All files have a similar basic format consisting of a single header record, two or more storm history records and one storm type record (maximum record length is 80 bytes). The number of storm history cards depends on the duration of the storm.
BASIC FORMAT OF ALL DATA SETSHeader record (always one record):
Sample storm listing for the North Atlantic Ocean:
86390 11/15/1985 M= 9 11 SNBR= 839 KATE XING=1 SSS=2 L
86400 11/15* * * *2110638 35 999
86410 11/16*2160639 45 998*2170642 50 996*2150648 55 993*2110653 70 987
86420 11/17*2070660 75 981*2040664 75 984*2070673 75 982*2110688 80 977
86430 11/18*2140700 80 976*2160718 80 975*2160733 80 975*2190751 85 972
86440 11/19*2210768 95 967*2210784 95 968*2270802 90 971*2320819 80 976
86450 11/20*2390835 85 972*2460845 95 968*2520853 105 956*2600860 105 955
86460 11/21*2680865 105 954*2750866 100 961*2830865 95 965*2920861 85 967
86470 11/22*3020851 80 975*3150835 65 983*3250815 50 990*3370792 45 996
86480 11/23*3470762 40 1003*3440735 35 1005*3400720 35 1006E3350705 35 1006
86490 HR FL2 079 083 085 145U149 151
COLUMN NUMBERS
00000000011111111112222222222333333333344444444445555555555666666666677777777778
12345678901234567890123456789012345678901234567890123456789012345678901234567890
In this example there are 9 storm history records (as specified in columns 20 and 21 of header record) extending from 1800UTC Nov. 15 thru 1800UTC Nov. 23. There are four possible entries per UTC day with one entry each for 0000,0600,1200 and 1800UTC. These entries are in columns 12 thru 28/29 thru 45/46 thru 62/ and 63 thru 79, respectively. Each UTC entry gives latitude(degs N in this case), longitude (degrees W in this case) wind in knots (1-minute average in this case) and pressure in whole millibars. Missing winds are entered as -999. Missing pressures are left blank.
Latitudes and longitudes have been multiplied by 10 such that 213 is read as 21.3 and 801 is read as 80.1. For the North Atlantic, latitudes are degs N and longitudes are degrees west. A few North Atlantic storms move somewhat east of Greenwich and are coded as a negative longitude (see, for example, storm FAITH, 1966). For other basins, see appropriate note on page 4.
Columns 12, 29, 46 and 63 give additional info on the storm at each UTC where:
Supplementary information on winds - Columns 24, 41, 58 and 75 give supplementary wind information. Possible entries are blank, 'E','C' or 'P'. For the North Atlantic, Eastern Pacific and Western Pacific basins, these columns are always blank and can be ignored. 'E' is used in the North Indian and the Southern Hemisphere basins while 'P' and 'C' are used in the Australian basin only.
A designation of 'E' indicates that the a wind was assigned based on the global National Climatic Data Center (NCDC) tropical cyclone data set. Except in very recent years, that data set does not contain winds but rather contains an indicator as to whether the storm was a tropical depression, tropical storm or a hurricane/typhoon at the given time. To maintain the format shown above, a wind of 65E knots was arbitrarily assigned for hurricanes, 35E for tropical storms, and 25E for depressions.
For the North Indian Basin, there is an additional special problem. Before 1980, the practice over that basin was to refer to all storms with 10-minute mean winds over 48 knots (equivalent to about a 55 knot wind averaged over 1-minute) as "severe cyclonic disturbances". This was the highest category recorded. Thus, a wind entry of 64E for that basin could be any wind above 48 knots.
Another limiting feature of the 'E' wind designation is that information was available at 0000 and 1200 UTC only. Linear interpolation for winds and quadratic interpolation for storm positions was used to obtain 0600 and 1800 UTC data.
Characters 'P' and 'C' apply only to the Australian Basin. The principal data source, as obtained from the Australia Bureau of Meteorology, contained pressures but no winds. For these storms, pressure was converted to winds using:
Whenever, the wind computed to less than 15 knots from equation (1) a value of '15C' was assigned.
