The prediction of tropical cyclone intensity (i.e. maximum surface wind speed) is perhaps the foremost challenge faced today by the operational weather prediction community. In this seminar, I will describe my work defining a new perspective on past progress in tropical cyclone (TC) intensity prediction, my research concerning the predictability of TC intensity, and recent advances at the Naval Research Laboratory (NRL) in observing and modeling TCs.
First, I will demonstrate that the accuracy of National Hurricane Center (NHC) Atlantic basin TC intensity forecasts has significantly improved over the last 20 years, in contrast to the prevailing view that there has been no improvement. This new interpretation of the NHC intensity forecast accuracy trend results from a normalization of the mean forecast error that accounts for changes in the statistical characteristics of the ‘best-track’ intensity observations used to verify the forecasts. Essentially, the average variability of a ‘best-track’ TC intensity time series has increased through the years, making it systematically more difficult for a forecast time series to match the observed time series. Further research is needed to understand why the average variability of ‘best-track’ intensity time series has increased, but I will provide some hypotheses.
While the new interpretation of the NHC intensity forecast accuracy trend gives hope that forecast improvements will continue, at some point the predictability of TC intensity will limit the expected accuracy of any deterministic forecast. In the second part of the talk, I will explore the issue of TC intensity predictability in the context of a simple dynamical model (the Coupled Hurricane Intensity Prediction System, CHIPS) by examining the sensitivity of model simulations to the initial state of the TC vortex and its environment. Amongst initial TC intensity, initial TC intensification rate, environmental vertical wind shear magnitude and oceanic mixed layer depth, it is found that CHIPS-simulated intensity is most sensitive to the initial intensity and shear magnitude. The sensitivity is highly state-dependent, in a manner such that CHIPS-simulated intensity is most sensitive to the initial intensities and shear magnitudes most likely to be observed for Atlantic basin TCs. Intriguingly, this result suggests that the most common TC intensity forecast situations are those in which the expected deterministic forecast accuracy is the lowest.
Finally, I will conclude with a brief overview of continuing development work at NRL pertaining to our new operational tropical cyclone prediction model, COAMPS-TC, and our role in the multi-year NASA Hurricane and Severe Storm Sentinel field program.