Using the HFIP high-resolution hurricane test dataset to diagnose interconnected track and intensity errors in hurricane forecasts

As part of the HFIP High-Resolution Hurricane (HRH) Test, diagnostic analyses were performed for forecasts of rapid intensification (RI) and rapid weakening (RW) events in Atlantic tropical cyclones. A principal interest in these analyses was to determine if high-resolution forecasts could better capture these episodes than lower-resolution versions of the same model. Concurrently with these studies, a similar comparison was made for possible improvements in cyclone track forecasts as resolution increased. One implication of these essentially independent studies was the importance of combining track and intensification analyses in order to try to disentangle the two sources of error. For instance, it was clearly suggested that intensification error statistics for short lead-time forecasts were different from longer lead times when the hurricane track errors had become large. In this paper, we present some initial qualitative diagnoses that seek to interpret the different impact of the two sets of errors. We first present a cursory geographic climatology of forecast and observed periods of rapid intensification and weakening for the dataset of Atlantic storm forecasts created for the HRH. By then collating track observations and forecasts with parallel RI and RW records, we will be able to analyze the relation between intensification and track error for various lead times. To demonstrate methods used to establish these relationships, cumulative frequency displays of RI and RW timing periods and representative track maps for individual models and for individual storms are presented. Since inter-run consistency of tracks and intensity changes is a critical source of uncertainty for forecast evaluation, metrics and qualitative displays that assess consistency will be presented.