Warm Season Lightning Distributions Over the Northern Gulf of Mexico Coast and Their Relation to Synoptic-Scale and Mesoscale Environments

AUGUST 2005 S M I T H E T A L . 415

Warm Season Lightning Distributions over the Northern Gulf of Mexico Coast and Their Relation to Synoptic-Scale and Mesoscale Environments

JESSICA R. SMITH* AND HENRY E. FUELBERG Department of Meteorology, The Florida State University, Tallahassee, Florida

ANDREW I. WATSON NOAA/National Weather Service Forecast Office, Tallahassee, Florida

(Manuscript received 30 January 2004, in final form 31 January 2005)

ABSTRACT

Cloud-to-ground lightning data from the National Lightning Detection Network are used to create a warm season (May–September) lightning climatology for the northern Gulf of Mexico coast for the 14-yr period 1989–2002. Each day is placed into one of five flow regimes based on the orientation of the low-level flow with respect to the coastline. This determination is made using the vector mean 1000–700-hPa wind data at Lake Charles and Slidell, Louisiana. Flash densities are calculated for daily, hourly, and nocturnal periods. Spatial patterns of composite 24-h and nocturnal flash density indicate that lightning decreases in an east-to-west direction over the region. Flash densities for the 24-h period are greatest over land near the coast, with relative maxima located near Houston, Texas; Lake Charles, Baton Rouge, and New Orleans, Louisiana; Biloxi, Mississippi; and Mobile, Alabama. Flash densities during the nocturnal period are greatest over the coastal waters. Lightning across the northern Gulf coast is closely related to the prevailing low-level synoptic flow, which controls the sea breeze, the dominant forcing mechanism during the warm season. Southwest flow, the most unstable and humid of the five regimes, exhibits the most flashes. In this case, sea-breeze-induced convection is located slightly inland from the coast. Northeast flow, the driest and most stable of the regimes, exhibits the least amount of lightning. The large-scale flow restricts the sea breeze to near the coastline. Geographic features and local mesoscale circulations are found to affect lightning across the region. Geographic features include lakes, bays, marshes, swamps, and coastline orientations. Thermal circulations associated with these features interact with the main sea breeze to produce complex lightning patterns over the area.

1. Introduction have used data from the National Lightning Detection Network (NLDN) (Cummins et al. 1998) to study light- Cloud-to-ground (CG) lightning causes injury and ning patterns across the contiguous United States. death, disrupts human activities, and damages property. Analyses by Orville (1991, 1994), Orville and Silver Understanding the mesoscale processes that lead to (1997), Huffines and Orville (1999), Orville and Huff- convective development and its resulting lightning is ines (2001), Orville et al. (2002), and Zajac and Rut- necessary to produce better forecasts. Many studies ledge (2001) show that Florida, particularly central Florida, has the greatest lightning flash densities in the nation. The Gulf of Mexico coast from Houston, Texas, * Current affiliation: NOAA/National Weather Service Fore- to Alabama exhibits secondary flash density maxima. cast Office, Birmingham, Alabama. Due to its abundant lightning, Florida has been the focus of several regional lightning studies (e.g., Maier et al. 1984; Hodanish et al. 1997; Reap 1994). Lericos et al. Corresponding author address: Henry E. Fuelberg, Dept. of Meteorology, The Florida State University, Tallahassee, FL (2002) recently created a 10-yr warm season lightning 32306-4520. climatology for the Florida peninsula as a function of E-mail: [email protected] the position of the subtropical ridge axis. They also

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

WAF870 416 WEATHER AND FORECASTING VOLUME 20 emphasized the role of coastal features in the distribu- lightning near Lake Charles, Louisiana. Several hy- tion of lightning. These previously mentioned Florida potheses have been proposed to explain the enhance- lightning studies focused on the peninsula; however, ment: sea-breeze effects, urban heat island effects, ur- Camp et al. (1998) examined lightning distributions in ban air pollution and its resulting modification of mi- the panhandle region of northern Florida. This area has crophysical processes, and saltwater effects (Steiger et a more complex coastline and is influenced by only one al. 2002; Orville et al. 2001; Steiger and Orville 2003). large body of water, the Gulf of Mexico. The current study creates a detailed lightning clima- The northern Gulf coast is a hotbed of lightning ac- tology for the northern Gulf of Mexico coast (Fig. 1). tivity, ranking second behind Florida (e.g., Orville and This region of major lightning activity has received little Huffines 2001). Steiger and Orville (2003) found re- previous attention. The area contains major contrasts in gions of enhanced lightning over portions of southern surface characteristics—forests, sandy beaches, exten- Louisiana. However, the northern Gulf region has not sive swamps, large lakes, and major cities. The thermal been examined in detail. contrasts between these surfaces can produce circula- The sea breeze is an important factor in producing tions that interact with each other and with the main thunderstorms in coastal regions, and many investiga- Gulf of Mexico sea breeze to form complex patterns of tors have examined it in detail. Wexler (1946) and convection and lightning. We seek to answer the fol- Simpson (1994) give complete descriptions of the sea lowing questions: 1) What are the areal distributions of breeze. Estoque (1962) was one of the first to model the lightning as a function of the low-level synoptic flow? 2) sea breeze in two dimensions, while Pielke (1974) stud- How do the number of flashes and their spatial patterns ied Florida’s sea breezes using an early three-dimen- vary as a function of time? 3) What are the physical sional model. Arritt (1993) used a two-dimensional nu- mechanisms leading to the formation, movement, and merical model to analyze how variations in the strength timing of lightning patterns in the area? of the onshore–offshore environmental flow influenced the strength and inland penetration of the sea breeze. These studies have shown that offshore environmental 2. Data and methodology flow produces a strong sea-breeze circulation whose a. Lightning data leading edge remains near the coast. Deep convection, if it occurs, also is confined near the coastline. Onshore In complete operation since 1989, the NLDN, owned environmental flow produces a sea breeze with differ- and operated by Vaisala, Inc., detects CG lightning ent characteristics—the thermal circulation is weaker flashes over the continental United States and immedi- but advances farther inland during the day. Any asso- ate coastal waters. Specifics concerning network meth- ciated convection tends to be weaker and spread over a odology and operations are described by Cummins et larger distance from the coast. al. (1998). The network consists of 106 ground-based The effects of coastlines on the sea breeze were ex- sensors across the United States. Although CG light- plored by McPherson (1970) using three-dimensional ning flashes may consist of several return strokes, only numerical modeling and by Gibson and Vonder Haar the first stroke’s data are retained by the flash grouping (1990) using satellite imagery. Florida’s sea breezes and algorithm. Data include the flash’s time, latitude, lon- their interactions with lake and river breezes have been gitude, polarity, strength, and multiplicity. studied by several researchers, including Blanchard and The detection efficiency and location accuracy of the López (1985), López and Holle (1987), Wakimoto and NLDN have improved greatly since its inception. Dur- Atkins (1994), and Laird et al. (1995). ing the early years of operation, detection efficiency Recent localized studies have described the occur- ranged between 65% and 85%, while location accuracy rence of lightning near major cities. Watson and Holle was 8–16 km (Cummins et al. 1998). System upgrades in (1996) and Livingston et al. (1996) examined flash pat- 1995 allowed a greater number of flashes to be de- terns across the southeastern United States, particu- tected, thereby improving the network’s capabilities. larly near Atlanta, Georgia, in preparation for the 1996 Since these upgrades, the NLDN has a detection effi- Summer Olympics. Enhanced warm season lightning ciency of 80%–90% and is accurate to within 0.5 km activity was observed over and downwind of 16 major (Cummins et al. 1998). cities in the Midwest (Westcott 1995). Along the Gulf Following the suggestion of Cummins et al. (1998), coast, Houston, Texas, has been an area of interest. we removed from the dataset weak positive flashes hav- Steiger et al. (2002) documented enhanced lightning ing strengths less than 10 kA. The dataset also was over the city, most of which was due to large lightning examined for duplicate flashes. When two or more events. Steiger and Orville (2003) noted enhanced flashes occurred within 10 km of one another within a

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 417

FIG. 1. Map of domain extending from 28° to 32°N and 87° to 96°W. Major cities and geographical features are labeled. The boundary at 91.5°W separates the domain into eastern and western zones of equal area. Radiosonde sites are located at Lake Charles (LCH) and Slidell (SIL), LA.

