- Email: [email protected]
Wildfires in eastern Texas in August and September 2000: Emissions, aircraft measurements, and impact on photochemistry
ARTICLE IN PRESS
Atmospheric Environment 39 (2005) 4983–4996 www.elsevier.com/locate/atmosenv
Wildfires in eastern Texas in August and September 2000: Emissions, aircraft measurements, and impact on photochemistry Victoria Junquera, Matthew M. Russell, William Vizuete, Yosuke Kimura, David Allen Center for Energy and Environmental Resources (R7100), The University of Texas at Austin, 10100 Burnet Road, Austin, TX 78758, USA Received 11 February 2005; received in revised form 26 April 2005; accepted 6 May 2005
Abstract The accuracy of wildfire air pollutant emission estimates was assessed by comparing observations of carbon monoxide (CO) and particulate matter (PM) concentrations in wildfire plumes to predictions of CO and PM concentrations, based on emission estimates and air quality models. The comparisons were done for observations made in southeast Texas in August and September of 2000. The fire emissions were estimated from acreage burned, fuel loading information, and fuel emission factor models. A total of 389 km2 (96,100 acres) burned in wildfires in the domain encompassing the Houston/Galveston-Beaumont/Port Arthur (HGBPA) area during August and September 2000. On the days of highest wildfire activity, the fires resulted in an estimated 3700 tons of CO emissions, 250 tons of volatile organic carbon (VOC) emissions, 340 tons of PM2.5, and 50 tons of NOx emissions; estimated CO and VOC emissions from the fires exceeded light duty gasoline vehicle emissions in the Houston area on those days. When the appropriate aircraft data were available, aloft measurements of CO in the fire plumes were compared to concentrations of CO predicted using the emission estimates. Concentrations estimated based on emission predictions and air quality models were within a factor of 2 of the observed values. The estimated emissions from fires were used, together with a gridded photochemical model, to characterize the extent of dispersion of the fire emissions and the photochemistry associated with the fire emissions. Although the dispersion and photochemical impacts varied from fire to fire, for wildfires less than 10,000 acres, the greatest enhancements of CO and ozone concentrations due to the fire emissions were generally confined to regions within 10–100 km of the fire. Within 10 km of these fires, CO concentrations can exceed 2 ppm and ozone concentrations can be enhanced by 60 ppb. The extent of photo-oxidant formation in the plumes was limited by NOx availability and isoprene emissions from forested areas downwind of the fires provided most of the hydrocarbon reactivity in the plumes. r 2005 Elsevier Ltd. All rights reserved. Keywords: Wildfires; Wildfire inventory; Wildfire emissions; Emissions inventory; Emissions modeling; TexAQS; Photochemical modeling; CAMx; Plume rise
Corresponding author. Tel.: +1 512 475 7842; fax: +1 512 471 7060.
E-mail address: [email protected] (D. Allen). 1352-2310/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2005.05.004
ARTICLE IN PRESS 4984
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
1. Introduction Outdoor fires can emit substantial amounts of particulate matter (PM), carbon monoxide (CO), nonmethane hydrocarbons (NMHCs), nitrogen oxides (NOx), and ammonia (NH3) into the atmosphere (Sandberg, 1999). In Texas, emissions of CO and fine PM (PM with diameters less than 2.5 mm, PM2.5) from fires account for 10% and 1–2% of total annual statewide emissions, respectively (Dennis et al., 2002). On days and seasons when large fire events occur, these percentages can be much higher, and fires can dominate emissions and impact air quality over substantial areas (Liu, 2004). The air quality impacts of fires are often estimated based on predicted emissions. Fire emission predictions are based on estimates of area burned, fuel mass burned per area, and emissions per mass of fuel combusted. Significant uncertainties can arise in estimating each of these parameters; thus the emission estimates can be uncertain. Dennis et al. (2002) estimated uncertainties of approximately a factor of 2 in area burned and fuel loadings. Additional uncertainties in assessing the air quality impacts of fires are due to uncertainties in emission factors, which depend on the nature of the combustion (smoldering versus flaming) and the plume rise of the fire. The primary goal of this work is to evaluate the uncertainties associated with estimating fire emissions by comparing predicted emissions to observations made during a large air quality field program, conducted in southeast Texas during the summer of 2000. A secondary goal is to characterize the photochemistry occurring in the wildfire plumes during the study period.
2. Methodology 2.1. Emissions inventory In this study, emissions from wildfires were estimated for the regional domain shown in Fig. 1. Wildfire emissions were estimated for the months of August and September 2000; the measurements to which the emission estimates will be compared were collected during the Texas Air Quality Study (TexAQS), which was conducted from 15 August–15 September 2000. August and September 2000 was a period of drought and intense wildfire activity in southeast Texas. As documented in the results section of this paper, during August and September of 2000 approximately 384 km2 (95,000 acres) burned in wildfires in the HGBPA domain (the red area shown in Fig. 1), 66% of which burned in Texas. In contrast, in August and September of 1996 wildfires consumed 16.0 km2 (3960 acres), and in August and September of 1997, 59 km2 (14,620 acres) burned in
Texas (Dennis et al., 2002). Because of the intense drought during the summer of 2000, most prescribed burns that would normally occur during August and September were delayed or not conducted (Freds, 2002; McCown, 2002; Taylor, 2002), and therefore wildfires dominated the fire emissions during the study period. The information used to estimate emissions from wildfires includes the location and date of the fire, the fire area, the type of vegetation burned, the density of the vegetation at the location of the fire, also termed the fuel loading factor (mass of fuel available per acre), the fraction of fuel consumed in the fire, the emission factor of the fuel (pounds of pollutant per ton of fuel), and the emission efficiency. The fuel loading factor and the fraction of fuel burned during the fire can be combined into one parameter, the fuel consumption factor (mass of fuel consumed per acre); the fuel consumption factors for specific locations are then combined with fuel-based emission factors, into a location specific composite emission factor, in units of emissions of pollutant per acre burned. Wildfire activity data, including acreage burned and fire location, were collected from incident reports available through the National Interagency Fire Management Integrated Database (NIFMID), the Texas Interagency Coordination Center, the US Department of Agriculture Forest Service (USFS), the Louisiana Interagency Coordination Center, and the Oklahoma Department of Agriculture, Food & Forestry. The information provided by each of these data sources is described in detail elsewhere (Junquera, 2004). All the data concerning individual fire incidents were integrated into a single database and duplicate reports of fires were identified through common parameters, such as the fire incident name and location. Only fires greater than 0.04 km2 (10 acres) were analyzed for duplications in this study, since most of the smaller fires lacked complete information in the databases. It was assumed that duplicated fires under 0.04 km2 would not add significantly to uncertainties in the calculation of emissions. Once fire locations were determined, fuel consumption was estimated for each fire incident. Vegetation type and density vary significantly over the domain of the study, so the fuel consumption can vary significantly from fire to fire. Relatively little information on vegetation type was available in the incident reports. The type of burned vegetation was reported directly in the fire incident reports for only 9% of the fires in Texas and 27% of the fires in the regional domain, or 31% of the burned area in Texas and 29% of the burned area in the regional domain. Therefore, in this work, fuel models, rather than reported types of burned vegetation, were used to estimate fuel loading for fires occurring on wildland and rangeland. Specifically, the First Order Fuel Effects Model (FOFEM), a publicly available computer model developed by the Intermountain Fire
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4985
Fig. 1. Horizontal (a) and vertical (b) structure of the modeling domain for the study. Fire emissions during the study period were estimated for the regional domain, with special attention paid to the Houston-Galveston, Beaumont Port Arthur (HGBPA) subdomain. Emissions were assumed to enter a variety of vertical layers in an air quality model (Comprehensive Air Quality Model with extensions, CAMx).
