JULY 2005 F ARFÁN AND CORTEZ 2069

An Observational and Modeling Analysis of the Landfall of (2003) in Baja ,

LUIS M. FARFÁN Centro de Investigación Científica y de Educación Superior de Ensenada B.C., Unidad La Paz, La Paz, Sur, Mexico

MIGUEL CORTEZ Servicio Meteorológico Nacional, Comisión Nacional del Agua, México, Distrito Federal, Mexico

(Manuscript received 20 July 2004, in final form 26 January 2005)

ABSTRACT

This paper documents the life cycle of Tropical Marty, which developed in late September 2003 over the eastern Pacific Ocean and made landfall on the Baja California peninsula. Observations and best-track data indicate that the center of circulation moved across the southern peninsula and proceeded northward in the . A network of surface meteorological stations in the vicinity of the storm track detected strong winds. Satellite and radar imagery are used to analyze the structure of convective patterns, and rain gauges recorded total precipitation. A comparison of Marty’s features at landfall, with respect to Juliette (2001), indicates similar wind intensity but differences in forward motion and accumu- lated precipitation. Official, real-time forecasts issued by the U.S. National Hurricane Center prior to landfall are compared with the best track. This resulted in a westward bias of positions with decreasing errors during subsequent forecast cycles. Numerical simulations from the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model were used to examine the evolution of the cyclonic circulation over the southern peninsula. The model was applied to a nested grid configuration with hori- zontal resolution as detailed as 3.3 km, with two (72- and 48-h) simulations. The output of the model provides a reasonable prediction of Marty’s motion during landfall and the circulation characteristics are consistent with the information derived from observations. Additionally, the model was used to estimate the role of the peninsular mountain ranges in a realistic simulation of the storm track over the Gulf of California.

1. Introduction second time in 15 yr when two storm systems moved into the area in the same season. The earlier two-storm The 2003 season of tropical delivered 16 landfall season occurred in 1993, making landfall as named storms in the northeastern Pacific Ocean, and 5 tropical storms that contributed with a moderate of them made landfall in Mexico. Among this group, amount of rainfall (Avila and Mayfield 1995). two hurricanes came ashore in the southern portion of For the years 1951–2000, the history of tropical cy- Baja California. Ignacio made landfall on 24 August clone landfalls in northwestern Mexico was studied by and Marty moved across the area on 22 September Jáuregui (2003). In that 50-yr time frame, the Baja Cali- (Beven 2004). These storms brought strong winds and fornia peninsula was the most exposed region and it heavy rainfall that was responsible for several deaths, was struck by 20 hurricanes and 33 tropical storms. Fur- enormous property damage, soil erosion and flooding, ther, the other west coast states had hurricane and and evacuation from coastal communities. This was the tropical storm landfalls that ranged from 1 in to 26 in . By comparison, a total of 31 systems made landfall on the Yucatán Peninsula during the Corresponding author address: Dr. Luis M. Farfán, Centro de same period. Investigación Científica y de Educación Superior de Ensenada B.C., Unidad La Paz, Miraflores #334, La Paz, B.C.S. 23050, Tropical cyclones are important in the climatology of México. summer precipitation over western Mexico because E-mail: [email protected] they provide a significant portion of the water re-

© 2005 American Meteorological Society

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC

MWR2966 2070 MONTHLY WEATHER REVIEW VOLUME 133

Ϫ FIG. 1. Time series of sustained wind speeds (circles; m s 1) at during the period (a) 1600 UTC 14 Aug–1600 UTC 16 Aug 2000, (b) 1600 UTC 8 Sep–1600 UTC 10 Sep 2000, (c) 0400 UTC 12 Sep–0400 UTC 14 Sep 2001, and (d) 0000 UTC 27 Sep–0000 UTC 29 Sep 2001. Time axis (day/hour) is centered at maximum speed and covers a period of 48 h. Time interval for speeds is 10 min and 1 h for barbs. Full barb represents 5.0 m sϪ1 and half barb 2.5 m sϪ1. Dotted lines mark the location of speeds at 19 and 33 m s Ϫ1. sources. In a study covering the period 1949–97, Engle- (NHC) showed that 44 named systems developed in the hart and Douglas (2001) documented the distribution eastern Pacific during this period. The year 2000 was of rainfall associated with the approach of these sys- selected because it was the first year that a significant tems. For three stations in Baja California, the maxi- number of automatic surface stations were operating in mum rainfall in the southern tip (Cabo San Lucas area) Mexico (Rosengaus 2001) and the first year of an op- was recorded during extreme cases while stations erating station in the southern part of the peninsula. northward received less rainfall because they are lo- For this study, the analysis of each season was limited to cated farther from the core of the storm activity in the July, August, and September. Although station density tropical Pacific. Documenting structure and intensity of and distribution are not ideal to analyze all cases in precipitation of significant cases enhances our ability to detail, these high-resolution data are useful to deter- calculate social, economic, and ecological impacts on mine storm impact over specific coastal areas. this area. Figure 1 shows sustained winds (10-min average) The impact of tropical cyclones on Baja California is with speed above 19.0 m sϪ1 during the 2000–02 tropi- illustrated by surface data at Cabo San Lucas for 2000– cal storm seasons. The horizontal (time) axis is cen- 02. Records from the U.S. National Hurricane Center tered at the occurrence of maximum speed for each

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2071

TABLE 1. Selected parameters recorded from the automatic surface station in Cabo San Lucas during the 2000–02 seasons.

Date (year/month/day) Tropical cyclone MAXSPD MAXGST MAXRNF MINDIS 00/08/15 Ileana 20.2 28.2 120 76 00/09/09 Lane 24.2 29.9 101 450 01/09/13 Ivo 19.1 24.3 113 225 01/09/28 Juliette 39.1 46.7 1011 100

