1.3: A GLOBAL CLIMATOLOGY
Global statistics on various aspects of tropical cyclones are presented. A Mercator map projection, similar in size and scale to that used in an earlier global climatology by Crutcher and Quayle (1974), is used to display the data in a homogeneous and global context, rather than by ocean basin. Additional charts, not amenable to the global format, are presented for individual basins. The visual format is used as much as possible, with minimal accompanying descriptive text. The data and discussion complement that found in Elsberry et al. (1987).
Figure 1.7: WMO CLASSIFICATIONS OF TROPICAL CYCLONES, TROPICAL DEPRESSIONS AND TROPICAL DISTURBANCES.
|Figure 1.8a: Comparison of classifications of tropical cyclones by the TCP regional bodies and JTWC. Definitions for the north Indian Ocean vary by country, the example is for India (see Fig. 1.8b) for complete list). The dashed line indicates the intensity threshold for naming or numbering tropical cyclones; note the lower value in the north Indian Ocean.|
|Figure 1.8b: Classification of cyclonic disturbances in the north Indian Ocean (WMO, 1986)|
The initial step in preparing the global climatology was to develop a global data set in a common format. Current tropical cyclones archives were obtained from the ocean-basin representatives cited under acknowledgments There were many inhomogeneities in these data that needed to be addressed in developing a common format. These, and associated matters are discussed in this section.
The data sets obtained from the various sources used widely different computerised or hard copy formats. These format dissimilarities could be circumvented, but it was more difficult to resolve inhomogeneities, such as: widely different periods of record, missing data, different documentation of the same tropical cyclone in adjacent basins, asynoptic observation times, different wind thresholds for stages of tropical cyclones, different practices for naming tropical cyclones and different wind averaging times in the various basins. In general, it was possible to completely or partially deal with all but the wind averaging heterogeneity (Section 1.3.3).
This lack of a standard for reporting and archiving tropical cyclone information continues to be a significant problem for research and inter-basin comparisons. WMO action to develop a global standard is highly recommended.
Preparation of a complete global climatology has only been possible since the satellite era began in the mid-1960s. Before then documentation of tropical cyclones in remote areas of the globe was very fragmented and mostly depended on chance encounters with ships or populated land areas (Holland, 1981c; Elsberry, 1987). Aircraft reconnaissance, which began in the mid-1940s, has provide excellent documentation, but only over portions of the North Atlantic and the western North Pacific basins.
Even with satellite data, a standardised method for interpretation, establishing suitable documentation procedures and formats, and dissemination of the data took several years to develop. Since tropical cyclones are relatively rare events that exhibit considerable interannual variations, at least 20 years of record is required for an adequate climatology. Thus, it is only recently that a comprehensive global climatology has become feasible.
Seven tropical cyclone basins (Table 1.2) are considered in this study. In Elsberry et al. (1987) the southeast Indian Ocean, Australia and the western South Pacific Ocean were considered as a single entity. However, there are sufficient longitudinal variations in tropical cyclone characteristic to justify separation into two basins at 142oE.
The principal sources used are specified in Table 1.2. Global tropical cyclone summaries through the year 1986 issued by the United States National Climatic Center in Asheville, North Carolina were used to fill in some missing data. The JTWC Annual Tropical Cyclone Reports, were used for most of the Southern Hemisphere basins for the three seasons 1986-87 through 1988-89. Data from neighbouring met services were utilised to develop the data base for the southern-hemisphere basin definitions used in Table 1.2.
Since the available periods and the quality and quantity of data were highly variable, different periods were used for each type of climatology. In general, intensity summaries were confined to the post-satellite years. In basins with aircraft reconnaissance (North Atlantic and the western North Pacific), climatologies extend from 1945.