Storm type record (always a single record):The final record for each storm gives information on the type of storm. This is contained in columns 7 and 8. There are 7 possible entries:
Note: Table 1 of Neumann et al. (1987) gives additional information on North Atlantic storm types.
The storm type record card also occasionally contains a remark about the data. This can be ignored. For the North Atlantic basin only, the years 1899 and beyond also contain the maximum Saffir-Simpson scale number reached in the United States which border the Gulf of Mexico and the North Atlantic Ocean.
For all except the North Atlantic basin, the remark, beginning in column 49, 'LAST STORM FOR SEASON', appears. This is the last storm for the season in that basin. For the Northern Hemisphere, the seasons go from 1 January to 31 December while in the Southern Hemisphere, they go from 1 July to 30 June. By convention, the 1990 season in the SH ends on 1 July, 1991.
The remark 'LAST STORM FOR SEASON' is not used for the North Atlantic basin. Rather, the character 'L' appears in column 80 of the header card of the last storm.
The North Atlantic storm type record contains U.S. "hit" information starting in column 9. Each 4-character "hit" code is of the format "XSSI", where SS is the 2-letter state code, I is the impact on the state as a Saffir-Simpson intensity (1-5), and X has the following meanings for Florida and Texas only:
XSS AFL = Northwest Florida BFL = Southwest Florida CFL = Southeast Florida DFL = Northeast Florida ATX = South Texas BTX = Central Texas CTX = Northeast Texas
Finally, the North Atlantic storm type record contains 6 coastal crossing indices in columns 57 through 79 and Saffir-Simpson intensity information in columns 9 through 44. These entries are explained below.
Note on latitudes and longitudes: Latitudes for Northern Hemisphere basins are N and those for southern hemisphere basins are S. Longitudes are specified as W for the North Atlantic and the Eastern Pacific and as E for all other basins. Except for the North Atlantic, this convention is given on the header card (starting in column 55). In some basins, it is possible for storms to move from W to E longitudes or from E to W longitudes. In these cases, a longitude of 10.0E in a W system would be coded as -010.0W while a longitude of 170.0W in an E system would be coded as 190.0E.
SPECIAL NOTES FOR EACH BASINThis data set was structured as a database for the preparation of a global climatology of tropical cyclones for the World Meteorological Organization. Approximately three months time was spent in structuring the data base from individual basin data bases. All WMO recommended meteorological services were contacted. Some provided data and others did not. In general, there were many sources of data available. Often, there were conflicts in formats,tracks, winds and dates, names of storms, etc. These differences were subjectively resolved using meteorological judgement consistent with the time available. Certainly, it is considered likely that further indepth research into the tracks may indicate that revisions, additions, deletions thereto may be necessary. This is particularly true in the Southern Hemisphere basins.
One troublesome aspect is the differences in wind averaging times in the various basins. The United States including the Joint Typhoon Warning Center use a 1-minute average while other nations use a 10-minute average. Maximum wind is inversely proportional to averaging time. Thus, a storm, classified as a tropical storm by JTWC may not be similarly classified by countries using the 10-minute system. The topic is discussed by Simiu and Scanlan (1978) and Krayer and Marshall (1992).
In the discussions below, NCDC refers to the National Climatic Data Center in Asheville, NC and JTWC refers to the Joint Typhoon Warning Center in Guam.
(The following information applies to storm records provided by Charles Neumann of SAIC)
The North Atlantic dataset is the most comprehensive and, in general, the data can be used for the entire period beginning in 1886. However, the observational systems gradually improved over the years and there are certain milestones while relate to the accuracy of the data. Also, the data set has been extensively used and most (but certainly not all errors--typographical or otherwise) have been removed from the North Atlantic data set.
Before 1899, very little was known about the storm intensity over the open water and the intensities were inferred from indirect observations and ships logs. Wireless telegraph from ships began in 1905 and this improved the timely receipt of observations from ships at sea. However, it was not until 1945 that aircraft reconnaissance was used. The coastal radar network was completed about 1955 while satellites appeared in the mid-1960's.
This data set is maintained by the National Hurricane Center. However, extensive changes have been by SAIC thereto. These changes include:
Allowances are made for up to two crossings. These six coastal crossing timing indices (three for each potential crossing) are contained on the final card of each storm in columns 57,58,59, columns 61,62,63, and columns 65,66,67 for the first potential crossing and in columns, 69,70,71, columns 73,74,75 and columns 77,78,79 for the second potential crossing. A negative index indicates that the storm either did not cross a major coast.
To use the indices, it is first necessary to interpolate the 6 hourly winds and positions given in the storm data cards to hourly values. The interpolation scheme given by Akima (1970) is recommended for the position interpolation and a linear interpolation is suggested for the winds. A non-linear wind interpolation scheme could incorrectly boost a 63 knot wind to, say a 64 knot wind at other than the six-hourly anchor points. Hour number 1 is assigned to the first position and wind which, in the example below, is at 1200UTC on July 30; hour 7 to the second position and wind, etc. In this example, the last hour would be hour 85 (0000 UTC on August 3). The indexing scheme indicates that the storm was offshore at hour 49 (29.5N, 94.3W with 70 knots of wind) and was onshore at hour 55 (30.2N, 95.2W with 50 knots of wind). The storm crossed the coast nearest to hour 50.
Hour 50 is the closest hour that the storm crossed the coast. Thus, the position could be slightly onshore or slightly offshore, depending on the timing. Coastal crossings were determined using a somewhat smoothed coastline which did not include barrier islands. The indicator 'U' in columns 60 or 72 indicates that the crossing was over the contiguous United States.
A suggested procedure for obtaining hourly winds as the storm crossed the coastline would be to hold the wind constant from hour 49 to hour 50 and then linearly decrease the winds from hour 50 to 55.
89240 07/30/1989 M= 5 03 SNBR= 867 CHANTAL XING=1 SSS=1
89250 07/30 *2250900 20 1011*2350902 25 1010
89260 07/31*2440905 30 1009*2540910 35 1004*2620917 50 0995*2710922 55 0993
89270 08/01*2790928 65 0991*2870935 70 0987*2950943 70 0984*3020952 50 0993
89280 08/02*3080961 35 1000*3150969 25 1004*3230975 20 1007*3330980 20 1008
89290 08/03*3450985 20 1009*
89300 HR TX1 049U050 055 -99 -99 -99
COLUMN NUMBERS
00000000011111111112222222222333333333344444444445555555555666666666677777777778
12345678901234567890123456789012345678901234567890123456789012345678901234567890
Saffir-Simpson scale:
The Saffir-Simpson (SS) index appearing in column 59 of the North Atlantic data indicates the maximum SS when the storm crossed the U.S. coastline. Possible values are 1 to 5 (see Table 5 in Neumann, et al, 1987). An SS entry of '0' indicates that the storm was classified as a hurricane over coastal waters but weakened to below hurricane strength as it crossed the coast and damage was less than SS=1. An entry of '9' indicates that SS numbers were not assigned (this applies to pre-1899 storms). The final record for each storm (columns 9-44) amplifies the SS information on a state-by-state basis. (See Table 6 of Neumann, et al. 1993 for description of these data). SS information is only applicable for the North Atlantic basin.
The National Hurricane Center is in the midst of an Atlantic Hurricane Database Re-Analysis project. Using this new information, the data file for the North Atlantic basin has been extended backward to 1851 and the years 1886-1914 have been replaced.
The years 1915-2000 contain storm records as provided by Charles Neumann at SAIC, except that Hurricane Andrew 1992 was upgraded to a category 5 hurricane as part of the re-analysis and all records for that storm have been replaced.
From 2001-on, the file is being updated from NHC's archives of "Hurricane Season Tropical Cyclone Reports", which contain a best track for each storm.
Before the beginning of the satellite era in the mid-1960's, very little was known about these storms. Hence the data are very fragmented before 1966. Use of pre-1965 data is not recommended unless allowances are made for the missing data. There was very little aircraft reconnaissance of these storms unless there was a threat to the Hawaiian Islands.