MAXSPD ϭ speed (m sϪ1). MAXGST ϭ Maximum peak gust (m sϪ1). MAXRNF ϭ Maximum storm rainfall (mm). MINDIS ϭ Minimum distance (km) between the best-track position and station location. identified event. Maximum speeds tend to occur in con- datasets during the landfall phase of Tropical Cyclone junction with wind shifts and a pressure minimum (not Marty (2003) in Baja California. Marty is an important shown) that suggests the approach of well-defined, case to analyze, in part, because it made landfall while moving circulations. In fact, all cases are associated sustaining hurricane intensity over major cities of the with tropical cyclones within 450 km of the station, southern part of the peninsula. Observations indicated which included Ileana and Lane in 2000 and Ivo and that the urbanized population experienced an impact Juliette in 2001 (Fig. 1). There were no cases of strong that was similar to that from Hurricane Juliette in 2001. winds or nearby tropical cyclones during 2002. Impor- Specific attention is given to the following topics: 1) tant parameters describing storm impact at the station documenting Marty’s maximum intensity so that a com- are given in Table 1, with the approach of Juliette as the parison with Juliette at Cabo San Lucas is possible, 2) only case having hurricane-intensity sustained wind describing Marty’s evolution from official forecasts and speeds (above 33 m sϪ1). Although Lane remained well model simulations that started before the storm made offshore, 24.2 m sϪ1 sustained wind speeds and 29.9 landfall, and 3) applying model simulations as a guide msϪ1 peak gusts were recorded at the station. For all to understanding changes in the storm track while it cases, peak gusts tend to be 19%–40% higher than the proceeded over the peninsular landmass. For this pur- corresponding sustained wind speeds. pose, the fifth-generation Pennsylvania State Universi- Tracks of systems that had an impact in the southern ty–National Center for Atmospheric Research Meso- peninsula are shown in Fig. 2 and their positions are scale Model (MM5) was used. derived from the NHC best-track dataset for the east- This paper is divided into five sections. In section 2, ern Pacific. In general, higher wind speed occurs when the area of study and sources of observational data are tropical cyclones are located south or southwest of described. In section 3, a description of the best track Cabo San Lucas. Most storm systems arrive from areas and observed data on the evolution of Marty are pro- south of 20°N and move along a northwestward direc- vided. Section 4 presents operational forecasts in real tion, with the exception of Juliette (2001), which ac- time, as well as MM5 simulations, which are focused on quired a large meridional component and crossed the describing changes of the storm’s track and structure at peninsula. While positioned west of the recording sta- landfall. The fifth section summarizes results and states tion, Juliette moved slowly northward, generating the conclusion. strong winds (20–39 m sϪ1; Fig. 1d) that lasted more than one day, including hurricane intensity for a few 2. Geography and data hours. These observations suggest more property dam- age compared to the other storms. Juliette’s strength The study area is limited to Baja California and ad- resulted in a critical situation, when maximum total jacent portions of the Pacific Ocean and Gulf of Cali- rainfall (Table 1), determined from a network of re- fornia. Since the main impact of Marty occurred over ports within 60 km from the site, reached 1000 mm. the southern peninsula, attention is focused on the state Since this area is in mountainous terrain (Fig. 2), such of (BCS), between 22.9° and 28.0°N heavy rainfall should generate excessive streamflows in (Fig. 2). Major topographic features are the southern the natural higher portions of the drainage basins and (23.0°–24.0°N) and central (24.8°–27.6°N) mountain across urban areas occupying the lower portions of ranges, with peaks typically about 1500 m (highest: 2100 these watersheds. m) and 1000 m (highest: 1700 m), respectively. The This paper documents observed and forecasted most populated areas in southern BCS are the commu-

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2072 MONTHLY WEATHER REVIEW VOLUME 133

FIG. 2. Best-track positions of tropical cyclones that resulted in sustained wind speeds greater than 19 m sϪ1 at Cabo San Lucas during the seasons of 2000–01. Positions are at 6-h time intervals and labels next to the name are dates (year/month/days) in which each system is located within this plot. Black dots indicate storm center while maximum winds were recorded at the station. Straight lines represent mountain ridges used throughout the paper, and the nested box depicts the location of Baja California Sur. nities of La Paz, Cabo San Lucas, and San José del We used a network of automated surface stations Cabo, containing 38.4%, 9.0%, and 7.3%, respectively, operated by the Servicio Meteorológico Nacional of the state’s population of 424 000 (INEGI 2000). Fig- (SMN) and Secretaría de Marina (SEMAR) agencies in ure 3 shows community locations. Mexico. Observations transmitted include sustained Data are extracted from a variety of operational winds, peak gusts, air temperature, relative humidity, sources that include satellite and radar imagery, surface and station pressure. The data are reported at 10- and observations, and numerical weather products. Satellite 15-min intervals for SMN and SEMAR stations, respec- imagery from the Geostationary Operational Environ- tively. Data from four automatic stations in Baja Cali- mental Satellite-10 (GOES-10) provides cloud cover fornia are analyzed in this study. One station is located structure and intensity derived from the infrared and in Cabo San Lucas, the same site that supplied the his- visible channels available at a temporal (spatial) reso- torical observations shown in Fig. 1. Additional meteo- lution of1h(4km)and15min(1km), respectively. rological data from airports, issued by the Servicios Onboard calibration capabilities of the satellite mea- a la Navegación en el Espacio Aéreo Mexicano sure convective intensity from brightness temperature, (SENEAM) agency, were used (Fig. 3). as detected by the infrared channel. Menzel and Pur- Daily precipitation data from the network of 120 rain dom (1994) discuss the imagery calibration process and gauge stations in BCS, managed by the SMN and Clark (1983) provides information to determine cloud- Comisión Nacional del Agua (CNA) agencies, are used top temperature from the image gray level. to determine the distribution and intensity of rainfall

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2073

FIG. 3. Best-track positions for Tropical Cyclone Marty, 18–26 Sep 2003. Plus signs are positions estimated in real time at 3-h intervals. Best track at 0000 UTC includes correspond- ing intensity: tropical depression (TD), tropical storm (TS), or hurricane (HR). The asterisk denotes landfall position. Reference locations used throughout the paper are indicated with horizontal line and station data sources are in parentheses. during the passage of the storm. Additionally, these also used as data for the MM5 simulations discussed in agencies operate a weather radar system in northwest- section 4. ern Mexico; their imagery data are used to document reflectivity structures associated with Marty. One is lo- 3. Observations cated in Cabo San Lucas at the automatic surface sta- tion. Another radar station is located at Guasave on the According to NHC records, Tropical Cyclone Marty mainland coast (Fig. 3). The data archives contain im- developed and progressed from 18 to 26 September agery with an effective range of 480 km, recorded at 2003. Figure 3 shows the corresponding best-track and 15-min intervals. The imagery is also used as a source of real-time positions from operational advisories (NHC data for reconstructing best tracks for tropical cyclones 2003). Real-time data are important primary elements generated by the NHC (Avila et al. 2003). for issuing timely watches and warnings to protect To examine the large-scale airflow, data are obtained population centers. The first advisory was issued early from operational analyses and forecasts of the Global on 19 September, showing a strengthening tropical de- Forecast System (GFS). This is a dynamic model oper- pression 380 km south of Cabo Corrientes, Mexico, ated by the National Centers for Environmental Pre- with a westward motion. There were some differences diction (NCEP), used to provide guidance for NHC between the best-track and real-time positions early in forecasts during the tropical cyclone season (Surgi et al. the tracking operations of the storm because the system 1998). The gridded version of the GFS used in this showed erratic motion. study has global coverage and a resolution covering The GFS analyses indicated an anticyclonic circula- approximately 100 km ϫ 100 km. The GFS fields are tion over the during the initial stages of

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2074 MONTHLY WEATHER REVIEW VOLUME 133

FIG.4.GOES-10 infrared imagery at (a) 1800 UTC 21 Sep, (b) 0000 UTC 22 Sep, (c) 0600 UTC 22 Sep, (d) 1200 UTC 22 Sep, (e) 1800 UTC 22 Sep, and (f) 0000 UTC 23 Sep 2003. Horizontal resolution of the imagery is 4 km and the grayscale indicates cloud-top temperatures (°C). Hurricane and tropical storm intensity from the best track are indicated with the hurricane and star symbol, respectively.

