Principle Investigators :

Submitted to the Office of the Federal Coordinator of Meteorological Services as a report under the National Post-Storm Data Acquisition Plan
17 August 1998

    The NOAA Paka Data Acquisition Team (NPDAT) of Arthur Chiu, Greg Forbes, and Sam Houston was assembled on Guam soon after Super Typhoon Paka's movement through the Marshall Islands in mid-December 1997. This was the first such group charged with gathering storm related data under the National Post-Storm Data Acquisition Plan (OFCM 1997). A full report of the findings and recommendations of this team is provided in NOAA (1998b). A primary concern for the team was to determine if the peak wind gust of 105 m s-1 (205 kt) reported at Andersen Air Force Base (AAFB) was reliable. The team also needed to determine whether or not the damage, which was greatest over the northern two-thirds of the island, was related to unusually extreme winds. The NPDAT spent the period from 22-26 December 1997 gathering all available wind data, while conducting ground and aerial surveys. In addition, an aerial survey was conducted over Rota, the island to the north of Guam. The wind data collected were digitized, and adjusted to a common framework for height, exposure, and averaging time (Powell et al. 1996; hereafter referred to as PHR). Then they were quality controlled and composited to produce surface wind analyses and swaths over Guam. The wind data gathered, the methodology, results, and some preliminary conclusions will be presented in this document.


    Paka's genesis occurred in the central Pacific Ocean during late November 1997, which is near the end of the official central Pacific hurricane season (NOAA 1998a). The tropical cyclone's track across the Pacific Ocean is shown in Fig. 1a. The strong El Nino Southern Oscillation (ENSO) event, which began earlier in 1997, was well underway at the time of Paka's formation. An equatorial westerly wind burst southwest of Hawaii led to the formation of twin tropical cyclones: Paka north of the equator and Pam to the south (JTWC 1997). The tropical depression which later became Paka formed on 28 November 1997 and was officially declared a tropical storm by the Central Pacific Hurricane Center (CPHC) at 1200 UTC 2 December. Pam became a southern hemisphere tropical cyclone on 6 December and moved toward the south and southeast before weakening to a tropical depression over the south Pacific Ocean on 10 December (personal communication, G. Padgett 1998). Paka crossed the International Dateline on 6 December and eventually strengthened to a tropical storm with 31 m s-1 (60 kt) sustained winds, before weakening on 8 December. The tropical storm moved into the Federated States of Micronesia on 11 December and strengthened to a typhoon. Typhoon Paka's winds increased to 59 m s-1 (115 kt) on 12 December, but later weakened as its forward motion increased. The typhoon raced westward at 8 m s-1 (16 kt) toward the southern Marshall Islands.
    A typhoon watch was issued by the NWSFO, Guam, at 2300 UTC 14 December for the Marshall Islands when Paka was over 1200 km (650 nm) east-southeast of Guam. The watch was upgraded to a typhoon warning for Guam, Rota, Tinian, and Saipan at 1530 UTC 15 December [at the same time that Paka was upgraded to a super typhoon with sustained winds of 72 m s-1 (140 kt)]. As Paka approached Guam and Rota, its sustained winds decreased to 64 m s-1 (125 kt) and its forward motion slowed to 5 m s-1 (9 kt). However, during its closest approach to Guam (Fig. 1b), the typhoon intensified to 67 m s-1 (130 kt), while its forward motion slowed to 3 m s-1 (6 kt). Paka continued to intensify after leaving the Marshall Islands and its winds reached maximum 1-min sustained values of 82 m s-1 (160 kt) on 18 December. Later the typhoon weakened rapidly and completely dissipated by 22 December.