Sciences Laboratory of the USFS, was used to predict fuel consumption (Reinhardt et al., 1997). The model incorporates vegetation types for the contiguous US, and default fuel loading values and consumption equations for each vegetation type (Dennis et al., 2002). The program, however, does not include information on the spatial distribution of these vegetation types. Dennis et al. (2002) selected 15 vegetation cover codes in FOFEM version 4.0, which corresponded to vegetation types in the study domain, and crossreferenced them with vegetation types allocated onto a land cover database of the state of Texas developed by Wiedinmyer et al. (2000, 2001). As a result, a land use–land cover (LULC) database with spatially resolved FOFEM 4.0 vegetation cover types and fuel consumption factors was created for Texas. The rest of the study domain was assigned fuel loading factors based on National Fire Danger Rating System (NFDRS) fuel
models, which were developed as a method to rate wildfire danger by the USDA Forest Service (Burgan, 1988). Emissions from wildfires were calculated by projecting the daily fire information (location and acreage) onto a map with spatially resolved composite emission factors. Details of the fuel consumption assignments have been reported by Junquera (2004) and the methods have been described by Dennis et al. (2002). 2.2. Ambient measurements A variety of ambient measurements were available to evaluate the performance of the emission estimates. These included total aerosol concentrations (as characterized by aerosol backscatter) as a function of elevation measured by a NOAA aircraft with downward-looking aerosol and ozone light detection and
ARTICLE IN PRESS 4986
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
ranging (lidar), and gas-phase air pollutant concentrations measured by a second NOAA operated aircraft. Details of the aircraft measurements and operation are described elsewhere (NOAA, 2003; NCAR, 2002) and are only summarized here. In the lidar measurements, the aerosol channel was set at a wavelength of 359 nm. The aircraft typically flew at an altitude of 3500 m above middle sea level (MSL), and the aerosol backscatter profiles extend approximately from 2500 m MSL to the surface with a vertical resolution of 15 m and a time resolution of 10 s (NOAA, 2003). The aerosol lidar backscatter data are not calibrated and therefore provide information about the aerosol distribution in a qualitative rather than quantitative manner (NOAA, 2003). The results of the measurements performed can be obtained from the NOAA web site (http://www.etl.noaa.gov/et2/data/data_ pages/texaqs/air_aerosol.html). The National Center for Atmospheric Research (NCAR) Electra aircraft, operated by NOAA, took 1-s interval measurements of various gaseous compounds and PM on several days of the modeling period. The full suite of measurements is described by Ryerson et al. (2003); data sets for all flights are available via ‘ftp’ or on magnetic tape upon request (NCAR, 2002). In this work, the primary focus will be on measurements of CO. Measurements from all of these platforms were available for 15 August to approximately 15 September 2000. Particular focus in this work will be on measurements made on 30–31 August and 3–6 September. A large fire that occurred on 30 August in Liberty Co., 80 km (50 miles) northeast of Houston, caused locally elevated PM and ozone concentrations. A very high regional wildfire activity characterized 3–6 September. 2.3. Photochemical modeling The wildfire emissions inventory was merged with an existing inventory of emissions from point, area, and mobile sources obtained from the Texas Commission on Environmental Quality (TCEQ) as part of the Houston/ Galveston Air Quality Science Evaluation (TCEQ, 2003). As described in the results section, biogenic emissions played a significant role in the photochemistry of the fire plumes; therefore, the details of the biogenic emission estimation methods are of interest here. The biogenic emission inventory was calculated with the Global Biosphere Emissions and Interactions System (GloBEIS) model. The source code and documentation for GloBEIS is publicly available and is described in detail by Yarwood et al. (1999a, b). Biogenic emissions are influenced primarily by vegetation type and density, solar radiation, cloud cover, and ambient temperature. GloBEIS therefore requires data characterizing LULC, ambient temperatures, and solar radiation or cloud cover across a region and time period of interest.
Vegetation characteristics were based on a landuse/ land-cover database developed by Wiedinmyer et al. (2001) for the state of Texas. Meteorological data for GloBEIS were extracted from several sources. Surface temperatures were developed by spatially interpolating temperatures measured by National Weather Service (NWS) and other weather stations throughout southeast Texas (Vizuete et al., 2002). The University of Maryland and the National Oceanic and Atmospheric Administration (NOAA) for the Global Energy and Water Cycle Experiment (GEWEX) Continent Scale International Project (GCIP) provided the estimates of PAR fluxes (TCEQ, 2004). All wind speed and humidity data were derived from the NCAR/Penn State Mesoscale Model version 5, MM5. The emissions and meteorological modeling data were input into the Comprehensive Air Quality Model with extensions (CAMx) version 3.11, a three-dimensional (3D) eulerian photochemical grid model (ENVIRON, 2000). The modeling domains have a horizontal resolution of 16 km 16 km (East Texas subdomain), 4 km 4 km (HGBPA subdomain), or 1 km 1 km (HG subdomain) (Fig. 1). The creation of the modelready file and the results of the modeling runs are briefly described in this section. CAMx includes two-way grid nesting, and a subgridscale Plume-in-Grid (PiG) module to treat the slow dispersion and the chemistry of large point source plumes. Unless the PiG module is turned on, emissions from each point source are instantly dispersed into an entire grid cell volume, given by the grid cell area and the height of the vertical layer. Wildfires were treated as point sources or ‘‘stacks’’. Each fire was modeled as a series of stacks with identical location and different heights, so that the emissions calculated for the wildfire would be uniformly distributed throughout the vertical rise of the plume. The PiG module was turned off for all fires, causing emissions to be instantly dispersed into each grid volume. Thus, the emissions from each fire were allocated homogeneously into a volume whose area was given by the resolution of the grid and whose height was given by the plume rise of the fire. The area of a cell in the 4 km 4 km and 16 km 16 km grids is equivalent to roughly 4000 and 63,000 acres, respectively. Thus, emissions from fires that are smaller than the grid cell area were over-diluted in the model. The fire emissions inventory was preprocessed using the Urban Airshed Model (UAM) Emissions Preprocessor System (EPS) version 2.0. Input into EPS2.0 includes an emissions data file in Aerometric Information Retrieval System (AIRS) Facility Subsystem (AFS) format (AFS file), a chemical split factor file (CHEMSPLIT), and a temporal split file (TMPRL). The CHEMSPLIT file includes information on the chemical speciation of NMHC emissions. For the fires, ethane,
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
ethylene, and propylene were the dominant organic emissions, comprising approximately half of the total NMHC emissions. The CHEMSPLIT file converted these speciated emissions into Carbon Bond-IV (CB-IV) species. In the CB-IV chemical mechanism, species with similar chemical properties are lumped into CB-IV groups that undergo the same chemical reactions. In the TMPRL file, the emissions were assumed to be uniform and constant throughout a day for each fire incident. For fire incidents lasting more than one day, emissions were allocated uniformly throughout the burning period. Details of the chemical speciation used in the model are provided by Junquera (2004). The plume rise for the fires, which is necessary for 3D photochemical modeling, was estimated based on calculations performed with FIREPLUME. The FIREPLUME model uses Briggs’ two-thirds law (Brown et al., 1999), and employs different calculation strategies for stable, neutral, and unstable atmospheric conditions. For fires smaller than 0.4 km2 (100 acres), a plume rise value corresponding to the top of CAMx layer 3
4987
(170.5 m) was assigned. For fires between 0.4 and 3.2 km2 (800 acres), and greater than 3.2 km2, plume rise values corresponding to the top of CAMx layers 4 (256.9 m) and 6 (431.7 m), respectively, were assigned. The rationale for these assignments is described by Junquera (2004).
3. Results and discussion 3.1. Emissions inventory In August and September 2000, 518 km2 (128,000 acres) were burned in wildfires in Texas and 389 km2 (95,000 acres) in the HGBPA domain (Table 1). In the HGBPA domain, 2% of the fires were larger than 3.24 km2 (800 acres) and accounted for 56% of the total area burned, and 74% of the fires were smaller than 0.405 km2 (100 acres) and burned only 5% of the total area. Fig. 2 shows wildfire locations and acreage burned from 22 August to 6 September. Estimated emissions of
Table 1 Burned area in km2 (acres) during the study period; 2–8 September was characterized by the highest wildfire intensity Period
Regional domain
Texas
HGBPA domain
August and September 2–8 September
971 (240,000) 386 (95,300)
518 (128,000) 251 (62,000)
389 (96,100) 262 (64,700)
Fig. 2. (a) Wildfires during the period from 22 August to 29 August 2000. (b) Wildfires during the period from 29 August to 6 September 2000.
ARTICLE IN PRESS 4988
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
Fig. 3. Emissions of CO, NMHCs, NOx, and PM2.5 from wildfires in the HGBPA domain during August and September 2000. Emissions of CO and NMHC from wildfires exceeded emissions from Light Duty Gasoline Vehicles, LDGV (indicated by horizontal line) on some days.
CO, NMHCs, PM2.5, and NOx in the HGBPA domain are shown in Fig. 3. Fig. 3 also shows the daily average emissions of CO, NMHCs, and NOx from light duty gasoline vehicles (LDGV) in the Houston-Galveston area. Fig. 3 shows that emissions of CO and NMHCs from fires exceeded emissions from LDGV on some days. The highest emissions during this period were approximately 3700, 250, 340, and 50 short tons day 1 for CO, NMHC, PM2.5, and NOx, respectively. 3.2. Comparison with aircraft data Aircraft measurements were used to assess the overall accuracy of the emission estimates. The data best suited for this comparison were collected on 6 September , in a plume from a multi-day fire in the McFaddin National
Wildlife Refuge (NWR) that consumed 13,000 acres. On 6 September , the McFaddin NWF fire plume was brought into the Houston area by a northeast wind. Fig. 4 shows the location of the fire, the flight track of the NOAA/NCAR Electra aircraft on 6 September, and the CO measurements made by the aircraft. Within the fire plume, CO concentrations of up to 490 ppbv were measured at 500 m above sea level (ASL). Background CO concentrations, or CO concentration that were not caused by the McFaddin NWR fire, were determined from aircraft measurements before and after the sharp CO peak associated with the fire, and ranged from 80 to 160 ppbv. The background concentration values were subtracted from the average CO concentration in the peak caused by the fire, 350 ppbv, yielding the concentration caused solely by the fire. Using an average value of the wind speed measured by the aircraft, 5.9 m s 1, the CO flux (mass/area time) was calculated. It was then assumed that the CO concentration was uniform throughout the mixing height. The mixing height was estimated with HYSPLIT (http://www.arl.noaa. gov/ready/hysplit4.html) for several locations along the flight section that intersected the fire plume, and an average value of approximately 500 m was obtained. This value is consistent with the coastal location of the fire and was in agreement with mixing height data estimated by Senff et al. (2002) for Galveston Bay area during the early afternoon (400–700 m). Thus, 500–700 m was used as the mixing height value. From the mixing height, the length of the CO plume due to the fire (14 km), the average CO concentration over background in the fire plume, and the wind speed, the CO flow (mass/time) was calculated. The McFaddin NWR was characterized as a marshland in newspaper articles (Houston Chronicle, 2000a, b; Associated Press Newswires, 2000; Baton Rouge Advocate, 2000) and NIFMID reports. The composite emission factor for Wet Grasslands in Texas (FOFEM 6) was therefore chosen to compute the burning rate (acres/time). The burned acreage was then calculated for a 24-h period. The fire started on 2 September (NIFMID), burned out between two levees on 4 and 5 September (Houston Chronicle, 2000a), was 95% contained by 7 September (Associated Press Newswires, 2000; Baton Rouge Advocate, 2000), and was completely mopped up by 9 September (NIFMID). From this information, the area that burned on 6 September was expected to fall somewhere between 4 and 20 km2 day 1 (1000 and 5000 acres day 1). This estimate of acreage burned per day can be compared to the acreage that would be required to generate the CO emission rate estimated from the aircraft data (mass/time). The burn rate required to generate the emission rate estimated from the aircraft data was 11–20 km2day 1 (2800–5000 acres day 1), depending on the CO background concentration and
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4989
Fig. 4. September, CO concentrations measured by NOAA’s Electra aircraft. (a) CO concentrations are represented in different colors, and wildfires sizes are represented by different dot sizes. The CO plume, from the McFaddin National Wildlife Refuge fire, was detected by the aircraft at approximately 11 AM. (b) Detail of the CO aircraft measurements with the fire plume boxed.
