Cloud-to-ground (CG) lightning detected by the National Lightning Detection Network (NLDN) indicates a relatively high flash density over Houston, Texas, for the 12-year period 1989–2000. A significant enhancement of 45% in the flash density is observed compared to the nearby surrounding areas. The strength of the enhancement varies on the basis of season and time of day, with the greatest increases occurring during the summer (58%) and during the 0900–1800 LT time periods in each season. Observations indicate that large lightning events (defined as days with >100 flashes in a geographic region that includes Houston and nearby rural areas) were responsible for the climatological lightning anomaly and that increased thunderstorm initiation was not the most significant cause of the enhancement. A decrease (−12%) in the percentage of positive flashes is observed over the city. Higher negative median peak currents along the coast and well into the Gulf of Mexico were also discovered. Several explanations for our observations are suggested. The urban heat island and increased cloud condensation nuclei concentrations, especially from industrial pollution, are speculated to be significant factors in creating lightning enhancement. Pollution effects are speculated to cause a change in a thunderstorm's charge distribution, which can affect the polarity of CG flashes. The potential effect of the nearby coastal Gulf salt water on the calculated peak current is examined. Variations in multiplicity values across the region are observed but not explained.
 A 12-year climatological analysis (1989–2000) of National Lightning Detection Network (NLDN) data has indicated a significant enhancement of lightning activity over the Houston, Texas, region as compared to nearby rural areas [Orville et al., 2001]. The Westcott  study was the first to use NLDN data to reveal the effect of several cities on enhancing cloud-to-ground (CG) lightning activity over and downwind of them, but Houston was not in that study. Orville et al. are the first to document the effect that Houston has on CG lightning. They showed the enhancement was evident during both the winter and the summer seasons throughout the 1989–2000 period. In this paper we extend the study of Orville et al. Please note that hereinafter, “lightning” and “flash” refer only to cloud-to-ground lightning.
Westcott  and Orville et al.  discussed several mechanisms for explaining the enhanced lightning activity. These include the urban heat island circulation, addition of thermal energy, frictional lift, and air pollution. Orville et al. also examined the sea breeze influence on lightning activity over and near Houston. The main problem that remains to be solved is to determine the significance of each contributing factor listed above to the climatological lightning enhancement over Houston. Assessing the impact of these urban effects on climate has been a significant research problem for meteorologists over the past several decades [e.g., Landsberg, 1981].
 The NLDN, in addition to measuring the flash location, also measures the polarity, peak current, and multiplicity of each detected flash. An analysis of these characteristics during the 1989–2000 period over the Houston region reveals some interesting results. The percent of flashes lowering positive charge to ground is observed to be less over and downwind of the Houston urban area than nearby upwind areas. In contrast, Lyons et al. [1998b] and Murray et al.  correlated the ingestion of forest fire smoke from Mexico in 1998 with a higher percentage of positive CG lightning, but the microphysical reasons of how pollution might affect the polarity of lightning were not suggested. Large negative peak current flashes were observed immediately off the Gulf of Mexico coastline during the 1989–2000 period. Lyons et al. [1998a] suggested surface conductivity as a possible cause because the conductivity of seawater is greater than that for land.
2. Previous Studies of Urban Weather Modification
 In the work of Westcott , 16 Midwestern cities (during the months of June, July, and August 1989–1992) were shown to have an enhancement of CG lightning frequency by as much as 40–85% over and downwind of the urban areas as compared to rural areas upwind. Westcott speculated the following causal factors for this finding: increased urban cloud condensation nuclei (CCN) concentrations, the urban heat island, and frictional lift. Unfortunately, there was no single factor found in her study to explain the increase in CG lightning, so a combination of the aforementioned factors was assumed. Westcott suggested that future research in this area would require CCN and cloud drop size spectra measurements in order to determine the most important city effects on lightning.