Marty’s life cycle. This circulation provided weak, tivity prior to, during, and after Marty’s progress over steering flow at midlevels that was directed along the the southern part of the peninsula. The imagery indi- southern coast of Mexico. While showing limited mo- cates that, while located south of Cabo San Lucas, tion, the depression intensified into a tropical storm at Marty steadily intensified; minimum cloud-top tem- 0600 UTC 19 September and was upgraded into a hur- peratures decreased from –70°C at 1800 UTC 21 Sep- ricane at 0000 UTC on 21 September. A northward tember (Fig. 4a) to –80°C at 0600 UTC on 22 Septem- track developed after 1200 UTC as the hurricane ap- ber (Fig. 4c). As the core approached land, some con- proached Baja California. Landfall occurred at about vective cells reached –85°C over the foothills east of the 0930 UTC on 22 September near Cabo San Lucas southern mountains. This is 15°C from the extreme (Franklin 2004). The corresponding estimate of sus- temperature (Ϫ100°C) in an intense tropical cyclone Ϫ1 tained winds was 43 m s and this was the maximum recorded from a geostationary satellite (Ebert and Hol- intensity during the entire best-track record. The sys- land 1992). tem moved offshore into the central Gulf of California, The storm strengthened during the next few hours was located near Santa Rosalía on 23 September, and, and at 1200 UTC (Fig. 4d) the area of intense convec- during the following few days, reached the northern tive activity reached a horizontal diameter of about 250 gulf with declining intensity. km. This time corresponds to the first assignment by NHC of the storm’s position over land; the center was a. Satellite imagery placed 60 km southeast of La Paz. Marty continued GOES-10 infrared imagery of Marty’s landfall is moving northward the rest of the day and was located shown in Fig. 4. This imagery represents convective ac- over the gulf at 1800 UTC (Fig. 4e). Convective cells

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2075 near the core weaken by 0000 UTC on 23 September (Fig. 4f). Minimum cloud-top temperatures were in the range from –60° to –65°C, which indicated a weakening trend over the central gulf. The visible imagery during the morning of 22 Sep- tember showed the core circulation moving northward from the southern part of the peninsula. The first image was available at 1345 UTC. Animation of subsequent images exhibit convective development centered 50 km north of La Paz. Figure 5 shows the image at 1445 UTC, with a reduced area of around the core and bands of clouds over the gulf and mainland. An inter- esting feature is the elongated cloud cluster north of the core (feature A in Fig. 5), which intensified during in- teraction with the land surface from 1500 to 2100 UTC. This interaction is associated with a distinct area of convective activity along the southern end of the moun- FIG.5.GOES-10 visible image at 1445 UTC 22 Sep 2003. The tains in central BCS. dot indicates center of circulation; A and B represent the northern and southern features of convection, respectively, discussed in the paper. Horizontal resolution of the imagery is 1 km. b. Radar imagery

Imagery from radar stations positioned close to the (Fig. 6b) and satellite (Fig. 5) images share several landfall region is shown in Fig. 6; the scans are taken features, including 1) the center of circulation occur- from a constant elevation angle, also known as a plan ring north of La Paz, 2) curved bands of convection position indicator (PPI). This imagery tends to under- (features A and B), and 3) few convective clusters estimate maximum rainfall rates during the storm’s ap- south of the core. The storm center from radar is proach. In spite of this, the imagery describes, in a slightly west of the real-time position assigned by the qualitative fashion, mesoscale features of the circula- NHC at 1500 UTC (Fig. 3). The storm maintained its tion associated with Marty. The radar reflectivity is convective structure during the next few hours and two treated as a source of data to outline the spatial dis- distinct areas of reflectivity are identifiable. One of tribution of convective patterns rather than precipita- these bands is north of the core near the peninsula and tion. another along the mainland coast. Marty was outside Figure 6a shows reflectivity from the Cabo San Lucas the effective radar range of the Guasave station at radar at 0626 UTC 22 September, when the satellite 2100 UTC, and therefore this does not permit further data placed the hurricane center about 100 km south- analysis. east of the station. The structures in this figure show the center of the system without precipitation and its loca- c. Rain gauge network tion consistent with the corresponding satellite image (Fig. 4c). The reflectivity covered an adjacent area in An estimate of rainfall during the storm passage is the Pacific, offshore in the western gulf, and the land based on the BCS rain gauge network from 1500 UTC area south of La Paz. This image shows two distinct on 19 September to 1500 UTC 25 September. This pe- bands over the northern quadrants: one on the central riod covers the approach, landfall, and departure of ring (feature B in Fig. 6a) and a second band (feature Marty over BCS. Figure 7 shows the horizontal distri- A) along the peninsula’s southern coast. Unfortunately, bution of precipitation and the position of the mountain this was the last image available from the radar because range above 500 m. The maximum precipitation oc- a power outage occurred during the storm’s presence curred in the southeastern part of the range, where the over land. satellite detected the coldest cloud tops. Recorded pre- The mainland radar station at Guasave provided cipitation was in the range of 380–410 mm from sites monitoring of Marty’s development during the early along the eastern foothills. This area agrees with the morning of 22 September. Reflectivity depicts north- area of deep convection detected by the infrared imag- ward motion over the southern gulf after 1200 UTC, ery on 22 September. Another area of high rainfall oc- and Fig. 6b shows the image at 1447 UTC. This is close curred over the central mountains, ranging from 340 to to the time of the visible image from satellite. The radar 400 mm. Since no stations are located along the gulf

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2076 MONTHLY WEATHER REVIEW VOLUME 133

FIG. 6. (a) Reflectivity from the Cabo San Lucas radar at 0626 UTC 22 Sep 2003, and (b) reflectivity from the Guasave radar at 1447 UTC 22 Sep 2003. In (a) and (b), the dot indicates the location of the center derived from image animations, and labels (A and B) mark refer- ence features discussed in the paper. The radar is located at the center of the domain.

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2077

FIG. 7. Total rainfall from the network of rain gauges from 1500 UTC 19 Sep to 1500 UTC 25 Sep 2003. Rainfall is in units of mm and shading is as shown in the grayscale. Plus symbols indicate stations used in this plot. Contour for terrain elevation (solid line; 500 m) outlines mountain location. Best-track data include time (day/hour) and intensity using the notation defined in Fig. 4. coast between 24.3° and 25.6°N, less reliable estimates est pixel to selected locations in BCS. These series also of rainfall are available there. depict hourly precipitation from nearby automatic sur- From Fig. 7, rainfall indicates distinct areas exceed- face stations, and as expected, precipitation tends to be ing 300 mm within a period of several days, represent- greatest as the core of the tropical cyclone approaches ing the typical annual rainfall budget received in the the station. For example, in Cabo San Lucas (Fig. 8a), region each year. In fact, long-term annual precipita- this characteristic occurred from 0600 to 1200 UTC 22 tion for most of these stations is between 50 and 450 September, with a peak rate of 84 mm hϪ1 and mini- mm (SMN 2004). Marty provided an extraordinary con- mum cloud-top temperature of –76°C. By contrast, tribution to the water resources that are normally avail- rates were lower in La Paz (Ͻ25 mm hϪ1; Fig. 8b), with able. The rainfall near Cabo San Lucas from this event the minimum cloud-top temperature at –80°C. Another is larger than amounts recorded during the passage of high rate (Ͼ50 mm hϪ1) occurred at Santa Rosalía (Fig. recent tropical cyclones, with the exception of Hurri- 8c), with a minimum cloud-top temperature near cane Juliette in 2001 (Table 1). –50°C. Precipitation in convective systems is known to fol- Figure 8 shows significant rainfall rates for 6- to 12-h low an empirical relationship with estimates derived periods. Total precipitation from automatic stations is from geostationary satellites (Vicente et al. 1998). To summarized in Table 2. Steep topography along the compare observations of rainfall against satellite- peninsula and the heavy rains from derived data, Fig. 8 shows time series of cloud-top tem- (2003) in late August were ideal conditions for gener- peratures from GOES-10 infrared imagery at the clos- ating flash in some watersheds. For example, the

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2078 MONTHLY WEATHER REVIEW VOLUME 133