1-s interval, only the first flash’s data were retained, sounding. As shown in previous studies (López and although their multiplicities were combined (Cummins Holle 1987; Camp et al. 1998; Lericos et al. 2002), the et al. 1998). No corrections were applied to compensate flow within this layer provides a good indication of sea- for the variations in detection efficiencies and location breeze and thunderstorm movement during the warm accuracies across the study area during the 14-yr pe- season. Two radiosonde sites were chosen to describe riod. This produces a slight underestimation in the flash the low-level flow in the region—Lake Charles, Loui- densities that follow. siana (LCH), in the western portion of the domain, and We used data from the months of May–September Slidell, Louisiana (SIL), in the domain’s eastern portion 1989–2002. These warm season months were chosen (Fig. 1). because of their enhanced convection and because syn- Radiosonde data from 1989 to 1999 were available on optic-scale forcing typically is weak, with little influence the Radiosonde Data of North America CD-ROM dis- from midlatitude systems. Instead, mesoscale phenom- tributed by the Forecast Systems Laboratory (FSL) and ena such as sea and lake breezes interact with their the National Climatic Data Center (NCDC) (FSL and environments, surface features, and each other to pro- NCDC 1999). Data for 2000 to 2002 were obtained duce complex patterns of convergence and convection. from FSL’s Web site (http://www.fsl.noaa.gov/docs/ Our study domain spanned 28°–32°N and 87°–96°W, data/fsl-data.html). encompassing the northern Gulf of Mexico coastline and adjacent waters (Fig. 1). As shown by Cummins et al. (1998), five NLDN sensors are located within the 3. Results area. Individual flashes were counted within a 2.5 km ϫ 2.5 km grid, corresponding to a 353 ϫ 178 array of To investigate general lightning patterns in the north- 6.25-km2 grid cells. ern Gulf coast region, all flashes were grouped together without any consideration of wind direction or time of day. Figure 2a shows a general increase in lightning b. Radiosonde data from west to east. There are enhanced flash densities Radiosonde data were used to categorize each day of along the entire Gulf of Mexico coastline, suggesting a the period according to the prevailing low-level flow in link to the sea breeze. However, the strongest maxi- the area. The vector mean wind in the 1000–700-hPa mum is in coastal Mississippi, near Biloxi, where flash layer was computed each day using the 1200 UTC densities exceed 8.0 flashes kmϪ2 yrϪ1, with “year” de-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 418 WEATHER AND FORECASTING VOLUME 20

Ϫ Ϫ FIG. 2. Composite lightning flash density maps (flashes km 2 yr 1) for all warm season days from 1989 to 2002, where “year” corresponds to the warm season from May to Sep. (a) The upper scale corresponds to the 24-h composite, while (b) the lower scale is for the nighttime lightning composite from 2200 to 0700 CDT (0300 to 1200 UTC). fined as the warm season from May to September. hanced lightning described by Steiger et al. (2002) and Other maxima along the northern Gulf coast are lo- Orville et al. (2001), while Steiger and Orville (2003) cated near major metropolitan and/or industrial areas. described the maximum near Lake Charles. A weaker Areas near Houston, Texas; Lake Charles and New maximum of 4–6 flashes kmϪ2 yrϪ1 is found near Baton Orleans, Louisiana; and Mobile, Alabama, exhibit flash Rouge, Louisiana. Each of these urban maxima may be density maxima of 5–8 flashes kmϪ2 yrϪ1. (City loca- caused or enhanced by several factors, including con- tions and geographic features are shown in Fig. 1.) vergence due to sea, lake, and river breezes (e.g., Pielke Houston’s maximum corresponds to the region of en- 1974; Wakimoto and Atkins 1994; Laird et al. 1995);

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 2 live 4/C AUGUST 2005 S M I T H E T A L . 419

FIG. 3. Average 1000-hPa heights for the composite case. Contours are in 5-m increments. convex coastlines; urban heat island effects; and air pol- To document the large-scale flow over the region lution. during the warm season, reanalysis data were obtained Reduced flash densities are found over the Atchafa- from the National Centers for Environmental Predic- laya Basin and Lake Pontchartrain as well as over the tion–National Center for Atmospheric Research (Kal- extreme northern and southern portions of the domain nay et al. 1996). Specifically, 1000-hPa geopotential (Fig. 2a). The Atchafalaya Basin and much of coastal height data for each warm season day during the 14-yr Louisiana is a region of swamps, marshes, and lakes. period were averaged to create a composite analysis Their reduced flash densities are consistent with the (Fig. 3). The result shows the subtropical ridge axis findings of Hodanish et al. (1997) who noted that wet- extending from central Florida into Louisiana. A lands are associated with relatively small amounts of trough is located over southern Alabama, with a rela- lightning due to weaker differential heating and result- tively strong height gradient over southeast Texas and ing thermal circulations. Flash densities over Lake southwest Louisiana. Pontchartrain are only 2–4 flashes kmϪ2 yrϪ1. Previous The diurnal distribution of flashes for the composite research has shown that large lakes in south Florida are (all days, all flows) period is shown in Fig. 4 for the associated with subsidence and diminished lightning entire study area in Fig. 1. Hourly values range from (e.g., Pielke 1974; Blanchard and López 1985). ϳ172 000 at 0000 central daylight time (CDT ϭ UTC – The northern portion of the study region (Fig. 2a) 5 h) to near 1.4 million at 1500 CDT. The afternoon has small flash densities because of its greater distance maximum is due to sea-breeze-induced convection and from the Gulf of Mexico. The sea breeze typically does to deep convection associated with other mesoscale not propagate this far inland (e.g., Arritt 1993), and forcing mechanisms whose effects are enhanced by af- moisture supplies are more limited. Over the extreme ternoon heating. The smaller, secondary peak during southern portion of the domain (the open waters of the the early morning (0700 CDT) probably is attributed to Gulf of Mexico), the number of flashes is thought to be the land breeze, which is associated with early morning low because of three factors. First, the sea breeze, the offshore convection. dominant forcing mechanism, is a daytime inland phe- To investigate the small number of flashes during the nomenon. Second, convection usually is weaker over nighttime, flash densities were calculated for the period water than over land, resulting in a smaller number of 2200–0700 CDT (Fig. 2b). Although flash densities for flashes (e.g., Orville and Henderson 1986). Third, since this period are much smaller than those of the 24-h the NLDN sensors are located only over land, detection composite (Fig. 2a), the decrease from east to west still efficiency and location accuracy decrease with increas- is evident. Enhanced nighttime flash densities stretch ing distance from the coastline (Cummins et al. 1998). from just offshore of Galveston Bay (southeast of

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 420 WEATHER AND FORECASTING VOLUME 20

FIG. 4. Diurnal distribution of all flashes (ϫ 105) in the region shown in Fig. 1. Hour 0100 CDT denotes flashes between 0100 and 0159 CDT.