    1. Anemometer observations

      Typhoon Paka's strong winds over the northern two-thirds of Guam resulted in significant damage and also caused the loss of most wind observation and recording equipment on the island [e.g., the Apra Harbor Handar, the National Ocean Service (NOS) Next Generation Tide gage (NGTG), etc.]. Other failures of wind instruments occurred when power sources failed during the typhoon. For example, wind-driven rain in Paka's outer eyewall caused the loss of power to the generator supplying power to the Andersen Air Force Base (AAFB) FMQ-13 "hot-film" wind measuring instrument and the Automated Surface Observing System (ASOS) located at the Guam NWSFO. The location of the equipment that recorded wind data on Guam during Paka is shown in Fig. 2.
      The NPDAT visited Guam during 22 - 26 December 1997 to collect all available meteorological data (primarily wind data). In addition to data collection, visits were made to all wind observing sites to document anemometer height, local terrain, exposure, proximity to buildings or vegetation, and other factors that might affect the wind measurements as a function of azimuth. The sites visited and the status of the wind observation records are listed in Table 1.
      Examples of some wind observation sites visited and documented are provided in Appendix B. Time-series of all available wind and pressure data are provided in Appendix C. The data from the various wind measurement sites were examined closely, because each location had unique characteristics and a variety of wind instruments. Some wind observations were compared with those made by official observing equipment that have known measurement standards, heights, averaging times, exposures (i.e., marine or over land) and local terrain effects. Two types of instruments, ASOS and NGTG, have provided high wind measurements in tropical cyclones at other locations in the past. For example, wind speed measurements in excess of 50 m s-1 (97 kt) were made before the power failed to the ASOS on St. Thomas, U.S. Virgin Islands in Hurricane Marilyn of 1995 (Powell and Houston 1998, hereafter referred to as PH98). Also National Ocean Service (NOS) Next Generation Tide Gages (NGTG) measured high winds in Hurricane Andrew's (1992) south Florida landfall (Powell and Houston 1996, hereafter referred to as PH96) and Hurricane Opal's (1995) landfall in north Florida described by PH98. The Guam ASOS and NGTG wind observation records were of relatively short duration and ended prior to the peak winds in Paka. They were compared with nearby stations that had longer records to determine how reliable the observations were until the time they failed. Figure 3 shows comparisons between the Guam ASOS and the NWSFO F-420 wind measuring equipment at the Guam International Airport. Also, the ASOS winds were compared with the wind record from an amateur station at Kuentos Communications, Inc. (KCI) in Fig. 3. A similar amateur wind measurement system measured the highest surface wind gust during Hurricane Andrew's south Florida landfall according to PHR. These comparisons indicated that both of these instruments had reliable wind observations while the ASOS was operating (the power failed to ASOS after 0753 UTC and the F-420 was destroyed after 0845 UTC). The KCI system had a complete wind record throughout the typhoon, although it suffered from approximately 2 h of significant blockage of the wind by an adjacent building. This limited the usefulness of the entire wind record at KCI, although the portion of the wind record from the south after 1200 UTC was considered reliable. The Guam University Handar site was slightly further away from the ASOS than the KCI site. The sustained wind speeds at the Guam University site were always lower than the ASOS winds. After the ASOS failed, the winds at this site were still low, even when the southerly wind exposure would have had an over water fetch after Paka's center passed northwest of the island. When this site was visited on 23 December, the anemometer cups were not rotating despite wind speeds estimated at 5-10 m s-1. The anemometer may have been failing on 16 December. For example, the gust factors at the Guam University were extremely large (ranging from 1.7 to over 3.0). The cups were possibly responding to brief strong wind gusts, but may have rotated so slow that the 1-min mean observations were underestimated. The gusts recorded at the Guam University site were considered reasonable values when compared with the other sites and an appropriate gust factor could be applied to estimate the maximum 1-min sustained wind from these wind gusts.
      The close proximity and marine exposure of the NOS NGTG and Handar sites located at Apra Harbor provided very useful comparisons (Fig. 4) during the onset of Paka's outer eyewall winds. The NOS NGTG failed after 1000 UTC and the Apra Harbor Handar instrument failed after 1250 UTC. The NGTG wind speeds were similar to those at the Handar site, even though the Handar site was located on a building 14 m above the ground with marine exposure from the north and northwest. The wind gusts from the Handar site were greater than those at the NGTG site. The Apra Harbor Handar site also had some spurious wind gusts that were too high to be considered real wind speeds (Guard 1998) prior to the destruction of the instrument after 1250 UTC (these recorded wind gusts may have been the result of spurious signals due to electronic noise as Paka's wind began to destroy the Handar instruments).
      Observations were also available from Rota's airport at a Handar instrument and a SAWRS site using an F-420 wind instrument. The observations from these two instruments were compared for the period that data were available (Fig.5). The SAWRS station at the Rota airport closed after 0850 UTC, so the primary surface data source during Paka's peak winds over Rota was the Handar site (the last observation available from this site was at 1200 UTC after the winds began to decrease). Note that the sustained winds and gusts recorded here were much weaker throughout the storm than those observed on the left side of the hurricane over Guam. This was likely due to the distance the island was from Paka's track (Fig.1b).

