The forecasting skill impact of verifying TC dissipation

Most conventional skill measures of TC intensity forecasting explicitly ignore TC forecasts (by models or humans) when there is no TC to verify against. Such a decision is largely based on the desire to have a dataset that is consistent between track and intensity forecasts, and also conditional on the existence of a TC. Once dissipation has occurred, there is no longer a position to verify against, and thus the prior intensity forecasts are also ignored. The research presented here quantifies the impact of such excluded intensity forecasts on overall seasonal intensity performance, using the period 2000-2009. When a 15kt verification intensity (a reasonable environmental tropical wind speed absent a TC) is used to verify human intensity forecasts post-dissipation, the resulting impact on TC intensity forecasts ranges from a few percent degradation of mean absolute error from 12-36hr to five to seven percent degradation out to 120hr. The use of 10kt instead of 15kt generally doubles these degradation values. A detailed breakdown of these statistics, and the geography of dissipation cases where impact on skill is large, is presented. Similar statistics will also be presented for model guidance, and also for the reverse situation: where the forecasts anticipated dissipation, but the actual TC did not cease to exist.

Given the marginal edge that human and deterministic numerical guidance have over statistical methods for TC intensity forecasting, the degradation values shown here may wish to be considered when quantifying the practical value of intensity forecasts.