Baixe Matematiacs financieras e outras Esquemas em PDF para Direito falimentar, somente na Docsity!
THE DVORAK TROPICAL CYCLONE
INTENSITY ESTIMATION TECHNIQUE
A Satellite-Based Method that Has Endured for over 30 Years
This insight, which expresses itself by what is called
Imagination, is a very high sort of seeing, which does not
come by study, but by the intellect being where and what
it sees, by sharing the path, or circuit of things through
forms, and so making them translucid to others.
—Ralph Waldo Emerson (1803–82)
T
HE DVORAK TECHNIQUE LEGACY.
The Dvorak tropical cyclone (TC) intensity estima-
tion technique has been the primary method of monitoring
tropical systems for more than three decades. The technique has likely
saved tens of thousands of lives in regions where over one billion people are
directly affected by TCs (commonly called hurricanes, typhoons, or cyclones).
The Dvorak technique’s practical appeal and demonstrated
skill in the face of tremendous dynamic complexity
Color-enhanced IR image of Hurricane Katrina, viewed from GOES-12 on 28 August 2006
BY C HRISTOPHER VELDEN , B RUCE HARPER , FRANK WELLS , J OHN L. B EVEN II, R AY Z EHR ,
TIMOTHY O LANDER , M AX M AYFIELD, C HARLES “C HIP ” G UARD, M ARK L ANDER ,
ROGER E DSON , L IXION AVILA , A NDREW B URTON , M IKE TURK , A KIHIRO K IKUCHI ,
A DAM C HRISTIAN , P HILIPPE C AROFF,
AND PAUL M CC RONE
AMERICAN METEOROLOGICAL SOCIETY SEPTEMBER 2006 |^1195
place it among the great meteorological innovations of our time. It is difficult to think of any other me- teorological technique that has withstood the test of time and had the same life-saving impact. The Dvorak technique has been more than a criti- cal analysis and forecast- ing tool. It has become the most important input to our highly valuable present-day TC archives. Historical TC best-track (post processed for archives) datasets are the cornerstone for the es- timation of risks from TCs in regions without routine aircraft reconnaissance, with applications in engineering, climate change assessments, insurance, and other fields. The future evolution of the Dvorak and similar satel- lite-based TC intensity es- timation methods is of vital interest to the meteorological and coastal communities, and the continued im- provement should be a top research priority in the atmospheric sciences. We examine the development of the Dvorak technique, review its basic assumptions, and relate them to the technique’s success. We also identify some limitations and common misapplications of the technique, and briefly discuss selected regional modifications and enhancements.
Laying the foundation. By the late 1960s, polar orbit- ing satellites with visible and limited IR capabilities were providing TC forecasters with coarse-resolution imagery several times a day. At this time there were
no enhancement or animation capabilities. Early work by Fett (1966), Fritz et al. (1966), and Hubert and Timchalk (1969) was generally unsuccessful in inferring TC intensity from this type of imagery. The National Oceanic and Atmospheric Administration (NOAA), National Hurri- c a n e C e nt e r ( N HC ; i n Miami, Florida), and the Joint Typhoon Warning Center (JTWC; until recently located on Guam) were primarily us- ing satellite imagery for TC positioning and directing weather reconnaissance air- craft to developing convective areas. No reliable intensity es- timation techniques existed. As the number of satellites increased and their capabili- ties improved (Table 1), it became clear that the science of deploying remote sensing in space was outpacing the ability of meteorologists to apply it. From his Washington, D.C., office in the Synoptic Analysis Branch of the Environmental Science Services Administration (the precursor to NOAA), scientist Vernon Dvorak developed his cloud pattern recognition technique based on a revolution- ary conceptual model of TC development and decay. Dvorak and his colleagues derived an empirical meth- od relating TC cloud structures to storm intensity using a simple numerical index [the current intensity (CI)], corresponding to an estimate of the maximum sustained (surface) wind (MSW), as shown in Table 2. The earliest internal NOAA reference to this work is Dvorak (1972), followed by an update (Dvorak 1973). Dvorak worked in an operational environment, and
AFFILIATIONS : VELDEN AND O LANDER —CIMSS, University of Wisconsin—Madison, Madison, Wisconsin; H ARPER —Systems Engineering Australia, Brisbane, Australia; WELLS —Santa Rita, Guam; B EVEN , AVILA , AND M AYFIELD —NOAA Tropical Prediction Center, Miami, Florida; Z EHR —NOAA/NESDIS/RAMM, Fort Col- lins, Colorado; G UARD AND E DSON —NOAA/NWSFO, Tiyan, Guam; L ANDER —University of Guam, Mangilao, Guam; B URTON —Bureau of Meteorology, Perth, Australia; TURK—NOAA/NESDIS Satellite Analysis Branch, Washington, D.C.; C AROFF —Meteo-France, RSMC-La Reunion, Reunion Island, France; K IKUCHI —RMSC-To- kyo Typhoon Center, Japan Meteorological Agency, Tokyo, Japan; C HRISTIAN —Joint Typhoon Warning Center, Pearl Harbor, Hawaii; M C C RONE —Air Force Weather Agency, Omaha, Nebraska
A supplement to this article is available online (DOI:10.1175/BAMS- 87-9-Velden) CORRESPONDING AUTHOR : Christopher Velden, CIMSS, University of Wisconsin—Madison, 1225 W. Dayton St., Madison, WI 53706. E-mail: chrisv@ssec.wisc.edu The abstract for this article can be found in this issue, following the table of contents. DOI:10.1175/BAMS-87-9- In final form 19 May 2006 ©2006 American Meteorological Society
Vernon Dvorak (circa: late 1970s).
1196 | SEPTEMBER 2006
The basic steps in the Dvorak technique can be summarized as follows (Fig. 1):
- Determine the TC center location.
- Make two quasi-independent estimates of the intensity of the TC.
- Choose the best intensity estimate.
- Apply selected rules to determine the final esti- mate of intensity.
Following Fig. 1, the Dvorak technique analyst first assigns T-numbers “tropical” (“T”) numbers (hereafter Tnum) and relates these to storm intensity (Dvorak 1975, 1984). One Tnum unit represents a typical one- day intensity change based on climatological data (see Fig. 2 for examples related to cloud patterns). Figure 3 illustrates where the primary patterns are typically assigned in relation to Tnum and TC in- tensity ranges. While the Tnum is generally a good first guess at the TC intensity, Dvorak observed that convection in some weakening TCs degenerates faster than the corresponding MSW. Thus, the Tnum does not always relate directly to TC intensity. Instead the Tnum is converted into the CI number. For developing or steady-state storms, the Tnum and CI are usually identical or close. A standard table is used to convert CI to MSW, and a wind–pressure relationship is then used to assign the corresponding estimated minimum sea level pressure (MSLP; Table 2). Very weak, pregen-
esis tropical disturbances are assigned T1.0 (Tnum = 1.0). Minimal tropical storm intensity (MSW of 35 kt) is T2.5. Minimal hurricane intensity (65 kt) is T4.0, T5.0 = 90 kt, T6.0 = 115 kt, and T7.0 = 140 kt. The rare T8.0 = 170 kt is the top of the scale. The Tnum and CI approach helps alleviate the problem of unreasonable intensity assignments due to poor-quality images or unrepresentative image analy- sis. The Tnum also is an effective tool for normalizing intensity change according to the current intensity of a storm. For example, a 5-kt increase of maximum wind speed from 30 to 35 kt is an equivalent change in terms of Tnum to a 15-kt increase from 140 to 155 kt. This normalization helps improve the evaluation of environmental forcing on intensity change.
The evolution of the technique. The Dvorak technique evolved significantly during the 1970s and 1980s, and has continued to be modified by regional centers since then. Originally the technique was largely reliant on pattern-matching concepts and the application of Dvorak’s development/decay model. Later revisions (Dvorak 1982, 1984) shifted the emphasis toward measurement of cloud features.
TABLE 2. Summary of the Dvorak (1984) Atlantic and WestPac wind–pressure relationships.
CI MSW (kt) (^) MSLP (hPa)Atlantic MSLP (hPa)WestPac
1.0 25 1.5 25 2.0 30 1009 1000 2.5 35 1005 997 3.0 45 1000 991 3.5 55 994 984 4.0 65 987 976 4.5 77 979 966 5.0 90 970 954 5.5 102 960 941 6.0 115 948 927 6.5 127 935 914 7.0 140 921 898 7.5 155 906 879 8.0 170 890 858
F IG. 1. The basic steps in the Dvorak technique.
1198 | SEPTEMBER 2006
By basing the technique on observed 24-h changes in cloud pattern and intensity, Dvorak (1975) ad- dressed the problem of short-term changes in cloud structure (i.e., diurnal cycles) that might be unrep- resentative of true intensity change. Images 24 h apart are used to determine if the TC has developed, weakened, or retained intensity. The observed trend in cloud features is then applied to the Dvorak model of development and decay to obtain an estimate of intensity. In Dvorak (1984) this was more rigorously quantified into the mode-expected T number (MET). Although the development/decay model remains integral to the Dvorak technique, the significant revisions of 1982 and 1984 shifted the emphasis of the technique toward direct measurement of cloud features. Dvorak (1984) states, “when the measure- ment (of cloud features) is clear-cut giving an inten- sity estimate that falls within prescribed limits, it is used as the final intensity.” If this measurement is not clear-cut, the analyst then relies on the development/ decay model in conjunction with pattern matching. Several innovations allowed the shift toward greater reliance on cloud feature measurement. First, the introduction of cloud pattern types such as the curved band (CB) and shear patterns allowed analy- sis of TCs without an eye or central dense overcast (CDO). The CB pattern has become the most widely used pattern type for TCs below hurricane strength. Second, IR imagery was applied for the first time. This brought about yet another pattern type—the embedded center (EMBC), or the IR equivalent to the
CDO. Third, this revision created the enhanced IR (EIR) eye pattern, in which TC intensity is related to the cold cloud-top IR temperatures surrounding the center and the warm IR temperatures in the eye. This is the most objective of all Dvorak measurements and has led to attempts to automate the inten- sity analyses (Zehr 1989; Velden et al. 1998). All versions of the Dvorak technique have featured a system of rules, or constraints. Many of these rules/constraints have changed somewhat unsystem- atically between versions. Others have evolved based on verification studies and practical application. For example, the original tech- nique had no set criteria for when a tropical disturbance should be classifiable. Criteria were subsequently introduced in Dvorak (1975) and more rigorously quantified in Dvorak (1984). Perhaps most controversial are the constraints on allowable intensity change over specific periods of 24 h or less. The technique has always had such limits. However, Dvorak (1984) quantified maximum allowable 6-, 12-, 18-, and 24-h changes in Tnum, with the maximum allowable 24-h change of 2.5 Tnums. These constraints usually work well, but experience has shown that rapidly intensifying TCs can change Tnum by 3.0 or more in 24 h, and systems in strong shear or moving over colder sea surfaces can weaken
F IG. 2. Examples of characteristic cloud patterns of developing TCs (from Dvorak 1973).
F IG. 3. Primary Dvorak cloud patterns in relation to Tnum and TC intensity ranges that they are typically assigned.
