Las Vegas Carbon Monoxide Study AWMA Presentation T and B Systems Environmental Research and Meteorological Consulting

ABSTRACT
This paper presents an overview of a recent (2001) "saturation" study to measure carbon monoxide (CO) levels in the Las Vegas Valley of Nevada. The study was performed during a seasonal period that historically experiences CO levels that exceed the 8-hour National Ambient Air Quality Standard (NAAQS). The purpose of the Las Vegas CO Saturation Study was 1) to identify geographic patterns of CO concentrations, including potential areas with higher concentrations not previously known, and 2) to determine if the routine CO monitoring station locations are representative of areas that experience the higher CO levels.
The study began with a network of 32 continuous CO samplers operating throughout the Las Vegas Valley. Stations were added during the program to refine the spatial distribution of higher concentrations, ending with 63 stations. The CO monitors utilized had never been operated in this type of study but proved to be very accurate, reliable, and easily operated as long as specific procedures were followed. A mobile van equipped with an EPA equivalent CO analyzer and position recording operated as well during periods when higher CO concentrations were likely to occur. An extensive quality assurance program added credibility to the results.
The primary result of the study was to confirm that the routine monitoring station locations were representative of the higher CO concentrations in the Valley. Maps with isopleths of the highest 8?hour average CO concentrations were developed based exclusively on the saturation network sites. It was found that the routine network sites were indeed located where the peaks levels occurred. Moreover, the gradients were for the most part well defined by the existing network
.
INTRODUCTION
The Clark County, Nevada, Department of Air Quality Management (DAQM) contracted with Technical & Business Systems (T&B Systems) for a "saturation" study of carbon monoxide (CO) concentrations during the seasonal period from before Thanksgiving to after New Year's Day, specifically, November 20, 2001 through January 5, 2002. Historically, this is the most likely period to have the highest CO concentrations measured by the County's air monitoring network. The study objectives were to:

· Obtain additional CO concentration information in areas where the exceedances of the NAAQS for CO have been experienced,

· Improve the conceptual understanding of the relationship between land use, meteorology, and ambient CO concentrations,

· Identify geographic areas with high CO concentrations to support risk assessment and potential population exposure to CO exceedances, and

· Evaluate the current Clark County DAQM monitoring network CO sampling locations in accordance with U.S. Environmental Protection Agency (EPA) siting criteria.

In short, the study was planned to answer two primary questions: are the DAQM monitoring sites adequately identifying the highest 8?hour average concentrations occurring in the metropolitan Las Vegas Valley area, and what geographic areas are affected by high CO concentrations? Prior to 1999, at least one 8?hour average per year exceeded the primary NAAQS for CO, 9 parts-per-million (ppm). Aggressive efforts to reduce CO emissions have contributed to the significant decline in exceedance occurrences; none have been recorded since 1998. The saturation study was planned to provide the information to assess the adequacy of the spatial distribution of County DAQM monitoring sites.
The technical approach to meeting the study objectives was to operate a network of CO monitoring sites throughout the Valley, with emphasis on the locales historically most likely to experience the highest CO concentrations and where rapid recent development has occurred. Increasing the number of sampling locations well above the number of sites in the DAQM monitoring network effectively "saturated" the community with sampling locations and provided the data needed to achieve the study objectives. The Saturation Network was achieved by a combination of fixed (called "Static") sites and moveable (called "Dynamic") sites operated during the anticipated worst-case meteorological conditions. The samplers ran nearly continuously at the monitoring sites, producing a complete record of CO concentrations at the selected locations. Another element in saturating the Valley with measurements was conducting mobile sampling using an instrumented van during select CO episodes. The van was driven around the Valley to observe CO loading in a real-time mode during nighttime and early-morning hours when meteorological conditions conducive to peak CO concentrations occur.
The network monitoring sites were located throughout the Las Vegas Valley, with a higher density east of the downtown area. Previous monitoring and modeling results indicated this to be the most likely area to encounter higher CO concentrations during periods of typical worst-case meteorological conditions and higher CO emission rates. The dynamic sites were located both to increase the density of sampling locations in the static network and to test other areas that might produce higher CO concentrations due to expected airflow pathways. As dynamic sites were installed and the preliminary data appeared useful in delineating spatial CO patterns, many were left operating continuously to maximize the amount of information collected.