 The Metropolitan Meteorological Experiment (METROMEX), conducted during the summer months in the early to mid-1970s in the St. Louis, Missouri, area, was one of the few thorough field programs to investigate urban effects on local weather phenomena such as rainfall and thunderstorms [Changnon, 1981b]. Precipitation maxima were observed to the northeast and east of the urban heat island centered on the commercial district of the St. Louis inner city [Semonin, 1981]. Hail and wind gusts had a maximum of occurrence in the region of the eastside of the rainfall maximum. Thunder day patterns for 1971–1975 showed a distinct maximum over and downwind from St. Louis, with a 40% increase in thunderstorms and their durations in these areas as compared to background rural values [Changnon, 1981a].
 Surface conditions recorded during METROMEX showed the following characteristics for the St. Louis area: a maximum urban heat island of greater than 2.0°C observed between 0000 and 0600 LT, an average (4 years) summer dew point deficit over the urban area of about −0.5°C, and significant low-level convergence (−3 to −4 × 10−5 s−1) and upward motion (3–7 cm s−1) over and downwind of the city during the daylight hours [Semonin, 1981; Braham, 1981; Braham et al., 1981]. Most of these conditions promote boundary layer convection, except the dew point deficit.
Huff and Changnon  specifically developed a climatology for Houston concerning its effects on rainfall, hail days, and thunder days. The climatology is based on at least 20 years of recorded data from 15 to 25 weather bureau stations within 80–120 km of the city. The effects on summer rainfall near the city center revealed a 9% maximum increase from nearby rural values. Air mass storms were indicated as the primary reason for this anomaly. A maximum point increase in thunder days was found to be 10% during the same season over the city and 15 km downwind but was not a statistically significant finding. The hail-day increase, however, was found to be significant. A maximum area increase of nearly 400% during the summer was found over an area of industrial growth 15 km downwind of the urban area.
 During a study conducted in New York City, where one may expect to find sea breeze and urban effects on the local weather, vertical velocity was found to be 2 times larger during midday than either in the early morning or in the evening [Landsberg, 1981, p. 143]. The sea breeze front was also found to steepen, caused by the city frictional effects [Anderson and Bornstein, 1979]. This type of development, we suggest, can particularly enhance convection during a typically warm, humid summer day over a city like Houston.
3. Data and Methods
 The NLDN data were obtained from Global Atmospherics, Inc., Tucson, Arizona, for the period 1989–2000. The network consists of 106 sensors across the United States [Orville and Huffines, 1999]. The NLDN was upgraded in 1994, which included a combination of Improved Accuracy From Combined Technology (IMPACT-combined DF and time-of-arrival (TOA)) and TOA sensors. A full description of the upgrade is given by Cummins et al. . The upgrade resulted in improving the median accuracy to within 500 m and expected flash detection efficiency to 80–90% (for events with peak currents above 5 kA). Positive flashes with median peak currents of less than 10 kA, believed to be primarily cloud flashes after the 1994 upgrade, were rejected on the basis of the recommendation from Cummins et al. The lightning data were plotted using 5 km resolution. The plots included county, urban, and suburban outlines (defined by population density) for the Houston area. The data, used to create these outlines, were in Shapefile format and came from the U.S. Geological Survey (available from the World Wide Web at http://www-atlas.usgs.gov/atlasftp.html, 2001).