|Basin Number||Basin name||Areal Extent||Principal Data
|1||North Atlantic||North Atlantic Ocean, Caribbean Sea and Gulf of Mexico||NHC|
|2||Eastern North Pacific||North America to 180oE||NHC|
|3||Western North Pacific||West of 180oE, including South China Sea||JTWC|
|4||North Indian||Bay of Bengal and Arabian Sea||India Met. Dept.|
|5||Southwest Indian||South Indian Ocean west of 100oE||Met.Service, Reunion|
|6||Southeast Indian/Australia||Southern Hemisphere 100-142oE||Bur. Met. Australia|
|7||Southwest Pacific/Australia||Southern Hemisphere east of 142oE||RSMC, Fiji|
|Season||North Atlantic||Eastern N Pacific||Western N Pacific||North Indian||Southwest Indian <1000E||Australia/ SE Indian <100-1420E||Australia/SW Pacific >1420E||Totals|
|% Global Total||11.6||12.0||19.8||19.7||30.7||35.7||6.5||5.6||12.4||9.9||8.2||7.6||10.8||9.5||100.0%|
Table 1.3 summarises the frequency of tropical cyclones by basin and is based on the data sources specified in Table 1.2. Alternate data sets that exist in some basins would result in slightly different statistics. For example, records maintained by the Central Meteorological Bureau of the People's Republic of China (CMB, 1972; Dong, personal communication, 1990), indicate slightly larger western North Pacific frequencies than those given by JTWC. Some of these inconsistencies are due to the different wind averaging times used.
The average annual frequency of tropical cyclones over the globe is 83.7. This is slightly higher than the value of 80.1 from Frank (1987) and Gray (1985) or 81.6 from Crutcher and Quayle (1974), but is consistent with the use of post-satellite data only. Also, considerable differences can be found for some years between Table 1.3 and Frank (1987, his Table 3.1). This is likely due to the use of very recently issued basin summaries here.
In Table 1.3 a distinction is made between maximum winds averaged over 10-min and 1-min periods. Following WMO guidelines all Regional Associations except RA IV (Northern and Central America) and the JTWC, use 10-min winds(1). In Table 1.3, 1-min averaging is indicated for the Southern Hemisphere basins in 1986-87 through 1988-89, when JTWC data were used.
The longer the averaging period used, the lower the maximum wind speed for a cyclone of a given intensity. Therefore, the cyclone frequency statistics in Table 1.3 are low compared to those that would be derived using 1-min averages. This is a heterogeneity that is difficult to address and has been generally ignored when comparing statistics from different basins.
Following Simiu and Scanlon (1978), a multiplication factor of 0.871 is applied to convert between 1-min and 10-min winds. Thus, a 63/118 km h-1 (34/64 kt) wind in the 1-min system would only be equivalent to about 55/103 km h-1 (30/56 kt) in the 10-min system. The effect on tropical cyclone frequency is estimated in Table 1.4 by converting the maximum wind observations to 10-min means then recalculating the statistics for the North Atlantic, eastern North Pacific and western North Pacific.
The general conclusion obtained from Table 1.4 is that the different averaging times have an inconsequential effect on tropical cyclone frequency. The 4% difference lies well within the variations arising from analysis methods for determining maximum intensity and is too small to justify any additional homogenisation of the statistics in Table 1.3. The 17% difference for systems of hurricane or typhoon strength is sufficiently large, however, to require care with interpretation. All statistics for severe systems in Table 1.3 and in the following sections should therefore be assigned an uncertainty of at least ±15%.
Figures 1.9 through 1.14 depict tropical cyclone frequency and motion on a global scale using the data sources in Table 1.2. Only systems of at least minimal tropical cyclone intensity (>63 km h-1 or 34 kt) are included in the analysis. The region from 10oW to 30oE is not included since tropical cyclones do not occur over these longitudes in either hemisphere. We also note that in the Atlantic basin the analysis includes the extratropical stages of storms (see comments in Section 1.3).
The tracks of all tropical cyclones for the 10-year period 1979-1988 are displayed in Fig. 1.9. These tracks were constructed from 6-h best-track positions interpolated to hourly positions using the method of Akima (1970).
Cyclone density is indicated by Fig. 1.10, following approach adopted by Neumann and Pryslak (1981) for the North Atlantic and by Xue and Neumann (1984) and PRC (1972) for the western North Pacific. It is based on the number of tropical cyclones (resolved into hourly positions) passing within 75 nm (139 km) from each of 77 x 163 grid points spaced along the equator at 2o intervals, but linearised in the meridional direction to the Mercator map shown. The advantage of using circular areas rather than the traditional 2½o or 5o Marsden squares is discussed by Taylor (1986). The data were normalised to a 100 year period and smoothed with the 9-point filter of Shuman (1957).
