[NCEP Logo]

Preliminary Report
Hurricane Roxanne
7 - 21 October 1995

Lixion A. Avila
National Hurricane Center
29 November 1995

Hurricane Allison
Tropical Storm Barry
Tropical Storm Chantal
Tropical Storm Dean
Hurricane Erin
Tropical Depression Six
Hurricane Felix
Tropical Storm Gabrielle
Hurricane Humberto
Hurricane Iris
Tropical Storm Jerry
Tropical Storm Karen
Hurricane Luis
Tropical Depression Fourteen
Hurricane Marilyn
Hurricane Noel
Hurricane Opal
Tropical Storm Pablo
Hurricane Roxanne
Tropical Storm Sebastien
Hurricane Tanya

[1995 Atlantic Hurricane Season]

Roxanne was the first October hurricane that formed and reached Category 3 intensity on the Saffir/Simpson Hurricane Scale (SSHS) in the western Caribbean Sea since Hurricane Hattie (Category 4) in October 1961 (113K GIF). After hitting the Yucatan peninsula, Roxanne meandered in the Bay of Campeche for several days causing death and destruction along the coast of Mexico from the States of Yucatan through Campeche and Tabasco.

a. Synoptic History

Roxanne formed from a complex combination of several synoptic-scale features (a broad low-pressure area, a tropical wave and an upper trough) which interacted over the western Caribbean. Formation of tropical cyclones from this combination has been commonly observed in the western Caribbean during previous decades, (e.g., 1930, 1940).

As early as 6 October, radiosonde data from the western Caribbean indicated a broad well-established low- to middle-level cyclonic circulation with cloudiness and showers between the Cayman Islands and Honduras. A distinct tropical wave, tracked from the coast of Africa on 26 September, became convectively active over the central Caribbean on 4 October. The wave reached the western Caribbean early on the 7th and interacted with the pre-existing area of disturbed weather. A slow-westward moving upper-level trough was at that time located over the Windward Passage, to the east of an upper-level anticyclone centered over the southeastern Gulf of Mexico and left by Hurricane Opal. This combination of the high and the trough aloft resulted in diffluent northerly winds over the low-level disturbance. Dunn (1960) and Riehl (1951) recognized this pattern as favorable for the development of incipient disturbances by. The 200 and 850 mb data for 1200 UTC 6 October and the schematic pattern suggested by Dunn are displayed in Fig. 1 (80K GIF). Note that the prevailing meteorological conditions on that day were similar to those suggested in Fig. 1c.

On 6 October, the National Hurricane Center (NHC) surface analysis showed a broad 1004 mb low not too far from the east coast of Nicaragua. Associated winds were only 10 to 15 knots. Satellite images indicated a gradual increase in organization and cloud banding features began to develop early on the 7th as the tropical wave reached the area. It is estimated that the system became a tropical depression at 1800 UTC 7 October just east of Nicaragua. The next day, a reconnaissance plane confirmed the presence of a tropical depression with a pressure center of 1004 mb and 30 knot winds. Satellite images and surface observations indicated a steady intensification. The depression became Tropical Storm Roxanne at 0000 UTC 9 October and a hurricane by 0600 on the 10th. During that period, data from reconnaissance planes indicated that the pressure dropped to 989 mb and by 1200 UTC on the 10th the pressure was down to 972 mb. Prior to intensification, the low-level center was located on the northern edge of the deep convection due to the northerly winds produced by the upper high over the Gulf of Mexico. However, the upper trough previously located over the Windward Passage became a cut-off low and moved west-southwest into Central America. This allowed the outflow to become established in all quadrants.

Initially, Roxanne was a threat to Cuba and the Cayman Islands as it moved northward in response to a weak trough over Florida and the eastern Gulf of Mexico. The trough moved eastward and was replaced by a high pressure system. Roxanne turned northwestward and then westward toward the Yucatan peninsula and intensified. A similar abrupt change in motion, not expected by climatology, was taken by Hurricane Hattie in 1961.