mixing height chosen. Thus, for the McFaddin fire, estimates of emissions and burn rates based on aircraft measurements (11–20 km2 day 1 burn rates) were consistent with the emission inventory data (4–20 km2 day 1), but both the inventory and the estimates of emissions based on aircraft measurements have uncertainties of at least a factor of 2. 3.3. Downward-looking lidar Lidar measurements provide a qualitative measure of the atmospheric concentrations and vertical distribution of PM. Lidar measurements made during the most
intense fire period, 2–8 September, indicated much higher PM backscatter than during other days of the study. As shown in Fig. 5, aircraft measurements on 3 September (not the same day as the day measurements were made by the Electra aircraft) immediately downwind of the McFaddin NWR fire (red rectangle in Fig. 5) indicate that elevated PM concentrations were present throughout the mixed layer. The fire plume was carried into the Gulf of Mexico and was not detected in subsequent measurements on 3 September. On 6 September (the day of the Electra measurements), the lidar data indicate that elevated PM concentrations were observed at all elevations and throughout the domain.
ARTICLE IN PRESS 4990
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
Fig. 5. (a) PM backscatter registered by the NOAA aircraft along its flight path with downward-looking lidar instrumentation on 3 September (top) and 6 September. (b) Corresponding aircraft trajectories and backscatter measured at 2000 m AGL. 3 September shows a sharp spike in PM backscatter (red rectangle), which was caused by the McFaddin NWR fire. The lidar detected the highest PM backscatter on 6 September, the day with the highest fire activity.
3.4. Photochemical modeling More detailed assessments of the air quality impacts of fires can be performed using 3D gridded photo-
chemical models. In this work, CAMx was used as the modeling tool and the modeling analyses examined the spatial dispersion of fire emissions, and the impact of fire emissions on ozone formation and other
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4991
Fig. 6. Difference plots (concentrations predicted by the simulation including fire emissions—concentrations predicted by simulation without fire emissions) for CO (a) and ozone (b) in layer 1 (ground level to 33.9 m above ground level). The figure shows, for each grid cell, the maximum difference in CO and ozone concentrations throughout the period from 22 to 31 August.
Fig. 7. Difference plots (concentrations predicted by the simulation without fire emissions—concentrations predicted by simulation including fire emissions) for CO (left) and ozone (right) in layer 1 (ground level to 33.9 m above ground level). The figures are for 6 September 2000 hour 12; the time of the largest increase in ground level ozone concentrations due to fires.
photochemical processes. In the analyses, CAMx simulations were performed both with and without the emissions from fires included. The difference between these simulations characterizes the impact of the fires. Figs. 6 and 7 show two difference plots (concentrations predicted by the simulation including fire emissions minus concentrations predicted by simulation without fire emissions) for CO and ozone in layer 1 (ground level to 33.9 m above ground level). Fig. 6 shows, for each grid cell, the maximum difference in CO and ozone concentrations throughout the period from 22–31 August. Fig. 7 shows the same data for 6 September. CO is shown because, as a slow-reacting species, it characterizes the dispersion of the fire emissions. A
4-km2 (1000-acre) fire in Liberty Co., 80 km (50 miles) northeast of Houston, caused a peak CO concentration of 856 ppb on 30 August. In Allen Co., Louisiana, a 16-km2 (4000-acre) fire burned on 31 August causing a peak CO concentration of 2.266 ppm. On 23 August, a 3-km2 (750-acre) fire in Brazoria Co., 97 km (60 miles) south of Houston, caused a 937-ppb peak CO concentration and a 60-km fire plume with northwest direction. The same fires caused localized increases in O3 concentrations of 8, 42, and 60 ppb, on 23, 30, and 31 August, respectively. The 23 August wildfire was located in a rural region where limited NOx sources were available resulting in lower ozone concentrations. The wildfire plumes on 30 and 31 August encountered more
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4992
Fig. 8. Location of process analysis control volumes on (a) 30 August 2000 and (b) 6 September 2000.
anthropogenic sources of NOx emissions leading to increased ozone productivity. The results of the simulations reported in Fig. 6 indicate that for wildfires less than 10,000 acres, the greatest enhancements of CO and ozone concentrations due to the fire emissions are confined to regions within 10–100 km of the fire. 3.5. Process analysis The CAMx model has the capability of performing a more detailed analysis of the physical and chemical processes influencing the formation and accumulation of ozone and other photochemical pollutants, using techniques referred to as process analysis. Process analysis tools have been described by Jeffries and co-workers (Jang et al., 1995a, b; Jeffries and Tonnesen, 1994; Jeffries, 1995; Tonnesen and Jeffries, 1994; Wang, 1997), and the set of tools used in this work have been described by Vizuete (2005). Process analysis methods were applied on two simulation days, 30 August 2000 and 6 September 2000. On 30 August an isolated rural wildfire northeast of Houston, shown in Fig. 6, was
advected over relatively limited number of NOx sources. The 6 September fires plumes, shown in Fig. 7, were advected toward the Houston urban core. These two days provide distinctly environmental conditions for ozone generation. The processes influencing ozone production in both scenarios were examined with the process analysis tool. The process analysis tool allows quantitative tracking of individual physical and chemical process that contribute to changing pollutant concentrations. The processes tracked include horizontal and vertical pollutant fluxes crossing cell boundaries, chemical production and consumption rates, emission rates, and deposition rates. These rates are aggregated over a collection of vertical and horizontal grid cells defined as a control volume. Fig. 8a and b shows the horizontal dimensions of the control volume used for this analysis. The location of each control volume was chosen to capture the largest increases in ground level ozone concentrations due to wild fire sources. The height of the control volume tracked the mixing height, as described in Vizuete (2005).
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4993
70
60
VOC / NOx
ppbC / ppbN
50
40
30
20
10
0 0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23
(a)
LST 70
60
VOC / NOx
ppbC / ppbN
50
40
30
20
10
0 0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 LST
(b)
Anthropogenic VOC / NOx
Total VOC / NOx
Fig. 9. Time series of the VOC to NOx ratios for the (a) 30 August 2000 and (b) 6 September 2000 control volumes.
VOC/NOx ratios for each control volume, for each hour, are shown in Fig. 9a and b. Ratios during daylight hours peaked at 64 at hour 17 on 30 August and 30 at hour 18 on 6 September. These values are typical of a photo-chemical environment where ozone production is limited by NOx availability. The increased number of non-wildfire (urban) NOx sources located in the 6 September control volume provided additional NOx and reduced VOC/NOx ratios. Fig. 10a and b identifies the specific VOCs that reacted to produce ozone within the control volumes, during hours 13–16 on 30 August
and hours 10–13 on 6 September. The time period selected for analysis for each simulation day was chosen to coincide with the hours of largest ozone productivity within the process analysis control volumes. The VOC classifications shown in Fig. 10 are based on the Carbon Bond mechanism version 4 (CBIV) with revised radical termination mechanism, used in the simulations (ENVIRON, 2000). The 30 August control volume generated 12 ppb of ozone compared to 27 ppb on 6 September. The 30 August control volume is located in a rural region where the majority of ozone formation
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4994
Ozone Produced (11.9 ppb) by each VOC
MEOH 1% CH4 4%
CO 3%
Ozone Produced (27.3 ppb) by each VOC ETOH 1% MEHO 2% CH4 7%
FORM 7%
CO 9% FORM 11%
ALD2 7% MGLY 1% PAR 6% ETH 1%
AL2D 12%
ISPO 33%
OLE 6% XYL 1%
MGLY 1%
ISOP 63%
(a)
(b)
XYL 2% OLE ETH 6% 2%
PAR 14%
Fig. 10. Ozone produced chemically by each volatile organic carbon (VOC) class used in the CBIV chemical mechanism for (a) 30 August 2000 hours 13–16, and (b) 6 September 2000 hours 10–13.
chemistry, 63%, is due to reactions of isoprene. In the emission inventory developed for this work, the fires had very low rates of isoprene emissions, compared to other reactive hydrocarbons, so the bulk of the ozone formation reactions are due to isoprene emissions from the forests downwind of the fires. On 6 September, the fire, and the control volume, is located closer to Houston and is impacted by anthropogenic sources. Isoprene reactions account for only 33% of total ozone production. A more detailed analysis of the ozone chemistry is provided by reviewing several additional parameters, shown in Fig. 11. Each parameter has been summed over the same hours used in the analyses presented in Fig. 10. The increased anthropogenic sources on 6 September increased the extent of VOC reactions by 12 ppb. The 6 September control volume also had 21 ppb more NO to NO2 conversions resulting in the production of 15 ppb more of ozone. Two additional indicators of the conditions for ozone chemical production are OH and NO cycle values. These values represent the average number of times each new OH radical or NO molecule reacts before being lost in termination reactions or before being advected out of the control volume. NO cycle values were 13.1 and 5.2 for the 30 August and 6 September simulation days. 30 August had a OH cycle value of 1.8 and the 6 September simulation day had a value of 2.4. The relatively larger NO cycle and smaller OH cycle values for 30 August suggest a NOx limited atmosphere. In this control volume only a limited amount of NOx is available for ozone production and NO reacts multiple times before
being lost in termination reactions. In contrast, there are sufficient amounts of NOx and VOC present in the 6 September control volume to triple the ozone production relative to the values observed on 30 August. Under the chemical conditions presented here, sources of NOx play a significant role in determining ozone productivity.
