Improving predictability of landfalling hurricanes through assimilation of surface mesonet observations using ensemble kalman filter methods

Accurate forecasts of the intensity and structure of a hurricane at landfall can save lives and mitigate social impacts. However, among recent efforts in hurricane forecast improvements, few studies have focused on landfalling hurricanes. This reflects the complexities of predicting hurricane landfalls and uncertainties in representing the near surface atmospheric conditions in numerical weather prediction models.

In this study, the impact of assimilation of surface mesonet observations on the predictability of landfalling hurricanes is examined using a mesoscale community Weather Research and Forecasting (WRF) model. Specifically, an ensemble Kalman filter method developed by NCAR Data Assimilation Research Testbed (DART) is employed to assimilate hourly surface mesonet observations in conjunction with other available conventional and satellite observations.

Numerical results show positive impact of assimilation of surface mesonet observations on the forecasts of landfalling time and location of Hurricane Katrina (2005). Various data assimilation configurations (e.g., localization, inflation and ensemble size etc.) are examined to evaluate their effectiveness on surface data assimilation. The complexity of near surface atmospheric processes, the influence of data assimilation on the structure of atmospheric boundary layer (ABL), and the interaction between ABL and Hurricane Katrina during its landfall are further discussed.