Impact of targeted dropsonde observations on the track forecast for SINLAKU (200813) using Ensemble Kalman Filter

Typhoons are the most severe phenomena due to the intense vortex and precipitations as well as socio-economic damages. However, there are difficulties on the understanding and prediction of typhoons because typhoons are mostly located on the oceanic area of relatively data-sparse regions. During August and September 2008, THORPEX Pacific Asian Regional Campaign (T-PARC) has been performed in the Western North Pacific to investigate the formation, structures, targeting, and extratropical transition of tropical cyclones. In this study, the impact of targeted dropsonde on the track forecast for typhoon SINLAKU (200813) is investigated using ensemble Kalman filter (EnKF).

A serial ensemble square-root filter (EnSRF), one of the deterministic algorithms of EnKF, is used for this study. Series of experiments are configured according to observation data types that are assimilated with EnSRF. The first experiment assimilates conventional observations and referred to as a control experiment. The conventional observations contain the surface, radiosonde, atmospheric motion vectors (AMVs), and aircraft observations. The second experiment includes the typhoon position assimilation as well as the conventional observations. The position observations are based on the best track data from the Regional Specialized Meteorological Center (RSMC) Tokyo. The third experiment assimilates the conventional, typhoon position, and targeted dropsonde observations. The targeted dropsonde observations are from 4 T-PARC aircraft: DOTSTAR ASTRA, DLR Falcon, NRL P-3, and USAF WC-130.

Related to the EnKF data assimilation, following experimental configurations are used: i) 36 ensemble members, ii) covariance localization with 1800 km due to the sampling error, iii) covariance relaxation of 0.8 prior weighting, and iv) multiple physical parameterizations for each ensemble member. The covariance relaxation and the use of multiple physics schemes are considered for maintaining the ensemble spread to avoid the filter divergence. As a forecast model, the Advanced Research WRF (ARW) modeling system version 2.2 is used. Two nesting domains of 45- and 15-km resolution are centered at 25°N, 125°E on the Western North Pacific region.

The ensemble mean fields are verified against radiosonde and dropsonde observations at contemporary levels. By the ensemble filtering, root mean square error (RMSE) of prior mean is reduced. In deterministic forecast initiated by ensemble mean analysis, the track forecast is improved with position assimilation and dropsonde assimilation. With position and dropsonde assimilations, the forecast track is greatly improved for ensemble mean as well as ensemble members. More detailed analyses will be presented in the extended abstract.