4. Conclusions Emissions from wildfires can have a significant impact on regional air quality. This work assessed the accuracy of wildfire emission estimation tools and used air quality models to assess the spatial distribution of the air quality impacts of fires in southeast Texas. Comparison of aircraft measurements and emission estimates demonstrated that, within the uncertainty limits of the tools, emission estimates are accurate. The modeling demonstrated that, for wildfires less than 10,000 acres, the greatest enhancements of CO and ozone concentrations due to the fire emissions were generally confined to regions within 10–100 km of the fire. Within 10 km of these fires CO concentrations can exceed 2 ppm and ozone concentrations can be enhanced by 60 ppb. The photochemistry occurring in the wildfire plumes is limited by NOx availability and, therefore, wildfire plumes that are advected over urban areas and anthropogenic NOx sources can have significantly different photochemical impacts than wildfire plumes that do not encounter additional NOx sources.
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4995
40 36.4 35
30
28.5
27.3
25
20 16.6 15.1 15
10
13.1
12.6
11.9
8.8 5.2
5 1.8 0
New OH
VOC Reacted
2.4
OH Cycle
NO to NO2 Conversion
NO Cycle
Ozone Chemical Generation
ppb August 30, 2000
September 6, 2000
Fig. 11. Key parameters for the chemistry occurring in the control volumes for 30 August 2000 (hours 13–16) and 6 September 2000 (hours 10–13).
References Associated Press Newswires, 2000. Fires continue to burn throughout Texas, September 7, 2000. Baton Rouge Advocate, The, 2000. Texas’ largest wildfire nearly under control, September 7, 2000, p. 4-A. Brown, D., Dunn, W., Lazaro, M., Policastro, A., 1999. The FIREPLUME model: tool for eventual application to prescribed burns and wildland fires, Proceedings from The Joint Fire Science Conference and Workshop, June 15–17, 1999. http://www.nifc.gov/joint_fire_sci/conferenceproc/4 Burgan, R., 1988. 1988 Revision to the 1978 national fire danger rating system. Research Paper SE-273, USDA Forest Service, Southeastern Region, Asheville, NC. Dennis, A., Fraser, M., Anderson, S., Allen, D.T., 2002. Air pollutant emissions associated with forest, grassland and agricultural burning in Texas. Atmospheric Environment 36, 3779–3792. ENVIRON, 2000. User’s Guide Comprehensive Air Quality Model with Extensions (CAMx) version 3.10. ENVIRON International Corporation, Novato, California. Freds, D., 2002. Kerr Wildland Management Area, Hunt, Texas, personal communication. Houston Chronicle, 2000a. In: Rice, H., Horsewell, C., Gonzalez, J.W., Stewart, R. (Eds.), Parched Conditions Keep East Texas Wildfires Alive, September 6, 2000, p. 1. Houston Chronicle, 2000b. In: Rice, H., Horsewell, C., Gonzalez, J.W., Stewart, R. (Eds.),Smoke from Area Wildfires may Revisit the City Today, September 7, 2000, p. 27.
Jang, J.C., Jeffries, H.E., Byun, D., Pleim, J.E., 1995a. Sensitivity of ozone to model grid resolution: part I. Application of high resolution regional acid deposition model. Atmospheric Environment 29 (21), 3085–3100. Jang, J.C., Jeffries, H.E., Tonnesen, S., 1995b. Sensitivity of ozone to model grid resolution: part II. Detailed process analysis for ozone chemistry. Atmospheric Environment 29 (21), 3101–3114. Jeffries, H.E., 1995. Photochemical Air Pollution. Chapter 9 in Composition, Chemistry, and Climate of the Atmosphere. Van Nostand, Reinhold, New York. Jeffries, H.E., Tonnesen, S., 1994. A comparison of two photochemical reaction mechanisms using a mass balance process analysis. Atmospheric Environment 28 (18), 2991–3003. Junquera, V., 2004. Inventory of emissions of gaseous compounds and particulate matter from wildfires in east Texas in August and September 2000 and comparison with aircraft measurements. M.S. Thesis, University of Texas at Austin. Liu, Y., 2004. Variability of wildland fire emissions across the contiguous United States. Atmospheric Environment 38, 3489–3499. McCown, M., 2002. Chambers County Texas Agricultural Extension Agent, Personal communication. NCAR, 2002. Project ]2000-828 AESOP/TEXAQS-2000, Texas Air Quality Study, NSF/NCAR L-188 Electra (N308D). National Center for Atmospheric Research, Atmospheric Technology Division, Research Aviation
ARTICLE IN PRESS 4996
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
Facility. Web site last updated August 2002. http:// raf.atd.ucar.edu/Projects/AESOP4 NOAA, 2003. Texas Air Quality Study 2000 Airborne Lidar Aerosol Data, National Oceanic and Atmospheric Administration, Environmental Technology Laboratory, Optical Remote Sensing Division. Web page last modified August 2003. http://www.etl.noaa.gov/et2/data/data_pages/texaqs/ air_aerosol.html4 Reinhardt, D.E., Keane, R.E., Brown, J.K., 1997. First Order Fire Effects Model: FOFEM 4.0, User’s Guide. US Department of Agriculture, Forest Service. Intermountain Research Station, Ogden, Utah. Ryerson, T.B., Trainer, M., Angevine, W.M., Brock, C.A., Dissly, R.W., Fehsenfeld, F.C., Frost, G.J., Goldan, P.D., Holloway, J.S., Hubler, G., Jakoubek, R.O., Kuster, W.C., Neumann, J.A., Nicks Jr., D.K., Parrish, D.D., Roberts, J.M., Sueper, D.T., Atlas, E.L., Donnelly, S.G., Flocke, F., 2003. Effect of petrochemical industrial emissions of reactive Alkenes and NOx on tropospheric ozone formation in Houston, Texas. Journal of Geophysical ResearchAtmospheres 108 (D8), 4249. Sandberg, D.V., Hardy, C.C., Ottmar, R.D., Snell, J.A.K., Acheson, A., Peterson, J.L., Seamon, P., Lahm, P., Wade, D., 1999. National Strategy Plan: Modeling and Data Systems for Wildland Fire and Air Quality. US Department of Agriculture, Forest Service, Pacific Northwest Research Station 60pp. Senff, C., Banta, R., Darby, L., Angevine, W., White, A., Berkowitz, C., Doran, C., 2002. Spatial and temporal variations in mixing height in Houston, Final Report, TNRCC Project F-20. http://www.etl.noaa.gov/et2/data/ data_pages/texaqs/Task_F20_final_report.html4 Taylor, B., 2002. Texas A&M University Agricultural Research Station, Sonora, Texas, Personal communication. TCEQ, 2003. Houston/Galveston air quality science evaluation. Air quality modeling files. CAMx modeling files. Files last updated December 2003.
Texas Commission on Environmental Quality (TCEQ), 2004. The state implementation plan for the control of Ozone pollution: attainment demonstration for the Houston/ Galveston Ozone non-attainment area. P.O. Box 13087, Austin, Texas 78711–3087, www.tceq.state.tx.us. Tonnesen, S., Jeffries, H.E., 1994. Inhibition of odd oxygen production in carbon bond IV and generic reaction set mechanisms. Atmospheric Environment 28 (7), 1339–1349. Vizuete, W., 2005. Ph.D. Thesis, University of Texas at Austin. Vizuete, W., Junquera, V., McDonald-Buller, E., McGaughey, G., Yarwood, G., Allen, D.T., 2002. Effects of temperature and land use on predictions of biogenic emissions in Eastern Texas. Atmospheric Environment 36, 3321–3337. Wang, Z., 1997. Comparison of three vertical diffusion schemes in the SARMAP air quality model with integrated process rate analysis method and continuous process composition and source receptor methodology, Ph.D. Dissertation. Chapel Hill, University of North Carolina at Chapel Hill. ftp://airsite.unc.edu/PDFS/ESE_UNC/Students/ PhDThesis/Wang/4 Wiedinmyer, C., Strange, I., Estes, M., Yarwood, G., Allen, D.T., 2000. Biogenic hydrocarbon emission estimates for North Central Texas. Atmospheric Environment 34, 3419–3435. Wiedinmyer, C., Guenther, A., Estes, M., Strange, I., Yarwood, G., Allen, D.T., 2001. A land use database and examples of biogenic isoprene emission estimates for the state of Texas, USA. Atmospheric Environment 35, 6465–6477. Yarwood, G., Wilson, G., Shepard, S., Guenther, A., 1999a. User’s Guide to the Global Biosphere Emissions and Interactions System—Version 3.1. 101 Rowland Way, Suite 220, Novato, California, available at: http://www.globeis.com. Yarwood, G., Wilson, G., Emery, C., Guenther, A., 1999b. Development of the GloBEIS—a state of the science biogenics emissions modeling system. Final Report to the Texas Natural Resource Conservation Commission, 12100 Park 35 Circle, Austin, Texas 78753.
