The impact of satellite-sensed winds on the intensity forecasts of tropical cyclones is evaluated by a simulation study with an axisymmetric numerical model. The parameterized physics in the forecast model are deliberately made different from those in the model that generates the observation. Model-generated “observations” are assimilated into forecasts by 12 h dynamic initialization.
A series of 24 h forecasts with and without assimilation of satellite-sensed winds are conducted and compared with the observations. Results indicate that assimilation with marine surface (or low-level) wind alone does not improve intensity forecasts appreciably, that a strong relaxation coefficient in the initialization scheme causes model rejection of the assimilation, and that an attenuating relaxation coefficient is recommended. However, when wind observations at the outflow level are included in the assimilation, forecasts improve substantially. The best forecasts are achieved when observations over the entire lower troposphere are assimilated.
Additional experiments indicate the errors in the satellite observations contaminate the forecast. But the assimilation of inflow and outflow winds still improve the intensity forecast if the satellite observation errors are less than or about the same magnitude of those in the initial wind field.