Mn-~ 1950 GABY, LUSHINE, MAYFIELD, PEARCE AND TORRES 587Satellite Classifications of Atlantic Tropical and Subtropical Cyclones: A Review of Eight Years of Classifications at MiamiD. C. G^B-, J. B. LUSHINE, B. M. MAYFIELD, S. C. PEARCE AND F. E. TORRESSatellite Field Services Station, National Environmental Satellite Service, NOAA, Miami, FL 33146(Manuscript received 29 October 1979, in final form 10 January 1980) ABSTRACT Estimates of the locations and maximum sustained wind speeds of all tropical and subtropical cyclonesin the North Atlantic Ocean, the Caribbean Sea and the Gulf of Mexico have been made at Miami since1971 using satellite techniques developed by Timchalk et al. (1965), Dvorak (1972) and Hebert and Poteat(1975). The estimates were compared with the National Hurricane Center's "best tracks" data to establishthe measure of accuracy achieved. These data are not entirely independent because the best tracks themselves are determined partly from the satellite estimates; however, comparisons were made only duringperiods when aerial reconnaissance was also available. The average difference between satellite-derivedmaximum sustained wind speeds and best track maximum sustained wind speeds has consistently been-7 kt with standard deviation of -8 kt. The average difference between satellite locations and best tracklocations has decreased to -17 n mi with standard deviation of ~14 n mi, which is believed to be an approximate lower limit for the present state of the art and technology. These results and other informationare provided for an 8-year period.1. Introduction and historical background For the past eight hurricane seasons, 'the MiamiSatellite Field Services Station (SFSS) of the National Environmental Satellite Service (NESS),formerly the Satellite Applications Section (SAS) ofthe National Weather Service's National HurricaneCenter (NHC), has provided "classifications" of alltropical and subtropical cyclones in the North Atlantic Ocean, the Caribbean Sea and the Gulf of Mexico.Classification includes fixing the location of thestorm circulation center, estimating the maximumsustained wind speed (MWS), and describing certaincharacteristics of the storm such as the trend ofdevelopment, mean motion and indications of probable future change. These classifications, togetherwith information from other sources, are used by theNHC's hurricane specialists to formulate theiradvisories and warnings. This report reviews theresults from classifications over an 8-year period ofeffort, the progress made and suggests reasonableexpectations for the future. The thrust of our efforts has been to provide theNHC with the best possible classifications fromsatellite imagery. The importance to the hurricane.forecast of well-determined storm circulation centers is treated in two papers by Neumann (1975a,b).The importance of an accurately estimated MWS isobvious and well recognized as an equally significantparameter for hurricane forecasts and warnings.The Miami SFSS meteorologists have up-to-thehalf-hour electronic image animation available to aidin locating circulation centers, together with enlargement prints of satellite pictures of the storms. From the 'beginning, an effort has been made toevaluate the accuracy of SFSS estimates of stormlocation and MWS. No absolute measure of accuracy in estimating the location and strength oftropical or subtropical cyclones is possible becausethere is no absolute "ground truth" i.e., neither theexact location nor the precise wind speed of thestorm is known at any given time. However, comparisons have been made each year with the NHC'sbest tracks and with data contained in individualstorm reports in the belief that these data representthe closest available approximation to the truth.These best tracks are determined after the fact fromcareful post analyses of all pertinent data. They arenot entirely independent data since the best tracksthemselves are determined partly from the satelliteestimates. However, the effect of this dependencywas minimized by evaluating satellite classificationsonly during periods when aerial reconnaissance andother data were also available. For a more detaileddiscussion of this question see Gaby et al. (1975).2. Evaluations by year Formal classifications of all tropical and subtropical Atlantic cyclones began at Miami duringthe 1971 hurricane season, with an SAS staff of twosatellite meteorologists. Visible imagery from the588 MONTHLY WEATHER REVIEW VOLUME 1081048 NUMBER OF CASES73 148 162 110 53 80I .f9.6 .. 