A pronounced bias in tropical cyclone minimum sea level pressure estimation based on the Dvorak technique

Abstract

A pronounced and highly significant bias is uncovered in tropical cyclone minimum sea level pressure (MSLP) estimates calculated using the Dvorak technique. The bias is present in operational estimates from each of the primary Atlantic tropical analysis centers (TACs). The bias can be approximated as a linear function of latitude and is caused by the dependence of tropopause temperature on latitude. On average, MSLP estimates from each TAC are consistently too high (compared to aircraft reconnaissance measurements) at higher latitudes and too low at lower latitudes. The latitude of zero bias is near 23°N. Because the relationship between tropopause temperature and latitude is fairly robust among the global ocean basins, the latitude-dependent bias that exists in Dvorak technique MSLP estimates of Atlantic basin tropical cyclones should extend to Dvorak technique estimates in all ocean basins. A simple linear fit is constructed between the Dvorak technique MSLP estimate errors and latitude, and this is applied as a latitude-dependent bias correction to the MSLP estimates. The correction has a significant effect on the error statistics of the samples from each TAC. Root-mean-square error is reduced by roughly 11%, 9%, and 10%, respectively, in the Tropical Analysis and Forecast Branch (TAFB), Satellite Analysis Branch (SAB), and Air Force Global Weather Center (AFGWC) samples. Using available wind data, it is shown that a much weaker latitude-dependent bias exists in Dvorak technique estimates of near-surface wind (Vmax). This is consistent with a recent study that used aircraft-based data from Atlantic tropical cyclones (TCs to demonstrate that for a given MSLP, the associated measured Vmax tends to be weaker at higher latitudes. The empirical relationship between MSLP and Vmax used in the Dvorak technique has no dependence on latitude, which indirectly introduces a bias in the estimated wind that counteracts the bias in the MSLP estimates. This suggests that historical best-track data formed using Dvorak technique estimates contain a systematic latitude-dependent MSLP bias and a systematic inconsistency in the relationship between MSLP and Vmax. Correction of the MSLP bias in past tropical cyclones that were estimated using the Dvorak technique may have measurable effects on the present tropical cyclone climatology. © 2004 American Meteorological Society.