 A statistical analysis of the lightning data was conducted to determine the significance and nature of the lightning variations over and near Houston. The mean lightning characteristic values in three 0.7° latitude by 0.85° longitude boxes were calculated (see Figure 1 for the location of the boxes with regard to Houston, Harris County, and Galveston Bay). One box (B) was centered on the Houston flash density anomaly (chosen to include the urban outline and the associated >6 flashes km−2 contour), and the other two boxes (A, C) were located to the southwest and northeast of the city (classified as environmental regions). These boxes were chosen to be about the same distance away from the coastline so that one box would not be more affected by the sea breeze than the others. The selection of the locations of boxes A and C, as representative of the Houston environment, was supported by the observed background southwest-to-northeast gradient of increasing CG lightning flash density over east Texas [see Orville and Huffines, 2001, Figure 3]. Calculating the mean lightning characteristic values for each box and comparing them enabled us to assess the quantitative strength of the anomaly. This procedure was conducted for each season (defined by month, e.g., summer is June, July, and August) and time of day (0900–1800 LT (late morning/afternoon), 1800–0900 LT (night/early morning)) combination. The statistical significance of the anomaly was determined by performing a one-sided student t-test of the differences in mean values [Milton and Arnold, 1995, pp. 351–352] both for boxes B versus A and for boxes B versus C.
 To understand whether it was many small lightning events or a few large events contributing to the enhancement, the amount of lightning that occurred in each of the aforementioned geographical boxes for every day of the 12-year period was calculated. The daily flash counts (in all three boxes combined), on a given day, were found to range from zero to several thousand flashes. These data were further analyzed by segmenting the number of flashes in all three boxes combined, on a given day, into intervals and determining for each flash interval what fraction of the total flashes within all three boxes combined were detected in the Houston box (expected value is 0.33 without enhancement).
4. Lightning Flash Characteristics Over the Houston, Texas, Region
4.1. Flash Density
Figure 2 shows the 12-year (1989–2000) climatological mean flash density (5 km resolution) for a portion of southeast Texas, centered on Houston. Over three million flashes were recorded in this region during this period. Values of over 6 flashes km−2 yr−1 dominate the Houston urban area (outlined in white), with an area of more than 7 flashes km−2 yr−1 located about 10 km to the northeast of the city. The shape of the anomaly was asymmetric, with greater values extending farther to the east and northeast of Houston. According to plots provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado (see http://www.cdc.noaa.gov/), mean steering level winds (700 mbar) over this period were west-southwesterly over Houston, which may have been responsible for the asymmetric shape of the lightning anomaly.
 Three 0.7° latitude by 0.85° longitude geographic boxes (Figure 1) were chosen to compare the amount of lightning in the Houston anomaly with that over the environment. Table 1 (column 2) shows the ratios for the amount of lightning over Houston compared to that for the average environment for different seasonal and time-of-day periods. To calculate these ratios, the mean density value in the Houston box was divided by the average of the mean density values in the environmental boxes. For all 12 years, the enhancement in flash density was near 45%. Table 1 also reveals that the enhancements were greater for the 0900–1800 LT period (late morning/afternoon). The season that had the greatest effect on lightning was the summer (58%), while the winter showed the weakest enhancement at 18%.
Table 1. Ratios in Mean Flash Density and Percent Positive Values for 1989–2000 Seasonal and Time-of-Day Periodsa
2B/(A + C)
2B/(A + C)
P-values for the differences in mean values, both between boxes B and A and between B and C (Figure 1), are shown in parentheses.
 The percentage of positive flashes was less over and near the city of Houston compared to the nearby environment for all 12 years (−12%, Table 1, column 5). Figure 3 displays the distribution of percent positive flashes throughout the period 1989–2000. Only 5% of the flashes recorded over Houston were positive. Table 1 shows that the summer had the greatest reduction in percent positive values over Houston (−20%), while the winter season values were enhanced by 17%. The effect of the time of day is dependent on the season.
4.3. Median Peak Negative Current
 The geographic distribution of the CG lightning median peak negative current over the Houston area for 1989–2000 showed a very distinct pattern (Figure 4). Over the Gulf of Mexico the calculated peak current was enhanced as compared to the values over land. The transition was quite sharp, positioned at the coastline. Figure 4 shows that the enhancement extended over Galveston Bay as well. Values over Houston averaged near 26 kA, while over the Gulf waters they approached 32 kA. Another interesting feature observed from Figure 4 was the extension of the enhanced current values over the land areas about 80 km to the east and northeast of Houston. Peak currents there were near 28 kA.
4.4. Mean Negative Multiplicity
 The mean negative multiplicity (number of strokes per flash) values for the 1989–2000 period across southeast Texas are plotted in Figure 5. Values generally ranged between 2.40 and 2.85 over the Houston vicinity. Higher values were located to the north of Houston, with a distinct minimum (2.25) observed 100 km to the east of the city. Also, a band of enhanced values near 2.85 existed offshore parallel to the coastline.
4.5. Median Peak Positive Current
Figures 4 and 6 show that the maximum values of median peak positive current (25 kA) were less than those for negative current (32 kA) within the area examined. The highest values for both occurred over the Gulf of Mexico. Also, an area of enhanced positive current (values near 21 kA) was located over the western half of Figure 6 (50 km west of Houston), which corresponds to a region of minimum negative current (24 kA) in Figure 4. The coastline transition to increased values of positive current over the Gulf was not so distinct as for negative current.
4.6. Mean Positive Multiplicity
Figure 7 shows the geographic distribution of mean positive multiplicity values over southeast Texas. The variations in positive multiplicity (1.00–1.25) are less than for negative flashes (2.25–2.85) (Figure 5). There is a gradual increase in positive multiplicity from the coast to about 100 km inland. Apparently, the city had no effect on this lightning characteristic.
5. Statistical Inferences on the Nature of the Houston Lightning Enhancement
 The fraction of total CG flashes (sum within boxes A, B, and C in Figure 1) which occurred in the Houston box (B) for different intervals of daily lightning flash counts is shown in Figure 8. The intervals of flash counts shown are the total number of flashes recorded in all three boxes for a given day in the 1989–2000 period. The expected value is 0.33 since there were three geographic boxes examined (two environmental, one Houston). Generally, the days with lower amounts of lightning (below 100 flashes) gave the expected value or less. However, for higher flash count intervals, the fraction of total lightning flashes over Houston became 0.4 or higher, with fractions above 0.5 in the highest interval (>10,000 flashes). These observations indicate that large lightning events (defined here as days with >100 flashes in all the boxes combined) were responsible for the climatological lightning anomaly observed over Houston.
 Another important consideration is to determine whether increased storm frequency was the significant cause of the observed climatological lightning anomaly over Houston. The number of days during the 1989–2000 period, which had at least one flash detected in each of the three geographic boxes (Figure 1) separately, was determined. The Houston area (box B) had 1530 such days, while the environmental boxes (A, C) had an average of 1424 lightning days. The average number of flashes per day in the Houston area and within the environmental boxes was 293 and 213 flashes/day, respectively, during these events. This shows that Houston had only 7% more days with lightning than over the environment during this period but had 38% more lightning on those days. Only 6% of the days with lightning detected in at least one box had at least one flash over the Houston box and none over either of the environmental boxes. The average flash rate was only 22 flashes/day for those events that occurred just over Houston. These results indicate that more thunderstorm initiation was not the most significant cause of the climatological lightning enhancement over Houston.
 There were 953 days with at least one flash recorded in all the boxes. This represents 49% (953/1957) of the total number of days with lightning recorded in at least one box throughout the period. The average flash rate for these days over Houston and the environment was 417 and 268 flashes/day, respectively. This information also indicates that enhancement in average daily lightning activity, not an increase in the number of thunderstorm days (which would indicate more initiation), was the more significant contributor to the climatological anomaly over Houston. Also, if more thunderstorm initiation was an important contributor, one would expect enhancement in all the flash intervals shown in Figure 8.
6.1. Flash Density
 As mentioned in section 2, several aspects of the Houston area may have contributed to the lightning enhancement. The complex sea breeze and urban heat island create areas of enhanced mesoscale low-level convergence and upward motion, conditions favorable for thunderstorm initiation. The urban area can also contribute to convection through the addition of thermal energy and mechanical lift (frictional effects). Finally, enhanced CCN concentrations affect the clouds over cities in a variety of ways. The main problem to solve regarding the lightning enhancement over Houston is to determine the significance of each factor contributing to the observed lightning pattern.
6.1.1. Sea breeze-urban heat island circulation effect
Orville et al.  simulated several cycles of the land-sea breeze with the MM5 numerical model over a portion of southeast Texas, including Houston. Their results for 1700 UTC (1100 LT) show that without the “city” breeze, low-level convergence would be minimal over Houston. Upward vertical velocity maxima due to the complex (because of the irregular coastline) sea breeze alone were never located over the Houston area throughout the sea breeze cycle. Orville et al. found that with a simulated city the circulation pattern is much stronger than in the no-city run, with enhanced low-level convergence directly over Houston. The convergence over Houston triggered convective cells during the following few hours of their model run. The no-city simulation did not produce any convection, indicating sea breeze effects alone were not strong enough to initiate convection over the city. From these results, the urban heat island plays a more important role than the sea breeze in initiating thunderstorms that may be responsible for the Houston lightning anomaly.
 According to Table 1 the most significant lightning enhancement was observed over the Houston area during the summer (58%). All of the seasons had greater enhancements in the late morning/afternoon than during the night/early morning time period. Maximum heating and instability occur over Houston during these times of greatest lightning enhancement.
 Previous studies of the urban heat island and sea breeze [Pielke and Segal, 1986] have indicated that these circulations are stronger when a weak synoptic-scale flow is present. Over Houston this is found to occur during the summertime (see average wind speed by month trace available from the World Wide Web at http://www.cdc.noaa.gov/∼cas/Climo/fsod/plot.pr.475.sub.html, 2001). Urban heat island intensity, defined as the surface temperature difference between the rural and the urban areas, has been observed to be greater at night. However, Yoshikado  performed model simulations of the interaction between these two systems (urban heat island and sea breeze) and found that the daytime urban heat island circulation is stronger than the nocturnal one in spite of generally smaller values of surface temperature difference. During the daytime, the heating associated with the heat island is distributed through a considerably deeper layer than that associated with the nighttime stable boundary layer, resulting in larger pressure perturbations and greater horizontal and vertical accelerations over the city [Vukovich and Dunn, 1978]. Yoshikado also suggests the urban temperature excess could be the same for both the nighttime and the daytime over large cities. Braham et al.  took measurements that show the convergent flow deepened and strengthened from 1200 to 1400 LT over St. Louis during the METROMEX study. Hence convective initiation by the urban heat island/sea breeze system can be expected to be more frequent during the warm season daytimes. The lightning observations from Table 1 give support for this hypothesis. Detailed measurements of the urban heat island circulation over Houston are required to confirm these inferences.
6.1.2. Pollution effects
 The city of Houston has become one of the most polluted areas in the United States during the past decade (see pollution maps available on the World Wide Web at http://www.epa.gov/air/data/mapview.html, 2001). Approximately 50% of the U.S. petroleum refinery capacity is in the Houston region (W. Read, National Weather Service, personal communication, 2001). The main concentration of PM-10 (particulate matter with a diameter less than 10 μm) point emissions is located in the eastern sections of Houston county (Harris) (see map available on the World Wide Web at http://www.utexas.edu/research/ceer/texaqs/images/pm10-1.gif, 2001). PM-10 may be used as a proxy to indicate where most of the pollution is emitted in the Houston area, but the type of pollutant is also important to know. Landsberg [1981, p. 193] presents results of a study on the effect of refinery effluents on clouds. Mainly sulfates and nitrates were found in the refinery plumes. Nitrates are suspected of being the more active kind of nucleus (they are larger and more hygroscopic than the sulfates). Sulfates, however, are very small particles (diameter <0.1 μm) and more likely to stabilize clouds. Seinfeld and Pandis [1998, p. 429] state that the typical urban aerosol mass distribution usually has two distinct modes, one in the submicron regime and the other in the coarse particle regime.