Houston) to the eastern edge of the domain. The off- layer from the 1200 UTC radiosonde releases at Lake shore nighttime flash densities may be enhanced by the Charles and Slidell (LCH and SIL). Of the possible warm, shallow waters in the region. Bathymetric data 2142 days during the period, soundings were available reveal that ocean depths less than 100 m extend from for 2069 days from LCH and 2108 days from SIL. 100 to 200 km off the coasts of Louisiana and Missis- Figure 5 shows the distribution of the vector mean sippi (information online at http://www.ngdc.noaa.gov/ wind directions at LCH (in the western part of the do- mgg/ibcca/images/). High-resolution satellite-derived main) and SIL (in the eastern portion). The distribution sea surface temperatures show these shallow areas to from LCH (Fig. 5a) is unimodal, with the peak number be relatively warm during the summer (information on- of days having southerly flow. Flow with a northerly line at http://fermi.jhuapl.edu/avhrr/gm/averages/). The component occurs infrequently. Wind directions at SIL maximum of nighttime flashes is located offshore of (Fig. 5b) are somewhat different from those at LCH, coastal Mississippi. The reason for this maximum is un- most likely due to variations in the generally east-to- certain; however, it may be due to a merger of land west-oriented subtropical ridge axis. The SIL distribu- breezes from the Mississippi and Louisiana coasts. Spe- tion is more uniform, and the peak is skewed slightly to cifically, the southward-moving land breeze from the right of that at LCH. The greatest number of days coastal Mississippi may be colliding with an eastward- occurs when the wind is from 225°. Days with southerly moving land breeze from Louisiana, leading to en- winds are the second most common, while northeast hanced convergence. Relatively small offshore flash flow is the least frequent. densities occur just southwest of New Orleans. This Based on the vector mean wind, each day was placed relative minimum may be due to the convex coastline, into one of five flow regimes. Since the flow can be which promotes localized divergence. weak, we employed a “calm flow” category to account for days when vector mean wind speeds were less than a. Individual flow regimes 2.5 m sϪ1. All directions were included in this regime. The low-level synoptic flow plays a significant role in For days with stronger winds, the orientation of the the formation and evolution of the sea breeze and its coastline was used to define four additional flow re- associated convection and lightning (e.g., Gentry and gimes that consider the onshore and offshore winds that Moore 1954; Arritt 1993). Therefore, it is informative have been shown to greatly influence the strength and to classify each day according to its wind direction and inland propagation of the sea breeze (e.g., Pielke 1974; to examine flash patterns as a function of the flow. For Arritt 1993). Unlike Florida’s relatively straight coast- every warm season day during the 14-yr period, we lines, the coastlines of Louisiana and Mississippi have calculated the vector mean wind in the 1000–700-hPa varying orientations. Using a different coastline orien-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 421

FIG. 5. Distribution of days according to their 1000–700-hPa vector mean wind directions for (a) LCH and (b) SIL. Directions are grouped into 5° bins. Flow regimes are labeled at the top of each histogram, with arrows denoting the divisions between regimes. tation and thus different flow regimes for each portion have equal directional ranges (90°) and are denoted of the region would have been impractical. Therefore, northeast (356°–86°), southeast (86°–176°), southwest an average coastline orientation of 86° was assumed, (176°–266°), and northwest (266°–356°). Thus, two of and days with vector mean wind speeds greater than 2.5 the quadrants represent onshore flow (the southeast msϪ1 were categorized into one of four quadrants or and southwest regimes), while the remaining two de- flow regimes based on this orientation. These regimes note offshore flow (the northeast and northwest re-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 422 WEATHER AND FORECASTING VOLUME 20

TABLE 1. Number of days associated with each combination of flow regimes at LCH and SIL. Flow regimes for SIL are listed vertically, and those from LCH are read horizontally.

SIL/LCH Calm Northeast Southeast Southwest Northwest Total Calm 119 25 73 103 20 340 Northeast 38 121 44 10 14 227 Southeast 32 42 264 72 2 412 Southwest 46 4 50 559 57 716 Northwest 66 21 11 111 137 346 Total 301 213 442 855 230 2041

gimes). Due to the complex coastal orientation in the lightning flashes over the western half of the domain area (bays, capes, etc.), these general onshore–offshore vary widely between flow regimes. The southwest cat- designations do not represent each specific location in egory has the largest number of days (840) and flashes the study area. The distribution of days in each flow (2 999 296). In fact, the maximum number of flashes on regime is shown in Fig. 5. a single day (44 846) occurs during southwesterly flow. We initially were unsure whether sounding data from Conversely, the northeast and northwest regimes have LCH and/or SIL should be used to classify the days the smallest number of days and flashes (approximately according to their low-level flow. To investigate, the 225 days and 675 000 flashes). Detailed discussions 1000–700-hPa vector mean wind directions from the about each regime are provided in later sections. 2041 days having data at both LCH and SIL were com- Figure 6b and Table 3 show results for the eastern pared (Table 1). The southwest regime contains the half of the domain (87.0°–91.5°W). Two thousand greatest number of days at both SIL and LCH, while ninety-five days out of the possible 2108 were classified the northeast regime has the smallest number. With the into one of the five flow regimes. The southwest regime exception of calm flow, approximately 66% of days at again has the largest number of days and flashes (719 the two locations are in the same category, and of the and 2 946 218, respectively), while northeast flow ranks remaining days, a large portion are in an adjacent cat- last among the regimes with only 245 days and 785 577 egory. However, there are some days when the low- flashes. level flow at the two stations is in opposite quadrants. Figure 2a shows a general decrease in flash density These differences suggested that vector mean winds from east to west, and this observation is confirmed in from both radiosonde sites should be used to classify Tables 2 and 3. Although the eastern and western do- individual days. Therefore, we divided the study area mains have equal areas, considerably more flashes oc- into eastern and western zones of equal area. Winds at cur in the eastern region (8 239 874) than the western LCH are assumed to represent the western half of the portion (6 639 535). The eastern region also has more region (91.5°–96.0°W), with sounding data from SIL days with lightning (1874 versus 1736) and a greater representing winds in the eastern half (87.0°–91.5°W). median number of flashes per day (1440 versus 918). The directional ranges for each regime were described These contrasts do not appear to result from the rela- previously (356°–86°, etc). tively minor differences in low-level environmental Figure 6 relates the daily flash count across each half wind in the two regions (Fig. 5). After examining the of the region to the mean low-level wind direction for thermodynamics of each region, we found that the con- that day. We also calculated statistical parameters from vective available potential energy (CAPE) for the west- the daily lightning and radiosonde data (Tables 2 and ern area actually is greater than in the east where more 3). The category “all days” describes lightning charac- lightning occurs (Table 4). Surface data revealed that teristics as a whole, regardless of the vector mean wind, LCH is slightly warmer and more humid than SIL, and is analogous to the composite case in Fig. 2. contributing to its greater CAPE. Precipitable water Figure 6a and Table 2 are based on wind data from values in the two areas generally are similar (Table 4), LCH and lightning data from the western half of the except for the northwest flow regime where the western domain (91.5°–96.0°W). Radiosonde data were avail- region is somewhat drier (1.46 versus 1.61 in.). Thus, able for 2069 days out of a possible 2142, with 2045 days the thermodynamics of the western and eastern regions classified into one of the flow regimes. The remaining does not appear to explain their differing flash densi- days could not be classified because of problematic ties. soundings, although they were included in the all-days We believe that the east–west contrast in lightning category. Results show that the number of days and mostly is due to the nature of the land–sea interface.

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 423

FIG. 6. Scatter diagram of number of lightning flashes (ϫ 104) vs 1000–700-hPa vector mean wind direction for (a) LCH (the west sector) and (b) SIL (the east sector). Flow regimes are labeled and denoted by solid black lines.

Specifically, the western two-thirds of the domain con- Thus, the land–sea temperature gradient is reduced, as tains extensive regions of swamps and marshes that ex- is the strength of the resulting sea-breeze circulation. tend inland 50 to ϳ100 km. Local climatic data reveal Conversely, coastal Mississippi and Alabama exhibit a that temperatures in this broad zone are intermediate more sharply defined coastline. Although some swamps to those farther inland and over the Gulf of Mexico. are present, they are relatively narrow. The result is a

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 424 WEATHER AND FORECASTING VOLUME 20

TABLE 2. Statistical parameters for each flow regime in the western portion of the region between 91.5° and 96.0°W. The mean 1000–700-hPa vector wind from LCH was used to categorize flow days. The all-days category includes every day, including those for which sounding data were unclassifiable. Thus, it is not the sum of the categories listed above.