    2. WSR-88D reflectivities and Doppler winds

      The Guam WSR-88D was repaired and became operational on 15 December prior to the arrival of Paka's outer eyewall over extreme northern Guam (NOAA 1998a). The WSR-88D has the ability to store data as Level II or Level IV products (Crum et al. 1993). One of the best tropical cyclone data sets ever captured in digital form would likely have been available if the Level II data recorder had functioned properly during Paka's passage over Guam. However, this system failed completely and the power to the WSR-88D also failed after 0721 UTC. Fortunately, the Level IV images and data were available and recorded at 4 locations (AAFB, Guam NWSFO, JTWC, and CPHC). The JTWC provided these Level IV data to the NWS Operational Support Facility (OSF) and the NPDAT. The Tropical Prediction Center (TPC) provided the resources necessary for the NPDAT to access these data.
      The WSR-88D reflectivity data defined the concentric eyewall pattern that existed during Paka's approach to Guam and the outer eyewall's crossing of the island's northern coastline (see Appendix D). The track of the typhoon's center was also available from these data. The most important quantity available from the Guam WSR-88D Level IV data were the Doppler winds. Images of the high-resolution radial wind velocities were available at approximately 6 min intervals for the 1.5o tilt (high-resolution 0.5o tilt data are not available on Guam due to beam blockage from a tall residential building that was constructed after the WSR-88D became operational).
      Doppler velocities (VD) from the WSR-88D were not easily used in constructing two-dimensional surface wind fields. Two-dimensional surface wind fields in tropical cyclones can not be easily constructed using VD data from a single WSR-88D. Therefore, it was necessary to use a technique for single ground-based radar wind retrieval. Appendix D contains details about the Ground Based Velocity Track Display (GBVTD) used for producing two-dimensional surface wind fields based on the Guam WSR-88D VD.