19.~ . iS.0 'AVERAGEStandard deviationAverage algebraic difference7.1 AVERAGE 7.34~ no wind estimate m~e 'Average absolute difference 1971 1972 1973 1974 19175 i 1976 1977 1978 YEARS FIG. 1. Miami SFSS kropical and subtropical cyclone maximum sustained wind speed"accuracy" based on average difference between satellite estimates and NHC best tracks(NHC value minus SFSS value).geostationary Advanced Technology Satellite No. 3(ATS-3) and from the Automatic Picture Transmission(APT) polar-orbiting satellite ESSA-8 were used toprovide two classifications per day in support ofNHC operations. During the 1971 season the technique developed by Timchalk et al. (1965) was usedmost of the time with occasional use of the newlydeveloped technique of Dvorak (unpublished at thattime). As expected, the ATS-3 pictures provedsuperior for this purpose. Results were much betterfor systems with well-defined circulation centersthan for systems which had no obvious centers ofcirculation. Interestingly, the average difference in'MWS was -6 kt, based on 77 cases, a figure thathas varied only slightly over the years (see, Fig. 1).The average difference in location was about 36 nmi, based on 96 cases2 In 1972 classifications were made using theDvorak (1972) technique exclusively. The new technique proved superior, especially for the weakersystems, and added considerable objectivity to thework. All but two classifications were made fromATS-3 visible images. The remaining two were madefrom ESSA-8 APT imagery. The season providedonly 48 cases. These gave an average location dif ~ Details are contained in a report by D. C. Gaby--Performance evaluations for the 1971 hurricane season. UnpublishedInternal Memo. SAS, NHC, Miami, November 1971.ference of 33 n mi with standard deviation of 17 n mi.In 1972, storm intensity was given in terms of minimum central pressure, rather than MWS, showing anaverage minimum pressure difference of 3.8 mb withstandard deviation of 2.7 mb. There was an increasein staff of one meteorologist late in the season as theNESS formed the Miami SFSS and the former SASwas dissolved.2 In 1973 classifications were again made using theDvorak (1973). technique with minor local improvements. Almost all classifications were made fromATS-3 visible imagery. The average location difference was 25.8 n mi with standard deviation of 19.2 nmi, based on 87 cases. Storm intensity was againgiven in terms of minimum central pressure (averagepressure difference of 4.7 mb with standard deviation of 3.8 mb) and MWS. The average absolute windspeed difference was 6.4 kt based on 73 cases,a The 1974 season marked a turning point in ouroperation with the launch of the geosta, tionary Synchronous Meteorological Satellite No. 1 (SMS-I), ~ Details are contained in a report by D. C. Gaby--Performance evaluations for the 1972 hurricane season. Presentedat the' NOAA/NWS Hurricane Warning Service EvaluationConference, Miami, November 1972. a Details are contained in a report by D. C. Gaby--MiamiSFSS tropical cyclone classifications for the Atlantic Ocean for1973. Presented at the Interdepartmental Hurricane WarningConference, Miami, January 1974.MAY 1980 GABY, LUSHINE, MAYFIELD, PEARCE AND TORRES 589prototype of the new Geostationary OperationalEnvironmental Satellite (GOES) series. Additionally, our staff increased to provide 24 hours per daycoverage. Most classifications were made with theSMS-1 satellite