FIG. 8. Time series of rainfall in mm (thick line; left axis) from automatic surface stations and cloud-top temperature in °C (dotted line; right axis) derived from GOES-10. Stations are (a) Cabo San Lucas for 0900 UTC 21 Sep–0900 UTC 23 Sep, (b) La Paz for 1330 UTC 21 Sep–1330 UTC 23 Sep, and (c) Santa Rosalía for 0000 UTC 22 Sep–0000 UTC 24 Sep 2003. Rainfall readings are from previous 60 min with respect to the assigned hour. Horizontal, dotted lines mark the location of the minimum temperature in the vertical axis. rainfall in Santa Rosalía was more than 2.5 times the Paz recorded 47% of the normal annual rainfall. This long-term, annual precipitation (79 mm; SMN 2004). moderate amount (82.8 mm) is supported by readings Cabo San Lucas received 91% of their normal annual from nearby network sites (Fig. 7). Stations located rainfall. In contrast, the SEMAR station north of La within a radius of 15 km of La Paz reported rainfall

TABLE 2. Surface observations from automatic surface stations in Baja California during the passage of Tropical Cyclone Marty, 22–23 Sep 2003.

Station MAXSPD MAXGST MINPRS TTLRNF Cabo San Lucas 41.1 52.3 967.0 184.0 SEMAR 17.0 23.9 973.6 82.8 Santa Rosalía 17.6 27.5 991.8 201.9 Bahía de los Angeles 16.1 22.5 995.0 22.1

MAXSPD ϭ Maximum sustained wind speed (m sϪ1). MAXGST ϭ Maximum peak gust (m sϪ1). MINPRS ϭ Minimum sea level pressure (mb). TTLRNF ϭ Total storm rainfall (mm).

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2079

FIG. 9. Time series of observations at (a) Cabo San Lucas for 0900 UTC 21 Sep–0900 UTC 23 Sep, (b) La Paz for 1330 UTC 21 Sep–1330 UTC 23 Sep, and (c) Santa Rosalía for 0000 UTC 22 Sep–0000 UTC 24 Sep 2003. Sustained wind speeds (m sϪ1; left axis) and sea level pressure (mb; right axis) are indicated with circles and plus symbols, respectively. Full wind barb represents 5.0 m sϪ1 and half barb 2.5 m sϪ1. Dotted lines mark the location of maximum speeds in the vertical axis and wind barbs use same notation as in Fig. 1.

ranging from 115 to 121 mm, with the largest amount Data from the SEMAR station, located approxi- falling on 22 September. mately 15 km north of La Paz (Fig. 12b), are used to analyze storm characteristics as it moved into the Gulf d. Automatic stations of California. The wind speed and direction at this site A time series of the sustained winds recorded at (Fig. 9b) changed between 1200 and 1500 UTC 22 Sep- Cabo San Lucas is shown in Fig. 9a and selected pa- tember, including a minimum speed of 2.5 m sϪ1 at 1330 rameters are listed in Table 2. These observations de- UTC, flanked by peaks of more than 15 m sϪ1, indicat- fine a point of maximum intensity at 0850 UTC 22 Sep- ing passage of the storm center near the station. Since tember with sustained speeds of 41.1 m sϪ1 and peak the pressure is not as low as that recorded at Cabo San gusts above 50.0 m sϪ1. These values are only 5%–12% Lucas, the reduced intensity was, in part, caused by the higher than those detected during the approach of Juli- interaction with the land surface. The minimum pres- ette (Table 1). Figure 9a includes pressure observations sure from the station’s data is 973.6 mb at 1345 UTC. A reduced to sea level, used to define a minimum pressure ship anchored in La Paz recorded 971.9 mb (Beven of 967.0 mb at 0920 UTC. This information supports the 2004). A station located 15 km west of La Paz, operated statement by the NHC (Franklin 2004) that landfall by Centro de Investigaciones Biológicas del Noroeste occurred at 0930 UTC with minimum pressure of 970 (CIBNOR) and San Diego State University (SDSU), mb and sustained speeds up to 43.3 m sϪ1. recorded 972.3 mb. This information is included in

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2080 MONTHLY WEATHER REVIEW VOLUME 133

Table 2 and helped locate the storm center at 1200 best-track positions and official forecasts that are avail- UTC to the west of the SEMAR site, over the city of La able over the Internet (NHC 2003). During the 2003 Paz. (Figure 12b shows station locations.) season, these forecasts assimilated information avail- The time series from observations in Santa Rosalía able at 0000, 0600, 1200, and 1800 UTC, and each prod- (Fig. 9c) reflects the storm’s approach between 1800 uct contains projections valid at 12, 24, 36, 48, 72, 96, UTC 22 September and 0600 UTC 23 September, and 120 h. For this study and to simplify the analysis of Ϫ reaching a maximum speed of 17.6 m s 1 and gusts up tracks prior to landfall, we limit the examination to the Ϫ to 27.5 m s 1. These data and a minimum pressure of 0000 UTC cycle from 19 to 22 September, roughly 84– 991.8 mb support gradual weakening of the storm, as 12 h before Marty was first placed over land in the the best track now classified the system as a tropical best-track data at 1200 UTC on 22 September. storm over the central part of the gulf. The period NHC tracks are based on guidance from a consensus 0120–0410 UTC contained missing data, which was of several numerical models (Goerss 2000). In the east- caused by technical problems during the broadcast. Ob- ern Pacific, this includes the GFS, as well as global servations from Bahía de Los Angeles (Fig. 3) suggest models from the Geophysical Fluid Dynamics Labora- passage of the storm about 1200 UTC on 23 September. tory (GFDL), the Navy Operational Global Atmo- The corresponding series (not shown) indicates a cy- spheric Prediction System (NOGAPS), and the U.K. Ϫ clonic wind shift with maximum speed of 16.1 m s 1 and Met Office (UKMET). According to Franklin et al. Ϫ peak gusts up to 22.5 m s 1 (Table 2), in agreement (2003), the consensus forecasts were found to provide a with the tropical storm intensity maintained by the reasonable degree of accuracy in the prediction of the NHC. Hurricane Kenna (2002) track at landfall over the west We compared the intensity from Marty’s passage at coast of Mexico. Because of close communication be- Cabo San Lucas to cases discussed in section 1. Maxi- tween the NHC and the SMN (Rosengaus 2001), the mum sustained wind speed during Marty (Fig. 9a) was consensus track is an important guidance tool that is slightly higher than the case of Juliette (2001; Fig. 1d). available to operational forecasters in Mexico. However, Marty had larger forward motion at landfall For the first official forecast issued at 0000 UTC 19 (Fig. 3) and winds in the range of tropical storm and September, Marty takes a westward track controlled by hurricane strength that lasted for a limited period. Ad- the presence of a high pressure system north of the ditionally, Marty had only a 6–12 h of intense rainfall storm and is associated with significant zonal motion (Fig. 8a) and the rain gauge network recorded a smaller (Fig. 10). The forecast of 0000 UTC 20 September amount across the southern part of the peninsula (Fig. showed a ridge over the Gulf of Mexico, resulting in a 7), about 25%–40% of the maximum recorded from moderate northward motion. Additionally, the associ- Juliette. ated discussion (NHC 2003) recognized the possibility of movement toward land and the SMN (2003) started 4. Forecasts and simulations issuing hurricane watches for BCS. Passage over the peninsula was expected from the forecast of 21 Septem- The previous section discussed evolution in track and ber: the hurricane was expected to make landfall near structure experienced by Hurricane Marty over Baja 25°N and later move over the northern gulf. Finally, California. Here, we provide a discussion of predictions northward motion was kept in the forecast of 22 Sep- prior to landfall. Data for the projected development of tember with landfall along the southwestern coast of the circulation are derived from real-time forecasts is- the peninsula, and this was expected to occur between sued by the NHC. Additionally, high-resolution meso- the next 12–24 h. In summary, the official forecasts ex- scale model (MM5) simulations examine general as- perienced significant changes during the 84–12 h prior pects of the evolution of the low-level winds and pre- to landfall. cipitation during the interaction of the incident storm with the landmass in the southern part of the peninsula. The spatial and temporal resolution of the model out- b. Model description put is also applied, in an attempt to enhance the limited Further information for the study of the evolution of number of in situ observations that are available in the Marty’s structure at landfall is derived from numerical area of interest. simulations performed with the MM5 (Grell et al. 1995). The model is successful in analyzing convective a. Official forecasts systems in tropical and subtropical areas. In particular, A sample of the NHC forecasts issued during the MM5 reproduces changes in motion and structure of development of Marty is shown in Fig. 10. This includes tropical cyclones at landfall over coastal areas of the

