Emanuel's (pers. comm. 2007) Power Dissipation Index, including a correction for overestimated intensities in the 1950s and 60s, shows unprecedented high values in recent decades in the context of the past ~60 yr, and correlates remarkably well with low-frequency tropical Atlantic SST variations. Tropical Atlantic SSTs have increased over the past century, and we believe that anthropogenic forcing and internal climate variability have both likely made a discernible contribution to the strong warming since 1970. A limitation of Emanuel's index is that during the pre-satellite era (pre-1965), aircraft reconnaissance did not fully cover the basin, increasing uncertainty in PDI during that period. Also, the index does not extend back to the late 1800s and early 1900s, when Atlantic SSTs were apparently ~0.7C cooler than at present.
Landsea (EOS, 2007) uses landfalling statistics to infer no significant increase in basin-wide tropical storm counts since 1900. His conclusion is based on trends computed between pairs of active or inactive epochs in the series. U.S. landfalling hurricane activity (frequency and PDI) show no increasing trend over the past century or so; hurricane counts have a slight negative trend since 1878. However, Landsea's critical assumption of a constant landfalling fraction over time limits confidence in this assessment.
Holland and Webster (Phil. Trans. R. Soc. A 2007) examine relationships between numbers of all Atlantic named storms and hurricanes, minor hurricanes, and major hurricanes, and conclude that tropical cyclone and hurricanes have increased dramatically during the past century, related to the rise in tropical Atlantic SSTs. While one might presume that hurricanes and major hurricanes are less likely to have been missed in pre-satellite years than tropical storms, their key assumption that the existing HURDAT data accurately portrays long-term basin-wide statistics for hurricanes and major hurricanes requires further substantiation. In addition, while raw (unadjusted) basin-wide hurricane counts show a significant rising trend beginning in years 1881, 1891, …, 1921, such trends are not statistically significant if one begins in 1851, 1861, or 1871 (R. Smith, pers. comm. 2007).
We use historical Atlantic ship track and storm track data to estimate the expected number of missing tropical storms each year in the pre-satellite era (1878-1965). After adjustment, the storm counts covary with tropical SSTs on multi-decadal time scales, but their long-term trend (1878-2006) is weaker than the trend in similarly normalized SSTs (though both are nominally positive). The linear trend in adjusted storm counts for 1900-2006 is strongly positive (+4.2 storms/century) and highly significant according to three tests which attempt to account for serial correlation. However, this trend begins near a local minimum in the time series and ends with the recent high activity, perhaps exaggerating the significance of the trend. The trend beginning from 1878 is weakly positive, and not statistically significant with p~0.3. While we generally prefer using as long a series as possible to assess trends, the uncertainty in the late 1800s is larger than that during the 1900s—an important caveat on the results using the earlier start date. The results also suggest that the average duration of Atlantic tropical cyclones has decreased for reasons as yet unexplained. Tropical cyclone occurrence rates appear to have decreased in the western part of the basin (consistent with declining U.S. landfalling hurricane counts) but may have increased slightly in the central and eastern basin, suggesting a structural change such as shifts in storm tracks. Important assumptions of our methodology, such as that all landfalling storms since 1878 which were not seen by ships were detected and reported upon landfall, will require further investigation.
We have developed a new regional modeling framework designed specifically for dynamical downscaling of Atlantic hurricane activity. The non-hydrostatic model has a grid spacing of 18km and is run without convective parameterization, but with internal spectral nudging toward observed large-scale (basin wavenumbers 0-2) atmospheric conditions from reanalyses. The model simulates the observed rise in Atlantic hurricane activity (numbers, Accumulated Cyclone Energy (ACE), Power Dissipation Index (PDI), etc.) over the period 1980-2006 fairly realistically, as well as ENSO-related interannual variations in hurricane counts. Annual simulated hurricane counts from a two-member ensemble correlate with observed counts at r=0.86. The model does not simulate hurricanes as intense as those observed, with minimum central pressures of ~937 hPa and maximum surface winds of ~47 m/s being the most intense simulated in these experiments.
To explore possible impacts of future climate warming on Atlantic hurricane activity, we are re-running the 1980-2006 seasons, keeping the interannual to multidecadal variations unchanged, but altering the August-October mean climate according to changes simulated by an 18-member ensemble of AR4 climate models (years 2080-2099, A1B emission scenario). The warmer climate state features enhanced Atlantic SSTs, and also enhanced vertical wind shear across the Caribbean (Vecchi and Soden, GRL 2007). A key assumption of this approach is that the 18-model ensemble-mean climate change is the best available projection of future climate change in the Atlantic. Some of the models show little increase in wind shear, or even a decrease, and thus there will be considerable uncertainty associated with the hurricane frequency results, which will require further exploration. Results from our simulations will be presented at the meeting.
Based on the available evidence, we cannot yet conclude with high confidence that anthropogenic forcing has caused a discernible anthropogenic influence on hurricane activity to date.
# Collaborators on this work at GFDL include: Gabriel Vecchi, Joe Sirutis, Steve Garner, Isaac Held, and Bob Tuleya.
Supplementary URL: