A Climatology Of Rapidly Intensifying Tropical Cyclones In The North Atlantic Basin, 1975-2000
25th AMS Conference on Hurricanes and Tropical Meteorology (2002)
Chris C. Robbins and Stacy R. Stewart
National Hurricane Center, Miami, Florida
1. Introduction
Rapidly intensifying tropical cyclones pose a significant forecast problem for operational meteorologists. Unfortunately, the climatological nature of such systems for the North Atlantic basin is poorly understood and the formal literature offers very little insight. To partially fill this gap, a 26-year climatology (1975-2000) is being developed using the Atlantic Hurricane Database (HURDAT) maintained by the National Hurricane Center (Jarvinen et al. 1984).
2. Dataset and statistical methodology
Prior to 1975, much of the HURDAT data are incomplete (especially before the satellite era). Thus, the year 1975 was chosen as a starting point for the climatology to help ensure a more complete data set derived from both reconnaissance- and satellite-derived data. The statistics compiled exclude subtropical and extratropical cyclones.
HURDAT contains many systematic and random errors (Landsea et al. 2000) and the relevant data extracted appear to be highly non-Gaussian in nature. Thus, in this study, the median was used to measure the central tendency of various quantities since robust methods are not sensitive to particular assumptions about the overall nature of the data (Wilks 1995). For similar reasons the median absolute deviation (MAD) was computed examine the spread, or variability, of the data:
where x, is the ith data value, and q0.5 is the median, or 50th percentile, of the distribution.
3. Definition of rapid intensification
The definition of rapid intensification (RI) is very subjective and difficult to determine. However, we feel that a two-category increase in hurricane strength is a significant forecast problem and would certainly be considered rapid intensification. A two-category increase in hurricane strength equates to an increase in maximum sustained wind of approximately 32 kt and a decrease in central pressure of approximately 25 hPa based on the Dvorak pressure-wind curves derived for the Atlantic basin (Dvorak 1975). Thus, we have defined RI as a 24-h deepening of at least 25 hPa. This is consistent with the methodology of Kaplan and DeMaria (2000) who define RI as a 24-h increase in maximum sustained wind of at least 35 kt.
Note: At the time of this study, the largest pressure drop in a 24-hour period (fastest intensification) ever recorded in the North Atlantic Basic was 72 hPa (mb) by Hurricane Gilbert in 1988. That record was broken in 2005 by Hurricane Wilma, which deepened 97 hPa (mb) in 24 hours.Holliday and Thompson (1979), in their study of rapidly intensifying typhoons, used the upper 25th percentile of all 24-hour deepening rates as their criterion for RI. This equated to 42 hPa (24 h)-1. However for Atlantic hurricanes, less than one percent of all 24-hour pressure falls exceeded 42 hPa. In fact, the upper 25th percentile for 24-h deepening rates in Atlantic hurricanes is only around 15 hPa (Fig. 1). Although this rate of deepening could be considered climatologically infrequent, a 24-h pressure fall of 15 hPa may not necessarily pose a significant operational problem, nor be considered a significant intensification of a tropical cyclone.
4. Statistical analyses
All of the Atlantic tropical cyclones that experienced RI eventually became hurricanes. Out of the 151 hurricanes that formed in the Atlantic during the period of study, 47 (31%) had at least one period of RI during their lifetime. Further, since the 24-h period is determined by a moving summation of all four consecutive 6-h periods rather than by calendar day, it is possible for some storms to have several “periods” of RI. Thus, 168 “periods” of RI were found. The frequency of RI (not shown) peaks in September with 77 periods of RI (46%) spanning 23 storms (48%).
The monthly relative frequency distribution of all storms with at least one period of RI is skewed toward the latter part of the hurricane season. Twenty-eight percent of all hurricanes during the month of July have undergone at least one period of RI. The relative frequency increases to 33% for both August and September, but decreases to 21% in October and only seven percent in November. RI has occurred as early as July 9 (Hurricane Bertha, 1996) and as late as November 17 (Hurricane Lenny, 1999).
For each RI period, the four 6-h components that make up the net 24-h pressure change were examined. It was found that most of the pressure falls (median of 71%) in a 24-h period were accomplished during two of the four 6-h bins. Further, the two 6-h periods occurred consecutively in 84% of the RI cases examined. The median (MAD) 6-h pressure change computed by combining the two highest 6-h deepening rates within each RI period was 14 hPa (2 hPa); the median (MAD) pressure change for the two 6-h reduced-deepening periods in each 24-h RI period was 6 hPa (2 hPa). The MAD values computed for these parameters are an order-of-magnitude lower than the medians, indicating relatively low variance. Thus, these results imply that “periods” of rapid intensification are often followed or preceded by periods of significantly slower deepening within a given 24-hour period and that sustained periods of rapid deepening greater than twelve hours is unlikely.
Initial pressures at the onset of an RI period varied from 934 hPa to 1013 hPa. Most (51%) RI periods commenced in the interval 970-994 hPa. Interestingly, this range is approximately 10-15 hPa higher than the 956-985 hPa interval determined for Northwest Pacific typhoons by Holliday and Thompson (1979), consistent with the higher background mean sea-level pressure field found in the Atlantic. Further, the data also indicate a strong relationship between the 24-hour deepening rate and the initial central pressure of the system when the deepening began, with lower initial pressures corresponding to larger deepening rates (Table 1).
The minimum central pressure that the hurricanes ultimately achieved, given at least one period of RI, ranged from 888 hPa (Hurricane Gilbert, 1988) to 987 hPa (Hurricane Bill, 1987). The conditional frequency histogram for minimum pressures attained shows a distribution skewed toward the lower minimum central pressures with a peak in the 950-959-hPa range (Fig. 2). The conditional probability that the minimum central pressure will fall within this range given that at least one period of RI occurs is approximately 23%.
5. Future work
These results are preliminary and a great deal of work needs to be done to better understand the climatological nature of RI. As this study progresses, we will also investigate the large-scale environments associated with RI, which should guide us toward a better understanding of rapid intensification, the processes which drive it, and those which sustain it.
6. References
Dvorak, V. F., 1975: TC intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103, No. 5, 420-430.
Holliday, C. and A. H. Thompson, 1979: Climatological characteristics of rapidly intensifying typhoons. Mon. Wea. Rev., 107: 1022-1034.
Jarvinen, 8. R., C. J. Neumann, and M. A. S. Davis, 1984: A tropical cyclone data tape for the North Atlantic Basin, 1886-1983: Contents, limitations, and uses. NOAA Tech. Memo. NOAA/NWS/NHC 22, Coral Gables, Florida, 21 pp.
Landsea, C, C. and Coauthors, 2000: The Atlantic hurricane database re-analysis project: results for 1851-85. Preprints, 24th Conference on Hurricanes and Tropical Meteorology, Fort Lauderdale, FL, Amer. Meteor. Soc., 542-543.
Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences. Academic Press, 467 pp.