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2081

FIG. 10. Official forecast tracks issued at 0000 UTC 19, 20, 21, and 22 Sep 2003. Large circles indicate best-track positions and small circles are forecasts valid at 12, 24, 36, 48, 72, 96, and 120 h. Tracks derived from MM5 simulations, initialized at 0000 UTC 21 Sep (S21; cross symbols) and 0000 UTC 22 Sep (S22; plus symbols and thick line), are shown at 6-h intervals. The asterisk indicates best-track landfall position. western Atlantic (Liu et al. 1997), western Pacific (Wu Miller 1986) and the 30-km grid (Kain and Fritsch et al. 2002), and eastern Pacific (Farfán and Zehnder 1993). 2001). These case studies used high-resolution compu- Initial fields are derived from the GFS analysis- tational grids, adequate physical parameterizations, and forecast system with a 6-h and 100-km resolution. This realistic large-scale fields at initialization. system is used as input for the MM5 simulations be- To simulate Marty’s landfall, model integrations are cause the GFS is one of the operational models used for computed with a 90-km-resolution grid and three addi- tropical cyclone guidance at the NHC, and it is avail- tional, nested grids with resolutions of 30, 10, and 3.3 able on the Internet in real time, which allows a variety km (Fig. 11). The nested configuration is used to im- of local modeling applications. In particular, we are in- prove the model resolution of the coarse grid. There terested in the generation of short-term (48–72 h) simu- are 26 vertical sigma levels and the top of the atmo- lations because they can be executed in a near-real-time sphere is set at the 100-mb level. The model applies the mode. This means processing of data immediately upon following schemes of physical parameterization: bound- GFS product availability, and the NHC storm location ary layer (Hong and Pan 1996), radiative transfer (Du- and intensity data are in the public domain. dhia 1989), and explicit moisture with simple ice physics To generate fields with a realistic representation (Dudhia 1989). Representation of cumulus convection of storm structure and motion, simulations are initial- is applied to the 90-km grid (Betts 1986; Betts and ized prior to the landfall of Marty. Landfall occurred at

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2082 MONTHLY WEATHER REVIEW VOLUME 133

FIG. 11. Configuration and geographical coverage of the grids applied in the MM5 simulations. The spatial resolution is 90 km (D01), 30 km (D02), 10 km (D03), and 3.3 km (D04).

0930 UTC 22 September (Franklin 2004), when the cir- since maximum low-level winds from the GFS are 18 culation center was near Cabo San Lucas (asterisk in msϪ1. This represents 54% of the best-track intensity Fig. 10). Initialization times are 0000 UTC 21 Septem- and justifies an adjustment to the strength of the initial ber and 0000 UTC 22 September, which correspond to circulation. 33.5 and 9.5 h before the event, respectively. As shown Figure 12b shows (SST) for in Fig. 10, the corresponding official tracks take the selected contour levels, along with a representation of initial vortex northward over the central and southern terrain elevations, in the area of landfall. The distribu- part of the peninsula. Another reason to choose these tion of SST is in agreement with estimates from satellite initial times is based on recommendations from forecast (not shown), as areas off southern BCS have values in centers and public management offices (e.g., Marks et al. the range of 27°–30°C and distinct maxima above 30°C 1998). According to these recommendations, useful in- occur over several sectors in the gulf. The S21 simula- formation to mitigate storm damage is achieved by hav- tion was run for 72 h and the track followed by the ing accurate forecasts within 48 h of landfall. This pro- circulation is shown in Figs. 10 and 12b. For the pur- vides timely warnings and, in some cases, evacuation poses of this study, the center of circulation of Marty is decisions before strong winds and heavy rainfall arrive. determined from an examination of the 10-m stream- lines in the model output. c. Model simulations In general, the S21 simulation track agrees with the best-track positions as the model circulation moves 1) INITIALIZATION ON 21 SEPTEMBER northward, makes landfall over the western peninsula, The first simulation (S21) was initialized at 0000 and then passes over the central gulf. At landfall, the UTC 21 September. Figure 12a shows the correspond- center of circulation is west of the best track and this is ing GFS analysis for midlevel winds as well as observed likely due to the steering flow associated with a data. Initially, the circulation associated with Marty was midlevel located over the northern gulf. approximately 500 km south of Cabo San Lucas and This feature is shown in Fig. 12a and remains in the area had best-track, sustained winds of 33.5 m sϪ1. The fig- during the first 48 h of the simulation. Position errors ure indicates a cyclonic circulation centered near the increase as the low-level circulation, with a diameter same location; however, a bogusing scheme (Low-Nam about 300 km, moves over land west of the mountain and Davis 2001) is required to modify the MM5 vortex, range in central BCS (Fig. 12b). However, the S21

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2083

FIG. 12. (a) Analysis winds from the GFS and observations at the 500-mb level, and best-track position (cross symbol) at 0000 UTC 21 Sep 2003; (b) model track (cross symbol; valid hour on 22 Sep), terrain elevations (m), and SST (°C); (c) 10-m streamlines and isotachs (m sϪ1) valid at 0500 UTC 22 Sep; and (d) accumulated precipi- tation (mm) during the period 0300–0900 UTC 22 Sep. Full barb represents 5.0 m sϪ1 and half barb 2.5 m sϪ1. SST contour interval is 1°C and terrain elevations are shown at 300, 600, 900, and 1200 m. Model simulation is initialized at 0000 UTC 21 Sep. simulation provides a better prediction of storm land- resolution grid over the southern part of the peninsula fall and reduced magnitude of position errors when (D04 in Fig. 11). At this time, the low-level circulation compared to the corresponding official forecast. In the has strong wind speeds (Ͼ20 m sϪ1) over water and the official forecast, the circulation associated with Marty flow is weaker over land. The vortex has maximum makes landfall near 25°N and reaches the Gulf of Cali- speeds of 30 m sϪ1 at 20 km east of Cabo San Lucas and fornia south of 30°N (Fig. 10). a minimum, central pressure of 981 mb. This is com- Figure 12c shows that landfall occurs near Cabo San pared with 41.1 m sϪ1 and 967 mb from the series at Lucas, at 29 h into the model simulation (0500 UTC 22 Cabo San Lucas (0850 UTC in Fig. 9a), which indicates September), approximately 4–5 h before the actual that the model is underpredicting the actual circulation event. This figure includes fields from the high- intensity at landfall by 11.1 m sϪ1 and 14 mb.