Mean No. of Percent Min Median Max Flow No. of No. of flashes lightning lightning flashes 25th flashes 75th flashes regime flashes flow days dayϪ1 days days dayϪ1 percentile dayϪ1 percentile dayϪ1 Calm 1 133 770 303 3742 257 85 0 75 1424 4846 37 279 NE 489 840 218 2247 152 70 0 0 78 1869 31 488 SE 1 084 615 450 2410 399 89 0 85 893 2971 36 151 SW 2 999 296 840 3571 718 85 0 41 1202 5288 44 846 NW 866 246 234 3702 188 80 0 10 753 5154 34 342 All days 6 639 535 2069 3209 1736 84 0 28 918 4313 44 846 stronger land–sea temperature gradient that produces a when midlatitude synoptic systems may increase low- better-defined sea breeze. The further reduction in level wind speeds. Across the eastern section, calm flow flash densities over the upper Texas coast (Fig. 2a) occurs most often in August and least often during May likely is due to their often experiencing relatively dry (Table 6). Calm flow in the east exhibits the second southwesterly flow from south Texas and Mexico. greatest mean and median flashes per day (4472 and However, that area is near the western boundary of our 1889, respectively) (Table 3). In summary, calm (light domain and was not examined in detail. wind) days occur frequently and are large lightning pro- ducers. b. Calm flow Figure 7a shows the mean 1000-hPa geopotential The calm (light wind) regime consists of days when height analysis for days having calm flow. Although the the 1000–700-hPa vector mean wind speed is less than figure is based on the LCH flow regimes, very similar 2.5 m sϪ1, regardless of wind direction. Three hundred analyses were obtained for the SIL regimes (not three (352) days were classified as calm in the western shown). The analysis shows the synoptic pattern that (eastern) portions of the domain (Tables 2 and 3). Of typically is associated with the calm regime. Specifi- these days, 85% (91%) contain lightning in the western cally, high pressure associated with the subtropical (eastern) portion. ridge dominates the southeastern United States, with The large percentage of days with lightning is consis- the ridge axis extending across Florida into east Texas. tent with conducive atmospheric conditions (Table 4). The weak height gradient over the Gulf coast is indica- Both halves of the domain are characterized by large tive of the light winds over Louisiana and Mississippi. K-index values, CAPE, and precipitable water. Thus, The 24-h composite flash density map for calm flow is convection is favored because of moist, unstable condi- shown in Fig. 8a. To facilitate comparisons between tions. regimes, flash densities have been normalized as in In the western portion of the study area (Table 2), Lericos et al. (2002). That is, flash counts for each grid calm flow days have the greatest mean and median cell were divided by the area of the cell and the number flashes per day of all five regimes (3742 and 1424, re- of days in the particular flow regime, giving units of spectively). Calm flow in the west is most common dur- flashes kmϪ2 (regime day)Ϫ1. The flash density map for ing July and least common in May and June (Table 5) calm flow reveals an active pattern that is similar to the

TABLE 3. As in Table 2 except for the eastern portion of the region between 87.0° and 91.5°W. Days were classified according to the mean 1000–700-hPa vector wind from SIL.

Mean No. of Percent Min Median Max Flow No. of No. of flashes lightning lightning flashes 25th flashes 75th flashes regime flashes flow days dayϪ1 days days dayϪ1 percentile dayϪ1 percentile dayϪ1 Calm 1 574 044 352 4472 321 91 0 233 1889 5854 43 992 NE 785 577 245 3206 201 82 0 3 402 3371 51 327 SE 1 064 706 421 2529 389 92 0 135 1022 3479 30 553 SW 2 946 218 719 4098 647 90 0 257 1962 5495 43 742 NW 1 788 237 358 4995 304 85 0 73 1445 6245 48 111 All days 8 239 874 2108 3909 1874 89 0 123 1440 5037 51 327

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 425

TABLE 4. Median 1200 UTC sounding parameters for the five TABLE 6. As in Table 5 except based on classifiable data from flow regimes on days with and without lightning. The K index has SIL. units of °C, CAPE has units of J kgϪ1, and precipitable water is expressed in inches. CAPE was computed using the convective Calm Northeast Southeast Southwest Northwest temperature as the surface temperature. These parameters are May 37 33 83 185 87 commonly used to assess the potential for warm season convec- Jun 58 26 83 192 55 tion along the Gulf coast. Jul 90 28 60 159 84 Aug 97 69 78 99 79 Precipitable Sep 70 89 117 81 53 K index CAPE water Total 352 245 421 716 358 LCH Calm 29 2318 1.72 Northeast 19 718 1.40 Southeast 28 2200 1.72 Enhanced flash densities in the coastal areas (Fig. 8a) Southwest 27 2813 1.72 Northwest 26 1493 1.46 are in a broken band that parallels the coastline. Hourly SIL flash density maps indicate that this band is caused Calm 29 1343 1.69 by the sea breeze penetrating inland only slightly. Northeast 23 637 1.41 Large-scale winds less than 2.5 m sϪ1 are too weak Southeast 27 1296 1.67 Southwest 29 1837 1.75 to move the sea breeze farther inland. The strong- Northwest 28 1341 1.61 est density maximum within the study area [greater than 0.08 flashes kmϪ2 (regime day)Ϫ1] is located in coastal Mississippi. A closer view of this maximum is all-days composite shown in Fig. 2a. Maxima are lo- shown in a series of hourly maps in Fig. 9. Convection cated in similar areas, with most near metropolitan cen- develops just west of Mobile Bay at 1200 CDT (Fig. ters. This similarity occurs because only vector mean 9b). At 1300 CDT (Fig. 9c), the greatest flashes are wind speed is considered when classifying days as calm; farther west and centered around a small bay. How- therefore, all wind directions are included. Density val- ever, by 1400 CDT (Fig. 9d), convection is widespread ues are smaller than the all-days category because they and indicative of the larger-scale sea breeze. Finally, by were normalized per regime day rather than per entire 1500 CDT, the flash density peaks, with a swath of large warm season. values extending from Slidell to Mobile (Fig. 9e). This There are several regions of diminished flash densi- peak corresponds to the maximum in cloudiness and ties during calm (light wind) flow (Fig. 8a). Minima are deep convection found by Gibson and Vonder Haar found over large bodies of water and in regions of (1990). marshes and swamps. Thus, Galveston Bay (southeast It is clear that complex and sometimes subtle forcing of Houston), the Atchafalaya Basin, Lake Pontchar- mechanisms are important in producing the flash pat- train, and Mobile Bay exhibit flash densities between terns that are observed during calm flow (Figs. 8a and 0.0 and 0.04 flashes kmϪ2 (regime day)Ϫ1. Flash density 9). One important mesoscale forcing mechanism is the patterns in the northern part of the study area, well large-scale sea breeze, and a second is enhanced con- inland from the coast, are rather disorganized; only sev- vergence/divergence due to lakes and the shape of the eral small maxima of 0.04–0.06 flashes kmϪ2 (regime Ϫ coastline. The flash density maximum near Biloxi also day) 1 are evident. Density patterns in this area at may be influenced by urban factors such as air pollution hourly intervals (not shown) do not reveal an organized and heat island effects (e.g., Orville et al. 2001; Steiger progression. Thus, subtle forcing mechanisms may be et al. 2002; Westcott 1995). Additional, weaker maxima responsible for that convection. (Fig. 8a) are noted in the vicinity of other large cities (i.e., Lake Charles and New Orleans). TABLE 5. Monthly distribution of days for which LCH sounding Diurnal flash distributions for calm (light wind) con- data were classifiable. ditions (Fig. 10) are similar in the western and eastern Calm Northeast Southeast Southwest Northwest portions of the domain. The western half’s peak of ϳ150 000 (Fig. 10a) occurs at 1700 CDT, one hour later May 42 22 77 224 64 ϳ Jun 35 25 102 201 48 than the peak of the eastern half ( 200 000 flashes, Fig. Jul 86 15 71 207 33 10b) at 1600 CDT. Both distributions exhibit smaller, Aug 75 48 105 120 45 secondary peaks between 0700–0900 CDT. Early morn- Sep 65 108 95 83 44 ing offshore convection is the main cause for these sec- Total 303 218 450 835 234 ondary peaks.