INSTRUMENTATION
Two types of CO sampling instrumentation were used to collect the CO data. The CO samplers used in the network study sites were manufactured by Onset Computer Corporation (Onset). The samplers are small, self-contained units that include the CO sampler and a data logger. The primary CO sampling in the mobile van used during the Intensive Operating Periods (IOP) was a Dasibi 3003 CO analyzer.
The Onset CO analyzer, shown in Figure 1, consists of an electrochemical sensor and on-board data logger housed in a rectangular plastic container about 4-inches long by 3-inches wide and 1.5 inches high. The sampler was housed in a polyvinyl chloride (PVC) circular pipe section closed on the top with a grate on the bottom and holes in the side for air circulation. EPA recognizes the virtual impossibility of operating a network of reference analyzers in the density needed for a saturation study. Hence, various saturation studies have utilized non-reference sampling methods with appropriate comparisons to ensure adequacy of the information for the intended purposes. Preliminary tests in sampling chambers and the ambient environment coupled with comparative results from samplers collocated with DAQM designated equivalent analyzers at monitoring sites and in the mobile sampling van all showed excellent justification for recognizing results from the samplers as adequate for the purposes of this study.

Figure 1. Onset CO Analyzer and Enclosure

Extensive testing of the samplers was performed prior to proposing their use in this study because the CO samplers had not yet been used for an ambient measurement application. During testing, individual sampler calibration responses were measured using a Dasibi 3003 CO analyzer. These responses were used in data processing. During operations, aggressive quality control checks and operating samplers at two sites collocated with DAQM sites typically receiving higher CO concentrations helped to provide good comparability of results from the Study samplers and the DAQM CO monitors.
Additional tests were performed to ensure that natural ventilation through the polyvinyl chloride cylindrical tube housings would provide an adequate time response to changing CO concentrations. The initial response to increasing CO concentrations was at least as quick as that of the Dasibi analyzer, but a small lag time in the Onset sampler response was noted during decreasing concentrations. This lag was shown to be inherent in the detector technology and not in the natural ventilation of the sampler housing.
Other tests also indicated a small temperature dependence in the Onset sampler response. This temperature relationship was further defined by both laboratory-controlled conditions and by comparing the collocated measurements made in the field (between the DAQM Dasibi analyzer and the Onset sampler) during field conditions occurring in the Study. The relationship was applied during data processing.
To aid in the spatial mapping of CO concentrations and assess the adequacy of the placement of the fixed site sampling network, a van was outfitted for mobile CO monitoring. The mobile van was a late model sport-utility vehicle known to have low CO emissions, minimizing the potential impact of the van on the measurements. The real-time observations mapped the horizontal extent of the CO plume and aided in understanding the areas of highest concentration for placement of fixed site samplers during episodes. The sampling van traversed the Valley in two general modes. One pattern was cross-section passes to identify possible new areas of higher concentrations not covered by the Saturation Network. The other pattern was a close inspection of the immediate area in the vicinity of DAQM and Saturation Network sites to identify possible higher CO concentrations.
Figure 2 shows several views of the vehicle with the sampling probe mounted on the roof. This inlet height allowed sampling above the level of the tailpipe emissions and in the region of generally well-mixed air. The sample was drawn through the inlet by a pump system supplying air both to a Dasibi 3003 CO analyzer and to two Onset CO analyzers configured in a series flow arrangement. The first Onset CO sampler was programmed to collect data at 5-second intervals to help identify possible "spikes" (rapid rise and fall in concentration) possibly due to nearby vehicles. A laptop computer polled the second Onset CO sampler at one-second intervals to compute one-minute averages. CO readings from the Dasibi 3003 were also collected at one-second intervals and stored as 1-minute averages on a Campbell CR10 data logger and subsequently by the laptop computer. The CR10 data logger also recorded the outside air temperature from a platinum RTD probe mounted on the support rod for the sample inlet. The temperature data were recorded at the end of each minute.
The mobile van position was monitored by a Garmin Etrex Global Positioning System (GPS) and recorded by a laptop computer. One-second position readings were calculated to 1-minute averages of latitude, longitude and altitude. Once each minute the Dasibi, Onset analyzer and GPS data files were merged and a graphical "strip chart" type display of the data was updated on two laptop computers. This real-time display allowed "scrolling" capabilities to look back in time at past data. A second GPS receiver provided the real-time position to a street map display on the same two laptop-computers as the strip chart display. One of the computers was in the front seat where a scientist evaluated the data in real time and kept a log of events. The three computers were networked through a LAN hub and the computer times synchronized prior to the start to assure proper updating of the real-time files. Finally, on selected routes, pictures from a digital camera mounted with a forward view were recorded on the laptop computer located in the front seat.

Figure 2. Views of Sampling Van and Instrumentation

QUALITY ASSURANCE
Quality assurance was a fundamental portion of all phases of the study. The activities included equipment tests and data validation following U.S. EPA monitoring guidance, and network siting following a standard operating procedure written for the study. In addition:

· The equipment and gas standards used in the study underwent acceptance tests prior to use in data collection.

· Independent audits were conducted of all major components of the study.

· Sites were visited and reviewed against EPA criteria for siting micro-, middle-, and neighborhood-scale CO monitoring scales.