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2084 MONTHLY WEATHER REVIEW VOLUME 133

Figure 12c also shows that a region of moderate 1500 UTC satellite (Fig. 5) and radar (Fig. 6b) imagery winds (15–20 m sϪ1) affects most coastal areas south of indicated the circulation centered over the gulf, north 24°N, and the region extends offshore to cover a por- of La Paz. This fact suggests that the model simulation tion of the western gulf and the Pacific. This range in has a westward bias and that the position in the official wind speed is important since it represents equivalent forecast is in better agreement with the best-track pat- tropical storm conditions over these areas. Among tern (Fig. 10). other features that are important in the S21 simulation In spite of the westward bias, it is interesting to ex- is the capability to provide a reasonable amount and amine some characteristics of the incident vortex in the distribution of precipitation consistent with the rain S22 simulation close to the time of landfall. At 0900 gauge observations (Fig. 7). Accumulated rainfall UTC (Fig. 13a), streamlines indicate a circulation with around the landfall time of the model’s simulation (Fig. maximum speeds in the range of 20–25 m sϪ1 were lo- 12d) indicates a well-defined maximum greater than 75 cated around the center of circulation and maximum mm, and this is a pattern that developed along the speeds occurred over water. The circulation moved southern mountains. In this area, the low- to midlevel northward, and at 1200 UTC (Fig. 13b) is located near flow is perpendicular to the mountain ridge and ac- the coastline, 50 km west of the best-track position, and Ϫ1 quires large vertical velocities (up to 4 m s ). These displays a narrow band of speeds above 25 m sϪ1.In features are associated with enhanced precipitation in addition to observations over land, Fig. 13b includes the model and are consistent with the area of deep low-level winds over the ocean that are derived from convection detected by satellite imagery (Fig. 4). the National Aeronautics and Space Administration’s (NASA) Quick Scatterometer (QuikSCAT), and these 2) INITIALIZATION ON 22 SEPTEMBER data are used to confirm the presence of a closed cir- culation over the southern peninsula with strong winds Further reduction in position errors is achieved dur- over the gulf. The maximum winds from QuikSCAT ing the second model simulation, initialized at 0000 are located around 24°N, 109.5°W and have a south- UTC 22 September. This simulation is identified as S22, Ϫ easterly direction and a magnitude of 25 m s 1. Near runs for 48 h, and its track is shown in Fig. 10. In this this region, model speeds are in the range of 20–23 case, Marty is represented by a vortex with an initial msϪ1 and a wide band in the tropical storm threshold location 210 km south of Cabo San Lucas and maxi- Ϫ1 Ϫ (15–20 m s ) covers the gulf to 25.5°N. These facts mum wind speeds of 44 m s 1. This is a reference in- indicate that the model is able to simulate an extensive tensity used to modify the original (GFS) circulation, Ϫ1 area under tropical storm conditions over the southern which had maximum, low-level winds of 20 m s . The model circulation moves west of the southern moun- gulf and that maximum wind speeds are consistent with tains, and in contrast with S21, reaches the gulf near La available observations. Paz, which is in better agreement with observations The center of circulation moved over land, followed from satellite and radar imagery. Additionally, the a path coinciding with lower terrain, and at 1500 UTC model track ends along the central and northern gulf, (Fig. 13c) was located near the gulf. The maximum Ϫ1 while the official forecast takes the circulation over the wind speeds were over the gulf and an area of 25 m s Ϫ1 mainland and into after 48 h. In this simula- was located north of La Paz, in contrast to the 15 m s tion, substantial northward motion in northwestern sustained winds recorded at the SEMAR site. Over the Mexico is most likely influenced by the steering flow of following few hours, the center of circulation moved an anticyclonic circulation (not shown) centered over along the east coast of the peninsula, and at 1800 UTC at the 500-mb level. (Fig. 13d), was located 50 km southwest of the best To illustrate structural characteristics of low-level track. At this time, the circulation had intensified and a Ϫ flow, the distribution of 10-m streamlines and isotachs zone with speeds above 25 m s 1 was located near at 9, 12, 15, and 18 h into the S22 simulation are shown 25.5°N. In contrast, a slower flow, with a significant in Fig. 13, along with observed winds from all available component from the south, is present over southern surface stations described in section 2. A close ap- BCS, and this agrees with reports issued by the network proach to Cabo San Lucas occurred at 0900 UTC 22 of surface stations. September (Fig. 13a) and at 1200 UTC (Fig. 13b). The The pattern of precipitation derived from fields in cyclonic circulation is located along the BCS coast, west S22 is consistent with those implied from satellite im- of the best-track position. Since there are a limited agery and rain gauge data. In Fig. 14, the distribution number of reports, it is difficult to determine the actual and intensity of precipitation during 6-h intervals (Figs. position of the center of the hurricane. However, by 14a-c) and total precipitation predicted during the

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2085

Ϫ FIG. 13. Model streamlines and isotachs (shaded; m s 1) at the 10-m level valid at (a) 0900, (b) 1200, (c) 1500, and (d) 1800 UTC 22 Sep 2003. Observed winds are from the surface stations and QuikSCAT. Wind barbs use same notation as in Fig. 12a. Dots indicate corresponding best-track positions, asterisk best-track landfall, and thick lines terrain elevations at 300, 600, 900, and 1200 m. Position at 1500 UTC is interpolated from 1200 and 1800 UTC. Model simulation is initialized at 0000 UTC 22 Sep and data are derived from the 3.3-km grid.

storm’s passage (Fig. 14d) are shown. As the center of associated with strong vertical flow that developed over circulation enters the 3.3-km grid, a core area of heavy the mountain ridge. Most of the precipitation near rainfall (Ͼ75 mm) is confined to an area south of Cabo Cabo San Lucas fell from 0600 to 1200 UTC, consistent San Lucas from 0000 to 0600 UTC 22 September (Fig. with the rainfall record from the automatic station 14a). Later, the center of circulation approached the (Fig. 8a). peninsula and precipitation maxima occurred off the The pattern of precipitation moved northward and southwestern coast and along the flanks of the southern covered the central mountains (Fig. 14c). During the mountains (Fig. 14b). Precipitation in the range of 75– next 6 h, intense rainfall bands developed with a maxi- 100 mm occurred at terrain elevations above 600 m. As mum near 25.5°N, 111.0°W over the gulf near the east- discussed in the S21 simulation, rainfall maximum was ern coast of the peninsula. The local accumulation was

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2086 MONTHLY WEATHER REVIEW VOLUME 133