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 426 WEATHER AND FORECASTING VOLUME 20

FIG. 7. Average 1000-hPa height contours in 5-m increments for the following flow regimes: (a) calm flow, (b) northeast flow, (c) southeast flow, (d) southwest flow, and (e) northwest flow. The flow categories are based on LCH data.

c. Northeast flow the atmosphere is relatively dry and stable. In fact, values of the K index, CAPE, and precipitable water for Days when the 1000–700-hPa vector mean wind di- both halves of the domain are the smallest of all flow rection is between 356° and 86° compose the northeast categories (Table 4). flow category. This regime has the fewest number of The subtropical ridge does not influence the domain days (218 and 245) and flashes (489 840 and 785 577) in on days with northeast flow. Instead, strong high pres- both halves of the domain (Tables 2 and 3). In the sure over Illinois and low pressure over the Yucatan western segment, only 70% of the 218 flow days have Peninsula are major factors (Fig. 7b). A ridge axis ex- lightning. The mean, median, and maximum flashes per tends into southeast Texas, while an inverted trough, day also are the smallest of all five regimes. Within the possibly associated with frontal boundaries, extends eastern sector, a greater percentage of days (82% of the from the eastern Gulf of Mexico to the coastal regions 245 flow days) produces lightning (Table 3). Although of the mid-Atlantic states. Northeast flow occurs most it is interesting that the northeast regime exhibits the frequently during September (Tables 5 and 6) since greatest number of flashes on a single day in the eastern post–cold frontal high pressure systems are more com- portion (51 327), this maximum is due to an individual mon during this month than during the traditional sum- event and is not a general characteristic of the regime. mer months. These fronts bring relatively dry, stable air Convection is less likely during northeast flow because into the region (Table 4).

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 427

Magnitudes and patterns of flash density differ con- 4). To summarize, southeast flow produces a large per- siderably over the western and eastern halves of the centage of lightning days, but those days do not yield domain during northeast flow (Fig. 8b). Values in the great amounts of lightning. western half are relatively small. The largest densities There is a diffuse lightning pattern across the region [0.03–0.06 flashes kmϪ2 (regime day)Ϫ1], in a broken during southeast flow (Fig. 8c). The sea breeze is weak band from Houston to Lake Charles, are associated and advances onshore quickly, often without producing with the sea breeze but are farther inland and weaker significant convection. However, a weak, broken band than expected. Since the coastal portions of southwest of enhanced densities does stretch from Houston to Louisiana consist of marshes, swamps, and lakes, the Mobile, slightly inland from the coast, apparently due thermal circulation constituting the sea breeze may be to the weak sea breeze. An area of marginally en- farther inland where the stronger land–water tempera- hanced flash densities [only 0.03–0.06 flashes kmϪ2 (re- ture gradients are located. There is little lightning over gime day)Ϫ1] also extends from Baton Rouge north- the coastal waters of the western portion. ward into Mississippi. This area may result from inter- Flash densities across the eastern half of the domain actions between small-scale circulations (due to lakes, are greater in magnitude and areal coverage (Fig. 8b). rivers, etc.) and the low-level flow, or urban influences Two areas of enhanced values over land are evident— (i.e., heat island effects, industrial pollution, etc.). Fi- along the Mississippi coastline and along the Louisiana nally, relative flash density minima are located over and coast south of New Orleans. Thunderstorms remain immediately northwest of Lake Pontchartrain and over near the coasts because the northward advance of the the Atchafalaya Basin. sea breeze is limited by the large-scale northeasterly e. Southwest flow winds. Enhanced densities also are found over the coastal waters south and southwest of New Orleans. The southwest flow regime consists of days with vec- Hourly flash density maps (not shown) indicate that tor mean 1000–700-hPa wind directions between 176° this lightning is associated with early morning offshore and 266°. The subtropical ridge axis on these days (Fig. convection. Additionally, northeast flow produces a 7d) typically extends from the Atlantic Ocean to south- density minimum southwest of Lake Pontchartrain due ern Louisiana, with a broad low pressure system over to the advection of cooler and more stable air into the western Texas and Oklahoma. May, June, and July con- region. Reduced densities also are noted over the tain the most southwest flow days, while September has Atchafalaya Basin. Lightning activity in the western a definite minimum (Tables 5 and 6). section peaks at 1700 CDT (Fig. 10a) but reaches a Southwest flow is the most unstable, most frequent, maximum 1 h earlier at 1600 CDT in the eastern por- and most active of all categories. In both halves of the tion (Fig. 10b). study area, the number of days (840 and 719) and flashes (2 999 296 and 2 946 218) far exceed those of the d. Southeast flow other categories (Tables 2 and 3). Eighty-five percent The southeast flow regime contains days having vec- (90%) of the days have lightning in the western (east- tor mean wind directions between 86° and 176°. The ern) portions. Sounding parameters explain the pro- typical synoptic situation for these days (Fig. 7c) con- pensity for convection during southwest flow (Table 4), tains a closed high pressure over North Carolina, with with values of CAPE in both halves of the area being the subtropical ridge axis extending from the Atlantic the greatest of all regimes. The K-index and precipi- Ocean across the Carolinas to northern Louisiana. The table water values also are relatively large. number of flow days and total number of flashes during The flash density pattern in Fig. 8d has similarities to the 14-yr period are similar for the two halves of the that of the composite case (Fig. 2a). There is an east- domain (Tables 2 and 3), with the number of flow days to-west decrease in densities, and maxima are located ranking second among the five regimes. near major population centers. In the western half of Lightning occurs on many southeasterly flow days, the region, areas of enhanced flash density [0.04–0.06 with 89% (92%) producing lightning in the western flashes kmϪ2 (regime day)Ϫ1] are over Houston and (eastern) portions (Tables 2 and 3). However, in the Lake Charles. As previously discussed, the swampy na- eastern half, southeast flow exhibits the smallest mean ture of the western coastal regions likely causes weaker and second smallest median flashes per day (Table 3). sea-breeze circulations that take longer to form, In the western section, the median is somewhat greater, thereby producing less convection and lightning. Also, ranking third out of the five flow categories (Table 2). compared to other flow regimes, the southwest cat- The stability and moisture content of southeast flow are egory has relatively little offshore activity. intermediate to those of the other flow regimes (Table Flash densities across the eastern portion of the re-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 428 WEATHER AND FORECASTING VOLUME 20

Ϫ Ϫ FIG. 8. Lightning flash density maps [flashes km 2 (regime day) 1] for the five flow regimes: (a) calm flow, (b) northeast flow, (c) southeast flow, (d) southwest flow, and (e) northwest flow. The solid black line indicates the division between the western and eastern components of the domain. gion are greater than farther west (Fig. 8d). Maxima of showed a maximum of convection on the north and 0.05–0.08 flashes kmϪ2 (regime day)Ϫ1 are over New northwest sides of Mobile Bay during southwesterly Orleans, in coastal Mississippi, and in a north–south flow. They stated that convergence in this region is en- line extending northward from Mobile. The Mississippi hanced by interactions between the low-level flow and and Alabama maxima are consistent with Medlin and locally higher elevations. Their convective maximum Croft (1998) who noted that southwesterly low-level corresponds to the maximum in flash densities seen in flow produced the greatest radar echo frequencies in Fig. 8d. those areas. The thermal circulation that develops be- The flash density maximum seen near New Orleans tween the warm, shallow water and the land along the in the composite case (Fig. 2a) is especially prominent sharply defined coastline forms, strengthens, and pro- during southwest flow (Fig. 8d). New Orleans is located gresses inland during the morning. Convection typically along the southern edge of Lake Pontchartrain and has begins near the time of maximum heating and instabil- the largest area and population of the cities along the ity. The convection is enhanced by interactions be- northern Gulf coast. Hourly analyses reveal that en- tween the sea breeze and local circulations (i.e., bay hanced flash densities during southwest flow first ap- breezes) and topography. Medlin and Croft (1998) also pear over New Orleans at 1100 CDT (Fig. 11a). Values

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 8a-b live 4/C AUGUST 2005 S M I T H E T A L . 429

FIG.8.(Continued)

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 8c-e live 4/C 430 WEATHER AND FORECASTING VOLUME 20

Ϫ Ϫ FIG. 9. Hourly flash density maps [flashes km 2 (regime day) 1] for calm (light wind) flow between 1100 and 1500 CDT. Each map represents a 1-h time period. The white circles denote the city centers of New Orleans and Biloxi.

