A Statistical and Modeling Approach to Improved Land Wind Forecasts Associated with Tropical Cyclones

Developing accurate forecasts of hazardous weather associated with landfalling tropical cyclones (TCs) remains a difficult challenge for tropical meteorologists and operational forecasters. In particular, sustained wind speed forecasts and gust forecasts over land are areas with significant opportunity for improvement. A recent survey of forecasters at various National Weather Service (NWS) Weather Forecasting Offices (WFOs) in the mid-Atlantic and Southeast indicates a strong disconnect between research and operations for these wind speeds. Currently, NWS offices using a tool known as the TCMWindTool that interpolates the National Hurricane Center's (NHC) official forecast advisory wind radii to a 5 km x 5 km grid hourly grid, assuming linear temporal decay and a modified Rankine vortex for interpolation. Forecasters are then tasked with modifying these grids to account for local effects, including the input of land wind reduction and gust factors to be used within the tool. Forecasters have noted a high level of subjectivity and lack of strong scientific principles in the selection of these land reduction and gust factors. Strong discontinuities in forecasts from neighboring WFOs often result due to the lack of scientific guidance in the development of these forecast grids.

In this study, we aim to bridge the gap between research and operations for sustained wind speed and gusts over land. We suggest improvements in the interpolation methods used within the TCMWindTool. These suggestions include alternatives to the linear decay assumption to interpolate six hourly forecast verification times to hourly grids and to the modified Rankine vortex approach currently used. A larger focus for the study is on improvements in the land decay factor and gust factors applied in the TCMWindTool for various locations and tropical cyclone impacting situations. Objective bias calculations are shown for recent past forecasts issued by local WFOs, comparing to observations as well as the Hurricane Research Division H*Wind products. The verification indicates a general overprediction in sustained wind speeds for recent TCs impacting the study region and a need to increase land reduction factors used in the TCMWindTool. An analysis of gust factors for recent storms in the study region suggests gust factors of 20% as a minimum threshold, increasing to up to 60% in cases of continental flow and in cases of weak sustained wind speeds. Statistical analysis of the land reduction and gust factors are being combined with high resolution large eddy simulations of recent landfalling TCs with different thermodynamic and boundary layer conditions to provide research results that operational forecasters can directly benefit from.