FIG. 14. Model precipitation during the intervals (a) 0000–0600, (b) 0600–1200, (c) 1200–1800, and (d) 0600–2100 UTC 22 Sep 2003. Amounts of precipitation are indicated in mm and thick lines are terrain elevations at 300, 600, 900, and 1200 m. Model simulation is initialized at 0000 UTC 22 Sep and data are derived from the 3.3-km grid. above 75 mm. Inspection of the model fields indicates 0600–2100 UTC 22 September is shown in Fig. 14d. that this results from enhanced, low-level convergence Lesser amounts fell over low-level terrain, especially along the coast and upslope flow across the mountain around La Paz and along the Pacific coast north of range. These conditions originated from a fetch of 24°N. In contrast, distinct areas of high rainfall were strong, easterly winds over water and located in the generated over the southern and central mountains, northeast quadrant of the vortex from 1500 to 1800 consistent with the records from the regional network UTC. Comparing this feature with Marty’s of rain gauges (Fig. 7). However, precipitation derived detected from satellite (Fig. 5) and radar imagery (Fig. from the simulation underpredicted by about 30%– 6b), during passage over the central gulf, the S22 simu- 75%. This deficiency may be due, in part, to model lation provided a reasonable representation of the dis- sensitivities in the representation of boundary layer tribution of precipitation associated with Marty. processes and cloud microphysics (e.g., Braun and Tao A summary of total rainfall, starting at landfall, for 2000). These aspects are not discussed in this paper and

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2087

FIG. 15. Model tracks of Tropical Cyclone Marty from MM5 simulations initialized at 0000 UTC 22 Sep 2003. Positions are given at 3-h intervals for cases with (S22; crosses) and without topography (N22; triangles). Best-track positions are represented with dots. Simulation time (hours into the run) and best-track times (UTC) are to the right of positions. Terrain eleva- tions (dashed lines) for S22 are shown at 300, 600, 900, and 1200 m.

are the subject of a subsequent study that will discuss along with the S22 and best-track positions. A compari- the physical mechanisms involved in the accurate rep- son of these data is used to estimate the impact of the resentation of storm structures at landfall in Baja Cali- peninsular mountains in the path of the incident, cy- fornia. clonic circulation associated with Marty. During the first 12 h, the N22 track is slightly east of 3) SENSITIVITY TO MODEL TOPOGRAPHY the one obtained from the S22 case, as the center of circulation moved northward. The circulation makes To evaluate the sensitivity of the track generated in landfall near Cabo San Lucas and did not experience the model, with respect to the topography in BCS, an the deflection to the Gulf of California that occurred in additional simulation (N22) was performed. The N22 S22. Instead, motion in N22 is over the western side of simulation used initial fields at 0000 UTC 22 September the peninsula to reach the Pacific coast at 25.5°N and and applied the same physical parameterizations as after 24 h into the simulation (0000 UTC 23 Septem- S22; however, it has a modified topography with eleva- ber). At this time, the storm movement in the simula- tions set at 1 m. Figure 15 shows the track from N22, tion with topography and best track are along the gulf,

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2088 MONTHLY WEATHER REVIEW VOLUME 133 north of 26°N. This result suggests that the central available data shows that, during Marty’s landfall, sev- mountains act as a barrier to westward motion and eral stations detected significant changes in winds, pres- makes S22 a reasonable simulation of the actual move- sure, and precipitation. The information derived from ment into the gulf. these stations (Figs. 8 and 9) is consistent with the best- The results from the above simulations are consistent track estimates for storm position and intensity. with findings reported in other studies. For example, Official forecasts from the NHC (Fig. 10) had a west- Daida and Barnes (2003) claim a minimal response on ward bias in the initial tracks and the errors decreased the track of Hurricane Paine (1986) during its approach during subsequent forecast cycles. The bias is related to close to the southern mountains of BCS. This range has a large-scale, anticyclonic circulation north of Marty. a length of about 100 km (Fig. 2) and was not able to The first track predicting landfall over Baja California significantly alter the track during Marty’s passage. In was issued on 21 September, more than 24 h before the contrast, the central mountains have a length of 300 km event. The prediction of the landfall site was improved and were associated with motion over the gulf in S22. over time, and at 9.5 h in advance, Marty was expected According to Fig. 10, the occurrence of a similar track to move over the southern part of BCS and enter the over these mountains was present in the first simulation central gulf, which is in agreement with the observed (S21). The circulation motion in S21 and S22 is consis- storm motion. tent with track changes, documented by Bender et al. To enhance spatial and temporal resolution of the (1987), during the interaction of tropical cyclones with NHC forecasts, we used output generated from the mountains that have horizontal scales of 300 km. In this MM5 simulations. These simulations made use of ad- study, the tracks are modified toward the north of the equate physical parameterizations and high-resolution, direction of storm motion for cases in the northwestern nested grids. The model configuration includes a 3.3- Pacific and the Caribbean. km grid over the southern peninsula (Fig. 11), and this was used to determine wind and rainfall distribution 5. Summary and conclusion during the storm approach. A couple of model simula- tions were initialized at 33.5 h (S21) and 9.5 h (S22) The purpose of this study is to characterize the track before landfall, with data from the GFS. Both cases and evolution of the circulation associated with Hurri- were run until the circulation was located over the cane Marty over Baja California in 2003. According to northern gulf. The simulations show the ability to pro- the NHC best track (Fig. 3), the tropical depression vide a reasonable prediction of storm motion and a originated south of the Mexican coast and reached the southern peninsula on 22 September. This resulted in circulation structure that is consistent with the informa- strong winds and heavy precipitation over land as the tion derived from available observations. Section 4c storm system moved toward the Gulf of California. The provides a discussion of the results derived from these observations during the approach and landfall of the simulations. storm are described in section 3. Among the benefits derived from the model simula- The impact of Marty in Cabo San Lucas, when com- tions, we were able to determine landfall time within 5 pared with Juliette (2001), indicates that 1) the center h of the actual occurrence. The S21 simulation showed of circulation associated with Marty moved closer to reduced position errors, with respect to official fore- the peninsula while Juliette passed farther away; 2) casts, placing the incident vortex along the southwest- maximum wind speeds were similar, but Marty’s large ern coast of BCS near Cabo San Lucas. Additionally, forward motion resulted in a shorter duration of hurri- S22 projected the circulation across the peninsula near cane and tropical storm conditions; and 3) intense pre- La Paz, and later into the central gulf. Even though cipitation from Marty lasted only 6–12 h (Fig. 8), while landfall was approximately 50 km west of that derived the rainfall from Juliette lasted several days and re- from the best track, these simulations are useful to ex- sulted in larger accumulations over the mountainous amine aspects beyond findings from observations. This terrain of southern BCS. includes the distribution of surface winds along coasts Analysis of observations from automatic stations and precipitation derived from interaction of the low- (SMN and SEMAR) is an important tool to study storm level flow with the topography. However, the model impact by providing high, temporal data near the sur- underpredicted maximum intensity, including winds face. This tool was first applied by Rosengaus (2001) near landfall and total rainfall over the mountains dur- to monitor Hurricane Keith (2000) in southeastern ing passage of the storm. Simulation N22 postulated flat Mexico and, later, by Farfán (2004) for Juliette’s land- topography, which was useful to show that the central fall over BCS in 2001. In this paper, the analysis of mountains are an essential element in the realistic rep-