AMERICAN METEOROLOGICAL SOCIETY SEPTEMBER 2006 |^1199
data from polar-orbiting satellites. Both WestPac and Atlantic datasets were considered, with the WestPac again being the larger set. Like Erickson, Sheets and Grieman strongly preferred the use of MSLP rather than MSW as a measure of storm intensity because of the variability in wind speed measure- ment techniques used for validation (e.g., sea state, radar, and inertial systems), differences in experience and application between aircraft crews, and also the inherent variability in MSW measurements due to convective-scale influences. Accordingly, only cen- tral pressures were considered in the analyses with the following conclusions about the Dvorak (1973) CI–MSLP relationship:
- In the Atlantic, there was a clear tendency (bias) to overestimate the intensity by 5–10 hPa.
- In the WestPac, strong storms (< 920 hPa) underes- timated by as much as 20 hPa, but the overall result was very close to the assumed curve (no bias).
The next significant contribution was a separate wind–pressure relationship for the WestPac based on Atkinson and Holliday (1975, 1977), whose 28-yr dataset was then considered the most significant climatological review of WestPac TC intensity. The decision to adopt the Atkinson and Holliday (AH) wind–pressure curve for operational application in the WestPac was apparently based on revised Dvorak technique validation studies by Lubeck and Shewchuk (1980) and Shewchuk and Weir (1980). They used 396 cases during 1978–79, covering the full range of intensities. The reference best-track dataset included
subjective sources such as the AH relationship (which, because it was used operationally by the JTWC to derive surface winds since 1974, influenced best-track data). The reports conclude that the mean absolute intensity error was less than one CI number and that the developing TC stages were more accurately estimated than the weakening stages. The final recommendation of Shewchuk and Weir (1980) was to replace the Dvorak (1975) wind–pressure relationship by AH for the WestPac (Fig. 5). This occurred in 1982 (Dvorak 1982, 1984). The original relationship was retained for the Atlantic. The final relationships are presented in Table 2. Significantly, no further changes were made to the Dvorak (1975) MSW–CI relationship, although Dvorak (1984) for the first time notes rather candidly that the archive tracks themselves may now have become biased by the applica- tion of the technique itself, especially in the WestPac. The importance of defining and validating the wind–pressure relationships is exemplified by the operational impact of the Dvorak intensity estimates on the JTWC warnings in the WestPac during the 1970s and 1980s (Fig. 6; Guard et al. 1992). Over a 16-yr period from 1972 to 1987, dedicated aircraft reconnaissance gradually declined, and was roughly balanced by increased operational reliance on satel- lite reconnaissance, thanks to the Dvorak intensity estimates. During this time the aircraft reconnais- sance was available to provide an invaluable measure of ground truth. The aircraft reconnaissance pro- gram, provided by the 54th Weather Reconnaissance Squadron based at Andersen AFB, Guam, was deacti- vated in the summer of 1987. This put added pressure on the robustness of the Dvorak technique in the
F IG. 5. The evolution of the adopted wind–pressure relationships (from Harper 2002). Arrows indicate trend over time.
FIG. 6. Reconnaissance platforms used for JTWC warn- ings in the WestPac (from Guard et al. 1992).
AMERICAN METEOROLOGICAL SOCIETY SEPTEMBER 2006 |^1201
WestPac. While the satellite technique development at JTWC helped offset the elimination of aerial recon- naissance and accelerate the exploitation of remotely sensed data, it should be emphasized that the satel- lite-derived estimates were (or are) not as accurate as aircraft in situ measurements. Atlantic basin recon- naissance information remains a vital and necessary component of the U.S. TC warning system. Additional validation studies were carried out in the 1980s. In a first comparison, Gaby et al. (1980) found that the average difference between satellite- derived and aerial reconnaissance-based best-track maximum sustained wind speeds was approximately 7 kt. During the late 1980s, in response to the with- drawal of reconnaissance aircraft from the western North Pacific, Sheets and Mayfield studied the ac- curacy of geostationary satellite data, while Guard studied polar orbiting satellite data (both summa- rized in OFCM 1988). The authors evaluated internal consistency and absolute accuracy of the Dvorak technique, using independent analysts to provide the sample of estimates on past TCs. Their findings on the accuracy of the technique were similar to those of Gaby et al., and their internal consistency results showed that 85% of the independent common fixes were within 0.5 Tnum. In summary, the basic principles of the Dvorak technique evolved during 15 yr of active experi- mentation (1969–84) and were empirically derived from several hundred WestPac and Atlantic basin TC cases. During this period, improved satellite im- agery, proficiency in its use, and, to a lesser extent, greater aircraft accuracy in measuring TC winds (for empirical development) all contributed to enhancing the technique’s effectiveness. Meanwhile, several significant verification studies contributed to refin- ing the method.