which was located above the equator at 45-W longitude for the GATE experiment. Afew were made while SMS-1 was moving westwardlate in the hurricane season, and for storms overthe extreme western Caribbean Sea or the Gulf ofMexico a few were made using ATS-3 pictures. Visible spectrum (0.55-0.70 /am) pictures with 4 kmresolution were used during daylight and 8 km resolution infrared (10.5-12.6/am) pictures were used atnight. The average location difference was 18.4 n miwith standard deviation of 15.5 n mi, based on 313cases. The average absolute MWS difference was8.4 kt, and the average algebraic MWS differencewas -5.2 kt with standard deviation of 9.6 kt, basedon 148 cases. Details are given in Gaby et al. (1975). The 1975 season saw several significant ~mprovements in our operational techniques and a more complete evaluation of performance was made (Gabyet al., 1976). Among improvements were the following: 1) The GOES-East satellite was located permanently at 75-W longitude and 2 km resolutionvisible imagery was used, for the first time, in making daytime classifications. 2) The Hebert-poteat (1975) technique for theclassification of subtropical cyclones was used successfully, especially with Hurricane Doris whichevolved smoothly from a subtropical to a tropicalcyclone.4 3) Two studies by Pike (1974a,b) allowed relaxingthe developmental constraints imposed by the Dvorak(1973) technique for better operational application ofthe MWS estimates at the NHC. 4) Pike also provided a nomogram for correctingthe apparent position of a cyclone as viewed obliquelyfrom geostationary altitude above 75-W.~ An evaluation of the classifications for 1975 showedan average location difference of 16.5 n mi withstandard deviation of 11.2 n m, based on 115 cases.The average absolute MWS difference was 5.9 kt andthe average algebraic MWS difference was -2.3 ktwith standard deviation of 6.0 kt, based on 162 cases.An updated version of the Dvorak technique wasavailable (Dvorak, 1975). Comparisons among thefour meteorologists doing the bulk of the classifications showed no significant differences betweenindividuals, indicating that the .Dvorak technique 4 An improved version of the Hebert-Poteat technique is unpublished but available from the senior author at the NHC. 5 During Hurricane Eloise the location correction at landfallwas -5 n mi in northwest Florida; even higher values could havebeen expected farther west on the Gulf coast or at higherlatitudes in the Atlantic.provides objective, consistently reliable results evenwhen used by meteorologists with widely differingexperience levels. The Miami SFSS meteorologistshave always used a confidence factor in estimatingboth tropical cyclone location and intensity?Statistical evaluations showed that the higher confidence factors were indeed associated with the moreaccurate estimates.7 The time of imagery was shownto have some influence on the estimates of locationand MWS. Somewhat better results were obtainedusing the high-resolution visible imagery duringdaylight hours. In 1975 we approached a plateau inour ability to estimate the location and intensity oftropical cyclones given the present state of the artand technology. An evaluation of performance during the 1976 season gave results similar to those for 1975. Detailsare contained in Gaby et al. (1977). The averagelocation difference was 17.0 n mi with standarddeviation of 14.3 n mi, based on 115 cases. Significantly better results were obtained for storms ofhurricane intensity or greater. The average absoluteMWS difference was 7.7 kt, and the average algebraicMWS difference was -3.9 kt with standard deviation of 9.1 kt, based on 110~cases. Less constrainedrates of weakening after a study by Lushine (1977)were used routinely. Comparisons were again madeamong the meteorologists doing the bulk of theclassifications and no significant difference-wasfound between individuals. Confidence factors wereagain found to be well-correlated with the