Atmospheric Environment 39 (2005) 4983–4996 www.elsevier.com/locate/atmosenv
Wildfires in eastern Texas in August and September 2000: Emissions, aircraft measurements, and impact on photochemistry Victoria Junquera, Matthew M. Russell, William Vizuete, Yosuke Kimura, David Allen Center for Energy and Environmental Resources (R7100), The University of Texas at Austin, 10100 Burnet Road, Austin, TX 78758, USA Received 11 February 2005; received in revised form 26 April 2005; accepted 6 May 2005
Abstract The accuracy of wildfire air pollutant emission estimates was assessed by comparing observations of carbon monoxide (CO) and particulate matter (PM) concentrations in wildfire plumes to predictions of CO and PM concentrations, based on emission estimates and air quality models. The comparisons were done for observations made in southeast Texas in August and September of 2000. The fire emissions were estimated from acreage burned, fuel loading information, and fuel emission factor models. A total of 389 km2 (96,100 acres) burned in wildfires in the domain encompassing the Houston/Galveston-Beaumont/Port Arthur (HGBPA) area during August and September 2000. On the days of highest wildfire activity, the fires resulted in an estimated 3700 tons of CO emissions, 250 tons of volatile organic carbon (VOC) emissions, 340 tons of PM2.5, and 50 tons of NOx emissions; estimated CO and VOC emissions from the fires exceeded light duty gasoline vehicle emissions in the Houston area on those days. When the appropriate aircraft data were available, aloft measurements of CO in the fire plumes were compared to concentrations of CO predicted using the emission estimates. Concentrations estimated based on emission predictions and air quality models were within a factor of 2 of the observed values. The estimated emissions from fires were used, together with a gridded photochemical model, to characterize the extent of dispersion of the fire emissions and the photochemistry associated with the fire emissions. Although the dispersion and photochemical impacts varied from fire to fire, for wildfires less than 10,000 acres, the greatest enhancements of CO and ozone concentrations due to the fire emissions were generally confined to regions within 10–100 km of the fire. Within 10 km of these fires, CO concentrations can exceed 2 ppm and ozone concentrations can be enhanced by 60 ppb. The extent of photo-oxidant formation in the plumes was limited by NOx availability and isoprene emissions from forested areas downwind of the fires provided most of the hydrocarbon reactivity in the plumes. r 2005 Elsevier Ltd. All rights reserved. Keywords: Wildfires; Wildfire inventory; Wildfire emissions; Emissions inventory; Emissions modeling; TexAQS; Photochemical modeling; CAMx; Plume rise
Corresponding author. Tel.: +1 512 475 7842; fax: +1 512 471 7060.
E-mail address: [email protected] (D. Allen). 1352-2310/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2005.05.004
ARTICLE IN PRESS 4984
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
1. Introduction Outdoor fires can emit substantial amounts of particulate matter (PM), carbon monoxide (CO), nonmethane hydrocarbons (NMHCs), nitrogen oxides (NOx), and ammonia (NH3) into the atmosphere (Sandberg, 1999). In Texas, emissions of CO and fine PM (PM with diameters less than 2.5 mm, PM2.5) from fires account for 10% and 1–2% of total annual statewide emissions, respectively (Dennis et al., 2002). On days and seasons when large fire events occur, these percentages can be much higher, and fires can dominate emissions and impact air quality over substantial areas (Liu, 2004). The air quality impacts of fires are often estimated based on predicted emissions. Fire emission predictions are based on estimates of area burned, fuel mass burned per area, and emissions per mass of fuel combusted. Significant uncertainties can arise in estimating each of these parameters; thus the emission estimates can be uncertain. Dennis et al. (2002) estimated uncertainties of approximately a factor of 2 in area burned and fuel loadings. Additional uncertainties in assessing the air quality impacts of fires are due to uncertainties in emission factors, which depend on the nature of the combustion (smoldering versus flaming) and the plume rise of the fire. The primary goal of this work is to evaluate the uncertainties associated with estimating fire emissions by comparing predicted emissions to observations made during a large air quality field program, conducted in southeast Texas during the summer of 2000. A secondary goal is to characterize the photochemistry occurring in the wildfire plumes during the study period.
2. Methodology 2.1. Emissions inventory In this study, emissions from wildfires were estimated for the regional domain shown in Fig. 1. Wildfire emissions were estimated for the months of August and September 2000; the measurements to which the emission estimates will be compared were collected during the Texas Air Quality Study (TexAQS), which was conducted from 15 August–15 September 2000. August and September 2000 was a period of drought and intense wildfire activity in southeast Texas. As documented in the results section of this paper, during August and September of 2000 approximately 384 km2 (95,000 acres) burned in wildfires in the HGBPA domain (the red area shown in Fig. 1), 66% of which burned in Texas. In contrast, in August and September of 1996 wildfires consumed 16.0 km2 (3960 acres), and in August and September of 1997, 59 km2 (14,620 acres) burned in
Texas (Dennis et al., 2002). Because of the intense drought during the summer of 2000, most prescribed burns that would normally occur during August and September were delayed or not conducted (Freds, 2002; McCown, 2002; Taylor, 2002), and therefore wildfires dominated the fire emissions during the study period. The information used to estimate emissions from wildfires includes the location and date of the fire, the fire area, the type of vegetation burned, the density of the vegetation at the location of the fire, also termed the fuel loading factor (mass of fuel available per acre), the fraction of fuel consumed in the fire, the emission factor of the fuel (pounds of pollutant per ton of fuel), and the emission efficiency. The fuel loading factor and the fraction of fuel burned during the fire can be combined into one parameter, the fuel consumption factor (mass of fuel consumed per acre); the fuel consumption factors for specific locations are then combined with fuel-based emission factors, into a location specific composite emission factor, in units of emissions of pollutant per acre burned. Wildfire activity data, including acreage burned and fire location, were collected from incident reports available through the National Interagency Fire Management Integrated Database (NIFMID), the Texas Interagency Coordination Center, the US Department of Agriculture Forest Service (USFS), the Louisiana Interagency Coordination Center, and the Oklahoma Department of Agriculture, Food & Forestry. The information provided by each of these data sources is described in detail elsewhere (Junquera, 2004). All the data concerning individual fire incidents were integrated into a single database and duplicate reports of fires were identified through common parameters, such as the fire incident name and location. Only fires greater than 0.04 km2 (10 acres) were analyzed for duplications in this study, since most of the smaller fires lacked complete information in the databases. It was assumed that duplicated fires under 0.04 km2 would not add significantly to uncertainties in the calculation of emissions. Once fire locations were determined, fuel consumption was estimated for each fire incident. Vegetation type and density vary significantly over the domain of the study, so the fuel consumption can vary significantly from fire to fire. Relatively little information on vegetation type was available in the incident reports. The type of burned vegetation was reported directly in the fire incident reports for only 9% of the fires in Texas and 27% of the fires in the regional domain, or 31% of the burned area in Texas and 29% of the burned area in the regional domain. Therefore, in this work, fuel models, rather than reported types of burned vegetation, were used to estimate fuel loading for fires occurring on wildland and rangeland. Specifically, the First Order Fuel Effects Model (FOFEM), a publicly available computer model developed by the Intermountain Fire
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4985
Fig. 1. Horizontal (a) and vertical (b) structure of the modeling domain for the study. Fire emissions during the study period were estimated for the regional domain, with special attention paid to the Houston-Galveston, Beaumont Port Arthur (HGBPA) subdomain. Emissions were assumed to enter a variety of vertical layers in an air quality model (Comprehensive Air Quality Model with extensions, CAMx).
Sciences Laboratory of the USFS, was used to predict fuel consumption (Reinhardt et al., 1997). The model incorporates vegetation types for the contiguous US, and default fuel loading values and consumption equations for each vegetation type (Dennis et al., 2002). The program, however, does not include information on the spatial distribution of these vegetation types. Dennis et al. (2002) selected 15 vegetation cover codes in FOFEM version 4.0, which corresponded to vegetation types in the study domain, and crossreferenced them with vegetation types allocated onto a land cover database of the state of Texas developed by Wiedinmyer et al. (2000, 2001). As a result, a land use–land cover (LULC) database with spatially resolved FOFEM 4.0 vegetation cover types and fuel consumption factors was created for Texas. The rest of the study domain was assigned fuel loading factors based on National Fire Danger Rating System (NFDRS) fuel
models, which were developed as a method to rate wildfire danger by the USDA Forest Service (Burgan, 1988). Emissions from wildfires were calculated by projecting the daily fire information (location and acreage) onto a map with spatially resolved composite emission factors. Details of the fuel consumption assignments have been reported by Junquera (2004) and the methods have been described by Dennis et al. (2002). 2.2. Ambient measurements A variety of ambient measurements were available to evaluate the performance of the emission estimates. These included total aerosol concentrations (as characterized by aerosol backscatter) as a function of elevation measured by a NOAA aircraft with downward-looking aerosol and ozone light detection and
ARTICLE IN PRESS 4986
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
ranging (lidar), and gas-phase air pollutant concentrations measured by a second NOAA operated aircraft. Details of the aircraft measurements and operation are described elsewhere (NOAA, 2003; NCAR, 2002) and are only summarized here. In the lidar measurements, the aerosol channel was set at a wavelength of 359 nm. The aircraft typically flew at an altitude of 3500 m above middle sea level (MSL), and the aerosol backscatter profiles extend approximately from 2500 m MSL to the surface with a vertical resolution of 15 m and a time resolution of 10 s (NOAA, 2003). The aerosol lidar backscatter data are not calibrated and therefore provide information about the aerosol distribution in a qualitative rather than quantitative manner (NOAA, 2003). The results of the measurements performed can be obtained from the NOAA web site (http://www.etl.noaa.gov/et2/data/data_ pages/texaqs/air_aerosol.html). The National Center for Atmospheric Research (NCAR) Electra aircraft, operated by NOAA, took 1-s interval measurements of various gaseous compounds and PM on several days of the modeling period. The full suite of measurements is described by Ryerson et al. (2003); data sets for all flights are available via ‘ftp’ or on magnetic tape upon request (NCAR, 2002). In this work, the primary focus will be on measurements of CO. Measurements from all of these platforms were available for 15 August to approximately 15 September 2000. Particular focus in this work will be on measurements made on 30–31 August and 3–6 September. A large fire that occurred on 30 August in Liberty Co., 80 km (50 miles) northeast of Houston, caused locally elevated PM and ozone concentrations. A very high regional wildfire activity characterized 3–6 September. 2.3. Photochemical modeling The wildfire emissions inventory was merged with an existing inventory of emissions from point, area, and mobile sources obtained from the Texas Commission on Environmental Quality (TCEQ) as part of the Houston/ Galveston Air Quality Science Evaluation (TCEQ, 2003). As described in the results section, biogenic emissions played a significant role in the photochemistry of the fire plumes; therefore, the details of the biogenic emission estimation methods are of interest here. The biogenic emission inventory was calculated with the Global Biosphere Emissions and Interactions System (GloBEIS) model. The source code and documentation for GloBEIS is publicly available and is described in detail by Yarwood et al. (1999a, b). Biogenic emissions are influenced primarily by vegetation type and density, solar radiation, cloud cover, and ambient temperature. GloBEIS therefore requires data characterizing LULC, ambient temperatures, and solar radiation or cloud cover across a region and time period of interest.