During the early afternoon of the 10th, a well-defined eye became apparent on satellite images. Roxanne reached its maximum intensity of 100 knots and a minimum pressure of 956 mb about 2152 UTC just to the southeast of Cozumel.

Figure 2 (43K GIF) shows the northern eyewall viewed from the Cancun radar, crossing Cozumel at 2340 UTC 10 October. Roxanne made landfall just north of Tulum, on the mainland, Mexico just to the southwest of Cozumel about 0200 UTC 11 October. The hurricane continued westward over the Yucatan peninsula and emerged over the Gulf of Campeche as a minimal hurricane. It then temporarily weakened to tropical storm status. Once the center of circulation was completely over the waters, the tropical cyclone regained hurricane intensity and maintained that status for about 60 hours. It gradually weakened to a tropical storm and then to a tropical depression.

The steering currents were weak when Roxanne was in the Bay of Campeche. Consequently, the hurricane meandered within an area of less than 250 n mi for almost a week. During that period, several shortwave troughs and ridges rapidly passed by to the north of Roxanne, forcing the tropical cyclone to swing either southeastward or northwestward. Feeder bands and waves of 15 to 20 feet were pounding the coastal section from the State of Campeche to Veracruz throughout that time. Eventually, Roxanne was forced to move southward into the area near Veracruz by an approaching strong cold front. The remnants of the tropical cyclone moved southwestward into Mexico.

Roxanne's track is shown in Fig. 3 (43K GIF). Table 1 is a listing, at six-hour intervals, of the "best-track" position, estimated minimum central pressure and maximum 1-min sustained surface wind speed.

b. Meteorological Statistics

The best track pressure and wind curves as a function of time, shown in Figures 4 and 5 (52K GIF), are primarily based on reconnaissance aircraft, satellite intensity estimates from the National Hurricane Center (NHC), Satellite Analysis Branch (SAB) and the Air Force Global Weather Central (AFGWC). The analysis used numerous observations taken from ships in the northwestern Caribbean and the Gulf of Mexico.

The maximum intensity occurred about 2152 UTC 10 October when a reconnaissance plane measured 956 mb and wind of 114 knots at the 700 mb level. Table 2 lists ship reports with 34 knot winds or higher and Table 3 lists selected surface observations associated with Roxanne. Information provided by the Servicio Meteorologico de Mejico indicated that an automatic station near Merida reported sustained winds of 65 knots with gusts to 109 knots at 1900 UTC 11 October. A weather station in Veracruz reported nearly 12 inches of rain and the Tabasco Observatory reported 9.5 inches. The other reports from the region affected by Roxanne indicate that the rainfall totals were of only one or two inches there.

Very interesting observations were collected by Mr. Dan Hartman during the passage of Roxanne. He was located at 20 deg 30'N 86 deg 57'W, (southwest of Cozumel and along the coast of Yucatan). Pressure was measured with a HUGER mercury barometer. Wind speed was obtained from a Rainwise wind speed/direction receiver located on the roof of Mr. Hartman's home which is the top floor of a two story dwelling. The rain gauge is an All-Weather made to "U.S Weather Bureau specifications" and was located on the roof in an unobstructed location. The Log and Comments of Mr. Hartman are:

"On Sunday Oct. 8 approximately 5-6 PM (All times are local time which are two hour earlier than EST) winds from the north 0-10 mph.
1:45 PM Tuesday pressure 1000 MB and winds NW 30-40 MPH
3:30 PM pressure falling 988 MB winds NW 30-40 MPH
4:30 PM 980 MB winds NW 40-74 MPH
5:00 PM 978 MB winds shifting to NE and for a short time the winds seemed to decrease, then wind gusts increased to over 100 MPH (limit of the wind gage)
6:00 PM 983 MB winds likely over 100 MPH coming from anywhere from NE to SE. At this point the wind gauge had blown over and it became very difficulty to tell true wind direction. From various windows we have I could estimate wind direction by the rain blowing into the house.
6:45 PM988 MB winds still decreasing from NE to SE.
7:10 PM991 MB winds dropping
8:00 PM995 MB
8:35 PM 998 MB
9:00 PM999 MB
12:00 AM Wednesday1002 MB
1:30 AM1003 MB
6:30 AM1010 MB overcast scattered showers, winds E-SE