qualityof the estimates provided. High-resolution visibleimagery again made possible better estimates oflocation by day, but no significant difference inwind speed estimates was found as a function oftime of day. No evaluation of performance for the 1977 hurricane season has been published although some of theresults were presented at the 32nd InterdepartmentalHurricane Warnings Conference of January 1978.Also, no evaluation of performance for the 1978season has been formally presented. Appendices Aand B contain the statistical results of our evaluations of performance for 1977 and 1978, respectively.For 1977 the average location difference was 16.1 n 6 Confidence factors 1, 3, and 5, for location, refer to welldefined eye with certain picture registration, well-defined circulation center with certain picture registration, and poorly definedcirculation center with certain picture registration, respectively.Confidence factors 2, 4 and 6 are similar and refer to uncertainpicture registration; however, these occur so seldom that theywere not evaluated. Confidence factors 1, 2 and 3, for intensity,refer to the meteorologist being completely certain as to currentintensity number used, tempted to vary up or down by - T orST number, or might vary up or down by 1 T or ST number, ormore, respectively. ? There is probably some feedback into the system as theresult of the hurricane specialist drawing his best track moreclosely to satellite location estimates of higher confidence.590 MONTHLY WEATHER REVIEW VOLUME I08 TABLE 1. Comparisons of Miami 1978 SFSS classification Tnumbers as made from visible and enhanced infrared images forthe same times (-+30 rain). T-numbers T-n~mbers ~ T-numbersT-numbers Number ofSTORN equal but < 1/3 differ bydiffer by Comparisons different 1/2 > 1/2COR& 2 1 2 1 6ELL& 8 0 3 1 12G*~EYA 2 4 0 0 6TOTALS 47 24 26 5 102mi with standard deviation of 15.1 n mi, based on85 cases. Significantly better storm locations wereobtained during daylight hours (1830 GMT) usinghigh-resolution visible imagery than during nighttime hours (0630 GMT) using only infraredimagery? Higher confidence values related tobetter storm locations. The average absolute MWSdifference was 9.9 kt, and the average algebraicMWS difference was -9.2 kt with standard deviation of 6.2 kt, based on 53, cases. In 1977 a new technique for estimating storm intensity from enhancedinfrared (EiR) satellite imagery (Dvorak, unpublished at that time) was used experimentally. Nos!gnificant difference in tl~e accuracy of the MWSestimates was shown as a function of .time of day.Higher confidence values related to better windspeed estimates. In 1978 the location difference was 16.9 n miwith standard deviation of 17..1 n mi, based on 167cases. Significantly better storm locations were obtained during ~daylight hour~ using high-resolutionvisible imagery than during nighttime hours usingonly infrared imagery. The higher confidence valuesrelated to better storm locations. The average absolute MWS difference was 7.1 kt, and the averagealgebraic MWS difference was -2.9 kt with stand s These 1830 and 0630 GMT location estimates were alwaysby daylight or during the night, respectively, regardless ofstorm location. A similar statement is not always true for the 1230and 0030 GMT location estimates because-the storms may be indaylight or darkness depending 'upon their longitude. For example, storms in far western longitutles are typically in darkness at 1230 GMT but well illuminated at 0030 GMT, while theopposite is true for storms in far eastern longitudes.ard deviation of 9.0 kt, based on 80 cases. In1978 the technique for estimating storms fromEIR imagery was used on a full-time basis? Nosignificant difference in the accuracy of the MWSestimates was shown as a function of time of day. Table 1 shows a comparison of Dvorak T numbersfor 1978 storms using both visible (VIS) and enhanced infrared (EIR) imagery received at Miami.Whenever possible, classifications were