Limitations of the Dvorak technique. The Dvorak tech- nique does not directly measure wind, pressure, or any other quantity associated with TC intensity. It infers them from cloud patterns and features. This primary limitation leads to two basic sources of er- ror. First, the technique is physically restricted due to natural variability between the remotely sensed cloud patterns and the observed wind speed. Second, the method is subject to analyst interpretation and/or misapplication (which has motivated the develop- ment of an objective version, that is addressed in a subsequent section). Perhaps the most limiting factor is the reliance on IR images [when the visible (VIS) is not available] in which cirrus can obscure TC organization. Often the
central dense overcast as presented in the IR will cover a weak eye and/or developing eyewall structure. This can lead to underestimates of the true TC intensity. The Dvorak embedded center scene type attempts to recognize this condition, but is difficult to apply and imprecise due to uncertainties in locating the exact center. Furthermore, concentric eyes and associated eyewall replacement cycles (Willoughby et al. 1982), which have been recently linked to intensity change, can also be obscured in the IR, and were not part of the original Dvorak model. The potential application of microwave imagery to addressing these issues is discussed in a later section. A common limitation in applying the technique using geostationary satellite imagery involves the scan angle, or the viewing angle from the satel- lite subpoint. At large scan angles, TCs with small eyes can be underestimated using the Dvorak EIR method because the eye and attending warm bright- ness temperature is partially or fully obscured by the eyewall. A good example was Atlantic Hurricane Hugo in 1989. The subjective Dvorak classification at 1800 UTC 15 September estimated MSW of 115 kt and a minimum pressure of 948 hPa (i.e., CI 6.0). Data from reconnaissance aircraft indicated 140 kt
F IG. 7. (top) A schematic illustration of a commonly observed flow pattern: the low-level circulation in the monsoon trough during the northern summer in the WestPac. The “C” symbols indicate cyclonic gyres, and the “G” symbol shows a likely place for TC genesis within the trough. (bottom) Satellite IR image example from August 2001.
1202 | SEPTEMBER 2006
or published results. A brief summary of some of the local modifications applied to the basic Dvorak technique by regional TC analysis centers can be found online (http://dx.doi.org/10.1175/BAMS-87- 9-Velden).
The period of automation. Dvorak (1984) describes an initial attempt to create an objective approach to his method. The intensity (Tnum defined to the nearest 0.1) for an eye pattern is assigned according to two IR temperature measurements, with an approach analo- gous to the EIR technique. The only measurements needed are the warmest IR pixel within the eye and the warmest IR pixels on a prescribed TC-centric ring (55-km radius in the original method). The warmer the eye and the colder the surrounding eyewall tem- perature, the more intense the TC estimate will be. The typical ranges of the two temperatures and their sensitivity to the intensity estimate are quite differ- ent. For example, as the surrounding temperature decreases from –64° to –75°C, the intensity increases 1.0 Tnum, while an eye temperature increase from –45° to +15°C is needed for a 1.0-Tnum increase. Several modifications to this method have im- proved the intensity estimation results (Zehr 1989; Dvorak 1995):
- The 55-km-radius ring can be well inside the coldest IR ring (eyewall convection) of TCs with large eyes. The method was modified to compute an average IR temperature for a range of ring sizes (R = ~25–125 km) and uses the coldest.
- In many situations, estimates fluctuate widely over a relatively short time, primarily due to localized or semidiurnal convective f lareups. Averaging the computations (over 3–12 h) can produce more realistic intensity trends.
- The original Dvorak (1984) digital IR table was amended to cover anomalous “cold” eyes; cases with no warmer “eye” pixels in the TC central overcast. In these events, the eye temperature is set equal to, or can even be colder than, the sur- rounding central overcast ring temperature.
Evaluation of the original Dvorak digital method showed that it did not perform as well prior to IR eye formation. In the late 1980s, Zehr (1989) developed an objective technique using enhanced IR satellite data. This digital Dvorak (DD) method laid the foundation for the more advanced algorithms of today. The primary motivation for developing an auto- mated, objective intensity estimation scheme was to lessen subjectively introduced estimate variability
due to analyst judgment from Dvorak (1984). The most prominent subjectivity involves cloud pattern (scene) typing. In the 1990s, additional incentive for automation was the increased availability of higher- resolution, real-time global digital satellite data, and improved computer processing resources capable of furnishing sufficient analysis capabilities. This led to the development of the objective Dvorak technique (ODT; Velden et al. 1998), which began with a care- ful assessment of the DD algorithm. It was found that the DD performance was satisfactory only for well-organized TCs of minimum hurricane/typhoon or greater intensity. Reasonably accurate intensity estimates were possible when the storm possessed an eye structure. Eventually, through a procedure involving a Fourier transform analysis of the center and surrounding cloud-top regions in the IR imagery, the ODT incorporated the four primary Dvorak scene types: eye, central dense overcast, embedded center, and shear. By using these four scene-type designa- tions, a proper branch in the basic Dvorak logic tree could be followed to more accurately, and objectively estimate TC intensity. Eventually, a history file containing previous intensity estimates and analysis parameters was implemented for subsequent image interrogations by the ODT algorithm. A time-averaged Tnum replaced the traditional Tnum, removing much of the fictitious short-term intensity variability. In addition, specific Dvorak (1984) rules, such as the rule controlling the weakening rate of a TC after maximum intensity, were implemented to more closely follow the govern- ing principles. Statistically, the ODT was shown to be competitive with TC intensity estimate accuracies obtained with the subjective technique at operational forecast centers such as the Satellite Analysis Branch (SAB), the Tropi- cal Analysis and Forecast Branch (TAFB), and the U.S. Air Force Weather Agency (AFWA; Velden et al. 1998). These statistics were only valid for Atlantic basin TCs where aircraft reconnaissance MSLP measurements were available. The ODT was tuned for WestPac TCs using cases in the 1980s when aircraft validation was available. With selected threshold adjustments, the method performs reasonably well. The original goal of the ODT was to achieve the accuracy of the subjective Dvorak (1984) EIR method using computer-based, objective methodology. This goal was accomplished, however, with important limitations. The ODT could only be applied to storms at or greater than minimal hurricane/typhoon strength (storms meeting EIR criteria). Also, the ODT still required manual selection of the storm center.
1204 | SEPTEMBER 2006
Thus, development continued and an advanced objective Dvorak technique (AODT; Olander et al.