The various times of the barometric reading are irregular due to problems dealing with the hurricane as it impacted our home. It is quite possible that the actual barometric pressure was lower than 978 MB that I recorded. I just was too busy to spend too much time taking notes every 5 minutes. Likewise the wind speeds. As a subjective observation the winds in the storm seem to rise and fall quite a bit. It was not a steady wind force of so many miles and hour rather than ebbs and flows which build up into a higher tempo as the storm approached. Rainfall was also not a large factor. My readings showed approximately 5.4 inches falling from Monday night till Wednesday morning when I took the final readings. I did record 2.4 inches of this total falling Monday night. The storm surge on the east side of the island appears to have been on the order of 10 feet. This is based on my estimate of the high between the water line and the high water mark of the storm as shown by the trash line".

c. Casualty and Damage Statistics

Mixed reports of damage have been received at the NHC. The death toll so far is 14 according to the Ft. Lauderdale Sun-Sentinel, 20 October, 1995. U.S. Coast Guard, a C-130 reconnaissance plane investigating Roxanne and Mexican helicopters searched for people missing from a petroleum work-barge that sank with 245 people on board. Five of these 14 deaths appear to be related to this event including one American. According to Mexican authorities, more than 40,000 homes were damaged by Roxanne in the States of Campeche, Quintana Roo, Tabasco Veracruz and Yucatan. Numerous crops were damaged, cattle drowned and roads were washed out or blocked by mud and rock slides. The road between the City of Carmen and Campeche was completely destroyed. Thousands of people were evacuated. There are unconfirmed reports that many hotel lobbies in Cancun and Cozumel were damaged from pounding waves. Extensive tree damage was observed in Cozumel. Storm tides and swollen rivers caused the worst flooding in Campeche and Tabasco since 1927. It appears the worst damage was produced by the pounding of 15 to 20 foot waves along the shore for several days. The waters from the Gulf of Mexico surged hundred of yards inland.

This area had previously been affected by Hurricane Opal a week or two before and it is difficult to separate the damage caused by Opal and Roxanne. Best estimate of the combined damage in the Yucatan peninsula is $1.5 billion, according to insurance company estimates.

d. Forecast and Warning Critique

Table 4 lists the watches and warnings associated with Roxanne. Initially when Roxanne was in the western Caribbean moving northward, watches and warnings were issued for portions of Cuba and the Cayman Islands. In the advisories, the NHC emphasized that Roxanne was expected to turn toward the northwest and west and that the threat was the greatest for the Yucatan peninsula. Numerical models provided excellent guidance when forecasting the turn to the left toward Yucatan, even when Roxanne was moving northward 4 to 8 knots (See Fig. 6) [27K GIF]. Track models generally kept the hurricane moving west or northwest south of 25°N and almost never suggested a U.S. landfall.

Due to the erratic motion of the tropical cyclone in the Bay of Campeche, watches and warnings had to be extended, discontinued and reissued for portions of the coast of Mexico. Overall, Mexico was under watches and warnings almost continuously for 10 days. A hurricane warning for the landfall area near Cozumel was issued 23 hours in advance.

The individual errors of each track model as well as the errors of the official forecast are shown in Table 5. The official forecast was generally comparable with the past 10-year average, but slightly better than average for the 72-h period. It is interesting to note that in the case of Roxanne, the models A90E and BAMD performed better than the usually more reliable GFDL at the 72-h period. Most of the largest model and official forecast track errors occurred when Roxanne was meandering in the Bay of Campeche and steering currents were weak.

As occurred in other storms this season, the tropical cyclone intensified faster than forecast during the early stage. There was a negative bias (underforecast) at the 72-h period throughout much of Roxanne's lifetime.