    All wind observations that were representative of Super Typhoon Paka's winds over Guam and Rota, including data from anemometers and the Guam WSR-88D derived GBVTD winds, were assimilated using analysis techniques developed in the reconstruction of Hurricane Andrew's wind field (PHR; PH96). The surface wind observations were composited relative to the tropical cyclone's circulation center over a 3-6 h period of the storm's movement after being processed to conform to a common framework for height (10 m), exposure (marine or open-terrain over land), and averaging period (maximum 1-min sustained wind speed) using the methods of PHR. In recent years, these types of analyses have been made available to the forecasters at the National Hurricane Center (NHC) /Tropical Prediction Center (TPC) on an experimental basis, including the extremely active 1995 Atlantic basin hurricane season (PH98).
    All wind observation sites on Guam were visited and documented (for some examples, see Appendix B). The characteristics of each site were used to adjust the reliable wind observations to a common framework of height and time. The adjustment of the available surface wind observations to 10 m was difficult in some cases, because of the rugged terrain and sharp rise of the land from sea-level to an escarpment along many portions of Guam's coastline (e.g., JTWC's wind record). These local conditions probably result in unusual local wind effects, such as those that were experienced in tropical cyclones effecting Caribbean islands having sharp coastal relief and rugged topography (e.g., Hurricane Marilyn of 1995 in the U.S. Virgin Islands documented by PH98).
    After all of the surface wind data were adjusted to a common framework and quality controlled, they were processed in a nested, scale-controlled objective analysis package (Ooyama 1987; Franklin et al. 1993). The resulting product was a gridded wind field which was displayed as surface wind observations used in the analyses were considered to be epresentative of open-terrain over-land exposures at 10 m.
    Therefore, winds experienced at sea and along the immediate coastline of Guam may have been much higher. Likewise, winds in locations sheltered by topographic features would likely have been overestimated, while areas with "speed up" effects due to flow over topography or channelization of flow would likely have received higher 1-min sustained wind speeds.
    The analyses for Paka were centered at 0630 UTC, 0930 UTC, and 1230 UTC. The 0632 UTC GBVTD 10 m, maximum 1-min sustained wind field (Fig. 6) created using the methods described in Appendix D was used as a background field for the surface wind analysis. These surface wind analyses are shown as snapshots in Figs. 7 (0630 UTC), 8 (0930 UTC), and 9 (1230 UTC). These analyses were projected along the track in Fig. 1b to construct a swath of peak values of maximum 1-min sustained winds (Fig. 10). The winds were in excess of 50 m s-1 over much of the northern two-thirds of Guam and and values of up to 65 m s-1 occurred over the extreme northeastern tip of the island near AAFB. Winds of at least typhoon force (maximum 1-min sustained winds of at least 34 m s-1) occurred for over 8 h across some of the northern portions of the island. This would have most likely occurred in those areas which remained in the outer eyewall throughout Paka's closest approach to Guam. AAFB did experience some of the weaker winds associated with the "outer eye" according to Dominguez et al. (1998), so the duration of typhoon force winds here was less than 8 h. Most of the southern portion of Guam north of Merizo with open terrain over land exposure at 10 m had typhoon force winds for at least one hour. In reality, the sheltering effects of the varying topography here would like have limited the duration of typhoon force winds observed.


    Despite the loss of all official wind recording equipment on the northern portion of Guam (one amateur wind record was available throughout the event), wind fields were produced for Super Typhoon Paka's passage over the island. This was made possible by the use of GBVTD to produce wind fields based on the Guam WSR-88D Level IV VD's prior to the loss of power at the radar. The wind fields for open terrain over land exposure look reasonable (prior to 1200 UTC) when there was a wider distribution of reliable surface wind observations. Although sustained winds over a small area of extreme northern Guam may have been in excess of 65 m s-1, these winds were not representative of the areas south of AAFB. The duration of typhoon force winds was greater than 8 h across some populated regions (> 7 h over most of the northern two-thirds of the island). This long period of strong winds and the wind direction shifts (> 180o) that occurred as the typhoon's eyewall moved over the island likely contributed to much of the observed damage. In addition, the extended period of strong winds and the report of the extreme wind gust at AAFB, which was determined by the NPDAT to be unreliable based on the evidence gathered (Appendix E), likely led to the perception that Paka was much more severe than it was in reality.
    The loss of valuable data, such as the wind data from the ASOS at the Guam WSFO and the loss of the Archive II data record from the Guam WSR-88D, is considered to be unacceptable by the NPDAT. This is a recurring problem in recent tropical cyclone landfalls. As the United States, through efforts of agencies such as the Federal Emergency Management Agency (FEMA), moves toward a more aggressive policy of mitigating disasters, it is essential to have accurate and reliable wind measuring systems which can provide engineers with the data they need in their efforts to improve structural resistance to wind forces. Whenever any official system that measures data that are considered essential for quantifying the destructive characteristics of a storm is installed, storm data acquisition should be a primary concern. These observing systems need to be ruggedized, have uninterrupted power sources, and adequate data storage and retrieval systems for access by local officials and those charged with post-storm service assessment and data acquisition. If only a few official systems on Guam had recorded data throughout Paka, many of the questions about the nature of Paka's destructive wind field could have been answered by the NPDAT immediately. Instead, the team was forced to rely on data recorded by non-standard equipment (including a complete record from an inexpensive amateur weather system with back up power), low-resolution Level IV WSR-88D products, and visual inspection of the damage.

