cyclone climatology and storm rainfall better than the global low-resolution model and therefore can be used for future research on regional tropical cyclone impacts.
Tropical cyclone (TC) characteristics, such as basin-wide count, landfalling count, intensity, outer size, and precipitation over the eastern United States, are compared between three variable-resolution configurations of the Community Atmosphere Model version 5 that have their high-resolution (~28 km grid spacing) domain over different extents of the North Atlantic basin. The results from the variable-resolution versions are also compared to a global low-resolution configuration (~111 km grid spacing) of the same model. All model simulations are historical simulations over the years 1985-2014 and use prescribed sea surface temperatures. To facilitate the TC analysis, a novel methodology is developed to track TCs and extract TC precipitation using a dynamic estimate of their outer sizes. Results from the models are compared to observations and reanalysis. The variable-resolution configurations are more comparable to observations and reanalysis than the low-resolution configuration for all of the metrics examined. However, the variable-resolution models underestimate TC landfall counts in the eastern US and therefore also demonstrate a low bias for TC precipitation over land.