MAXIMUM WIND REACHING AT LEAST 63 km h-1 (34 kt)
|Number of occurrences|
MAXIMUM WIND REACHING AT LEAST 118 KM H-1 (64KT)
|Number of occurrences|
Care is needed in the interpretation of these data. A frequency of 100 cyclones over 100 years indicates an average of 1 per year. This should not be interpreted as a 100% probability of a cyclone occurring on that date. Rather, use of the Poisson distribution (Xue and Neumann, 1984) indicates a 37% chance of no tropical cyclone occurring. This distribution provides an excellent estimate of occurrence probability for small numbers of cyclones in limited regions. If a long period of accurate record is available Neumann, et al. (1987) found that the use of relative frequencies provide a better estimate of event probability.
|1-10||Cyclone identification code (if any) and name (if any) separated by 1 space if applicable. If the 'code and name' exceeds 10 characters, then use an abbreviation of the name (first letters only) instead of the full name in positions 1-10 and the complete name should then be entered in the expansion fields beginning in position 53. Not more than one name should be entered for the cyclone during the period shown in positions 11-20, that is the name most used for regional or international purposes by the centre indicated in positions 51-52.|
|19-20||Hour -universal time (at least every 6 hourly position - 00Z, 06Z, 12Z and 18Z)|
|21||Latitude indicator: 1 = North latitude; 2 = South latitude|
|22-24||Latitude (degrees and tenths)|
|25-26||Check sum (sum of all digits in the latitude)|
|27||Longitude indicator: 1 = West longitude; 2 = East longitude|
|28-31||Longitude (degrees and tenths)|
|32-33||Check sum (sum of all digits in the longitude)|
1 = good (<30 nm; <55 Km)
2 = fair (30-60 nm; 55-110 km)
3 = poor (>60 nm; > 110 km)
9 = unknown
Note*: Confidence in the centre position; Degree
of confidence in the centre position of a tropical cyclone expressed as the radius of the
smallest circle within which the centre may be located by the analysis.
|Dvorak T-number (99 if unknown or unvailable)
Dvorak C1-number (99 in unknown or unavailable)
|39-41||Maximum average wind speed (whole values) (999 if unknown or unavailable)|
|42||Units of wind speed. 1 - knots; 2 - meters per second|
|43-44||Time interval for averagin surface wind speed (minutes for measured or derived wind speeds or 99 if unknown (e.g. estimated)|
01 = tropics; disturbance (no closed isobars)
02 = a < 34 knot winds, < 17 n/s winds and at least one closed isobar
03 = 34-63 knots, 17-32 m/s
04 = > 64 knots, > 33 m/s
05 = extratropical
06 = dissipating
07 = subtropical cyclone (frontal, low pressure system that comprises initally barclinic circulation developing over subtropical water)
08 = overland
09 = unknown
|47-50||Central pressure (nearest hectopascal) 0000 if unknown or unavailable)|
|51-52||Source code (2-digit code to represent the country or
organization that provided the data to NCDC USA, WMO Secretariat, TCP applied to assign
number to additional participation centres, organizations)
RSMC Miami-Hurricane Center
Cyclone identification code and name; 11-20 Date time group; 21-34 Best track positions;
35-46 Intensity; 47-50 Pressure; 51-52 Source code; 53-80 Expansion.
Note: The 80 character format allows for data to be digitized on cards, diskettes, or magnetic tape.
The mean cyclone motion was obtained by averaging zonal and meridional components, together with scalar speed of the tropical cyclones within the 140 km circular domain around each grid point (Figs. 1.11-1.13). To avoid a bias towards slower moving tropical cyclones, the average motion of each cyclone per grid point was archived, regardless of the number of hours it remained within the domain. Only those grids which contained at least 7 cyclones were used in the analysis.