Vegetation characteristics were based on a landuse/ land-cover database developed by Wiedinmyer et al. (2001) for the state of Texas. Meteorological data for GloBEIS were extracted from several sources. Surface temperatures were developed by spatially interpolating temperatures measured by National Weather Service (NWS) and other weather stations throughout southeast Texas (Vizuete et al., 2002). The University of Maryland and the National Oceanic and Atmospheric Administration (NOAA) for the Global Energy and Water Cycle Experiment (GEWEX) Continent Scale International Project (GCIP) provided the estimates of PAR fluxes (TCEQ, 2004). All wind speed and humidity data were derived from the NCAR/Penn State Mesoscale Model version 5, MM5. The emissions and meteorological modeling data were input into the Comprehensive Air Quality Model with extensions (CAMx) version 3.11, a three-dimensional (3D) eulerian photochemical grid model (ENVIRON, 2000). The modeling domains have a horizontal resolution of 16 km 16 km (East Texas subdomain), 4 km 4 km (HGBPA subdomain), or 1 km 1 km (HG subdomain) (Fig. 1). The creation of the modelready file and the results of the modeling runs are briefly described in this section. CAMx includes two-way grid nesting, and a subgridscale Plume-in-Grid (PiG) module to treat the slow dispersion and the chemistry of large point source plumes. Unless the PiG module is turned on, emissions from each point source are instantly dispersed into an entire grid cell volume, given by the grid cell area and the height of the vertical layer. Wildfires were treated as point sources or ‘‘stacks’’. Each fire was modeled as a series of stacks with identical location and different heights, so that the emissions calculated for the wildfire would be uniformly distributed throughout the vertical rise of the plume. The PiG module was turned off for all fires, causing emissions to be instantly dispersed into each grid volume. Thus, the emissions from each fire were allocated homogeneously into a volume whose area was given by the resolution of the grid and whose height was given by the plume rise of the fire. The area of a cell in the 4 km 4 km and 16 km 16 km grids is equivalent to roughly 4000 and 63,000 acres, respectively. Thus, emissions from fires that are smaller than the grid cell area were over-diluted in the model. The fire emissions inventory was preprocessed using the Urban Airshed Model (UAM) Emissions Preprocessor System (EPS) version 2.0. Input into EPS2.0 includes an emissions data file in Aerometric Information Retrieval System (AIRS) Facility Subsystem (AFS) format (AFS file), a chemical split factor file (CHEMSPLIT), and a temporal split file (TMPRL). The CHEMSPLIT file includes information on the chemical speciation of NMHC emissions. For the fires, ethane,
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
ethylene, and propylene were the dominant organic emissions, comprising approximately half of the total NMHC emissions. The CHEMSPLIT file converted these speciated emissions into Carbon Bond-IV (CB-IV) species. In the CB-IV chemical mechanism, species with similar chemical properties are lumped into CB-IV groups that undergo the same chemical reactions. In the TMPRL file, the emissions were assumed to be uniform and constant throughout a day for each fire incident. For fire incidents lasting more than one day, emissions were allocated uniformly throughout the burning period. Details of the chemical speciation used in the model are provided by Junquera (2004). The plume rise for the fires, which is necessary for 3D photochemical modeling, was estimated based on calculations performed with FIREPLUME. The FIREPLUME model uses Briggs’ two-thirds law (Brown et al., 1999), and employs different calculation strategies for stable, neutral, and unstable atmospheric conditions. For fires smaller than 0.4 km2 (100 acres), a plume rise value corresponding to the top of CAMx layer 3
4987
(170.5 m) was assigned. For fires between 0.4 and 3.2 km2 (800 acres), and greater than 3.2 km2, plume rise values corresponding to the top of CAMx layers 4 (256.9 m) and 6 (431.7 m), respectively, were assigned. The rationale for these assignments is described by Junquera (2004).
3. Results and discussion 3.1. Emissions inventory In August and September 2000, 518 km2 (128,000 acres) were burned in wildfires in Texas and 389 km2 (95,000 acres) in the HGBPA domain (Table 1). In the HGBPA domain, 2% of the fires were larger than 3.24 km2 (800 acres) and accounted for 56% of the total area burned, and 74% of the fires were smaller than 0.405 km2 (100 acres) and burned only 5% of the total area. Fig. 2 shows wildfire locations and acreage burned from 22 August to 6 September. Estimated emissions of
Table 1 Burned area in km2 (acres) during the study period; 2–8 September was characterized by the highest wildfire intensity Period
Regional domain
Texas
HGBPA domain
August and September 2–8 September
971 (240,000) 386 (95,300)
518 (128,000) 251 (62,000)
389 (96,100) 262 (64,700)
Fig. 2. (a) Wildfires during the period from 22 August to 29 August 2000. (b) Wildfires during the period from 29 August to 6 September 2000.
ARTICLE IN PRESS 4988
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
Fig. 3. Emissions of CO, NMHCs, NOx, and PM2.5 from wildfires in the HGBPA domain during August and September 2000. Emissions of CO and NMHC from wildfires exceeded emissions from Light Duty Gasoline Vehicles, LDGV (indicated by horizontal line) on some days.
CO, NMHCs, PM2.5, and NOx in the HGBPA domain are shown in Fig. 3. Fig. 3 also shows the daily average emissions of CO, NMHCs, and NOx from light duty gasoline vehicles (LDGV) in the Houston-Galveston area. Fig. 3 shows that emissions of CO and NMHCs from fires exceeded emissions from LDGV on some days. The highest emissions during this period were approximately 3700, 250, 340, and 50 short tons day 1 for CO, NMHC, PM2.5, and NOx, respectively. 3.2. Comparison with aircraft data Aircraft measurements were used to assess the overall accuracy of the emission estimates. The data best suited for this comparison were collected on 6 September , in a plume from a multi-day fire in the McFaddin National
Wildlife Refuge (NWR) that consumed 13,000 acres. On 6 September , the McFaddin NWF fire plume was brought into the Houston area by a northeast wind. Fig. 4 shows the location of the fire, the flight track of the NOAA/NCAR Electra aircraft on 6 September, and the CO measurements made by the aircraft. Within the fire plume, CO concentrations of up to 490 ppbv were measured at 500 m above sea level (ASL). Background CO concentrations, or CO concentration that were not caused by the McFaddin NWR fire, were determined from aircraft measurements before and after the sharp CO peak associated with the fire, and ranged from 80 to 160 ppbv. The background concentration values were subtracted from the average CO concentration in the peak caused by the fire, 350 ppbv, yielding the concentration caused solely by the fire. Using an average value of the wind speed measured by the aircraft, 5.9 m s 1, the CO flux (mass/area time) was calculated. It was then assumed that the CO concentration was uniform throughout the mixing height. The mixing height was estimated with HYSPLIT (http://www.arl.noaa. gov/ready/hysplit4.html) for several locations along the flight section that intersected the fire plume, and an average value of approximately 500 m was obtained. This value is consistent with the coastal location of the fire and was in agreement with mixing height data estimated by Senff et al. (2002) for Galveston Bay area during the early afternoon (400–700 m). Thus, 500–700 m was used as the mixing height value. From the mixing height, the length of the CO plume due to the fire (14 km), the average CO concentration over background in the fire plume, and the wind speed, the CO flow (mass/time) was calculated. The McFaddin NWR was characterized as a marshland in newspaper articles (Houston Chronicle, 2000a, b; Associated Press Newswires, 2000; Baton Rouge Advocate, 2000) and NIFMID reports. The composite emission factor for Wet Grasslands in Texas (FOFEM 6) was therefore chosen to compute the burning rate (acres/time). The burned acreage was then calculated for a 24-h period. The fire started on 2 September (NIFMID), burned out between two levees on 4 and 5 September (Houston Chronicle, 2000a), was 95% contained by 7 September (Associated Press Newswires, 2000; Baton Rouge Advocate, 2000), and was completely mopped up by 9 September (NIFMID). From this information, the area that burned on 6 September was expected to fall somewhere between 4 and 20 km2 day 1 (1000 and 5000 acres day 1). This estimate of acreage burned per day can be compared to the acreage that would be required to generate the CO emission rate estimated from the aircraft data (mass/time). The burn rate required to generate the emission rate estimated from the aircraft data was 11–20 km2day 1 (2800–5000 acres day 1), depending on the CO background concentration and
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4989
Fig. 4. September, CO concentrations measured by NOAA’s Electra aircraft. (a) CO concentrations are represented in different colors, and wildfires sizes are represented by different dot sizes. The CO plume, from the McFaddin National Wildlife Refuge fire, was detected by the aircraft at approximately 11 AM. (b) Detail of the CO aircraft measurements with the fire plume boxed.
