The committee met to hear presentations concerning the re-analysis of Hurricane Andrew, and to begin discussion on this topic. Summaries (Franklin, Black, Landsea) and background papers (Franklin, Powell part1, Powell part 2, Black, Landsea) were received by the committee during the previous week, as requested. Ed Rappaport provided copies of his original Preliminary Report. A copy of the Haag Engineering Co. storm damage survey (part 1, part 2) was also provided to the committee. Mr. Saffir, who agreed to server as the non-NOAA observer (see coordination letter) was present for all of the proceedings. All of the materials before the committee were also provided to him (his reply).
Presentations were heard from Ed Rappaport, Jimmy Franklin, Mark Powell, Peter Black, and Chris Landsea. (It should be noted that the presentations were allowed to run longer than scheduled, and in fact took most of the day. These minutes are not an attempt to replicate all of the material presented, much of which is contained in the linked presentations above, but rather to summarize key points and discussion.)
Ed Rappaport reviewed the circumstances at NHC during the night of south Florida landfall (24 August 1992), and his recollections as the forecaster on duty. He noted that the 162 kt flight-level wind report was received after the 0800 UTC conference call, and very near or just after the release of the advisory, and that in his opinion the timing of the observation may have been a factor in possibly revising the operational wind estimate at the time. Due to the relatively warm tops in the satellite imagery there was some sentiment to actually lower the wind estimates, but this was not done due to the magnitude of the flight-level winds. Rappaport also noted that Robert Sheets, NHC Director at that time, mentioned that the disappearance or relative absence of rainbands during the last few hours as another reason for lowering the operational wind estimate. As far as post-analysis, Rappaport felt that during that era, accepted practice was such that it might have been somewhat more difficult to change the operational wind estimates without new data.
Jimmy Franklin reviewed the derivation of his GPS dropsonde guidelines and their application to Andrew. The guidelines (including the 90% reduction under consideration) were based on 429 eyewall soundings. He noted that the proposed increase to category 5 was not based on just one observation, but observations over a three-minute period. In response to a question, James stated that he felt the statistics were stable.
Franklin then discussed several new radar echo tracking estimates provided by Peter Dodge of HRD, who used data from the Miami WSR-57 just prior to failure. These echo-track vectors were taken as representative of winds at the top of the boundary layer. The peak speeds found reduce to either 134 kt at the surface using a mean eyewall profile representative of all hurricanes, or 146 kt at the surface using the equivalent profile for only high wind cases. Franklin favored the 146 kt estimate for this case.
The intensity changes proposed by Franklin were shown to create pressure-wind pairs which fall within the range of previous estimates. They are, however, on the 'high wind side' of the operational pressure-wind curve. He felt this was appropriate for a small, intense hurricane.
Franklin also pointed out that the system was deepening as it made landfall and that the Dvorak estimates had been consistently low. He felt that these additional factors also supported the proposed increase in intensity at landfall to 145 kt.
Mark Powell then presented his modified analysis for Andrew. The old boundary layer model used previously he now felt was under-estimating the winds. The old model has therefore been replaced with a new model incorporating the dropsonde results discussed by Franklin. Powell stated that only 13% of the drops in intense hurricanes (> 55 m/s) survive to 10 m. Franklin responded that the increase in failure occurred at about 30 m, so that most of the profile was captured.
In his revised analysis, Powell found a 150 kt maximum in the northern portion of the eyewall. However, he pointed out that the same analysis over-estimated the winds at Fowey Rocks at the time of instrument failure. Specifically, his analysis gave 143 kt versus the 109 -111 kts observed. There was quite a bit discussion about whether the maximum wind had been recorded at Fowey Rocks (also see questions below), and how this affected the comparison. Powell stated that he does not believe this was the maximum wind. He also stated that his 1996 analysis was closer to this observation. Powell felt that prior to landfall the sea-surface roughness had probably increased in the area from Fowey Rocks to the coast due to enhanced wave heights, which would act to decrease wind speeds. This is in contrast to the open water condition in which high winds act to smooth out the wave action, and thus decrease the roughness. It should be noted that this point engendered much discussion during and following the meeting, and became a principal point in subsequent committee deliberations. Powell's opinion was that the GPS dropsonde data was valid over open-water, but not at the immediate coast or inland.
Powell also pointed out that his re-analysis over-estimated the 'Fairbank' measurement - a gust calibrated to near 154 kt. Again, the time of failure became a key point of discussion, as was the distance inland (about 1 mile). It was not clear that the peak gust had been measured before failure. After much further discussion, it appeared that these issues could not be resolved to everyone's satisfaction.
In summary, Powell concluded that Andrew had made landfall as a Category 4 hurricane (125-129 kt at the immediate coast) but had in fact been Category 5 over open water.
