A major focus of our current research project is to improve prediction of the inland impacts of landfalling tropical cyclones. Important tools for forecasting tropical cyclones are numerical weather prediction (NWP) models, which in principle can provide accurate predictions of conditions several days in advance. However, in order for these models to generate reliable inland forecasts for tropical cyclones, an accurate representation of the intensity and structure of the storm at the initial time is required. Unfortunately, it is very difficult to obtain observational measurements in these dangerous storms while they are offshore, and operational NWP models have not been designed specifically for tropical cyclone prediction. Therefore, a focus of our research is to find efficient and accurate ways to represent a tropical cyclone at an initial time, so that this information can be provided as an initial condition for an NWP model.