made usingboth types of imagery. Note that 46% of theclassifications made with both VIS and EIR imageryat the same time gave exactly the same T number(MWS), and that 95% were no more than - Tnumber different. Thus, the classifications werenearly equally acci~rate whether by day or by night.This equally good performance regardless of thetime of day is attributed to the improved objectivity of the latest Dvorak technique.3. Summary of eight years of classifications There are a number of factors known to limit location accuracy, both for individual fixes and ascompared with the smoothed best tracks. An accepted rule for the limit to which any feature maybe located in an image constructed from 'a sequence of scan lines is approximately three timesthe "resolution" or width of the scan line. Thisresolution is degraded with increased distance fromthe satellite subpoint. Thus using 8 km resolution(at nadir) infrared images permits a location accuracy for individual fixes of about 24-30 km(16-18 n mi),'and using 2 km resolution (at nadir)visible images permits a location accuracy for individual fixes of -6-8 km (4-5 n mi) across theAtlantic hurricane belt. At very high latitudes theaccuracy would be less. The quality of availablegeographic grids is also a factor, but these arebecoming accurate to within 1-2 scan line widths.How well the storm circulation center is defined bythe cloud pattern is a factor, with poorly definedsystems being more difficult to locate, especiallywith only infrared imagery at night. Finally, comparisons of individual fixes with a smoothed besttrack often include an inherent lower limit of accuracy because of the small-scale oscillations bfthe eye or storm circulation center within the largerenvelope of the storm circulation. A rather pronounced trochoidal motion of tl)e eye has been observed in a number of storms and has been documented by radar in Hurricanes Carla 1961 (Anonymous, 1961) (see Fig. 2) and Dora 1964 (Anonymous,1964). It was first, observed in 24 hours per daysatellite observations with Tropical Storm Alma 9 This improved version of Dvorak's technique, "Tropicalcyclone classification procedures using visible and enhancedinfrared satellite imagery," is unpublished.M~.- 1980 GABY, LUSHINE, MAYFIELD, PEARCE AND TORRES 591~RO~/NSVILLE+ FIG. 2. The track of the eye of Hurricane Carla 1961 as observed by coastal radars. (Adapted from Weatherwise, October 1961.) The heavy line simulates a smoothed track displaced from the detailed eye track for clarity.1974. Lawrence and Mayfield (1977) determinedsuch a trochoidal path for the eye of Hurricane Belle 1976. Our experience and considerationof the above factors indicate that we are not likelyto achieve a better average overall location accuracythan -13-15 n mi compared to the smoothed besttracks with the present state of .the a~l and technology. This is illustrated in Fig. 3. For. the pastseveral years we have shown an average locationdifference (compared to the best tracks) of 1617 n mi, which is very close to the best that mightbe expected. Comparisons of daytime fixes usinghigh-resolution visible imagery to nighttime fixesusing low-resolution (8 km at nadir) infrared imageryhave shown that the daytime fixes are an average6-10 n mi more accurate than those at night. Animprovement in the resolution of the IR imageryfrom geostationary satellites to 2 km should providean improvement in location accuracy on average of-3-5 n mi. Hopefully, such an improvement willbe achieved at least in part with the series ofgeostationary satellites to be launched in the late1980's. Fig. 1 shows our ability to estimate accurately theMWS in tropical or subtropical cyclones. One maynote a rather consistent performance over the years.The average algebraic difference in maximum sustained wind speed compared to the NHC's besttracks has been about -5 kt and has never exceeded10 kt in any year. Although there has been littlesignificant difference from year to year in