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 9 live 4/C AUGUST 2005 S M I T H E T A L . 431

lightning enhancements in cities adjacent to major lakes. The hourly distribution of flash counts for the western half (Fig. 10a) peaks at 1500 CDT, with a secondary maximum between 0600 and 0800 CDT. Analyses of hourly flash density maps (not shown) reveal that this early morning lightning is associated with nearshore convection. As with the previous regimes, the distribution for the eastern portion (Fig. 10b) peaks 1 h earlier than the western distribution. f. Northwest flow Days with 1000–700-hPa vector mean wind directions between 266° and 356° are classified in the northwest flow regime. Northwest flow is a prodigious lightning producer, although the statistics vary considerably between the two halves of the region. The eastern portion (Table 3) has approximately 50% more days and double the number of flashes of the western portion (Table 2). Specifically, there are only 234 days with 866 246 flashes in the western portion, with 80% of the days producing lightning. On the other hand, the eastern section has 358 days and 1 788 237 flashes (Table 3), with a larger percentage (85%) of days producing lightning. In addition, of all the directional regimes, northwest flow has the second greatest (greatest) mean flashes per day [3 702 (4 995)] for the western (eastern) portion, although median rankings are somewhat lower. This prolific amount of lightning also is indicated ϫ 4 FIG. 10. Hourly total flash distribution ( 10 ) by flow regime by the daily flash counts in Fig. 6. Twelve days in the for (a) the western portion of the region between 91.5° and 96.0°W and (b) the eastern portion between 87.0° and 91.5°W. western portion (Fig. 6a) and 23 in the eastern half (Fig. 6b) exceed 20 000 flashes per day. Days with northwesterly flow are not governed by then increase over New Orleans and the other sides of the subtropical ridge (Fig. 7e). Instead, they are influ- Lake Pontchartrain (Figs. 11b and 11c) between 1200 enced by midlatitude synoptic systems, with a trough of and 1300 CDT. These areas of enhanced activity prob- low pressure, most likely associated with individual ably are due to locally increased convergence from the wave cyclones, located over Alabama and Georgia. lake breeze and its interaction with the southwest flow. Northwest flow is most common during May and least The sea breeze likely is a factor as well, although it is common during July (western portion) and September poorly defined in the hourly maps. By 1400 CDT (Fig. (eastern portion) (Tables 5 and 6). Due to its postfron- 11d), values are greatest over New Orleans [0.009– tal nature, northwest flow is not as moist and unstable 0.015 flashes kmϪ2 (regime day)Ϫ1]. Three flows appear as its southerly counterpart (southwest flow) (Table 4). to anchor this maximum over the city: the sea breeze, Conversely, it is not as dry and stable as its easterly the lake breeze, and the large-scale flow. Due to its counterpart (northeast flow). large population, there is the potential for urban light- The flash density map for northwest flow (Fig. 8e) ning enhancement. However, calculation of the total shows very active regions of lightning. Sea-breeze- flash count in the immediate New Orleans area (not induced maxima are confined to the coastline by the shown) reveals values that are similar to those of large-scale flow. The greatest densities, exceeding 0.08 smaller cities (Baton Rouge and Lake Charles). This flashes kmϪ2 (regime day)Ϫ1, extend from Biloxi to may be due to New Orleans’s proximity to Lake Pont- Mobile. In addition, there is considerable lightning off- chartrain, which gives the atmosphere a relatively mari- shore. Hourly maps (not shown) reveal that the off- time (stable) flavor. This hypothesis is consistent with shore convection occurs during the night and early the results of Westcott (1995) who noted weaker urban morning hours. Relative minima are noted over Lake

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 432 WEATHER AND FORECASTING VOLUME 20

Ϫ Ϫ FIG. 11. Hourly flash density maps [flashes km 2 (regime day) 1] for New Orleans and surrounding areas for southwest flow between 1100 and 1500 CDT. Each map represents a 1-h time period. The white circle denotes the approximate location of the city’s center.

Pontchartrain and its shadow region, as well as over the west flow (Fig. 8e). As noted earlier, the land–sea in- Atchafalaya Basin. terface is sharply defined, likely producing strong ther- The enhanced flash densities in coastal Mississippi mal circulations compared to areas farther west. Hourly and Alabama are especially well defined during north- flash density maps first indicate inland penetration of

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 11 live 4/C AUGUST 2005 S M I T H E T A L . 433

Ϫ Ϫ FIG. 12. Hourly flash density maps [flashes km 2 (regime day) 1] for Biloxi and surrounding areas for northwest flow between 1200 and 1600 CDT. Each map represents a 1-h time period. The white circle denotes the approximate location of the city’s center. the sea breeze at 1300 CDT (Fig. 12b). Flash densities From 1500 to 1600 CDT (Figs. 12d and 12e), values associated with the sea-breeze-related convection across coastal Mississippi and Alabama still are quite quickly intensify, reaching peak values of 0.015–0.024 large. The band of enhanced densities never advances flashes kmϪ2 (regime day)Ϫ1 at 1400 CDT (Fig. 12c). far inland since the sea breeze is restrained by the op-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 12 live 4/C 434 WEATHER AND FORECASTING VOLUME 20 posing large-scale flow. Since the Biloxi area is growing mum of 22 264 at 1200 CDT, a time when flash counts rapidly, localized urban and industrial effects likely are normally are increasing to the afternoon peak. becoming increasingly important, in addition to influ- g. Discussion ences of sea-breeze circulations and the large-scale flow. The results have shown that warm season lightning The evolution of flash densities near Baton Rouge over the northern Gulf of Mexico coast is closely re- during northwest flow also reveals interactions between lated to environmental and geographical features in the several circulation systems. Baton Rouge is located region. The direction of the large-scale environmental relatively far inland of the Gulf of Mexico and is not flow greatly influences the humidity and stability of the adjacent to any large bodies of water. The only signifi- area and, therefore, the amount of convection and cant geographic feature is the Mississippi River, which lightning that occur. Days with southwest flow off the flows along the city’s western side. Baton Rouge is a Gulf of Mexico are the most humid and unstable of the major industrial area, with many petrochemical plants four directional regimes, and these days exhibit the along the Mississippi River. greatest median number of flashes per day. Conversely, Hourly flash densities in the Baton Rouge area dur- days with northeast flow from the continent produce ing northwest flow are given in Fig. 13. At 1500 CDT the smallest median number of flashes per day since (Fig. 13a), values are greatest on the northwest side of this air is the driest and most stable. These results are Lake Pontchartrain (located in the lower-right corner). consistent with those of Camp et al. (1998) and Fuel- Convergence in this region is enhanced by interactions berg and Biggar (1994) who studied lightning patterns between lake breezes from Lakes Pontchartrain and and the preconvective environment, respectively, of the nearby Florida panhandle. Days with southeasterly or Maurepas (just west of Lake Pontchartrain) and the northwesterly flow produce intermediate median num- large-scale northwesterly flow. Flash densities over Ba- bers of flashes, consistent with their intermediate val- ton Rouge are minimal at this time. By 1600 CDT, flash ues of humidity and stability. densities increase around Lake Pontchartrain and near The sea breeze is the dominant forcing mechanism Baton Rouge (Fig. 13b). The area of enhanced flash for warm season deep convection over the northern densities continues to strengthen and increase, and by Gulf coast, with numerical modeling (e.g., Pielke 1974; 1800 CDT, a maximum of 0.018–0.021 flashes kmϪ2 Arritt 1993) showing that the large-scale flow affects (regime day)Ϫ1 is located over the city (Fig. 13d). En- both the strength and inland penetration of the sea hanced convergence from outflow boundaries of con- breeze. Current findings are consistent with those mod- vection near Lakes Maurepas and Pontchartrain, as eling results. Days with mean low-level vector wind well as urban enhancements, may be responsible for speeds less than 2.5 m sϪ1 (calm days) produce large these large flash densities. During the next hour, den- numbers of flashes because the sea breeze and its as- sities decrease dramatically over the area, and by 1900 sociated convection do not propagate far inland. Al- ϳ CDT, values in the area are only 0.0–0.009 flashes though offshore flow has been shown to produce a Ϫ2 Ϫ1 km (regime day) (Fig. 13e). To summarize, lake strong sea breeze that remains near the coastline (e.g., and urban processes appear to largely influence and Pielke 1974; Arritt 1993), the relative dryness and sta- enhance lightning in the Baton Rouge area. The sea bility of the environment appear to limit the amount of breeze does not appear to be a significant factor. convection and lightning along the northern Gulf coast. Hourly flash counts for the two halves of the domain Conversely, onshore flow has been associated with rela- differ considerably from one another (Fig. 10). Each of tively weak sea breezes that advance far inland. How- the previously discussed regime distributions peaked ever, current results suggest that the relatively large during midafternoon and exhibited an overnight mini- instability and humidity of the advancing maritime air mum. The hourly distribution for northwest flow in the counters the weaker sea breeze to produce large flash eastern section (Fig. 10b) follows this trend. The peak counts. Thus, the large-scale flow affects both the ther- (169 546) occurs at 1500 CDT, while a secondary peak modynamics and degree of forcing that produce con- (69 709) occurs at 0900 CDT. This smaller peak corre- vection and lightning, with these two influences possi- sponds to the offshore lightning activity discussed ear- bly counteracting each other. lier. However, the distribution for northwest flow in the The nature of the underlying land surface is observed western portion of the area differs from the others. It is to influence the strength of the land–sea temperature more uniform, varying by only ϳ46 000 flashes between gradient and the resulting sea-breeze circulation. maximum and minimum values (Fig. 10a). The peak Greatest flash densities in our study area are located flash count of 68 370 occurs at 1700 CDT and the mini- along coastal Alabama and Mississippi where the land–