mixing height chosen. Thus, for the McFaddin fire, estimates of emissions and burn rates based on aircraft measurements (11–20 km2 day 1 burn rates) were consistent with the emission inventory data (4–20 km2 day 1), but both the inventory and the estimates of emissions based on aircraft measurements have uncertainties of at least a factor of 2. 3.3. Downward-looking lidar Lidar measurements provide a qualitative measure of the atmospheric concentrations and vertical distribution of PM. Lidar measurements made during the most
intense fire period, 2–8 September, indicated much higher PM backscatter than during other days of the study. As shown in Fig. 5, aircraft measurements on 3 September (not the same day as the day measurements were made by the Electra aircraft) immediately downwind of the McFaddin NWR fire (red rectangle in Fig. 5) indicate that elevated PM concentrations were present throughout the mixed layer. The fire plume was carried into the Gulf of Mexico and was not detected in subsequent measurements on 3 September. On 6 September (the day of the Electra measurements), the lidar data indicate that elevated PM concentrations were observed at all elevations and throughout the domain.
ARTICLE IN PRESS 4990
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
Fig. 5. (a) PM backscatter registered by the NOAA aircraft along its flight path with downward-looking lidar instrumentation on 3 September (top) and 6 September. (b) Corresponding aircraft trajectories and backscatter measured at 2000 m AGL. 3 September shows a sharp spike in PM backscatter (red rectangle), which was caused by the McFaddin NWR fire. The lidar detected the highest PM backscatter on 6 September, the day with the highest fire activity.
3.4. Photochemical modeling More detailed assessments of the air quality impacts of fires can be performed using 3D gridded photo-
chemical models. In this work, CAMx was used as the modeling tool and the modeling analyses examined the spatial dispersion of fire emissions, and the impact of fire emissions on ozone formation and other
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4991
Fig. 6. Difference plots (concentrations predicted by the simulation including fire emissions—concentrations predicted by simulation without fire emissions) for CO (a) and ozone (b) in layer 1 (ground level to 33.9 m above ground level). The figure shows, for each grid cell, the maximum difference in CO and ozone concentrations throughout the period from 22 to 31 August.
Fig. 7. Difference plots (concentrations predicted by the simulation without fire emissions—concentrations predicted by simulation including fire emissions) for CO (left) and ozone (right) in layer 1 (ground level to 33.9 m above ground level). The figures are for 6 September 2000 hour 12; the time of the largest increase in ground level ozone concentrations due to fires.
photochemical processes. In the analyses, CAMx simulations were performed both with and without the emissions from fires included. The difference between these simulations characterizes the impact of the fires. Figs. 6 and 7 show two difference plots (concentrations predicted by the simulation including fire emissions minus concentrations predicted by simulation without fire emissions) for CO and ozone in layer 1 (ground level to 33.9 m above ground level). Fig. 6 shows, for each grid cell, the maximum difference in CO and ozone concentrations throughout the period from 22–31 August. Fig. 7 shows the same data for 6 September. CO is shown because, as a slow-reacting species, it characterizes the dispersion of the fire emissions. A
4-km2 (1000-acre) fire in Liberty Co., 80 km (50 miles) northeast of Houston, caused a peak CO concentration of 856 ppb on 30 August. In Allen Co., Louisiana, a 16-km2 (4000-acre) fire burned on 31 August causing a peak CO concentration of 2.266 ppm. On 23 August, a 3-km2 (750-acre) fire in Brazoria Co., 97 km (60 miles) south of Houston, caused a 937-ppb peak CO concentration and a 60-km fire plume with northwest direction. The same fires caused localized increases in O3 concentrations of 8, 42, and 60 ppb, on 23, 30, and 31 August, respectively. The 23 August wildfire was located in a rural region where limited NOx sources were available resulting in lower ozone concentrations. The wildfire plumes on 30 and 31 August encountered more
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4992
Fig. 8. Location of process analysis control volumes on (a) 30 August 2000 and (b) 6 September 2000.
anthropogenic sources of NOx emissions leading to increased ozone productivity. The results of the simulations reported in Fig. 6 indicate that for wildfires less than 10,000 acres, the greatest enhancements of CO and ozone concentrations due to the fire emissions are confined to regions within 10–100 km of the fire. 3.5. Process analysis The CAMx model has the capability of performing a more detailed analysis of the physical and chemical processes influencing the formation and accumulation of ozone and other photochemical pollutants, using techniques referred to as process analysis. Process analysis tools have been described by Jeffries and co-workers (Jang et al., 1995a, b; Jeffries and Tonnesen, 1994; Jeffries, 1995; Tonnesen and Jeffries, 1994; Wang, 1997), and the set of tools used in this work have been described by Vizuete (2005). Process analysis methods were applied on two simulation days, 30 August 2000 and 6 September 2000. On 30 August an isolated rural wildfire northeast of Houston, shown in Fig. 6, was
advected over relatively limited number of NOx sources. The 6 September fires plumes, shown in Fig. 7, were advected toward the Houston urban core. These two days provide distinctly environmental conditions for ozone generation. The processes influencing ozone production in both scenarios were examined with the process analysis tool. The process analysis tool allows quantitative tracking of individual physical and chemical process that contribute to changing pollutant concentrations. The processes tracked include horizontal and vertical pollutant fluxes crossing cell boundaries, chemical production and consumption rates, emission rates, and deposition rates. These rates are aggregated over a collection of vertical and horizontal grid cells defined as a control volume. Fig. 8a and b shows the horizontal dimensions of the control volume used for this analysis. The location of each control volume was chosen to capture the largest increases in ground level ozone concentrations due to wild fire sources. The height of the control volume tracked the mixing height, as described in Vizuete (2005).
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4993
70
60
VOC / NOx
ppbC / ppbN
50
40
30
20
10
0 0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23
(a)
LST 70
60
VOC / NOx
ppbC / ppbN
50
40
30
20
10
0 0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 LST
(b)
Anthropogenic VOC / NOx
Total VOC / NOx
Fig. 9. Time series of the VOC to NOx ratios for the (a) 30 August 2000 and (b) 6 September 2000 control volumes.
VOC/NOx ratios for each control volume, for each hour, are shown in Fig. 9a and b. Ratios during daylight hours peaked at 64 at hour 17 on 30 August and 30 at hour 18 on 6 September. These values are typical of a photo-chemical environment where ozone production is limited by NOx availability. The increased number of non-wildfire (urban) NOx sources located in the 6 September control volume provided additional NOx and reduced VOC/NOx ratios. Fig. 10a and b identifies the specific VOCs that reacted to produce ozone within the control volumes, during hours 13–16 on 30 August
and hours 10–13 on 6 September. The time period selected for analysis for each simulation day was chosen to coincide with the hours of largest ozone productivity within the process analysis control volumes. The VOC classifications shown in Fig. 10 are based on the Carbon Bond mechanism version 4 (CBIV) with revised radical termination mechanism, used in the simulations (ENVIRON, 2000). The 30 August control volume generated 12 ppb of ozone compared to 27 ppb on 6 September. The 30 August control volume is located in a rural region where the majority of ozone formation
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4994
Ozone Produced (11.9 ppb) by each VOC
MEOH 1% CH4 4%
CO 3%
Ozone Produced (27.3 ppb) by each VOC ETOH 1% MEHO 2% CH4 7%
FORM 7%
CO 9% FORM 11%
ALD2 7% MGLY 1% PAR 6% ETH 1%
AL2D 12%
ISPO 33%
OLE 6% XYL 1%
MGLY 1%
ISOP 63%
(a)
(b)
XYL 2% OLE ETH 6% 2%
PAR 14%
Fig. 10. Ozone produced chemically by each volatile organic carbon (VOC) class used in the CBIV chemical mechanism for (a) 30 August 2000 hours 13–16, and (b) 6 September 2000 hours 10–13.
chemistry, 63%, is due to reactions of isoprene. In the emission inventory developed for this work, the fires had very low rates of isoprene emissions, compared to other reactive hydrocarbons, so the bulk of the ozone formation reactions are due to isoprene emissions from the forests downwind of the fires. On 6 September, the fire, and the control volume, is located closer to Houston and is impacted by anthropogenic sources. Isoprene reactions account for only 33% of total ozone production. A more detailed analysis of the ozone chemistry is provided by reviewing several additional parameters, shown in Fig. 11. Each parameter has been summed over the same hours used in the analyses presented in Fig. 10. The increased anthropogenic sources on 6 September increased the extent of VOC reactions by 12 ppb. The 6 September control volume also had 21 ppb more NO to NO2 conversions resulting in the production of 15 ppb more of ozone. Two additional indicators of the conditions for ozone chemical production are OH and NO cycle values. These values represent the average number of times each new OH radical or NO molecule reacts before being lost in termination reactions or before being advected out of the control volume. NO cycle values were 13.1 and 5.2 for the 30 August and 6 September simulation days. 30 August had a OH cycle value of 1.8 and the 6 September simulation day had a value of 2.4. The relatively larger NO cycle and smaller OH cycle values for 30 August suggest a NOx limited atmosphere. In this control volume only a limited amount of NOx is available for ozone production and NO reacts multiple times before
being lost in termination reactions. In contrast, there are sufficient amounts of NOx and VOC present in the 6 September control volume to triple the ozone production relative to the values observed on 30 August. Under the chemical conditions presented here, sources of NOx play a significant role in determining ozone productivity.