 The Rosenfeld hypothesis [Rosenfeld and Lensky, 1998; Williams et al., 1999] is the basis for explaining how pollution can affect lightning production in storms that develop or move over cities like Houston. Analyses of data from the Advanced Very High Resolution Radiometer (AVHRR) by Rosenfeld and Lensky show that clouds forming over polluted areas were found to have a narrow or no coalescence zone, a deep mixed-phase zone, and glaciation occurring at higher levels when compared to rural clouds. The higher CCN concentration over cities acts to reduce the mean cloud droplet size, which decreases the droplet collision efficiency and the process of coalescence [Rogers and Yau, 1989, Figure 8.2]. These observations imply that more supercooled water is able to exist at greater depths in clouds that develop in a polluted environment. The noninductive charge separation process in thunderstorms has been found to be dependent on the amount of supercooled liquid water [Saunders, 1993]. More supercooled water may create larger graupel, which leads to more collisions with ice particles, and enhanced storm electrification. More cloud water in the mixed phase zone can also enhance cloud buoyancy through the freezing process.
 The observation that the greatest lightning enhancements occurred during the warm season afternoons also supports the pollution hypothesis. Weaker synoptic-scale winds during these time periods allow more pollution to be concentrated over the Houston industrial complexes. The urban heat island circulation, which is more intense during the summer and afternoons, has been observed to prevent the dispersion of urban pollutants and delay their inland transport [Yoshikado, 1992]. Mean 2000 seasonal PM-2.5 (particulate matter less than 2.5 μm) concentrations from two monitoring stations in Houston (Texas Natural Resource Conservation Commission (TNRCC) sites C01 and C15) show the summer had the highest concentration (13.54 μg m−3), while the winter had the lowest (9.42 μg m−3) (data available on the World Wide Web at http://www.tnrcc.state.tx.us/cgi-bin/monops/select_month/region12.gif, 2001). Higher climatological flash densities over eastern Harris County, directly over the major industrial areas, also suggest that pollution played a key role in the Houston lightning enhancement.
6.1.3. Other effects
 From the same MM5 simulation used by Orville et al. , it was determined that the convection that had developed entrained urban air with an equivalent potential temperature 3°K higher than in the no-city run, despite lower surface dew points. However, observations over St. Louis during METROMEX resulted in urban theta-e deficits between 2° and 4°K throughout much of the boundary layer. Studies of urban rain enhancement [Braham et al., 1981] used this result to explain their findings by claiming that the theta-e deficits over cities lead to weaker storm updrafts, which allowed previously suspended precipitation particles to fall to the ground. On the basis of these and other measurements of buoyancy over cities, Braham et al. concluded urban areas to be thermodynamically less favorable for storms than nearby rural areas and that the convergence field is more important in initiating cloud development over cities. These contradictory views on whether or not urban areas have theta-e excesses or deficits indicate that more measurements are needed to determine if Houston can not only initiate storms but also intensify them by this effect.
 Frictional lift [Westcott, 1995] may play a role in increasing storm initiation and lightning activity over Houston. The rough urban landscape and heat island can act as a barrier to the synoptic or mesoscale wind flow, enhancing upward motion on the windward side of the city. Frictional effects are greater with increasing wind speed. The observation that lightning enhancement was stronger during the times when mean surface wind speeds were lower (summer) does not support the friction hypothesis.