the average accuracy of the MWS estimate~, there hasbeen a real and significant improvement in the objectivity with Which the classifications are madeand a considerable improvement in our ability toestimate accurately the MWS at night. Indeed,most Miami satellite meteorologists now prefer theDvorak technique for the intensity estimates oftropical cyclones from EIR imagery even during thedaylight hours when high-resolution visible imageryis also available. From an operational point ofview, it is significant that those hurricane advisoriesissued in early morning and therefore based inpart on IR imagery will have the benefit of satellite MWS estimates just as accurate as by day.We believe it is also significant for operationsthat our ability to estimate accurately the MWS isappreciably better for storms of at least hurricane592 VOLUME 108uJUJ403O MONTHLY WEATHER REVIEW NUMBER OF CASES96 48 87 313 115 115 85 167~33(17) 18405.5) '2OI0 1971 1972 - 1973 1974 1975 1976 1977 1978 YEARS FIG. 3. Miami SFSS tropical and subtropical cyclone location "accuracy" based onaverage difference between satellite fixes and NHC best tracks. Figures in parentheses arestandard deviation (none for 1971)..intensity (65 .kt), i.e., for those storms potentiallymost damaging (Gaby et al., 1977).4. Conclusions An ability to locate tropical or subtropical cyclones with an overall average accuracy of -17 n mias compared to the NHC's smoothed best trackshas been demonstrated. Little overall improvementshould be expected in the near future. Some improvement may be expected with the higher resolution infrared imagery planned for the geostationarysatellites in the late 1980's. An ability to estimate the maximum sustainedwind speed in tropical or subtropical cyclones withan average accuracy of <10 kt has been demonstrated. This is probably close to the best thatmay be expected considering the natural variabilityof these storms. Continued improvement may beexpected in the objectivity of the techniques usedand these techniques may even be partially automated in time. Acknowledgments. Many of the classificationsconsidered in this report, especially during theearly years, were made by Kenneth O. Poteat, nowretired. Others were made' by Arthur C. Pike andDonald C. Cochran, now with the NESS SatelliteField Services Station, Honolulu, Hawaii. A fewwere made by Paul M. Duernberger, NOAA CorpsOfficer, on temporary duty assignment. We thankCharles J. Neumann, Chief, Research and Development, NHC, and John R. Hope and Paul J. Hebert,hurricane specialists, NHC, for their reviews of theinitial draft and many valuable suggestions. Ourthanks go also' to Teresa Barker for typing themanuscript and tables.APPENDIX AStatistical Summary of 1977 Performance Evaluations TABLE 2. Miami 1977 SFSS satellite vortex locations compared with NHC best tracks (for periods with aerial reconnaissance only); minimum central pressure below 1000 mb,Current Intensity (CI) number 2.0 or higher, or Subtropical (ST)number 1.5 or higher.STOPS4 AVENGE STANDARD ~ANGE 6F NLg~ER OF DIFFERENCE DEVIATION DIFFERENCE CASES (n.mi.) (n.mi.) (n.mi.)ANITA 12.0 1 h - h 0-55 31BABE 23.8 16.8 5-51 9CLARA 17.~ 16.6 3-62 2~DOROTHY 23.6 21.1 0-60 12EVELYN 9.0 3.8 5-1~ 9FRIEDA * 11.1 2.5 8-1" 6ALL STORM~ 16.1 15.1 0-62 85 # Did not reach T2.0 or < 1000 mb (not included in all stoma).MAY 1980 GABY, LUSHINE, MAYFIELD, PEARCE AND TORRES 593 TABLE 3. Comparisons of Miami 1977 SFSS satellite vortexlocation differences (compared to NHC best tracks) as a function of time of imagery. Note: time 0630 GMT representsinfrared imagery, and time 1830 GMT represents visible imagery.Times 0030 or 1230 GMT may represent either infrared or visibleimagery, depending on storm longitude.PICTURE TIME AVERAGE STANDARD RANGE OF NUMBER OF(GMT) * DIFFERENCE DEVIATION DIFFERENCE CASES (n.ml.) (n.ml,) (n.mi.)0030 20.0 18.6 2-62 220630 18.3 17.4 0-53 191230 18.2 19.6 0-60 171830 10.9 8.4 2-37 20 TABLE 7. Comparisons of Miami 1977 SFSS