Peter Black then presented over-land wind estimates from several independent data sources. First, an 'anywhere-to-anywhere' ratio was used to estimate over-land winds from flight-level winds. This ratio was developed by Ulhorn using Stepped Frequency Microwave Radiometer data, and relates the maximum wind anywhere along a flight leg to the maximum surface wind anywhere along the same flight leg. This ratio (about .88) is interpreted as yielding a 10-min over-water wind. Black then applied a conversion from water to land, and then an additional factor to obtain a 1-min sustained wind. The end result is .93. Thus the 162 kt flight level maximum is converted to a 152 kt over-land estimate.
Black also presented results using radar echo tracking from the Tampa radar. Tracking cells north of the center that developed along Dade County coastline and were advected inland, he obtained speeds of 170-180 kts. Although the beam was quite high (27000 ft) Black felt that due to vertical momentum transport these vectors could be taken as representative of motion at the top of the boundary layer. Using the same reduction factor as above, he obtained a surface 1-min sustained wind estimate of 169 kt, plus or minus 10 kt.
Black obtained a third wind estimate through a pressure gradient calculation. Pressure observations from Rappaport (Preliminary Report) were fit with a cubic spline. Assuming a gradient wind relationship, a calculation was performed along a north-south line through the center. The peak wind found along this line was 150 kt (after reduction from the top of the boundary layer) to the north of the center.
Black also described some meso-scale effects, and felt that his damage analysis suggested that a southwesterly wind burst within the eye was responsible for some of the worst damage. This analysis was consistent with his other estimates, and also supported an upgrade to Category 5.
Chris Landsea was the final speaker, and as planned blended together some of material previously presented. Landsea commented that something akin to Black's 'anywhere-to-anywhere' ratio was needed in obtaining over-land wind estimates. In his opinion, two conversions were needed; first, flight-level to the top of the boundary layer, and then to the surface. However, he stated that the result must take into account the increased roughness over land, and it was not clear if Blacks's .93 ratio had completely accounted for this.
Landsea briefly reviewed Franklin's GPS Dropsonde results, and Powell's modified H*wind analysis. As above, both were found to support category 5 intensity over water. He found good agreement between Dodge's radar echo-track vectors and the aircraft winds, when these were reduced to the surface. Landsea felt that the storm surge model results provided by Jarvinen indicated category 5 winds, as did several damage surveys. Landsea concluded that these disparate sources put the maximum sustained winds in the range of 136 to 155 kts at the time of landfall. Landsea and Franklin had coordinated on the revised track, thus the explicit track presented in this segment was the same as that of Franklin, with winds of 145 kt at landfall.
After the presentations, there was a period of open discussion among the committee and the presenters. Most of this discussion concerned the need to specify the correct roughness parameter in the H*wind program, and how this might be accomplished. The meaning of 'open terrain' was also debated. At this point there did not seem to be any clear resolution to the roughness issue. At the close of this portion of the meeting, the presenters agreed to respond to further questions which might arise during the coming week (see below).
The meeting concluded with a short committee-only session. Members were asked to express preliminary opinions. The consensus seemed to be that Jimmy Franklin's work was applicable in this case, and that there was good agreement on the open water estimate. After some discussion, it was agreed that the three 10-s aircraft observations supporting category 5 intensity were sufficiently representative. However, the boundary layer transition near the coast was problematic. There was agreement that better profiles in the near-offshore boundary layer (and to the extent possible over land) were needed to clarify the issue. It was pointed out that Landsea had also seemed undecided about the implications of this boundary layer transition during his presentation, and it was thus agreed that he be given the opportunity to resubmit his report.
Finally, it was decided that each member should form an opinion on the
best-track intensities based on all of the material presented and return
to the next meeting (8 August) prepared with a modified track table.
Addendum: Following the meeting, the discussion continued between the committee and the presenters, via electronic mail. The following documents this exchange. The parenthetical entries (1 through 4) serve to group the threads of the discussion.
One of key points which ensued from the above was the nature of hurricane boundary layer near the coast. James Franklin did some additional work attempting to stratify the dropsonde profiles by distance from the coast and reported on this (1) with two accompanying images - Georges offshore (1-offshore) and Georges nearshore (1-nearshore). Franklin's further email (2), with one accompanying image (2).
Pete Black's response to this work (2)
Franklin's answer (2).
Chris Landsea's response to this work (2)
The committee agreed on a set of questions to be posed to the presenters (3)
Black's response to the questions (3)
Franklin's response to the questions (3)
Powell's response to the questions (3)
Jim Gross made an additional request concerning the reduction factors (4).
Further clarification on this request from Pasch (4-1), (4-2), (4-3).
Franklin's response (4)
with plots for old reduction (4-old) vs.
new reduction (4-new).
Landsea's revised report (submitted Aug 8)