The comments on coastal crossings under ATLANTIC, above, also apply to Eastern Pacific storms except the crossing is over the west coast of North America, typically over Mexico.
Before about 1972 (this date is uncertain), the data were archived by the U.S. Navy. From about 1972 through 1988, the tracks were prepared by the NOAA, Eastern Pacific Hurricane Center in San Francisco and the Central Pacific Hurricane Center in Honolulu. Thereafter, all data were prepared by the NOAA National Hurricane Center in Coral Gables, FL.
These data are taken from the published annual summaries of the Joint Typhoon Warning Center on Guam. The data are usable throughout the period of record except that some users have noted winds appear about 10% too high before 1970. This is likely due to the use of a different wind/pressure relationship before that date. If needed, additional information is available on this item.
Use of these data had disclosed some apparent errors in storm positions near the coastal regions of Southeast Asia. These have been corrected after checking storm track maps published by the People's Republic of China Central Meteorological Bureau in Beijing. A comment about the correction appears on the final card of the respective storm. See, for example, storm number 13, 1954.
From 1877 through 1944, all data are from NCDC. From 1945 through 1980, the data are a combination of JTWC and NCDC, with NCDC data used only to indicate intensities (25E, 35E, and 65E) for a storm when no wind speeds are provided for the storm by JTWC. Beyond 1980, all data are from JTWC.
From 1877 through 1944, all data are from NCDC. From 1945 through 1981, the data are a combination of JTWC and NCDC, with NCDC data used only to indicate intensities (25E, 35E, and 65E) for a storm when no wind speeds are provided for the storm by JTWC. Beyond 1981, all data are from JTWC.
Note that the Southern Hemisphere data file contains Australia area cyclones, but the track and intensity data do not always agree with the data provided by the Australian BOM (see below).
Data for this area is provided by the Australian Bureau of Meteorology. The BOM data set contains tropical cyclone data back to the 1906-1907 season. All 0000, 0600, 1200, and 1800 UTC data have been extracted from the BOM data set. Additionally, location and intensity data that are one hour off from the 6-hourly times have also been used. Locations were linearly interpolated to the 6-hourly times and intensities (wind and pressure) were assigned to the nearest 6-hourly time. Data that are more than one hour off from the standard 6-hourly times are not currently extracted.
For the current North Atlantic and Eastern North Pacific seasons, preliminary storm records will be added in real-time to the data files throughout the season. These records will be created from the 03Z, 09Z, 15Z, and 21Z NHC "Forecast Advisories" and will be marked with a 'P' in column 6 (note that the hours on the preliminary records are different than for the final best tracks, which are hours 00Z, 06Z, 12Z, and 18Z). As the NHC makes the best track information available for each storm, the preliminary records will be replaced with the final best track records.
REFERENCES
Akima, H., 1970: A new method of interpolation and smooth curve fitting based on local procedures. J. Assoc. Computing Mach., 17, 589-602.
Ho, F.P., J.C. Su, K.L. Hanevich, R.J. Smith and F.P. Richards, 1987: Hurricane Climatology for the Atlantic and Gulf Coasts of the United States. NOAA Technical Report NWS 38, National Weather Service, Silver Spring, MD, 195 pp.
Krayer, W.R. and R.D. Marshall, 1992: Gust Factors Applied to Hurricane Winds. Bulletin of the American Meteorological Society, May, 1992, pp. 613-617.
Simiu, E. and R.H. Scanlan, 1978: Wind Effects on Structures. Wiley Interscience, New York, NY, 458 pp. (Later edition available)
Neumann, C.J., B.R. Jarvinen, C.J. McAdie and J.D. Elms, 1993: Tropical Cyclones of the North Atlantic Ocean, 1871-1992. NCDC Historical Climatology Series, 6-2, 190 pp. (publication expected December, 1993)
Neumann, C.J., 1987: The National Hurricane Center Risk Analysis Program (HURISK), NOAA Technical Memorandum NWS NHC 38, 56 pp.
C.J. Neumann, SAIC, November, 1993