confidence factorswith satellite maximum sustained wind speed differences (compared to NHC best tracks). Lower confidence factor numbersmean higher confidence. (knots) (knots) (knots) (knots)1 8.6 - 7.4 5.0 -22 to +9 22 TABLE 4, Comparisons of Miami 1977 SFSS confidence factorswith satellite vortex location differences (compared to NHCbest tracks). Low confidence factor numbers mean higherconfidence. (n.mi.) (n.mi.) (n.mi,)1 5.0 1.8 2-8 103 13.5 11,9 0-60 475 28.2 19,8 0-62 23 TABLE 5. Miami 1977 SFSS estimated maximum sustainedwind speeds minus NHC best track data. Comparisons are forperiods with aerial reconnaissance only; minimum centralpressure below 1000 rob, CI number 2.0 or higher, ST number 1.5or higher. Standard deviations are based on ~dgebraic averagedifference.STORE AVERAGE ABSOLUTE AVERAGE ALGEBRAIC STANDARD RANGE OF NUMBER OF DIFFERENCE DIFFERENCE DEVIATION DIFFERENCECASES (knots) (knots) (knots) (knots)ANITA 10.8 -10,8 7.1 -22 to 0 21~ABE 2.7 - 0.3 4.9 -10 to +5 7CLARA 9.2 - 9.2 6.6 -20 to 0 13DOROTI~ 17.9 -17.9 4,5 -25 to -10 8EVELYN 4.0 + 0.5 6.2 - 6 to +9 4FRIEDA * 1.3 - 1.3 2.2 - 5 to 0 6ALL STORMS 9.9 - 9.2 6.2 -25 to+9 53 * Did not reach T2.0 or ~ 1000 mb (not included In all storms). TABLE 6. Comparison of Miami 1977 SFS.S maximum sustained wind speed estimate differences (relative to NHC besttracks) as a function of time of imagery.(knots) (knots) (knots) (knots)0030 8.7 -8.1 9.1 -20 to +4 130630 11.3 -9.8 9.0 -25 to+9 121230 8,8 -8.8 7.6 -20 Go 0 I01830 9.5 -8.5 9,0 -20 to +5 11 * Picture times as sho~m or + 30 minutes. APPENDIX BStatistical Summary of 1978 Performance Evaluations TABLE 8. Miami 1978 SFSS satellite vortex locations compared with NHC best tracks (for periods with aerial reconnaissance only); minimum central pressure below 1000 mb, CInumber 2.0 or higher or ST number 1.5 or higher.STORM AVERAGE STANDARD RANGE OF NUMBER OF DIFFERENCE DEVIATION DIFFERENCE CASES (n.mI.) (n.mi,) (n.ml.)AMELIA 44.3 ...... 1BESS 13.4 8.4 3 to 33 12CORA 19.8 13.5 6 to 53 14DESRA 12.6 4.0 9 tO 18 4ELLA 11.8 8,9 0 to &l 43FLOSSIE 32.3 33,5 0 to 113 18GRETA 14.4 15,0 0 to 59 53HOPE 23.1 0.4 0 2IRMA * 10.3 6.5 0 to 22 9JULIET 14.0 11.1 3 to 38 8KE~DRA 23.3 1~.0 5 to 51 12ALL STORMS 16.9 17.1 0 to 113 167 ~ No aerial reconnaissance available (not included in all stoma). TABLE 9. Comparisons of Miami 1978 SFSS satellite vortexlocation differences (compared to NHC best tracks) as a function of time of imagery. Note: time 0630 GMT representsinfrared imagery, and time 1830 GMT represents visible imagery.Times 0030 or 1230 GMT may represent either infrared or visibleimagery, depending on storm longitude.(n.ml.) (n.mt.) (n.mi.)0030 17.1 13.7 0 to 46 380630 24.8 31.4 0 to 113 241230 18.9 16.5 2 to 54 311830 14.6 12.7 0 to 50 45 * Picture time as shown or ~ 30 minutes.594 MONTHLY WEATHER REVIEW VOLUME 108 T^BL- 10. Comparisons of Miami 1978 SFSS confidence factorswith satellite vortex location differences (compared to N. HC besttracks). Low confidence factor numbers mean higher confidence.CONFIDENCE AVERAGE STANDARD RANGE OF NUMBER OFFACTOR DIFFERENCE DEVIATION DIFFERENCE CASES (n.mi.) (n.mi.) (n.mi.)1 7.8 6.7 0 to 29 ~93 12.3 li.O 2 to 59 h95 27.9 20.8 0 to 113 65 T^BLE 11. Miami 1978 SFSS estimated maximum sustainedwind speeds minus NHC best track data. Comparisons are forperiods with aerial reconnaissance only; minimum central pressure below 1000 mb, CI number 2.0 or higher, ST number 1.5 orhigher. Standard deviations are based on algebraic averagedifference.ARE~AGE ABSOLUTE AVERAGE ALGEBRAIC STANDARDRANGE OF NUMBER OFDIFFERENCE DIFFERENCE DEVIATION DIFFERENCE CASES(k~ots) (knots) (knots) (knots)BESS 5.7 - 5.7 3.3 -10 to 0 6~O~A 2.0 + 0.3 3.2 - 5 to +6DEBEA 8.0 - 8.0 2.8 ,-10 to -6 21~L055I~ 0.6 + 0.6 1.7 0 to +5 9GRI~A 9.0 + 0.1 11.4 -20 to +23 2&~ * 3.0 - 3.0' 4.5 -10 to 0 5JOI, IET 9.8 - 9.8 3.3 -12 to -5I~RA 6.0 - 5.4 5.7 -15 yo +2 7 7.1 -2.9 9.0 -20 to +23 80aerial reconnaissance available (not included in