- emerged. The Dvorak (1984) curved band analysis is the primary tool for prehurricane/typhoon intensity TCs. The CB relates TC intensity to the amount of curved cloud banding surrounding the storm cen- ter. This amount is measured using a 10°log spiral (manually rotated). Defining the cloud field region over which the spiral is placed is quite subjective, but after discussions with numerous TC forecasters, and considerable trial and error, initial skepticism was overcome and an objective scheme was incorporated into the AODT. The final remaining subjective element was the manual determination/positioning of the TC center location. This proved to be the most challenging aspect of the AODT transition. A method was devel- oped to utilize a short-term track forecast (provided by NHC or JTWC) as a first guess for the storm center location. Then objective center-determination schemes search for curvature patterns and strong, localized gradients in the image brightness tem- perature (BT) field surrounding the interpolated forecast position (Wimmers and Velden 2004). Such BT gradient fields are typically associated with TC eyes, but can also be applied to EMBC and some CB scene types. If the objective center-estimation scheme locates a region that exceeds empirically determined thresholds, the region’s center is used as the AODT storm center location. On average, the AODT inten- sity estimates produced using the automated center location routine are only slightly worse than those obtained using manual storm center placements (Olander et al. 2004; Olander and Velden 2006). Additional Dvorak (1984) rules were incorporated into the AODT, including the constraint on TC inten- sity estimate growth/decay rate over set time periods. This modification reduced the averaging period for the AODT intensity calculation from 12 to 6 h. Another recent addition followed the discovery by Kossin and Velden (2004) of a latitude-dependent bias in the Dvorak estimates of MSLP. This bias is related to the slope of the tropopause (and corresponding cloud-top temperatures) with latitude. With the introduction of a bias adjustment into the AODT, the MSLP estimate errors were reduced (Olander et al. 2004). Interestingly, Kossin and Velden (2004) found that no such latitude-dependent bias exists in the Dvorak-estimated MSW. The most recent version of the objective algorithm progression is the advanced Dvorak technique (ADT). Unlike ODT and AODT, which attempt to mimic the
subjective technique, the ADT research has focused on revising digital IR thresholds and rules, and extending the method beyond the original application and con- straints (Olander and Velden 2006). The ADT, which has its heritage in Dvorak (1984), Zehr (1989), Dvorak (1995), Velden et al. (1998), and Olander et al. (2004), is fully automated for real-time analysis.
Today. As a testament to its success, the Dvorak technique continues to be used today at TC warn- ing centers worldwide. In addition to the regional TC analysis centers mentioned earlier, other centers such as those in Fiji, India, and the Central Pacific Hurricane Center in Hawaii, also employ the method as their primary TC intensity analysis tool. Even at the TPC in Miami where reconnaissance aircraft obser- vations are often available, the technique continues to be the chief method for estimating the intensity of TCs when aircraft data are not available, including most storms in the eastern Pacific, and TCs in the Atlantic east of ~55°W. Given the global applications, and the local modi- fications to advance the technique, it is informative to ask how accurate the current Dvorak estimates are. Brown and Franklin (2004) took a fresh look at the technique's accuracy. Figure 9 shows the error frequency distribution of Dvorak MSW estimates compared to intensities derived from reconnais- sance-based best-track data for Atlantic TCs between 1997 and 2003. Half of the errors were 5 kt or less, 75% were 12 kt or less (0.5 Tnums at tropical storm intensities), and 90% were 18 kt or less. Thus, the Dvorak-estimated MSW in the Atlantic basin are on the whole quite good, although Brown and Franklin (2002, 2004) mention occasional large outliers do exist. These recent studies along with those mentioned earlier attest to the consistency of the technique over its 30-yr life span, but remind us that the technique is far from perfect, and still suffers from its limitations. Hurricane Charley (2004) in the Atlantic was a good example of a storm whose intensity was significantly underestimated due to an eyewall replacement cycle and contraction of the eye to dimensions below the viewing capability of the Geostationary Operational Environmental Satellite (GOES). Cases like this compel Tropical Prediction Center (TPC) forecast- ers to rely heavily on aircraft reconnaissance data in landfalling TC situations. In TC basins outside of the Atlantic, evaluation of the Dvorak technique performance is hindered some- what by the lack of in situ validation. Organized field campaigns in these basins would provide a significant
AMERICAN METEOROLOGICAL SOCIETY SEPTEMBER 2006 |^1205
niques take advantage of the tropospheric profiling capability of the AMSU to depict TC warm cores, and statistically relate these measurements through hydrostatic assumptions to intensity. AMSU-based methods currently perform on par with the Dvorak technique, and could in time supplant it. However, microwave instruments are currently aboard low-earth- orbiting satellites that have limited data availability and timeliness. Thus, the Dvorak technique, which uses abundantly available geostationary satellite im- agery, will likely be em- ployed well into the twenty- first century. Ultimately, the optimal approach to satellite-based TC monitoring will likely be a consensus algorithm that exploits the advantages of each individual tech- nique, whether VIS/IR or MI based (Fig. 11). Initial
attempts at such an algorithm are showing great promise (Velden et al. 2004). As an example, TC MSLP estimates from the AODT and the AMSU tech- niques have been weighted by their situational perfor- mance into a consensus estimate for a large sample of TC cases (Herndon and Velden 2006). Preliminary results show that the weighted consensus is superior in performance to either of its individual elements. It is anticipated that by adding new technique members to the consensus, the accuracies will further improve. It also seems clear that any efforts to modernize the Dvorak technique approach should attempt to retain the basics of the method and be used in combination with the microwave methods. This consensus ap- proach should improve satellite-based TC intensity estimates, and also make possible reliable analyses of the entire TC surface wind field.