Considerable basin variability in the poleward extent of tracks is evident in Fig. 1.9. Some of this variability could be an artifact of different observational and documentation procedures. For example, some of the statistics in the North Atlantic are due to the inclusion of the early portions of the extratropical stage in the tropical cyclone archives. Further processing of these Atlantic data for the 10-year period shown in Fig. 1.9 indicate that the proportion of extratropical systems poleward of 42.5, 47.5 and 52.5oN is 50, 75 and 100%, respectively. However, the ability of tropical cyclones to maintain warm core characteristics as they move out of the tropics also is a function of a number of factors including sea surface temperatures (SSTs), environmental flow characteristics (such as vertical shear and equatorial extent of westerlies), and land/water configuration.
In the eastern North Pacific tropical cyclones typically encounter cold SSTs (<22.50C) at low latitudes (eg Frank, 1987) and dissipate before recurvature into the westerlies. By comparison, relatively warm SSTs in the western North Pacific and the North Atlantic allow tropical cyclones to be sustained well into the subtropics, even after recurvature into the westerlies.
Southern Hemisphere SSTs, on the average, are lower than those of the Northern Hemisphere for the same latitude and tend to be intermediate between the two extremes cited above. This, together with the more equatorward extent of westerlies typically result in poleward moving tropical cyclones losing their warm core characteristics at lower latitudes than for similar systems in the Northern Hemisphere.
The tracks in Fig. 1.9 indicate that the eastern North Pacific basin experiences the maximum number of tropical cyclones per unit area followed by the western North Pacific; but the latter has a much larger areal extent. Tropical cyclone densities in the remaining basins are seen to be reasonably similar. This is quantified in Fig. 1.10 using longer periods of record. The minimum contour of 5 cyclones per 100 y provides an approximate bound to tropical cyclone activity.
The maximum tropical cyclone density (303 per 100 y) occurs in the eastern North Pacific Ocean near 16oN, 112oW. Another substantial maximum (238 per 100 y) is found east of Luzon in the Philippines. The highest frequency in the Southern Hemisphere is 124 per 100 y off the Australian west coast. The local maxima of 88 in the south Indian Ocean near 90oE and 71 in the Australian Gulf of Carpenteria do not appear on earlier global climatologies (Fig. 2.2 of WMO, 1979; Crutcher and Quayle, 1974). This is interpreted as lack of detection of cyclones in these regions in the pre-satellite era (Holland, 1981c; Foley, 1989).
A useful estimate of the number of years having discrete tropical cyclone occurrence in a particular area (the number of years to expect no cyclones, 1 cyclone, etc) may be obtained by use of the Poisson distribution. Discussion on this application is given by Xue and Neumann (1984).
The mean directions of motion (Fig. 1.11) show that the classical recurvature patterns occur over the North Atlantic and the western North Pacific basins, and to a lesser extent in the southwestern Indian Ocean. As discussed earlier, cyclones over the eastern North Pacific typically dissipate before recurvature into the westerlies. In both the north Indian Ocean and northern and western Australia regions cyclones often encounter land and dissipate before or during recurvature. The near-equatorial approach of mid-latitude westerlies in the southwest Pacific leads to a predominantly eastward motion of tropical cyclones. Section 22.214.171.124 contains additional comments and details on recurvature.
|Figure 1.9: Tracks of tropical cyclones (maximum winds >63 km h-1, 34 kt) for the period 1979-1988|
|Figure 1.10: Frequency of tropical cyclones per 100 years within 140 km of any point. Solid triangles indicate maxima, with values shown. Period of record used is shown in boxes for each basin.|
|Figure 1.11: Mean direction of tropical cyclone motion over the periods indicated.|
|Figure 1.12: Mean scalar speeds (km h-1) of tropical cyclones over the indicated periods. Light shading indicates speeds <20 km h-1, dark shading >45 km h-1. Minimum number of cases, 7.|
|Figure 1.13: Mean vector speeds (km h-1) of tropical cyclones over the indicated periods. Light shading indicates speeds <20 km h-1, dark shading >45 km h-1. Minimum number of cases, 7.|
|Figure 1.14: Steadiness of tropical cyclone motion over the specified periods. The index is defined by 100x(vector speed)/(scalar speed). Light shading <65, darks shading >90, minimum number of cases, 7.|
The means of cyclone motion are presented in Figs. 1.12 and 1.13, respectively. Wide ranges in tropical cyclone translational speeds are observed, both within and across ocean basins. Some of the more significant aspects of translational speeds are:
Analyses of tropical cyclone position forecast errors by Neumann and Pelissier (1981) and Jarrell et al. (1978) demonstrate that they are highly dependent on the tropical cyclone translational speeds and variability thereof. The greatest forecast errors are typically associated with rapidly moving or highly variable tropical cyclones. Thus, indications of the steadiness of cyclone motion, combined with a knowledge of the mean translational errors provides a useful means of comparing forecast errors.
