Tropical cyclones in the North American Regional Reanalysis: An assessment of spatial biases in location, intensity, and structure

Abstract

AbstractLong‐term reanalysis data sets, such as the North American Regional Reanalysis (NARR) and Climate Forecast System Reanalysis (CFSR), are data‐rich resources for weather and climate research. However, investigations into tropical cyclone (TC) structures are lacking. This study examines position, intensity, and structure of U.S. landfalling TCs in these data sets during 1998–2012. TC positions are determined using three dynamic and thermodynamic parameters. In NARR, positions are problematic near the domain boundary but show statistically significant improvements compared with CFSR in regions over the U.S. and near its coastlines. TC intensity is universally underestimated in both data sets, yet the pressure‐wind relationship is reasonably maintained. NARR and CFSR intensities are highly correlated (r = 0.93), suggesting that these data sets are limited in a similar fashion in their generation of intensity and that advanced data assimilation techniques are unable to overcome the shortcomings of low resolutions. To investigate TC structure, a 10‐storm composite is constructed using the most intense model TCs at peak intensity. Both models develop robust warm cores (T′ ~ 6–7 K), but the radius of maximum winds (>100 km) is too broad compared with storms in nature. Secondary circulation features in NARR are more realistic, particularly the shallow inflow layer, vertical motion in the inner core, and confined outflow in the upper troposphere. Results indicate that, over the U.S. and near its coastline, NARR positions and structure are competitive with or slightly improved upon CFSR. Away from the domain boundary, NARR is a suitable resource for evaluating TC positions and certain aspects of the secondary circulation.