· Performance audits of the Onset monitors and the sampling van monitors were conducted. The audits consisted of exposing the monitors to three CO concentrations (0, 10.35 and 25.2 ppm). The audits were conducted using the same standards used during the audit the County's CO monitoring network.

· A systems audit was conducted of the study's data processing effort at the T&B Systems facility in Santa Rosa, CA. The audit consisted of interviews with key data processing personnel, and included tracking raw data points through the data processing effort to verify that procedures were being followed.

· During the course of the field program, there were routine zero and span quality control checks performed on the CO measurement equipment.

· Comparative testing.
Comparative testing consisted of both multiple Onset CO analyzers collocated, and Onset CO analyzers collocated with standard (Dasibi) CO monitors at DAQM sites that used FRM equipment. The latter comparisons were conducted at two sites selected because they typically experienced high CO levels and were easily accessible. Measurements were made continuously for the extent of the field study.
Both 1-hour and 8-hour averaged Onset data were compared with the DAQM monitor data set. The initial comparison was conducted using Onset measurements corrected only for calibration span (slope). It was noted that differences between the collocated instruments showed a distinct diurnal pattern, i.e., the Onset analyzers read lower than the Dasibi units during the nighttime and greater during the daytime. This was consistent with results of our cold-chamber tests of the Onset CO analyzer in which the calibrations appeared to be sensitive to ambient temperature. The relationship between monitor differences at the two sites and ambient temperature can be observed in Figure 3, which shows the differences observed at one of the sites, Freedom Park, as a function of ambient temperature.
The results of applying an adjustment to the Onset measurements for this bias improved the correlation between the two methods and reduced the standard deviation of the differences significantly. After adjustment for ambient temperature, 92 percent of the differences were less than 1 ppm and all were less than 2 ppm at the second site (Crestwood).
Peak daily 8-hour averages using the adjusted and validated data sets were next compared with the County's collocated measurements. The average difference between the instruments was 0.26 and 0.29 ppm for Freedom Park and Crestwood, respectively, well within the accuracy of the equipment. The standard deviation of the differences was 0.55 and 0.53 ppm, respectively which was well within the requirements for the study.
All the individual Onset samplers were calibrated in a chamber using "zero" (pure) and NIST traceable concentrations of CO gas. The sampler responses were checked throughout the program to ensure continued performance within the quality control guidelines. Sampler responses to ambient conditions were recorded on a 1-minute interval as the raw data. The raw data were transferred to a standard computer database at frequent intervals and immediately reviewed by the field personnel to ensure proper operation. In this manner, problems were identified and corrected in a timely manner.

Figure 3. Differences Between Onset and Dasibi CO Analyzers vs. Ambient Temperature - Freedom Park

RESULTS
Comparisons between the Onset samples and collocated CO measurements made by DAQM reference analyzers at two select sites indicated a small diurnal cycle bias of the Onset readings, generally less than 0.5 ppm. This bias apparently depends on ambient air temperature. Thus, in addition to factoring the individual sampler multipoint calibration responses, a correction was also made for the apparent temperature bias. The corrected data files were then used in the statistical analysis to determine the comparability of the Saturation Network results and the DAQM routine network results.
The analysis task consisted of two major elements: the Saturation Network of fixed sites and the mobile sampling with the monitoring van. The geographic distribution of CO concentrations throughout the Valley as defined by the Saturation Network of fixed sites was determined for those periods of observed high levels and for the overall maximum levels experienced during the field study. The maximum impacted areas were then compared with the locations of the DAQM sites. The mobile sampling measurements mapped the CO distribution to verify that the Saturation network was correctly configured to identify peak CO concentrations, and provided justification for Dynamic sites that were installed during the course of the field study.
Saturation Network Sampling
The major objective of this study was to determine if the DAQM CO Network captures peak CO ambient concentrations, and is representative of the CO exposure levels in the Las Vegas Valley. As can be seen from Figure 4, this appears to be the case. In the figure, isopleths of the peak running 8-hour averaged CO levels generated from the Saturation Network data for the period of the field study (November 20, 2001 to January 5, 2002) are shown. The locations of the existing DAQM sites that measure CO are shown as well. Of particular note is that there is a concentration of DAQM sites in the most severely impacted area. The Sunrise Acres (SA) site was ideally located to measure maximum levels as defined by the Saturation Network. Generally low CO concentrations (peaks <2.5 ppm) were experienced in the fringe areas outside the inner core. A sense of the CO gradient between the core and the outer areas in the Las Vegas Valley is provided by the sample of peak concentrations at select locations in the Valley shown in Table 1. The latter four sites in the table represent suburban areas and were exposed to significantly lower CO levels than in the urban core.