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 435

Ϫ Ϫ FIG. 13. Hourly flash density maps [flashes km 2 (regime day) 1] for Baton Rouge and surrounding areas for northwest flow between 1500 and 1900 CDT. Each map represents a 1-h time period. The white circle denotes the approximate location of the city’s center.

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC

Fig 13 live 4/C 436 WEATHER AND FORECASTING VOLUME 20 sea interface is sharply defined, yielding a strong hori- imposed on the larger-scale trend. Greatest densities zontal temperature gradient and sea-breeze circulation. were located near Biloxi, Mississippi, with other rela- Conversely, the swampy nature of coastal Louisiana tive maxima near Houston, Texas; Lake Charles, Baton yields a weaker land–sea temperature gradient, weaker Rouge, and New Orleans, Louisiana; and Mobile, Ala- sea-breeze forcing, and smaller flash densities. Other bama. During the nocturnal period (2200–0700 CDT), geographic features that create or modify mesoscale flash densities were greatest over the coastal waters, circulations that produce deep convection include lakes decreasing from east to west. The nocturnal lightning (e.g., Lake Pontchartrain), large rivers (e.g., the Missis- was enhanced by the warm shallow Gulf of Mexico sippi and Atchafalaya), and complex-shaped coastlines waters, by land breezes, and by the advection of land- (e.g., Mobile Bay). Finally, flash density maxima are forming convection onto the water. observed near the major industrial centers of Houston, Results showed the impacts of geographic features Lake Charles, Baton Rouge, and New Orleans. These on flash density patterns. The numerous lakes, bays, flash enhancements may be due to the thermal circula- marshes, and swamps of the region, as well as the shape tions described above, together with local effects such of the coastline, influenced density magnitudes. Con- as urban heat islands or altered cloud microphysical vex-shaped coastlines, such as Mobile Bay, enhanced processes due to air pollution (e.g., Steiger and Orville lightning development, as evidenced by the maximum 2003). in that area. Conversely, concave coastlines were asso- It is clear that the complex flash density patterns ob- ciated with nearby diminished flash densities. Large served along the northern Gulf of Mexico coast are due lakes, especially Lake Pontchartrain, suppressed con- to interactions between the large-scale flow and a va- vection due to cooler, more stable air over them. The riety of sometimes subtle mesoscale circulations. A large-scale flow also advected this air downstream of great deal of additional data and study will be required the lake, resulting in a smaller number of flashes, that to unravel these complex interactions. is, a lake shadow. Enhanced lightning in other areas near lakes likely was due to lake breezes interacting 4. Summary and conclusions with other thermal circulations. Marshes and swamps also retarded lightning activity, especially in the coastal Cloud-to-ground lightning data from the National region of western Louisiana. Temperature contrasts be- Lightning Detection Network have been used to deter- tween land and water were smaller in that area, pro- mine lightning patterns along the northern Gulf coast ducing weaker thermal circulations and weaker forcing and to study their relation to mesoscale forcing mecha- for convection. nisms. Although this area contains the second greatest Flash patterns along the northern Gulf coast were flash densities in the nation, it has received little previ- found to depend greatly on the low-level flow. This ous attention. The study period was the warm seasons relation occurs because the large-scale flow regulates (May–September) of the 14-yr period 1989–2002. Ra- the strength and movement of the sea breeze, the dom- diosonde data from Lake Charles and Slidell, Louisi- inant warm season forcing mechanism for convection. ana, were used to calculate the 1000–700-hPa vector The calm, southwest, and northwest flow regimes ex- mean wind each day. Based on that wind, the day was hibited the largest flash densities, while northeast and placed into one of five flow regimes. Four of the re- southeast flow produced the smallest values. Many of gimes were based on the wind components onshore or the regimes exhibited flash density maxima near the offshore of the coastline. The fifth regime, calm flow, aforementioned urban areas, as well as minima over included days when the mean low-level flow was less Lake Pontchartrain and the Atchafalaya Basin. than 2.5 m sϪ1, regardless of direction. Flash densities Selected sounding parameters were examined for were calculated on a 2.5 km ϫ 2.5 km grid and com- each flow regime using sounding data from the Lake posited for each flow regime at hourly, daily, and noc- Charles and Slidell radiosonde sites. There were clear turnal increments. differences in atmospheric moisture and stability be- Results for the “all days, all flows” case indicated that tween flow categories. Days with calm or southwest the northern Gulf coast is a very active lightning region. flow were the most humid and unstable. Consequently, Flash densities were found to decrease in an east-to- they had the greatest mean and median flashes of all west direction from Alabama to western Louisiana, five regimes. Conversely, northeast flow days were the probably due to the swampy nature of coastal Louisi- driest and most stable, producing the least amount of ana that reduces the horizontal temperature gradient lightning. The northwest and southeast flow days were producing the sea breeze. Complex patterns of smaller- intermediate in terms of stability, humidity, and the scale maximum and minimum flash density were super- median number of flashes per day. Northwest flow oc-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC AUGUST 2005 S M I T H E T A L . 437 curs most often during May and is typically a postfron- REFERENCES tal scenario, not associated with the subtropical ridge. Arritt, R. W., 1993: Effects of the large-scale flow on character- Diurnal distributions revealed the distinct cyclic na- istic features of the sea breeze. J. Appl. Meteor., 32, 116–125. ture of lightning in the region. Flash counts were small- Blanchard, D. O., and R. E. López, 1985: Spatial patterns of con- est during the overnight period. During the early morn- vection in south Florida. Mon. Wea. Rev., 113, 1282–1299. ing hours, the number of flashes began to increase, Camp, J. P., A. I. Watson, and H. E. Fuelberg, 1998: The diurnal reaching a maximum that corresponded to the time of distribution of lightning over north Florida and its relation to the prevailing low-level flow. Wea. Forecasting, 13, 729–739. maximum heating and instability. Southwesterly flow Cummins, K. L., M. J. Murphy, E. A. Bardo, W. L. Hiscox, R. B. had the most flashes and peaked earliest in the after- Pyle, and A. E. Pifer, 1998: A combined TOA/MDF technol- noon. On the other hand, the drier and more stable ogy upgrade of the U.S. National Lightning Detection Net- northeasterly flow had the smallest number of flashes work. J. Geophys. Res., 103, 9035–9044. and was one of the last flow regimes to reach its after- Estoque, M. A., 1962: The sea breeze as a function of the prevail- noon peak. Flash counts in the western half of the study ing synoptic situation. J. Atmos. Sci., 19, 244–250. FSL and NCDC, 1999: Radiosonde Data of North America 1946– region generally peaked 1 h later than those in the east- 1999. Version 1.0, CD-ROM. [Available from DOC/NOAA/ ern half. This may due to the weaker western thermal OAR, Forecast Systems Laboratory, R/FSL, 325 Broadway, circulations taking slightly longer to mature. Boulder, CO 80305.] Flash density maxima were observed near Lake Fuelberg, H. E., and D. G. Biggar, 1994: The preconvective envi- Charles, Baton Rouge, New Orleans, Biloxi, and Mo- ronment for summer thunderstorms over the Florida pan- bile. These urban maxima resulted from interactions handle. Wea. Forecasting, 9, 316–326. Gentry, R. C., and P. L. Moore, 1954: Relation of local and gen- between the prevailing synoptic flow and one or more eral wind interaction near the sea coast to time and location smaller-scale thermal circulations such as sea and lake of air mass showers. J. Meteor., 11, 507–511. breezes that are influenced by the presence of marshes Gibson, H. M., and T. H. Vonder Haar, 1990: Cloud and convec- and swamps as well as convex or concave coastlines. tion frequencies over the southeast United States as related Enhanced frictional convergence, heat island effects, to small-scale geographic features. Mon. Wea. Rev., 118, and altered cloud microphysics due to air pollution also 2215–2227. Hodanish, S., D. Sharp, W. Collins, C. Paxton, and R. E. Orville, are believed to contribute to the lightning enhance- 1997: A 10-yr monthly lightning climatology of Florida: 1986– ments seen in these areas. 95. Wea. Forecasting, 12, 439–448. This study has described complicated flash density Huffines, G. R., and R. E. Orville, 1999: Lightning ground flash patterns along the northern Gulf coast and has high- density and thunderstorm duration in the continental United lighted complex interactions between the synoptic-scale States: 1989–96. J. Appl. Meteor., 38, 1013–1019. flow and various mesoscale forcing mechanisms that Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re- analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471. appear to produce this convection and lightning. Much Laird, N. F., D. A. R. Kristovich, R. M. Rauber, H. T. Ochs III, additional study will be necessary to fully understand and L. J. Miller, 1995: The Cape Canaveral sea and river the causes for the flash density patterns. High- breezes: Kinematic structure and convective initiation. Mon. resolution numerical modeling will be especially impor- Wea. Rev., 123, 2942–2956. tant to fully understand the interplay between the com- Lericos, T. P., H. E. Fuelberg, A. I. Watson, and R. L. Holle, 2002: plex processes occurring in the region during the warm Warm season lightning distributions over the Florida peninsula as related to synoptic patterns. Wea. Forecasting, 17, season that produce deep convection. 83–98. Livingston, E. S., J. W. Nielsen-Gammon, and R. E. Orville, 1996: Acknowledgments. We appreciate the assistance of A climatology, synoptic assessment, and thermodynamic Todd Lericos (NWS Spokane) who answered many evaluation for cloud-to-ground lightning in Georgia: A study questions about computing procedures. Similarly, Ron for the 1996 Summer Olympics. Bull. Amer. Meteor. Soc., 77, Holle of Vaisala Inc. answered questions about the 1483–1495. NLDN data. Kent Kuyper (NWS Lake Charles) pro- López, R. E., and R. L. Holle, 1987: The distribution of summertime lightning as a function of low-level wind flow in central vided insights into the meteorology of the study area. Florida. NOAA Tech. Memo. ERL ESG-28, National Severe Chris Kiley at the Florida State University kindly as- Storms Laboratory, Norman, OK, 43 pp. sisted with figure preparation. Maier, L. M., E. P. Krider, and M. W. Maier, 1984: Average diur- This research was funded by the National Oceanic nal variation of summer lightning over the Florida peninsula. and Atmospheric Administration’s (NOAA) Coopera- Mon. Wea. Rev., 112, 1134–1140. tive Program for Operational Meteorology, Education McPherson, R. D., 1970: A numerical study of the effect of a coastal irregularity on the sea breeze. J. Appl. Meteor., 9, and Training (COMET) under Grant S00-19127 and 767–777. by the NOAA CSTAR program through Grant Medlin, J. M., and P. J. Croft, 1998: A preliminary investigation NA03NWS4680018. and diagnosis of weak shear summertime convective initia-