6.2. Percent Positive Flashes
Figure 3 reveals a significant reduction in percent positive CG flashes over the Houston area. Microphysical effects are speculated. Jayaratne et al.  showed in their experimental studies of the charging of soft hail (graupel) growing by riming during ice crystal interactions that impurities in the cloud water have a significant effect on the sign and magnitude of charge transfer to the graupel target. They ignited various materials (not given what the materials consisted of) in a cloud chamber, producing a large amount of CCN that caused the drop size spectrum to shift to smaller sizes. In this experiment, the charges acquired by the graupel target were negative at all temperatures (−6°C to −25°C). The magnitude of the negative charging increased when the droplets possessed stronger solutions of the contaminants. The charge reversal temperature of −20°C moved to higher temperatures when the droplets carried small traces of the most commonly occurring contaminants in nature. If negative graupel charging occurs at warmer temperatures due to an increase in impurities in the cloud water, this can extend the main negative charge region lower in the cloud, suppressing the positive charge center below [Pruppacher and Klett, 1997, Figure 18-2]. The expanded main negative charge region of the thunderstorm tripolar charge distribution model [MacGorman and Rust, 1998, Figure 3.2] may produce more negative CG flashes, decreasing the percentage of positive flashes. The lower-percent positive CG flashes observed over the Houston area may be explained on the basis of these experimental results and the above inferences drawn from them.
 Experimental results from Avila et al.  show that target graupel charged positively over most of the temperature range studied (−10°C to −25°C), when a smaller-droplet spectrum (one that had more than 30% of the droplets with sizes greater than 13 μm) was used, but negatively at temperatures below −18°C for the larger-droplet spectrum (had more than 50% of the droplets greater than 13 μm) during ice-ice collisions in the presence of supercooled water. This implies a deeper positive charge center in the lower region of a thunderstorm containing a smaller-droplet spectrum. This may further imply higher-percent positive values over an area that is likely to produce this type of droplet spectrum (Houston, more CCN, smaller mean droplet size). These contradictory hypotheses on how pollution may affect the polarity of CG flashes need to be further examined.
 The summer had the strongest reduction (−20%, Table 1) in percent positive values over Houston (−24% during the late morning/afternoon period), while the winter showed an enhancement (17%) in values (21% during the night/early morning period). Pollution was indicated as having a potential impact on the polarity of CG flashes, and the summer afternoons are most likely to have greater concentrations of aerosols over Houston because of weaker synoptic-scale winds and a stronger urban heat island circulation recirculating the pollutants. Hence the hypothesis, based on the experimental results of Jayaratne et al. , is supported by the observation that the greatest reduction in percent positive values occurred during the summer afternoon time period.
 CCN distributions and cloud-droplet spectra over the city of Houston must be measured and studied to further support the explanations given for the observed lightning characteristic patterns. The observations over Houston and those by Lyons et al. [1998b] indicate that aerosols may have a major impact on the lightning flash polarity, and the results of Jayaratne et al.  and Avila et al.  are the only basis for explaining these effects thus far.
6.3. Median Peak Currents for Negative and Positive Flashes
Figure 4 shows a strong tendency for high values of median peak negative current to be primarily over the salt waters of the Gulf of Mexico and Galveston Bay. The peak positive current also increased offshore (Figure 6) but not so distinctly at the coastline as the negative current.
Lyons et al. [1998a] noticed similar results with the negative current over the Gulf of Mexico. Their reasoning for this observation is perhaps the higher conductivity of the underlying salt water as compared to land. This would act to decrease the attenuation of the electromagnetic signal from a lightning discharge as it traveled over the water, making the flash appear stronger [Stratton, 1941, pp. 520–521]. There are a couple of problems with this hypothesis, however. First, the enhancement was observed to extend inland a few tens of kilometers, suggesting that not only surface effects were important. Secondly, the same pattern (a sharp transition at the coastline) was not observed for peak positive current, and it should if the attenuation explanation is correct.
6.4. Mean Negative and Positive Stroke Multiplicity
Figures 5 and 7 show observations that support previous studies [Orville and Huffines, 2001] that stroke multiplicity is a strong function of polarity. Negative multiplicity values ranged from 2.25 to 3.00, while most positive multiplicity values were near 1.00 over the examined area.