Table 1. Preliminary best track, Hurricane Roxanne, 7 - 21 October, 1995.
Position Pressure
Wind Speed
Lat. (°N)Lon. (°W)
7/180014.082.1100525 Tropical Depression
8/000014.282.7100525" "
060014.883.0100525" "
120015.083.2100525" "
180015.783.2100430" "
9/000016.583.1100235 Tropical Storm
060017.283.0100140" "
120017.982.999945" "
180018.482.999550" "
10/000018.983.798960" "
120019.485.097280 "
180019.986.096695 "
060020.088.197090 "
120019.989.198375 "
180019.790.098770 "
12/000019.590.899065 "
060019.591.599265 "
120019.592.299455Tropical Storm
180019.792.699555" "
13/000020.192.899355" "
060020.593.199355" "
120021.093.699255" "
180021.493.998855" "
14/000021.893.998255" "
060022.293.998160" "
180022.393.398070 "
15/000021.892.997970 "
060021.492.497975 "
120021.091.998075 "
180020.691.698075 "
16/000020.491.598175 "
060020.391.798470 "
120020.392.098565 "
180020.492.298765 "
17/000020.692.499165 "
060020.992.699560Tropical Storm
120021.292.999855" "
180021.393.1100050" "
18/000021.593.3100345" "
060021.893.6100440" "
120022.294.2100535" "
180022.394.7100835" "
19/000022.395.1100930 Tropical Depression
060022.495.4100930" "
120022.395.5100930" "
180022.395.6100930" "
20/000022.495.6100930" "
060022.095.5100930" "
120021.595.5100925" "
180020.595.5101025" "
21/000019.596.0101125 Dissipating
10/215220.086.7956100 Minimum Pressure
11/020020.087.6958100 Landfall
(near Tolum)

Table 2. Tropical Cyclone Winds (Ship encounters of 34 Knots or more)
Position Wind(kn)
C6VC10/09060016.987.5 310/361010.0
WNJG10/09140018.083.1340/42 998.4
WNJG10/09180018.783.5020/35 1001.6
WNJG10/10000019.884.1070/35 1000.0
NRDC10/10060016.985.6320/58 1007.2
SHIP10/10060017.184.4220/37 1004.2
ELFT810/10120018.086.7300/40 1006.5
3EWJ910/10120018.086.9280/35 1013.0
OUZW10/10120020.083.8160/38 1002.3
ELFT810/10180016.985.6230/45 1009.0
OYGK210/10180018.586.8250/40 1002.6
C6KJ510/10180021.786.1XXX/35 1000.5
OUZW210/10180022.085.1100/50 1004.5
OYGK210/11000018.986.1240/48 1001.0
PJNL10/11000022.185.1090/40 1004.0
C6KJ510/11060021.685.3140/35 1004.5
V2EM10/12060019.296.0350/37 1006.0
V2EM10/12120019.296.0360/40 1004.0
***YH10/13180019.984.5120/54 1009.0
ELSF810/15000021.693.0280/37 986.0
C6IK810/15000021.792.4180/47 985.0
10/15000026.296.1 360/351014.5
V2EM10/15120021.290.5 140/48998.0
10/15120021.995.0 360/401015.5
V2EM10/15210019.693.2 270/601005.0
DP3YK310/16060022.892.4030/47 1011.6
VSBZ510/16120023.392.5040/42 1011.7
VSBZ510/16180023.092.5040/40 1012.6
DP3YK310/16180023.792.8050/40 1015.0
SBZ510/17000022.893.2040/45 1009.2
SBZ510/17060023.193.1050/38 XXXX.X
DP3YK310/17060023.593.4050/40 1013.6
VSBZ510/17120023.592.8050/40 1009.0
VORX10/17120023.892.7060/35 1008.7
DP3YK310/17180023.991.5070/34 1013.0
VSBZ510/18000022.992.1090/35 1009.5
DP3YK310/18060021.591.7150/34 1011.0