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC 438 WEATHER AND FORECASTING VOLUME 20

tion for extreme southwest Alabama. Wea. Forecasting, 13, tributions associated with synoptic map types over Florida. 717–728. Mon. Wea. Rev., 122, 1698–1715. Orville, R. E., 1991: Lightning ground flash density in the con- Simpson, J. E., 1994: Sea Breeze and Local Wind. Cambridge Uni- tiguous United States—1989. Mon. Wea. Rev., 119, 573–577. versity Press, 234 pp. ——, 1994: Cloud-to-ground lightning flash characteristics in the Steiger, S. M., and R. E. Orville, 2003: Cloud-to-ground lightning contiguous United States: 1989–1991. J. Geophys. Res., 99, enhancement over southern Louisiana. Geophys. Res. Lett., 833–841. 30, 1975, doi:10.1029/2003GL017923. ——, and R. W. Henderson, 1986: Global distribution of midnight ——, ——, and G. Huffines, 2002: Cloud-to-ground lightning lightning: September 1977 to August 1978. Mon. Wea. Rev., characteristics over Houston, Texas: 1989–2000. J. Geophys. 114, 2640–2653. Res., 107, 4117, doi:10.1029/2001JD001142. ——, and A. C. Silver, 1997: Lightning ground flash density in the Wakimoto, R. M., and N. T. Atkins, 1994: Observations of the sea contiguous United States: 1992–95. Mon. Wea. Rev., 125, breeze front during CaPE. Part I: Single-Doppler, satellite, 631–638. and cloud photogrammetry analysis. Mon. Wea. Rev., 122, ——, and G. R. Huffines, 2001: Cloud-to-ground lightning in the 1092–1114. United States: NLDN results in the first decade, 1989–98. Mon. Wea. Rev., 129, 1179–1193. Watson, A. I., and R. L. Holle, 1996: An eight-year lightning cli- ——, and Coauthors, 2001: Enhancement of cloud-to-ground matology of the southeast United States prepared for the lightning over Houston, Texas. Geophys. Res. Lett., 28, 2597– 1996 Summer Olympics. Bull. Amer. Meteor. Soc., 77, 883– 2600. 890. ——, G. R. Huffines, W. R. Burrows, R. L. Holle, and K. L. Cum- Westcott, N. E., 1995: Summertime cloud-to-ground lightning ac- mins, 2002: The North American Lightning Detection Net- tivity around major midwestern urban areas. J. Appl. Meteor., work (NALDN)—First results: 1998–2002. Mon. Wea. Rev., 34, 1633–1642. 130, 2098–2109. Wexler, R., 1946: Theory and observations of land and sea Pielke, R. A., 1974: A three-dimensional numerical model of the breezes. Bull. Amer. Meteor. Soc., 27, 272–287. sea breezes over south Florida. Mon. Wea. Rev., 102, 115– Zajac, B. A., and S. A. Rutledge, 2001: Cloud-to-ground lightning 139. activity in the contiguous United States from 1995 to 1999. Reap, R. M., 1994: Analysis and prediction of lightning strike dis- Mon. Wea. Rev., 129, 999–1019.

Unauthenticated | Downloaded 09/27/21 07:17 PM UTC