 Climatological cloud-to-ground lightning data from the NLDN have been studied for portions of southeast Texas, centered on Houston, Texas, for the years 1989–2000. A significant lightning enhancement (45%) was discovered over Houston as compared to background rural values. Analysis by season and time of day (late morning/afternoon, overnight/early morning) indicated that significant enhancement occurred throughout the year, with the summer and fall late morning/afternoon periods having the largest Houston enhancements (60 and 74%, respectively).
 Statistical analysis shows that the enhancement in CG lightning was associated with large lightning events, especially days when more than 100 flashes were recorded in the three geographic boxes combined. If increased thunderstorm initiation over the Houston region was the main cause of the climatological anomaly, the enhancement should be observed with all sizes of lightning events. The results from Westcott  showed that the urban area did not initiate new lightning storms, suggesting that existing thunderstorms were the most strongly affected by cities.
 Several hypotheses for lightning enhancement over Houston were presented. The sea breeze-urban heat island effect was supported by the fact that the greatest lightning enhancements occurred during the warm seasons, times when the sea breeze and urban heat island circulations are strongest due to the increase in heat and instability and weak synoptic-scale winds. Increased thermal energy input from the heat island would intensify urban storms, but there is evidence [Braham et al., 1981] that cities may be thermodynamically less favorable for thunderstorm development. Finally, pollution is an important consideration when discussing the Houston environment due to the strong oil refinery and automobile presence there. The lightning enhancements were most significant when the urban air pollution concentration was the highest (summer). Recent satellite measurements show that the coalescence process is greatly reduced in urban convective clouds due to the large number of CCN present [Rosenfeld and Lensky, 1998]. Reduced coalescence, according to the Rosenfeld hypothesis, can enhance charge separation (hence lightning) by increasing the amount of supercooled liquid water available in the thundercloud.
 The city of Houston also had the effect of decreasing the percentage of positive flashes. Only microphysical effects could be speculated, and they were based on the experimental results of Jayaratne et al. . According to their results, if supercooled cloud droplets contain a higher concentration of contaminants, graupel charges negatively at warmer cloud temperatures. This should act to expand the main negative charge region to lower portions in the cloud, producing more negative CG discharges and decreasing the percentage of positive flashes. However, Avila et al.  showed that a smaller-droplet spectrum causes graupel to charge positively in the temperature range of −10°C to −25°C. Noting that cloud droplet sizes should be smaller in urban clouds, this finding suggests that the lower positive charge region should extend to higher altitudes in the cloud, contradicting the hypothesis for reduced percent positive lightning over Houston based on the Jayaratne et al. results.
 Finally, peak negative current estimates show a strong coastal dependence, with higher values located over the Gulf of Mexico and Galveston Bay. The higher conductivity of the underlying salt water is the only speculated cause, but we have no complete explanation for this observation.
 The results from this study have shown that Houston, and thus perhaps other urban areas, has a significant effect on cloud-to-ground lightning. Several hypotheses were presented to explain these effects, and each one was given some support by the observations. However, more research is required to test these hypotheses and assess their relative significance. The aerosol effect on lightning is not well established. CCN and cloud droplet measurements of the air entering and within urban clouds are needed to show strong evidence for the Rosenfeld hypothesis and the hypothesis relating urban percent positive values to the measurements of Jayaratne et al.  and Avila et al. . Modeling studies will also be beneficial. The ability to study factors associated with the urban heat island, sea breeze, and pollution separately will help to show the relative importance of each to the observed lightning enhancement.
 The lightning data were obtained from Global Atmospherics, Inc., Tucson, Arizona. Data handling at Texas A&M University is under the direction of Jerry Guynes, and we thank him for his help. Our research is part of a lightning program supported by the National Science Foundation (ATM-9806189 and ATM-0119476) and the National Oceanic and Atmospheric Administration (cooperative agreement NA87WA0063 and NA17WA1011).