SUMMARY. For the past three decades, Vernon Dvorak’s practical insights and tools for estimating TC intensity from satellite data have proven to be invaluable in forecast applications. Despite the in- herent limitations to an empirical method, and the opportunities for misapplication, the Dvorak tech- nique remains the most widely applied TC intensity estimation method in the world. A U.S. Air Force (1974) report identifies a key to the longevity of the technique: “The [Dvorak] model will provide reliable estimates with data of poor quality, with conflicting evidence, inexperience on the part of the analyst, and
F IG. 10. Example of concentric eyewalls in SSM /I imagery for Typhoon Dianmu (2004). (Courtesy of Naval Research Lab TC Web site.)
F IG. 11. Conceptual model of a satellite-based TC intensity analysis algorithm based on multispectral observations (Velden et al. 2004).
AMERICAN METEOROLOGICAL SOCIETY SEPTEMBER 2006 |^1207
variations in satellite camera system.” However, as identified in that report, a fundamental issue remains for the science to address: “There has been no satisfac- tory theoretical basis developed to explain intensity changes predicted by the [Dvorak] model, or depar- tures from expected changes which are observed in rapidly developing or weakening storms.” The Dvorak method has also been an extremely important tool for the development of our highly valuable TC archives. The increasing demand for greater certainty of environmental risk from TCs has pushed the commissioning of critical reviews of the possibility for systematic bias in the historical TC datasets (e.g., Harper 2002). These reviews show a potentially contentious emerging issue: the historical datasets have unavoidably inherited any biases that are implicit in the Dvorak technique. Those respon- sible for national data archives should make every effort to ensure that all possible storm parameters are documented and retained for future reanalysis (e.g., Dvorak T and CI numbers and landfall data). The World Meteorological Organization (WMO) has addressed this need in the western North Pacific basin with the development of the Extended Best Track Database [compiled and maintained at the Regional Specialized Meteorological Center (RSMC), Tokyo, Japan]. The practical appeal and demonstrated skill in the face of tremendous dynamic complexity place the Dvorak technique for estimating TC intensity from satellites amongst the greatest meteorological innovations of our time. Empirical techniques such as the Dvorak method intrinsically rely on new generations of analysts as champions to ensure they continue to improve. Dvorak’s greatest gift may have been to give us time to continue providing reasonably consistent daily products to operations while we try to understand the underlying physics of tropical cyclone intensity. The future of the Dvorak method is of vital interest to the meteorological community, and its continued evolution ranks as a global priority in tropical weather analysis and forecasting.
ACKNOWLEDGMENTS. The inspiration for this paper derives from the inventor of the technique itself, Vern Dvorak. The authors’ gratitude on behalf of the meteoro- logical community and TC-prone populations cannot be overstated. There is little question that the achievements of Mr. Dvorak have paved the way for advancements in applications of remote sensing to tropical cyclone research and forecasting. We would also like to acknowledge the many researchers and forecasters whom have contributed to the evolution and advancement of the Dvorak technique
over the years. Finally, the U.S. Air Force and the NOAA Aircraft Operations Center have provided crucial TC air- craft reconnaissance data from which the Dvorak method development and evaluation has benefited. The “Storm Trackers” and “Hurricane Hunters” of the USAF Weather Reconnaissance Squadrons C-130s and NOAA WP-3s are to be commended not only for providing invaluable in situ data for operational forecasting, but also ground truth measurements for remote sensing applications. The views, opinions, and findings contained in this report are those of the author(s) and should not be construed as an official National Oceanic and Atmospheric Administration or U.S. Government position, policy, or decision.
REFERENCES Atkinson, G. D., and C. R. Holliday, 1975: Tropical cyclone minimum sea level pressure–maximum sus- tained wind relationship for western North Pacific. FLEWEACEN Tech. Note JTWC 75-1, U.S. Fleet Weather Central, Guam, 20 pp. — , and — , 1977: Tropical cyclone minimum sea level pressure/maximum sustained wind relation- ship for the western North Pacific. Mon. Wea. Rev., 105, 421–427. Bankert, R. L., and P. M. Tag, 2002: An automated method to estimate tropical cyclone intensity using SSM/I imagery. J. Appl. Meteor., 41, 461–472. Brown, D. B., and J. L. Franklin, 2002: Accuracy of pressure–wind relationships and Dvorak satellite inten- sity estimates for tropical cyclones determined from re- cent reconnaissance-based “best track” data. Preprints, 25th Conf. on Hurricanes and Tropical Meteorology, San Diego, CA, Amer. Meteor. Soc., 458–459. — , and — , 2004: Dvorak TC wind speed biases determined from reconnaissance-based best track data (1997–2003). Preprints, 26th Conf. on Hurri- canes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 86–87. Brueske, K. F., and C. S. Velden, 2003: Satellite-based tropical cyclone intensity estimation using the NOAA–KLM series Advanced Microwave Sounding Unit (AMSU). Mon. Wea. Rev., 131, 687–697. Cocks, S. B., I. Johnson, R. Edson, M. Lander, and C. P. Guard, 1999: Techniques for incorporating SSM/I imagery into Dvorak tropical cyclone intensity Estimates. Preprints, 23rd Conf. on Hurricanes and Tropical Meteorology, Dallas, TX, Amer. Meteor. Soc., 584–587. Demuth, J. L., M. DeMaria, J. A. Knaff, and T. H. Vonder Haar, 2004: Validation of an Advanced Microwave Sounding Unit tropical cyclone intensity and size es- timation algorithm. J. Appl. Meteor., 43, 282–296.