The vector speeds, displayed in Fig. 1.13 are always less than the scalar speeds and the proportion of speed reduction compared with the scalar means in Fig. 1.12 indicates the degree of variability in tropical cyclone motion. Thus, the western part of the eastern North Pacific basin has highly consistent tropical cyclone motions, whereas the recurvature latitudes of the North Atlantic and the western North Pacific as well as the equatorial regions of the Southern Hemisphere experience highly variable motion. A climatological measure of speed variability therefore can be obtained by algebraically dividing vector speeds by scalar speeds. Such a measure is often referred to as a steadiness or constancy index (Crutcher and Quayle, 1974; Hope and Neumann, 1971; Miller et al., 1988).
The global variation in tropical-cyclone motion steadiness, as defined by the above index multiplied by 100 and rounded off the nearest integer value, is presented in Fig. 1.14. Note that the higher the index, the more consistent the motion, and perfectly steady cyclones would rate 100. Three ranges of steadiness have been arbitrarily defined with indices greater than 90 being rated high, 60-90 indicate average steadiness and systems below 60 are rated as erratic.
Noteworthy in Fig. 1.14 are the regions of erratic motion (some less than 40) in the Australian/southwest Pacific region and the remarkably consistent tracks in the eastern North Pacific basin. Relatively low steadiness values also are found over the recurvature latitudes of the North Atlantic and the western North Pacific basins.
Meridional profiles of zonal mean parameters for each of the seven ocean basins (Table 1.2) are presented in Figs. 1.15-1.21. These profiles were prepared from the same data used in Section 1.3.4. The data for each basin were averaged over 2o lat. bands centred at odd latitudes from 1-55o. The total number of tropical cyclones included in the analyses are given in the inset as "NSTMS = nnn", while the number of cyclones included in the various latitude bands is given under the column labeled "number of cyclones". Thus, for the eastern Pacific Ocean basin in Fig. 1.16, 401 tropical cyclones were used and 306 of these passed through the 2o lat. band centred on 15oN. Only those bands for which the number of tropical cyclones exceeded 10 were included in the analysis.
With some minor variations, tropical cyclones in all basins experience an increase in translational speed, both poleward and eastward, as they move into the subtropics. Variations include small speed maxima in the equatorial region (for example, near 10oN in western North Pacific) and at high latitudes (for example, near 50oN in the North Atlantic basin). Overall, the lowest translational speeds (see figure insets) occur in the north Indian basin and the fastest movement occurs in the North Atlantic and the western North Pacific basins.
The inset on each figure indicates the latitude of zero mean zonal motion. In basins which exhibit the classical recurvature patterns, such as the western North Pacific (Section 126.96.36.199), this can be interpreted as being near the average latitude of recurvature into the westerlies. However, for regions such as Australia and the southwest Pacific this latitude simply separates regions of mean westward and eastward motion; the few recurving cyclones in this region do so at 17-20oS rather than at the 9.9oS lat. of zero zonal motion.
Figs. 1.15 through 1.21 also include an index of the expected maximum cyclonic winds in each 2o lat. belt, relative to the average cyclonic maximum wind for the entire basin. This value was used in preference to the actual wind to avoid some of the wind heterogeneity problems among basins as discussed in Section 1.3.1.