Acronyms used in this report

Examples of Meteorological Observing Sites' Documentation

AAFB wind instrument
AAFB site exposure
AAFB aerial view
Apra Harbor Handar site
KCI tower instrument
KCI tower N & NW exposure
KCI tower W & S exposure
University of Guam instrument
University of Guam N & W exposure
University of Guam S exposure and aerial view

Meteorological data plots

AAFB wind trace
Apra Harbor Handar wind trace
Dan Dan, Handar wind trace
JTWC, Guam wind trace
KCI tower wind trace
Merizo, Guam wind trace
NOS Tide Guage wind trace
ASOS NWSFO, Guam wind trace
F-420 NWSFO, Guam wind trace
University of Guam, Handar wind trace
Airport SAWRS, Rota wind trace
Airport, Rota wind trace

Guam WSR-88D observations and application of the GBVTD technique


    Figures D-1 and D-2 show the radar reflectivities from the Guam WSR-88D at 0632 UTC. Paka's concentric eyewalls are evident, even though the convection in the inner eyewall does not completely surround the eye. The outer eyewall moved over most of the northern portion of Guam and was associated with heavy rain and high winds (over 8 h in some locations).


    The GBVTD technique developed by Lee et al. (1998) was used to retrieve wind observations from the Guam WSR-88D Archive IV data. This method uses least squares methods to fit the primary tropical cyclone circulation onto the Doppler velocities on series of rings (Fig. D-3), each having a constant radius from the tropical cyclone center. The mean velocity (VM), which is the component of the mean wind along the line joining the radar and the storm center, in addition to the symmetric radial (VR) and tangential (VT) wind components are retrieved for each radius. The asymmetric radial flows and the cross-beam VM are not resolved, but are instead aliased into the tangential winds. The GBVTD technique retrieves wind maxima that are not directly observed (i.e., the velocities perpendicular to the beam), because the GBVTD technique uses the Doppler velocity gradient, not the observed maxima to retrieve wind maxima. In the Lee et al. (1998) study, the GBVTD method retrieved good total wind in nearly all cases tested.
    In the case of the Guam WSR-88D data for Paka, Doppler velocities were digitized manually from Level IV products for the 1.5o tilt at 0533, 0632, and 0721 UTC ( Fig. D-3) on 16 December. The GBVTD technique was applied to the VD for each time. The GBVTD winds were assumed to be representative of mean boundary layer winds even though they ranged in height from 650 m to over 3000 m at radii ranging from 4 to 75 km, respectively. The GBVTD wind data for each time were adjusted to 10 m and maximum 1-min sustained surface winds and were used to create a background field for use in the surface wind analyses. These winds were first adjusted to 10 m based on the methods described by PHR. The gust factors used for the adjusted GBVTD winds were based on the volume of atmosphere that WSR-88D Doppler data were sampling. The major contribution to the WSR-88D VD is the horizontal component of the wind. Therefore, the horizontal area of each WSR-88D VD sample was used for estimation of a gust factor for the VD (the horizontal area varies from 66 to 1243 m2 at 4 to 75 km, respectively, radius from the WSR-88D). The sizes of these areas yielded averaging times that were > 1-min. The gust factor described by PHR was applied to each GBVTD wind field to adjust it to maximum 1-min sustained values at 10 m height. Once this was done for each time, the peak value of the wind at each grid point in the 10 m, maximum 1-min sustained GBVTD wind field centered on Paka was chosen to produce the final GBVTD field centered at 0632 UTC (Fig. 6). This wind field was then used as the background for the surface wind analyses.