Table 3. Hurricane Roxanne selected surface observations, October 1995.
Minimum sea-level
Maximum surface wind
speed (kt)
Storm surge b
Storm tide b
(storm total)
(UTC) a
Merida  35 d 4011/1200   
Merida a. station  65 10911/1900   
Valladolid992.7 28 11/0900    
Coatzacoalcos  35 4215/1900   
Paraiso  6570 15/1800   
Champoton  5665 15/1800   
Tuxpan  4451 16/1800   
Villahermosa  5570 15/xxxx   

a Time of 1-minute wind speed unless only gust is given.

b Storm surge is water height above normal tide level. Storm tide is water height relative to National Geodetic Vertical Datum (NGVD) which is defined as mean sea level in 1929.

c One-minute averaged wind.

d Unknown averaging period.

e A more extreme value could have occurred.

f WMO standard 10-min wind

g 12-hour total

Table 4. Tropical Cyclone watch and warning summary, Hurricane Roxanne
9/1500Tropical Storm Warning issued Cayman Islands
9/2100Tropical Storm Warning and Hurricane Watch issued Pinar del Rio and the Isle of Youth, Cuba
Hurricane Watch issued Chetumal to Cabo Catoche, Yucatan, Mexico
10/0300Hurricane Warning issued Chetumal northward and westward to Progreso
10/1500Hurricane Watch issued West of Progreso to City of Carmen
Tropical Storm warning and Hurricane Watch discontinued Cayman Islands and Cuba except for extreme western Cuba
10/2100 Tropical Storm Warning issued Belize City northward to just south of Chetumal
Tropical Storm Warning discontinued Extreme western Cuba
11/0600Tropical Storm Warning discontinued Belize City to just south of Chetumal
11/1500 Hurricane Warning issued City of Carmen to Coatzacoalcos
Tropical Storm Warning and Hurricane Watch issuedWest of Coatzacoalcos to Tuxpan
Hurricane Warning discontinuedChetumal to Progreso
11/2100Hurricane Warning issued West of Coatzacoalcos to Tuxpan
12/1500Tropical Storm Warning and Hurricane Watch City of Carmen to Tampico
Hurricane Warning discontinued West of Coatzacoalcos to Tuxpan
13/2100Tropical Storm Warning and Hurricane Watch issued North of Tampico to La Pesca
14/0300 Tropical Storm Warning and Hurricane Watch discontinued City of Carmen to Veracruz
14/1500Tropical Storm Warning and Hurricane Watch issued Veracruz to Progreso
Tropical Storm Warning and Hurricane Watch discontinued North of Tampico
14/1800Hurricane warning issued Progreso to Tampico
16/0300Hurricane Warning replaced by Tropical Storm Warning Progreso to Tampico
18/1500Tropical Storm Warning discontinued Progreso to Tampico
18/2100Tropical Storm Warning issued Tuxpan to San Fernando
19/1500Tropical Storm Warning discontinued Tuxpan to San Fernando

Table 5
(Errors in nautical miles for tropical storm and hurricane stages with number of forecasts in parenthesis)
Forecast TechniquePeriod (hours)
GFDI55 (34)113 (32)173 (30)240 (28)363 (17)
GFDL *48 (18)100 (17)158 (16)225 (15) 315 ( 9)
VBAR *54 (37)129 (35)210 (33)292 (31) 456 (27)
AVNI45 (35)82 (33)130 (31)190 (25)392 (25)
BAMD49 (38)96 (36)147 (34)194 (32)266 (28)
BAMM55 (38)121 (36)196 (34)277 (32)417 (28)
BAMS79 (38)170 (36)274 (34)377 (32)539 (28)
A90E43 (38)72 (36)114 (34)164 (32)251 (28)
CLIP47 (38)93 (36)151 (34)214 (32)354 (28)
NHC Official43 (38)83 (36)138 (34)198 (32) 269 (27)
NHC Official
(1985-94 10-yr average)
5098 194 296

* GFDL output not available until after forecast issued. VBAR output sometimes not available until after forecast issued.

Brian Maher
Jack Beven

Last updated Janaury 8, 1999