Tropical Cyclone Data Impact Studies: Influence of Model Bias and Synthetic Observations

Abstract

he impacts of assimilating dropwindsonde data and enhanced atmospheric motion vectors (AMVs) on tropical cyclone track forecasts are examined using the Navy global data assimilation and forecasting systems. Enhanced AMVs have the largest impact on eastern Pacific storms, while dropwindsonde data have the largest impact on Atlantic storms. Results in the western Pacific are mixed. Two western Pacific storms, Nuri and Jangmi, are examined in detail. For Nuri, dropwindsonde data and enhanced AMVs are at least as likely to degrade as to improve forecasts. For Jangmi, additional data improve track forecasts in most cases. An erroneous weakening of the forecasted subtropical high appears to contribute to the track errors for Nuri and Jangmi. Assimilation of enhanced AMVs systematically increases the analyzed heights in this region, counteracting this model bias. However, the impact of enhanced AMVs decreases rapidly as the model biases saturate at similar levels for experiments with and without the enhanced AMVs after the first few forecast days. Experiments are also conducted in which the errors assigned to synthetic tropical cyclone observations are increased. Moderate increases in the assigned errors improve track forecasts on average, but larger increases in the assigned errors produce mixed results. Both experiments allow for reductions in innovations and residuals when compared to dropwindsonde observations. These experiments suggest that a reformulation of the synthetic tropical cyclone observation scheme may lead to improved forecasts as more in situ and remote observations become available.