Short-term dynamic forecast bias and its impact on vortex scale data assimilation

Significant progress has been made recently on the tropical cyclone (TC) intensity forecast problem through research and development of high-resolution numerical model and data assimilation systems on the vortex scale. One of the major results from the experiments with different forecast systems is the positive impact on the forecasts through 5 days from the assimilation of airborne and other vortex-scale observations using dynamic (ensemble-based) data assimilation methods. This is encouraging and suggests that a synergy between vortex-scale observations and high-resolution forecast models enhances the ability to predict TC intensity. The results also show, however, that that the initial conditions at the vortex scale may not yet be optimal because their impact on the intensity forecast error tends to be smaller at shorter (up to 24 hours) than at longer forecast times, contrary to the expected predictability limits. This property suggests the influence of a systematic forecast error at short time scales which may not be strongly affected by the current improvements to the initial conditions. In this study we investigate the impact of short-term forecast bias on the data assimilation results using the HWRF (Hurricane Weather Research and Forecast model) Ensemble Data Assimilation System (HEDAS) that was developed at the Hurricane Research Division of the Atlantic Oceanographic and Atmospheric Laboratory in NOAA. The observations assimilated in this system at 1-h time intervals are airborne Doppler radial wind speed, Stepped-Frequency Microwave Radiometer surface wind speed, and flight-level and dropwindsonde measurements that are transmitted in real time from all research and operational aircraft . The diagnostic analysis of HEDAS results for 10 cases of hurricane-intensity storms shows that, during the assimilation, the wind velocity increments by the observations are strongly negatively correlated with the change in wind speed in the 1-h forecasts, implying that the improvement to the wind speed analysis due to the assimilation of the observations is effectively lost in the short-term forecast. The results suggest that this problem may be caused by a dynamical imbalance between the primary and secondary circulations in the vortex that leads to an adjustment during time scales longer that the data assimilation window.