4. Conclusions Emissions from wildfires can have a significant impact on regional air quality. This work assessed the accuracy of wildfire emission estimation tools and used air quality models to assess the spatial distribution of the air quality impacts of fires in southeast Texas. Comparison of aircraft measurements and emission estimates demonstrated that, within the uncertainty limits of the tools, emission estimates are accurate. The modeling demonstrated that, for wildfires less than 10,000 acres, the greatest enhancements of CO and ozone concentrations due to the fire emissions were generally confined to regions within 10–100 km of the fire. Within 10 km of these fires CO concentrations can exceed 2 ppm and ozone concentrations can be enhanced by 60 ppb. The photochemistry occurring in the wildfire plumes is limited by NOx availability and, therefore, wildfire plumes that are advected over urban areas and anthropogenic NOx sources can have significantly different photochemical impacts than wildfire plumes that do not encounter additional NOx sources.
ARTICLE IN PRESS V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
4995
40 36.4 35
30
28.5
27.3
25
20 16.6 15.1 15
10
13.1
12.6
11.9
8.8 5.2
5 1.8 0
New OH
VOC Reacted
2.4
OH Cycle
NO to NO2 Conversion
NO Cycle
Ozone Chemical Generation
ppb August 30, 2000
September 6, 2000
Fig. 11. Key parameters for the chemistry occurring in the control volumes for 30 August 2000 (hours 13–16) and 6 September 2000 (hours 10–13).
References Associated Press Newswires, 2000. Fires continue to burn throughout Texas, September 7, 2000. Baton Rouge Advocate, The, 2000. Texas’ largest wildfire nearly under control, September 7, 2000, p. 4-A. Brown, D., Dunn, W., Lazaro, M., Policastro, A., 1999. The FIREPLUME model: tool for eventual application to prescribed burns and wildland fires, Proceedings from The Joint Fire Science Conference and Workshop, June 15–17, 1999. http://www.nifc.gov/joint_fire_sci/conferenceproc/4 Burgan, R., 1988. 1988 Revision to the 1978 national fire danger rating system. Research Paper SE-273, USDA Forest Service, Southeastern Region, Asheville, NC. Dennis, A., Fraser, M., Anderson, S., Allen, D.T., 2002. Air pollutant emissions associated with forest, grassland and agricultural burning in Texas. Atmospheric Environment 36, 3779–3792. ENVIRON, 2000. User’s Guide Comprehensive Air Quality Model with Extensions (CAMx) version 3.10. ENVIRON International Corporation, Novato, California. Freds, D., 2002. Kerr Wildland Management Area, Hunt, Texas, personal communication. Houston Chronicle, 2000a. In: Rice, H., Horsewell, C., Gonzalez, J.W., Stewart, R. (Eds.), Parched Conditions Keep East Texas Wildfires Alive, September 6, 2000, p. 1. Houston Chronicle, 2000b. In: Rice, H., Horsewell, C., Gonzalez, J.W., Stewart, R. (Eds.),Smoke from Area Wildfires may Revisit the City Today, September 7, 2000, p. 27.
Jang, J.C., Jeffries, H.E., Byun, D., Pleim, J.E., 1995a. Sensitivity of ozone to model grid resolution: part I. Application of high resolution regional acid deposition model. Atmospheric Environment 29 (21), 3085–3100. Jang, J.C., Jeffries, H.E., Tonnesen, S., 1995b. Sensitivity of ozone to model grid resolution: part II. Detailed process analysis for ozone chemistry. Atmospheric Environment 29 (21), 3101–3114. Jeffries, H.E., 1995. Photochemical Air Pollution. Chapter 9 in Composition, Chemistry, and Climate of the Atmosphere. Van Nostand, Reinhold, New York. Jeffries, H.E., Tonnesen, S., 1994. A comparison of two photochemical reaction mechanisms using a mass balance process analysis. Atmospheric Environment 28 (18), 2991–3003. Junquera, V., 2004. Inventory of emissions of gaseous compounds and particulate matter from wildfires in east Texas in August and September 2000 and comparison with aircraft measurements. M.S. Thesis, University of Texas at Austin. Liu, Y., 2004. Variability of wildland fire emissions across the contiguous United States. Atmospheric Environment 38, 3489–3499. McCown, M., 2002. Chambers County Texas Agricultural Extension Agent, Personal communication. NCAR, 2002. Project ]2000-828 AESOP/TEXAQS-2000, Texas Air Quality Study, NSF/NCAR L-188 Electra (N308D). National Center for Atmospheric Research, Atmospheric Technology Division, Research Aviation
ARTICLE IN PRESS 4996
V. Junquera et al. / Atmospheric Environment 39 (2005) 4983–4996
Facility. Web site last updated August 2002. http:// raf.atd.ucar.edu/Projects/AESOP4 NOAA, 2003. Texas Air Quality Study 2000 Airborne Lidar Aerosol Data, National Oceanic and Atmospheric Administration, Environmental Technology Laboratory, Optical Remote Sensing Division. Web page last modified August 2003. http://www.etl.noaa.gov/et2/data/data_pages/texaqs/ air_aerosol.html4 Reinhardt, D.E., Keane, R.E., Brown, J.K., 1997. First Order Fire Effects Model: FOFEM 4.0, User’s Guide. US Department of Agriculture, Forest Service. Intermountain Research Station, Ogden, Utah. Ryerson, T.B., Trainer, M., Angevine, W.M., Brock, C.A., Dissly, R.W., Fehsenfeld, F.C., Frost, G.J., Goldan, P.D., Holloway, J.S., Hubler, G., Jakoubek, R.O., Kuster, W.C., Neumann, J.A., Nicks Jr., D.K., Parrish, D.D., Roberts, J.M., Sueper, D.T., Atlas, E.L., Donnelly, S.G., Flocke, F., 2003. Effect of petrochemical industrial emissions of reactive Alkenes and NOx on tropospheric ozone formation in Houston, Texas. Journal of Geophysical ResearchAtmospheres 108 (D8), 4249. Sandberg, D.V., Hardy, C.C., Ottmar, R.D., Snell, J.A.K., Acheson, A., Peterson, J.L., Seamon, P., Lahm, P., Wade, D., 1999. National Strategy Plan: Modeling and Data Systems for Wildland Fire and Air Quality. US Department of Agriculture, Forest Service, Pacific Northwest Research Station 60pp. Senff, C., Banta, R., Darby, L., Angevine, W., White, A., Berkowitz, C., Doran, C., 2002. Spatial and temporal variations in mixing height in Houston, Final Report, TNRCC Project F-20. http://www.etl.noaa.gov/et2/data/ data_pages/texaqs/Task_F20_final_report.html4 Taylor, B., 2002. Texas A&M University Agricultural Research Station, Sonora, Texas, Personal communication. TCEQ, 2003. Houston/Galveston air quality science evaluation. Air quality modeling files. CAMx modeling files. Files last updated December 2003.
Texas Commission on Environmental Quality (TCEQ), 2004. The state implementation plan for the control of Ozone pollution: attainment demonstration for the Houston/ Galveston Ozone non-attainment area. P.O. Box 13087, Austin, Texas 78711–3087, www.tceq.state.tx.us. Tonnesen, S., Jeffries, H.E., 1994. Inhibition of odd oxygen production in carbon bond IV and generic reaction set mechanisms. Atmospheric Environment 28 (7), 1339–1349. Vizuete, W., 2005. Ph.D. Thesis, University of Texas at Austin. Vizuete, W., Junquera, V., McDonald-Buller, E., McGaughey, G., Yarwood, G., Allen, D.T., 2002. Effects of temperature and land use on predictions of biogenic emissions in Eastern Texas. Atmospheric Environment 36, 3321–3337. Wang, Z., 1997. Comparison of three vertical diffusion schemes in the SARMAP air quality model with integrated process rate analysis method and continuous process composition and source receptor methodology, Ph.D. Dissertation. Chapel Hill, University of North Carolina at Chapel Hill. ftp://airsite.unc.edu/PDFS/ESE_UNC/Students/ PhDThesis/Wang/4 Wiedinmyer, C., Strange, I., Estes, M., Yarwood, G., Allen, D.T., 2000. Biogenic hydrocarbon emission estimates for North Central Texas. Atmospheric Environment 34, 3419–3435. Wiedinmyer, C., Guenther, A., Estes, M., Strange, I., Yarwood, G., Allen, D.T., 2001. A land use database and examples of biogenic isoprene emission estimates for the state of Texas, USA. Atmospheric Environment 35, 6465–6477. Yarwood, G., Wilson, G., Shepard, S., Guenther, A., 1999a. User’s Guide to the Global Biosphere Emissions and Interactions System—Version 3.1. 101 Rowland Way, Suite 220, Novato, California, available at: http://www.globeis.com. Yarwood, G., Wilson, G., Emery, C., Guenther, A., 1999b. Development of the GloBEIS—a state of the science biogenics emissions modeling system. Final Report to the Texas Natural Resource Conservation Commission, 12100 Park 35 Circle, Austin, Texas 78753.