| Description |
Tropical cyclone (TC) prediction, especially the forecast of TC intensity changes, is a challenging problem in the current operational and research community. Part of the difficulty in improving the current generation of operational forecast models is the uncertainty in initial conditions due to the lack of high-resolution TC inner-core observations. This dissertation focuses on incorporating new high-resolution TC inner-core observations, including both satellite and TC field campaign data, into the NCEP operational Hurricane Weather Research and Forecasting (HWRF) model using the Gridpoint Statistical Interpolation (GSI)-based ensemble-variational hybrid data assimilation (DA) systems. The influences of the following data on HWRF analyses and forecasts of hurricanes are first evaluated: 1) ocean surface winds derived from the National Aeronautics and Space Administration (NASA) Cyclone Global Navigation Satellite System (CYGNSS); 2) the new version of geostationary satellite-derived enhanced atmospheric motion vectors (AMVs), developed by the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin-Madison; and 3) the remote sensing and in-situ observations collected during the Tropical Cyclone Intensity (TCI) 2015 field experiment. Results show that the assimilation of the above new observations improves the simulation of three hurricanes [one nature-run hurricane, and Hurricanes Gonzalo (2014) and Joaquin (2015)] over the Atlantic Ocean in HWRF during their intensity changes. In addition, the degree of data impact depends on the DA configurations, DA methods and vortex initialization (VI) before DA in HWRF. Moreover, the implications of the DA results for the processes associated with hurricane intensification and weakening are examined. It is found that assimilation of CYGNSS winds and HDSS dropsonde observations leads to a better simulation of TC intensity changes, as it results in improved depiction of dynamical and physical processes within the hurricane inner-core. Considering the problems in the current HWRF initialization framework as observed in other studies, a revised initialization framework with GSI-4DEnVar and high-resolution, self-consistent HWRF ensemble background error covariance is proposed at the end of this dissertation. Experiments with Hurricanes Joaquin, Patricia and Matthew reveal not only a better representation of model initial conditions, but also improved track and intensity forecasts during the hurricane intensity change phase. |
