The Experimental HWRF-HEDAS system: Using satellite and airborne observations from GRIP/PREDICT/IFEX to evaluate the model and to assess the impact of data assimilation

Hurricane track forecast skill of the global models has improved significantly over the last two decades, largely due to data assimilation of satellite observations outside precipitation. However, forecasting hurricane intensity changes remains a challenge for the large-scale models due to their parameterized representation of convection. Recently, some progress has been shown on the intensity forecast by regional models due to increased resolution and advancements in parameterizations of small scale processes (e.g. parameterizations of the microphysical processes, the boundary layer processes and the ocean-atmosphere fluxes), but it has been limited. The limitations result from large uncertainties in initial conditions and remaining deficiencies in the sub-scale parameterizations.

To facilitate the evaluation of the forecast models, we are developing the JPL Tropical Cyclone Information System (TCIS- http://topicalcyclone.jpl.nasa.gov) containing satellite and airborne observations. We used the data compiled for cases of 2010 season to evaluate performance of HWRF (Hurricane Weather Research and Forecasting) model , focusing on the evolution of hurricane Earl. Comparison to microwave satellite observations revealed that the model forecasts show a great deal of realism in capturing the essential asymmetry of precipitation. It was found, however, that the skill increases with forecast range over first 1-2 days of forecast. In particular, we found that it takes18h to even 24h for the model to develop realistic structures. By that time the forecasted storm position has likely drifted from the observed and the modeled storm is developing in conditions different from that of the observed.

The long spin-up time might be the result of sub-optimal initial conditions that do not contain the observed precipitation structures. If so, this strongly implies that during the model initialization we should make every effort to assimilate airborne and satellite information inside the precipitating hurricane core in order to incorporate the precipitation-related thermodynamics and the important vortex asymmetries.

In this study we evaluate the HWRF forecasts that were initialized with improved initial conditions by application of Hurricane EnKF Data Assimilation System (HEDAS) that was developed at NOAA's Hurricane Research Division. Our methods include 2D maps as well as statistical tools such as CFADs, joint PDFs, time aggregates and Principal Component analysis.