(^1208) | SEPTEMBER 2006
—, —, and J. P. Kossin, 2004: The Advanced Objec- tive Dvorak Technique (AODT)—Latest upgrades and future directions. Preprints, 26th Conf. on Hur- ricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 294–295. Sheets, R. C., and P. G. Grieman, 1975: An evaluation of the accuracy of tropical cyclone intensities and locations determined from satellite pictures. NOAA Tech. Memo. ERL WMPO-20, 36 pp. Shewchuk, J. D., and R. C. Weir, 1980: An evaluation of the Dvorak technique for estimating TC intensi- ties from satellite imagery. JTWC Tech. Note 80-2, JTWC, Pearl Harbor, HI, 16 pp. Spencer, R., and W. D. Braswell, 2001: Atlantic TC monitoring with AMSU-A: Estimation of maxi- mum sustained wind speeds. Mon. Wea. Rev., 129, 1518–1532. U.S. Air Force, 1974: Tropical cyclone position and intensity analysis using satellite data. Tech. Rep. 1WWP 105-10, Department of the Air Force, Head- quarters First Weather Wing (MAC), APO San Francisco, CA, 88 pp.
Velden, C. S., T. Olander, and R. M. Zehr, 1998: Devel- opment of an objective scheme to estimate tropical cyclone intensity from digital geostationary satellite imagery. Wea. Forecasting, 13, 172–186. —, and Coauthors, 2004: Toward an objective satel- lite-based algorithm to provide real-time estimates of TC intensity using integrated multispectral (IR and MW) observations. Preprints, 26th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 280–281. Willoughby, H. E., J. A. Clos, and M. G. Shoreibah, 1982: Concentric eye walls, secondary wind maxima, and the evolution of the hurricane vortex. J. Atmos. Sci., 39, 395–411. Wimmers, A., and C. Velden, 2004: Satellite-based center-fixing of TCs: New automated approaches. Preprints, 26th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. Meteor. Soc., 82–83. Zehr, R., 1989: Improved objective satellite estimates of tropical cyclone intensity. Preprints, 18th Conf. on Hurricanes and Tropical Meteorology, San Diego, CA, Amer. Meteor. Soc., J25–J28.
1210 | SEPTEMBER 2006
U P P L E M E N T
This document is a supplement to “The Dvorak Tropical Cyclone Intensity Estimation Technique: A Satellite-Based Method That Has Endured for over 30 Years,” by Christopher Velden, Bruce Harper, Frank Wells, John L. Beven II, Ray Zehr,Timothy Olander, Max Mayfield, Charles “Chip” Guard, Mark Lander, Roger Edson, Lixion Avila, Andrew Burton, Mike Turk, Akihiro Kikuchi, Adam Christian, Philippe Caroff, and Paul McCrone ( Bull. Amer. Meteor. Soc., 87, 1195–1210) • ©2006 American Meteorological Society
- Corresponding author: Christopher Velden, UW-CIMSS, 1225 W. Dayton St., Madison, WI 53706. • E-mail: chrisv@ssec.wisc. edu • DOI: 10.1175/BAMS-87-9-Velden
THE DVORAK TROPICAL CYCLONE
INTENSITY ESTIMATION TECHNIQUE
A Satellite-Based Method that Has Endured for over 30 Years
BY C HRISTOPHER VELDEN , B RUCE H ARPER , F RANK WELLS , J OHN L. B EVEN II, R AY Z EHR , TIMOTHY O LANDER ,
M AX M AYFIELD, C HARLES “C HIP ” G UARD, M ARK L ANDER , ROGER E DSON , L IXION AVILA , A NDREW B URTON ,
M IKE TURK , A KIHIRO K IKUCHI , A DAM C HRISTIAN , P HILIPPE C AROFF, AND PAUL M C C RONE
T
he success of the Dvorak technique has been enhanced by numerous local modifications in the last quarter-century. We will now describe many of these modifications from around the world. Each of the three Australian Tropical Cyclone Warning Centers (TCWC) uses a different wind– pressure relationship. While they recognize there is little scientific justification for this, the absence of aircraft reconnaissance data has made it difficult to resolve the differences. The Perth and Darwin TCWCs also use the pressure–wind relationships in their Δp form. The changes in ambient pressure from one tropical cyclone (TC) to the next, and the changes that occur as a TC moves into higher latitudes are seen as small but significant sources of variance that can be explicitly accounted for by this simple modification. A number of Australian TC forecasters have come to the conclusion that the Embedded Center (EMBC) scene type temperature ranges generally give higher- than-warranted data tropical numbers (Tnums). In the absence of reconnaissance data this perception has arisen from noting discontinuities in Tnums as
the scene type changes, and the lack of agreement between the Tnum derived from EMBC cloud tem- perature measurements and that obtained by applying the Dvorak development/decay model. Given that the Dvorak technique is otherwise noted for its internal consistency, this has lead forecasters to speculate on possible reasons for the poorer performance of the EMBC scene type in the Southern Hemisphere. Two factors relating to tropopause temperatures have been identified. Tropical cyclones tend to occur at lower latitudes (higher tropopause) in the Southern Hemi- sphere, leading to colder cloud-top temperatures. Additionally, the Southern Hemisphere Tropics have colder warm season tropopause temperatures than the Northern Hemisphere Tropics, particularly in the Australian region (Kossin and Velden 2004). In re- sponse to this, Australian forecasters are prepared to weight the final estimate toward the model-expected T number (MET) when using the EMBC pattern (Burton 2005). At the Japanese Meteorological Agency (JMA) Regional Specialized Meteorological Center (RSMC) in Tokyo, the original Dvorak (1982, 1984) relation-
S6 | SEPTEMBER 2006