all stor~s). TABLE 12. ComparisOn of Miami 1978 SFSS maximum sustained wind speedestimate differences (relative to NHC best tracks) as a function of time ofimagery.PICTURE TIME AVERAGE ABSOLUTE A~ERAGE ALGEBRAIC STANDARD RANGE OF 'NUMBER OF (GMT) # DIFFERENCE DIFFERENCE DEVIATION DIFFERENCE CASES (knots) (knots) knots) (knots)0030 8.2 - 2.4 10.8 -20 to +23 180630 7.3 - 4.9 8.57 -20 to +10 131230 6.8 - 5.4 7.3 -18 to +10 161830 6.7 - 3.0 8.4 -18 to +15 23 i Picture times as show~ or ~ 30 minutes. ~ TABLE 13. Comparisons of Miami 1978 SFSS confidence factors with satellitemaximum sustained speed differences (compared to NHC best tracks). Lowerconfidence factor numbers mean higher confidence.CONFIDENCE AVERAGE ABSOLUTE AVERAGE ALGEBRAIC STANDARD RANGE OF NUMBER NU~tBERS DIFFERENCE DIFFERENCE DEVIATION DIFFERENCE CASES (knots) (knots) (knots) (knots)1 7.4 - 1.1 9.9 -18 to +19 272 6.7 - 3.7 8.5 -20 to +23 523 10.0 -10.0 ....... 1REFERENCESAnonymous, 1961: Hurricane Carla. Weatherwise, 14, 192 196. , 1964: Hurricane Dora. Preliminary Report, U.S. Weather Bureau.Dvorak, V. F., 1972: A technique for the analysis and forecasting of tropical cyclone intensities from satellite pictures. NOAA Tech. Memo. NESS 36, National Environ mental Satellite Service, NOAA, Washington, DC, 15 pp. , 1973: A technique for the analysis and forecasting of tropical cyclone intensities from satellite pictures. NOAA Tech. Memo. NESS 4.5 (Revision of NOAA TM NESS 36), National Environmental Satellite Service, NOAA, 19 pp.--, 1975: Tropical cyclone intensity analysis and forecasting from satellite imagery. Mon. Wea. Rev., 103,. 420-430.Gaby, D. C., D. R. Cochran, J. B. Lushine, S. C. Pearce, A. C. Pike and K. O. Poteat, 1975: Atlantic tropical cyclone classifications for 1974. NOAA Tech. Memo. NESS 68, Na tional Environmental Satellite Service, NOAA, Washington, DC, 6 pp.--, J. B. Lushine, B. M. Mayfield, S. C.. Pearce and .K.O.M^- 1980 GABY, LUSHINE, MAYFIELD, PEARCE AND TORRES 595 Poteat, 1976: Atlantic tropical and subtropical cyclone classifications for 1975. NOAA Tech. Memo. NESS 75, Na tional Environmental Satellite Service, NOAA, Washington, DC, 14 pp. .... and F. E. Torres, 1977: Atlantic tropical and subtropical cyclone classifications for 1976. NOAA Tech. Memo. NESS 87.~ National En vironmental Satellite Service, NOAA, Washington, DC, 13 pp.Hebert, P. J., and K. O. Poteat, 1975: A s~atellite classification technique for subtropical cyclones. NOAA Tech. Memo. NWS SR-83, Fort Worth, TX, 23 pp.Lawrence, M. B., and B. M. Mayfield, 1977: Satellite ob servations oftrochoidal motion during Hurricane Belle 1976. Mort. Wea. Rev., 105, 1458-1461.Lushine, J. B., 1977: A relationship between weakening of tropical cyclone cloud patterns and lessening of wind speed. NOAA Tech. Memo. NESS 85, National Environ mental Satellite Service, NOAA, Washington, DC, 12 pp.Neumann, C. J., 1975a: The effect of initial data uncertainties on the performance of statistical tropical cyclone predic tion models. NOAA Tech. Memo. NWS SR-81, National Weather Service, NOAA, Fort Worth, TX, 15 pp.Neumann, C. J., 1975b: A statistical study of tropical cyclone positioning errors with economic applications. NOAA Tech. Memo. NWS SR-82, National Weather Service, NOAA, Fort Worth, TX, 21 pp.Pike, A. C., 1974a: Suggested limits on short term increases in Atlantic tropical cyclone T numbers based on statistics of observed storm intensification. Satellite Field Services Station, National Environmental Satellite Service, NOAA, Miami, FL, 3 pp. (unpublished).---, 1974b: Application of suggested limits on short-term T number increase to the Florida Keys hurricane of September 1935. Satellite Field Services Station, National Environmental Satellite Service, NOAA, Miami, FL, 1 pp. (unpublished).Timchalk, A., L. Hubert and S. Fritz, 1965: Wind speeds from TIROS pictures of storms in the tropics. Meteorological Satellite Laboratory Rep. No. 33, U.S. Weather Bureau, Washington, DC, 33 pp.