Analysis and Forecast Impact of Direct Tropical Cyclone Observations with a Coupled Global-limited-area Data Assimilation System

The recent improvement of tropical cyclone (TC) track and intensity forecast improvement has been attributed to increased model resolution, improved data assimilation, and the rapid increase in the number of routinely assimilated observations over oceans. In this study we focus on the assimilation of observations that have a direct impact on tropical cyclones: 10 m scatterometer winds, Tropical Cyclone Vitals (TCVitals) minimum sea level pressure and position estimates, and reconnaissance dropwindsondes.

Each of the observations types that we investigate carries with them valuable information about the cyclone and its environment, but each also has its limitations in moderate-to-low resolution analysis systems. The Ensemble Kalman Filter is an excellent choice for the analysis of TCs because it uses a fully flow-dependent estimate of the error statistics. However, the ensemble is not robust to large observation errors, including the representativeness errors introduced by assimilating direct observations of the TC into a moderate-resolution grid that cannot accurately represent the pressure gradient within a TC.

The assimilation of QuikSCAT wind measurements has its own challenges. Excessive variance of TC location within the ensemble will cause large differences in the observed and background estimates of the low-level flow. Large innovations in wind vectors can have several unintended effects on the analysis including weakening the TC, rejecting the trusted observations, and creating a secondary vortex. In this study, we attempt to reduce the variance in the location of the analyzed TC by assimilating the TCVitals estimated minimum central pressure and DOTSTAR dropwindsondes, along with the QuikSCAT winds to allow for a more accurate analysis of the position and intensity of Typhoon Sinlaku (2008).

While the assimilation of intensity can lead to large analysis impacts, the stability of the system is adversely affected. Clipping the innovations of MSLP and DOTSTAR wind speeds to a prescribed maximum cutoff and assimilating all of the QuikSCAT observations should provide more stability in the forecasts and decreased variance in the background estimate of the intensity and position.