identifiable center but insufficient convection for a Dvorak classification. During extratropical transition phases, classifica- tion using the Dvorak method can yield unrepresenta- tive intensity estimates. A JTWC rule of thumb is to compare the Dvorak technique intensity to the extra- tropical (XT) technique intensity as the TC transitions to a baroclinic low. At some point during the transition the two values will be the same, usually near T3.5, and from that point on the system is classified using the XT technique (Miller and Lander 1997). Satellite analysts in the Tropical Analysis and Forecast Branch (TAFB) of the TPC in Miami have been using the Dvorak technique for over 30 years and continue to mostly adhere to the basic constraints originally developed by Dvorak in 1984 for both developing and weakening systems. Two of the more significant modifications to the original constraints are as follows: 1) A study by Lushine (1977) limits the rate of weakening of a system by holding the CI num- ber up to the highest Tnum attained during the past 12 h but never more than 1 Tnum above the current Tnum. 2) A more recent study by Brown and Frank- lin (2002) suggests that this rule does not weaken systems fast enough. They suggest that this rule be applied for a period of only 6 h. This modification also allows the Tnum to change by up to 1.0 Tnum over 6 h, 1.5 Tnums over 12 h, 2.0 Tnums over 18 h, and up to 2.5 Tnums over 24 h. A follow-up study by Brown and Franklin (2004) further suggests that an intensity based on the average of the Tnum and CI also reduces the bias during weakening systems. These modifications are especially applicable to sys- tems that weaken rapidly, such as are common in the east Pacific basin. The SAB in Washington, D.C., has derived the position and intensity for tropical disturbances in all basins for over 30 years. The SAB employs the Brown and Franklin (2002) modifications in all basins on a case-by-case basis. Even with these adjustments there are still anomalous TC cases exhibited by intensity change at steep rates. In such cases, it is left to the discretion of the analyst as to whether the situation warrants breaking all constraints in an effort to ar- rive at the current intensity. It is important to note, however, that the use of these local rules may at times be the result of a previous erroneous estimate. For example, if the estimated intensity 6 h previous to the current analysis was held the same when it should have been increased (only confirmed after postanalysis), the next analyst may need to employ local rules (or break all constraints) to accurately as- sess the current intensity.
At the U.S. Air Force Weather Agency (AFWA) in Omaha, Nebraska, the primary concern in the Meteorological Satellite (METSAT) Applications Branch is the use of the Dvorak (1984) constraint system to modify TC intensity estimates (a rather consistent theme among global tropical analysis cen- ters). During cases with a clear sign of rapid intensity change, AFWA analysts often temporarily suspend the implementation of the Dvorak (1984) constraints. The constraints appear to be particularly problematic when a TC intensifies from the T2.5 (weak storm stage) to T4.0 (initial hurricane intensity). As with other TC analysis centers, rapidly weakening sys- tems also present a significant challenge to AFWA analysts. After initial trials from 1977 to 1979, the Dvorak technique was officially adopted at the RSMC- La Reunion in 1981. The wind–pressure relation- ship associated with the Dvorak intensity estimates chosen for the southwestern Indian Ocean (SWIO) basin was the one originally designed for the West- Pac. However, as the 10-min-averaged wind was adopted at La Reunion, a conversion factor of 0. was applied to the original MSW scale, while a gust factor of 1.5 was applied to the MSW to estimate peak gusts. These conversion factors were later reconsidered based on a small sample of observed SWIO TC winds by synoptic reports at or near landfall. Starting with the 1999–2000 TC season in the SWIO, these conversion factors were modified to 0.88 (this conversion factor is now more or less being used by all agencies in the Southern Hemisphere) and 1.41, respectively. These latest modifications lead to a ~10% increase in the average MSW. This issue of wind-averaging periods and related con- version factors is still a question in the SIO region, and an important matter directly connected to the regional use of the Dvorak technique. As for MSLP, La Reunion analysts recently began to take into account the size of the TC and adjust the Dvorak- estimated MSLP, raising it for small systems, similar to the Australian approach. Procedures at La Reunion also include a modifica- tion for small systems at strong intensity: the inertia lag time of 12 h before lowering the winds may be reduced to 6 h. Also, during weakening phases as- sociated with extratropical transitioning TCs, it has also been recognized that Dvorak estimates are not appropriate. The regional applications and modifications to the basic Dvorak methods are perhaps a precursor to an evolutionary transition of satellite-based TC analyses from a single technique to a multispec-
S8 | SEPTEMBER 2006
tral approach. Observations from passive microwave sensors often reveal more accurate positions than just visual or infrared imagery alone. At AFWA, for ex- ample, the approach has been to take advantage of this multispectral information and incorporate it into the Dvorak intensity estimation process. Experience has shown that if an eye feature is seen in microwave, the pattern Tnum can be selected to reflect that occurrence, in lieu of the data Tnum, to indicate a greater intensity. On a broader scale, it must be noted that while Dvorak intensity estimates remain an important input to ar- chived best-track analyses, other satellite-based sources are having an increasing influence. The optimal fusion of all-available and emerging satellite observations as a direction for TC analysis and intensity estimation is discussed further in the section on looking ahead.
REFERENCES Burton, A. D., 2005: Notes on the application of the Dvorak technique. Meteorological Note 225, Bureau of Meteorology, National Meteorological Library, 8 pp. Koba, H., S. Osano, T. Hagiwara, S. Akashi, and T. Kikuchi, 1989: Determination of intensity of typhoons passing through the Philippine Islands (in Japanese). J. Meteor. Res., 41,^ 157–162. —, T. Hagiwara, S. Osano, and S. Akashi, 1990: Relationship between the CI-number and central pressure and maxi- mum wind speed in typhoons (in Japanese). J. Meteor. Res., 42,^ 59–67. Lushine, J., 1977: A relationship between weakening of TC cloud patterns and lessening of wind speed. NOAA Tech. Memo NESS 85, 12 pp.
AMERICAN METEOROLOGICAL SOCIETY SEPTEMBER 2006 |^ S