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC JULY 2005 F A R F Á N AND CORTEZ 2089 resentation of Marty’s track over the Gulf of California Beven, J. L., II, cited 2004: Eastern North season (Fig. 15). of 2003. [Available online at http://www.vos.noaa.gov/MWL/ In conclusion, the analysis of Marty’s landfall repre- april_04/np_hurr.shtml.] Braun, S. A., and W.-K. Tao, 2000: Sensitivity of high-resolution sents a contribution to the current knowledge of tropi- simulations of Hurricane Bob (1991) to planetary boundary cal cyclone impacts on Baja California. A previous layer parameterizations. Mon. Wea. Rev., 128, 3941–3961. study by Farfán (2004) used the network of surface Clark, J. D., 1983: The GOES user’s guide. NOAA/NESDIS, 169 observations to document the evolution of a system pp. with a limited number of stations over land, and, to Daida, S. K., and G. M. Barnes, 2003: Hurricane Paine (1986) grazes the high terrain of the Baja California peninsula. Wea. analyze storm evolution over areas that lack data, Forecasting, 18, 981–990. Farfán and Zehnder (2001) applied mesoscale model- Dudhia, J., 1989: Numerical study of convection observed during ing. This study of Marty suggests that mesoscale model winter monsoon experiment using a mesoscale two-dimen- simulations are an additional tool that may be consid- sional model. J. Atmos. Sci., 46, 3077–3107. ered by regional forecast centers to improve guidance Ebert, E. E., and G. J. Holland, 1992: Observations of record cold cloud-top temperatures in Tropical Cyclone Hilda (1990). used to formulate short-term predictions. In particular, Mon. Wea. Rev., 120, 2240–2251. the spatial distribution of low-level flow and precipita- Englehart, P. J., and A. V. Douglas, 2001: The role of Eastern tion during landfall should improve by performing ad- North Pacific tropical storms in the rainfall climatology of equate simulations. This guidance tool should supple- western Mexico. Int. J. Climatol., 21, 1357–1370. ment the track predictions derived from the consensus Farfán, L. M., 2004: Regional observations during the landfall of Tropical Cyclone Juliette (2001) in Baja California, Mexico. forecasts issued by the NHC. Mon. Wea. Rev., 132, 1575–1589. ——, and J. A. Zehnder, 2001: An analysis of the landfall of Hur- Acknowledgments. This study was supported by ricane Nora (1997). Mon. Wea. Rev., 129, 2073–2088. CICESE (L. M. Farfán) and SMN (M. Cortez). Addi- Franklin, J. L., cited 2004: Hurricane Marty: 18–24 September tional support was provided under CONACYT Grant 2003. Tropical cyclone report. [Available online at http:// SEP-2003-C02-42942. The authors thank Direcciónde www.nhc.noaa.gov/2003marty.shtml.] ——, L. A. Avila, J. L. Beven, M. B. Lawrence, R. J. Pasch, and Meteorología Marina (SEMAR, Capt. Juan M. Aguilar S. R. Stewart, 2003: Eastern North Pacific hurricane season Morales) for providing automatic station observations, of 2002. Mon. Wea. Rev., 131, 2379–2393. Dirección de Meteorología y Telecomunicaciones Goerss, J. S., 2000: Tropical cyclone track forecasts using an en- Aeronáuticas (SENEAM, Joaquín H. Rodríguez semble of dynamical models. Mon. Wea. Rev., 128, 1187– Hernández) for airport reports, and Global Change Re- 1193. search Group at San Diego State University (Steven J. Grell, G. A., J. Dudhia, and D. R. Stauffer, 1995: A description of the fifth-generation Penn State/NCAR Mesoscale Model Hastings) for the CIBNOR/SDSU data. The MM5 is (MM5). NCAR Tech. Note NCAR/TN-398ϩSTR, 122 pp. maintained by NCAR, which is funded by the National Hong, S.-Y., and H.-L. Pan, 1996: Nonlocal boundary layer ver- Science Foundation. Julián J. Delgado Jiménez pro- tical diffusion in a medium-range forecast model. Mon. Wea. vided technical assistance with the model simulations at Rev., 124, 2322–2339. CICESE’s facility in Ensenada. Comments and sugges- INEGI, cited 2000: General census of population and housing (in Spanish). [Available online at http://www.inegi.gob.mx/est/ tions by the anonymous reviewers helped to improve default.asp?cϭ701.] the manuscript. Jáuregui, E., 2003: Climatology of landfalling hurricanes and tropical storms in Mexico. Atmósfera, 16, 193–204. REFERENCES Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain–Fritsch scheme. The Repre- Avila, L. A., and M. Mayfield, 1995: Eastern North Pacific hurri- sentation of Cumulus Convection in Numerical Models, Me- cane season of 1993. Mon. Wea. Rev., 123, 897–906. teor. Monogr., No. 46, Amer. Meteor. Soc., 165–170. ——, R. J. Pasch, J. L. Beven, J. L. Franklin, M. B. Lawrence, Liu, Y., D.-L. Zhang, and M. K. Yau, 1997: A multiscale numeri- S. R. Stewart, and J.-G. Jiing, 2003: Annual summary: East- cal study of Hurricane Andrew (1992). Part I: Explicit simu- ern North Pacific hurricane season of 2001. Mon. Wea. Rev., lation and verification. Mon. Wea. Rev., 125, 3073–3093. 131, 249–262. Low-Nam, S., and C. Davis, 2001: Development of a tropical cy- Bender, M. A., R. E. Tuleya, and Y. Kurihara, 1987: A numerical clone bogusing scheme for the MM5 system. Preprints, 11th study of the effect of island terrain on tropical cyclones. Mon. PSU/NCAR Mesoscale Model Users’ Workshop, Boulder, Wea. Rev., 115, 130–155. CO, National Center for Atmospheric Research, 130–134. Betts, A. K., 1986: A new convective adjustment scheme. Part I: Marks, F. D., L. K. Shay, and PDT-5, 1998: Landfalling tropical Observational and theoretical basis. Quart. J. Roy. Meteor. cyclones: Forecast problems and associated research oppor- Soc., 112, 677–691. tunities. Bull. Amer. Meteor. Soc., 79, 305–323. ——, and M. J. Miller, 1986: A new convective adjustment Menzel, W. P., and J. F. W. Purdom, 1994: Introducing GOES-I: scheme. Part II: Single column tests using GATE wave, The first of a new generation of geostationary operational BOMEX, ATEX, and Arctic air-mass data sets. Quart. J. environmental satellites. Bull. Amer. Meteor. Soc., 75, 757– Roy. Meteor. Soc., 112, 693–709. 781.

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC 2090 MONTHLY WEATHER REVIEW VOLUME 133

NHC, cited 2003: Hurricane Marty: Advisory archive. [Available Surgi, N., H.-L. Pan, and S. J. Lord, 1998: Improvement of the online at http://www.nhc.noaa.gov/archive/2003/MARTY. NCEP global model over the Tropics: An evaluation of shtml.] model performance during the 1995 hurricane season. Mon. Rosengaus, M., 2001: Analysis of the 2000 tropical cyclone season Wea. Rev., 126, 1287–1305. in Mexico (in Spanish). Ing. Hidraul. Mex., 16, 5–23. Vicente, G. A., R. A. Scofield, and W. P. Menzel, 1998: The op- SMN, cited 2003: Survey of Hurricane Marty in the Pacific Ocean erational GOES infrared rainfall estimation technique. Bull. (in Spanish). [Available online at http://smn.cna.gob.mx/ Amer. Meteor. Soc., 79, 1883–1898. ciclones/tempo2003/pacifico/marty/marty.html.] Wu, C.-C., T.-H. Yen, Y.-H. Kuo, and W. Wang, 2002: Rainfall ——, cited 2004: Standard and provisional climatological normals, simulation associated with Typhoon Herb (1996) near Tai- 1961–1990 (in Spanish). [Available online at http://smn.cna. wan. Part I: The topographic effect. Wea. Forecasting, 17, gob.mx/productos/normales/estacion/normales.html.] 1001–1015.

Unauthenticated | Downloaded 10/02/21 10:39 AM UTC