Statement by the Super Typhoon Paka Data Acquisition Team regarding the peak wind gust recorded at Andersen AFB on 16 December 1997

In regards to the recorded wind gust of 205 knots (i.e., 236 miles per hour) during Typhoon Paka at Andersen AFB on 16 December 1997, based on the entire wind record at the site, Guam WSR-88D data, a site survey, and ground and aerial damage assessments, we consider the peak gust to be unreliable.
--- 23 February 1998
NOAA Typhoon Paka Data Acquisition Team:


The Super Typhoon Paka (1997) post storm data acquisition could not have been completed by the NPDAT without the assistance of the local officials, some of whom were only beginning to recover from the typhoon's effects on their own personal property and lives. In particular, the staff of the Guam National Weather Service Forecast Office and the Joint Typhoon Warning Center were extremely helpful during and after the NPDAT's visit. In addition, NOAA's Paka Service Assessment Team assembled on Guam aided the efforts of the NPDAT considerably. Coordination of the efforts between these two Teams, which have complementary missions, in future post-storm events should be considered a priority.


Crum, T. D., R. L. Alberty, and D. W. Burgess, 1993: Recording, archiving, and using WSR-88Ddata.
Bull. Amer. Meteor. Soc., 74, 645-653.

Dominguez, R. and staff, 1998: Events log for Super Typhoon Paka, Weather Office, AAFB, 5 pp.

Franklin, J. L., S. J. Lord, S. E. Feuer, and F. D. Marks, 1993: The kinematic structure of Hurricane Gloria (1985) determined from nested analyses of dropwindsonde and Doppler data, Mon. Wea. Rev., 115, 2433-2451.

Guard, C. P., 1998: A preliminary assessment of the maximum wind speeds associated with Typhoon Paka over Guam. University of Guam, Mangilao, Guam, 5 pp.

JTWC, 1998: A synopsis of Super Typhoon Paka, Joint Typhoon Warning Center, 9 pp

Lee, W., B. Jou, P. Chang, and S. Deng, 1998: Tropical cyclone structure retrieved from single Doppler radar observations. Part I: Interpretation of Doppler velocity patterns and the GBVTD technique on analytic tropical cyclones, Mon. Wea. Rev., (Accepted for publication).

NOAA, 1998a: Super Typhoon Paka service assessment report. NOAA, 25 pp.

NOAA, 1998b: Super Typhoon Paka data acquisition team report. NOAA, (In preparation).

OFCM, 1997: The National Post-Storm Data Acquisition Plan (Draft). Office of the Federal Coordinator of Meteorological Services and Supporting Research, Silver Spring, MD, 51 pp.

Ooyama, K. V., 1987: Scale controlled objective analysis, Mon. Wea. Rev., 115, 2479-2506.

Powell, M. D., 1980: Evaluations of diagnostic marine boundary layer models applied to hurricanes, Mon. Wea. Rev., 108, 757-766.

Powell, M. D., 1993: Wind measurement and archival under the Automated Surface Observing System (ASOS): User concerns and opportunity for improvements. Bull. Amer. Meteor. Soc., 74, 615-623.

Powell, M. D., S. H. Houston, and T. Reinhold, 1996: Hurricane Andrew's landfall in south Florida, Part I: Standardizing measurements for documentation of surface wind fields, Wea. Forecasting, 11, 304-328.

Powell, M. D., and S. H. Houston, 1996: Hurricane Andrew's landfall in south Florida, Part II: Surface wind fields and potential real-time applications. Wea. Forecasting, 11, 329-349.

Powell, M. D., and S. H. Houston, 1998: Surface wind fields of 1995 Hurricanes Erin, Opal, Luis, Marilyn, and Roxanne at landfall. Mon. Wea. Rev., 126, 1259-1273.

Last revised : August 20,1998