Tropical motion is a vector quantity, it is best described in the bivariate sense and Crutcher (1971) recommends use of the bivariate normal distribution. However, specification of the sizeable bivariate normal statistics(2) for the various basins is beyond the scope of this document. Such basin statistics, using mostly pre-satellite data, are given by Crutcher and various co-authors; for example, the North Atlantic Ocean basin by Crutcher and Quinlan (1971) and the southwest Indian Ocean basin by Crutcher and Nicodemus (1973).
Tropical cyclone scalar speeds, which have a zero lower bound and no upper bound, can be described by the Gamma distribution. Examples for the western North Pacific basin are given by Xue and Neumann (1984) and for the North Atlantic by Neumann (1987). Both publications describe methods (moment estimates and maximum likelihood) for estimating the two parameters needed to define the Gamma distribution. In addition, the latter publication describes other distributions useful in tropical cyclone climatology. This includes the Weibull distribution for describing maximum winds in tropical cyclones and the log-normal distribution for describing the radii of maximum wind (Hahn and Shapiro, 1967).
|Figure 1.15: Meridional profiles of specified tropical cyclone parameters for the North Atlantic basin averaged over 2o lat. bands (not computed for less than 10 cyclones in a band). The inset gives period of record, total number of cyclones, basin average cyclone motion components (u,v), translational speed (s), and the latitude where the averaged zonal motion component changes sign (uo). Vector arrows have lengths proportional to speed.|
|Figure 1.16: Meridional profiles of specified tropical cyclone parameters for the eastern North Pacific basin averaged over 2o lat. bands (details as in Fig. 1.15).|
|Figure 1.17: Meridional profiles of specified tropical cyclone parameters for the western North Pacific basin averaged over 2o lat. bands (details as in Fig. 1.15).|
|Figure 1.18: Meridional profiles of specified tropical cyclone parameters for the north Indian basin averaged over 2o latte bands (details as in Fig. 1.15).|
|Figure 1.19: Meridional profiles of specified tropical cyclone parameters for the southwest Indian basin averaged over 2o latte bands (details as in Fig. 1.15).|
|Figure 1.20: Meridional profiles of specified tropical cyclone parameters for the Australian/southeast Indian basin averaged over 2o latte bands (details as in Fig. 1.15).|
|Figure 1.21: Meridional profiles of specified tropical cyclone parameters for the Australian/southwest Pacific basin averaged over 2o latte bands (details as in Fig. 1.15).|
In basins where classical recurvature occurs, the average maximum intensity tends to occur just prior to recurvature. This confirms the findings of Riehl (1972), however, there is considerable variation from one tropical cyclone to another and it is inadvisable to use this as a forecasting rule for individual systems.
The maximum wind in tropical cyclones is related to a number of factors including sea surface temperatures, low level inflow and advection, high level outflow, topographic factors, vertical wind shear etc. (Merrill, 1987; Emanuel, 1986; Holland, 1987). Since a lag can be expected in tropical cyclone response to these factors, the average latitude of maximum wind can be interpreted as the region slightly poleward of where the factors related to intensification are maximised. For example, in the eastern North Pacific basin, maximum winds occur while tropical cyclones are embedded in the easterlies. Here, the cold SSTs at low latitudes cause rapid weakening as cyclones move poleward.
The steadiness index defined in Section 188.8.131.52 indicates the degree of variability of tropical cyclone motion in each latitude band (Figs. 1.15-1.21). As a general rule for the Northern Hemisphere basins, the least variability occurs in equatorial regions and the most variability is near the mean latitude of recurvature. Interestingly, this also is the region of highest intensity and thus of greatest forecast priority. In the Southern Hemisphere, and particularly in the Australian region, considerable variability is found in all but high latitude storms.