Figure 4. Peak 8-hour Average CO (ppm) for Period November 20 to January 5 Showing DAQM CO Sites

Table 1. Peak 8-hour CO Concentrations (ppm) from Select Sites in the Saturation Network

Mobile Sampling
The sampling van operated during two episodic periods-- from December 16 to 20 and December 26 to 29. Seven sampling events were conducted during the first episode, and four during the second one. Space does not allow a discussion of even the highlights of the monitoring. However, sampling on the evening of December 27-28 provides an interesting example of our findings.
The overall CO observations during the sampling period are depicted in Figure 5. In the figure, concentrations are shown color-coded relative to the key on the left side. As can be seen concentrations over essentially the entire Las Vegas Valley were mapped during each run. Key findings from the run include:
· Mapping performed on the west side around the new developments and casinos showed no significant concentrations of CO. Although not initially expected, this observation is consistent with the fleet of low emissions new cars generally operating in the so-called Las Vegas Strip.
· As generally supposed by the Clark County DAQM, the most significant concentrations of CO were not on either the extreme east or west side of the Valley, they were found back in the inner core region around the intersection of Charleston Blvd., Eastern Avenue, and Fremont Street. Figure 6 shows the numeric concentration values in this area with some of the highest values seen off of the major streets. Given the lack of automobile sources, the CO may be more related to home or water heating related sources. Mobile sampling repeatedly observed high CO levels in neighborhoods away from the major streets and was not related to adjacent vehicle sources.
· Traverses made down Boulder Hwy showed the plume extending almost to Henderson. Tracing the plume out along East Lake Mead Drive and down into the low areas where drainage would occur showed slightly elevated concentrations in the low areas that would be draining toward Lake Mead. Figure 7 shows the observed CO concentrations along the southeast portion of the sampling van track.

Figure 5. Observed CO Concentrations (ppm) During December 27 to 28 Evening/Night Period

Figure 6. Observed Numeric CO Concentrations (ppm) During December 27 to 28 Evening/Night Period in the Region of the Sunrise Acres Site. The DAQM sites are shown in blue triangles.

Figure 7. Observed Numeric CO Concentrations (ppm) During December 27 to 28 Evening/Night Period. The concentrations illustrate the drainage flow. The DAQM sites are shown in blue triangles.

KEY FINDINGS
Six of the Clark County DAQM CO monitoring sites are located in the parts of Las Vegas Valley that experience the higher CO levels, and the Sunrise Acres site is centered in the area of maximum impact. This observation leads to the conclusion that the DAQM CO monitoring sites are suitably located to identify the peak CO concentrations and the corresponding general area in which the higher levels occur.
The highest CO concentrations measured during the Saturation Study occurred in the vicinity of the Sunrise Acres DAQM site. This site experienced the highest concentrations in the DAQM network during recent years. The Saturation Network revealed that relatively high CO levels extend beyond the major streets in the area and into residential neighborhoods. The peak 8?hour average CO concentrations indicate that the higher concentrations (>5 ppm) extend about 1-mile to the north, south, and east of the peak area. This area includes six of the DAQM CO monitoring sites: CC, CW, JD, FP, SA and MC (Figure 4). Most of the area covered by the higher CO concentrations is east of Interstate 15.
The peak 1-hour average CO concentration measured at the Saturation Study sites was 18.3 ppm; this occurred at a micro-scale exposure site near Casino Center and Fremont Street. The peak 1-hour averages at the remaining sites were all less than 10 ppm. These levels are all well below the 1-hour NAAQS of 35 ppm.
The peak 8-hour average CO concentration measured during the program was 7.2 ppm. This occurred at the U16 site, which is two blocks south of the DAQM monitoring site at the Sunrise Acres School near the intersection of Eastern Avenue, Charleston Blvd. and Fremont Street. This CO level is below the 8?hour NAAQS of 9 ppm. The peak 8?hour average occurring at the Sunrise Acres DAQM site (from November 20 through December 2001) was 6.0 ppm, which was during the same 8-hour time period as the maximum at U16.
At least two secondary CO peak areas were observed in the Saturation Network that are not as well covered by the existing DAQM. One area is just northwest of the I15/95 interchange. The other is along the Boulder Highway where elevated concentrations of CO entrained in the nocturnal drainage were observed.
In the outlying areas of the Las Vegas Valley where DAQM monitoring is sparse, measured levels within the Saturation Network were confirmed to be very low.
Higher than expected CO levels were observed in residential neighborhoods that are not immediately adjacent to major thoroughfares. These elevated concentrations may not be related to traffic patterns but instead, may be the result of other sources of CO, such as space or water heating. The areas where this was observed are high population density residential neighborhoods where CO exhaust from older heaters and water-heating appliances may be significantly higher than in newer neighborhoods.

ACKNOWLEDGEMENTS
This program was funded and administered by the Clark County Department of Air Quality Management