|Figure 1.22: Seasonal profiles of tropical cyclone frequency for the North Atlantic (top), the eastern North Pacific (middle) and the western North Pacific (bottom) basins. Upper and lower bounds, respectively, refer to winds of at least minimal tropical cyclone intensity and at least hurricane intensity. Data have been smoothed over a 15 day period.|
|Figure 1.23: Seasonal profiles of tropical cyclone frequency for the north Indian (top), southwest Indian (middle) and Australia/southeast Indian (bottom) basins. Upper and lower bounds, respectively, refer to winds of at least minimal tropical cyclone intensity and at least hurricane intensity. Data have been smoothed over a 15 day period.|
|Figure 1.24: Seasonal profiles of tropical cyclone frequency for the Australia/southwest Pacific basin (top), together with combined statistics all basins (bottom). Upper and lower bounds, respectively, refer to winds of at least minimal tropical cyclone intensity and at least hurricane intensity. Data have been smoothed over a 15 day period.|
Figures 1.22-1.24 depict intraseasonal aspects of tropical cyclone frequency for each of the seven basins (Table 1.2). Combinations of tropical cyclones in all basins also are presented in Fig. 1.24. The time scale on these charts was adjusted to begin on 1 December and to end 14 months later, on 31 January to avoid splitting the Southern Hemisphere maxima.
Preparation of these figures required use of pre-satellite data, with the problems discussed in Section 1.3.1. Further, hurricane frequency(3) was not specifically documented over the north Indian basin prior to 1980 (Mandal, 1990, 1991). These problems were partially addressed for the north and the southwest Indian basins by computing daily frequencies both for the satellite era, 1980/1989, and for the earlier period of available data. The daily hurricane frequencies for the earlier years were then uniformly scaled upwards to match the satellite era. Tropical cyclone frequencies for the southwest Indian basin were similarly adjusted but no tropical cyclone scaling was needed for the north Indian basin.
Although no adjustments were made to the data for other basins, Holland (1981c) and Murphy (1988) indicate that estimated surface pressures in pre-satellite and early satellite years may have been too high in and around the Australian area. Accordingly, daily hurricane frequencies (and perhaps, tropical cyclone frequencies) may be somewhat higher than those depicted for these regions.
The data were smoothed using a linear moving average period of 15 days. For very long periods of record, Neumann et al. (1987) found that a 9 day smoothing period is optimal for removing random data fluctuations whilst preserving known seasonal variations. However, the longer 15 day period was necessary here to maintain homogeneity with some of the basins of short historical record. The figures also are presented in terms of the number of tropical cyclones per 100 years to be expected on any given day (discussed in Section 1.3.4).
Figs. 1.22-1.24 reveal considerable details in the character of the tropical cyclone season over the various basins. For example, some basins exhibit a sharp single maximum of tropical cyclone occurrence while others basins show multiple maxima. The western North Pacific has experienced tropical cyclones throughout the year, whilst most other basins have several months of little, or no tropical cyclone occurrence.
No attempt is made here to describe the meteorological reasoning for the various basin patterns, which is covered by Frank (1987). Considerable information relating to the Australian area and vicinity may be found in the Australian Tropical Cyclone Forecasting Manual (Bureau of Meteorology, 1978) and for the North Atlantic in the US NHC Trackbook (available from the NHC).
Some of the more important features depicted in Figs. 1.22-1.24 are:
The bottom panel of Fig. 1.24 contains the total tropical cyclone activity for all basins combined. Minimum tropical activity over the globe as a whole occurs during May; this is followed by a maximum in September and a secondary minimum is experienced in December. There also is a suggestion that the number of tropical cyclones reaching hurricane intensity is greater over the Northern than over the Southern Hemisphere. However, the magnitude of the difference is probably less than that shown in Fig. 1.24, since: 1. all Southern Hemisphere basins use 10-min average winds, which may introduce up to 20% bias (Section 1.3.3); and 2. considerable underestimation of southern cyclone intensity may have occurred in early and pre-satellite years (Murphy, 1988; Holland, 1981c).
1. Further confusion is caused by the lack of a uniform system of wind observations in the north Indian basin; the countries in this region use a combination of 1, 3 and 10-min averaging; for synoptic observations 3-min averaging is the most common.
2. Note that five parameters are necessary for each sampling space: the mean and standard deviation of the zonal and meridional components and the linear correlation coefficient between components.
3. Hurricane force winds are relatively infrequent in the north Indian basin and the early practice was to distinguish between tropical cyclones of 63-88 km h-1 maximum winds (34-48 kt) and severe tropical cyclones with maximum winds exceeding 88 km h-1 (48kt). Current practice is to also include the hurricane stage of winds in excess of 118 km h-1 (64 kt).
Contents Chapter 2