Effects of Residential Wood-Burning Emissions during Winter Months in a Northern San Diego County Location

Residential wood burning is common throughout San Diego County in the wintertime. Some areas get signi!cantly colder than others, and a higher quantity of wood is burned during these colder times. Residential wood burning results in higher ambient concentrations of emissions including particulate matter (PM10 and PM2.5), carbon dioxide (CO2), carbon monoxide (CO), nitrogen oxides (NOx), Polycyclic Aromatic Hydrocarbons (PAH’s), and aldehydes [1,2]. Maenhaut also saw seasonal variations in PM10 levels in Belgium, and separated elemental and organic carbon to speci!cally contribute these varying emission levels to wood burning in the region [3].


Introduction
Residential wood burning is common throughout San Diego County in the wintertime. Some areas get signi cantly colder than others, and a higher quantity of wood is burned during these colder times. Residential wood burning results in higher ambient concentrations of emissions including particulate matter (PM 10 and PM 2.5 ), carbon dioxide (CO 2 ), carbon monoxide (CO), nitrogen oxides (NO x ), Polycyclic Aromatic Hydrocarbons (PAH's), and aldehydes [1,2]. Maenhaut also saw seasonal variations in PM 10 levels in Belgium, and separated elemental and organic carbon to speci cally contribute these varying emission levels to wood burning in the region [3].
Some domestic wood burning emissions are of public concern because they are detrimental to the environment and human health. It was found that in San Jose, California, 42% of the PM 10 generated during winter months originated from wood smoke [4]. In Portugal, 18% of PM 10 emissions were due to residential wood burning, and eliminating wood burning in a particular area reduced emissions by 46% [5]. is suggests that the overall content of PM 10 in the air could have a signi cant e ect on the population's health during winter months. In addition, avocado wood, one of the dirtiest-burning woods available in the San Diego area produces more particulate matter than other wood types [6]. e physical conditions at the time the wood is burned can also impact the chemical composition of the released emissions, and can a ect human health [7]. Also, the levels of PM 10 and PM 2.5 , as well as CO present in wood burning emissions will theoretically be di erent from each other depending on whether they are a product of actively burning res or smoldering res. Di erences in temperature of the re positively correlate with the e ciency of the combustion process. Fires begin burning with a high temperature and are barely e cient in terms of turning fuel into products. In other words, high temperature res release more gases, such as CO and hydrocarbons, than particulates, which could be of concern if they are released in a poorly ventilated area. en, as the re continues with less fuel and becomes more of a smoldering re, the temperature lowers, and the re releases more particulates and water than gases [2].

e area of concern
Escondido, an incorporated area in the northern part of San Diego County, is an ideal location to study air quality because of its geographic and demographic characteristics. Tian et al. [8] suggests that demographics such as age and number of people in each household, income, and urban or rural infrastructure can have an e ect on the measurements of particulate matter in a particular area. Escondido has a population of 143,389 (2008 census), and the largest percentage of the population is between the ages of 15 and 64. e average number of people in each household was 3.08 in 2008. In 2010, 36% of households earned less than $30,000 per year, but 10% of households earned more than $100,000 per year [9].
is suggests that people in the same geographic area do have access to di erent sources of heat in the home.
times of the year. A nocturnal inversion causes the wood burning emissions to be trapped within the shallow surface layer above the valley during the night and early morning hours. Because the region is mountainous, the possibility for the polluted air to disperse becomes di cult. As a result, the pollutants linger over the valley in colder, wintertime temperatures, thus increasing the city's exposure to their e ects [6].

Measuring San Diego's ambient air pollutant concentration
Ambient air quality is measured at nine di erent locations throughout the county by the San Diego Air Pollution Control District (SDAPCD). In 2009, SDAPCD reported that Escondido ranked among the cities with the highest annual levels of PM 2.5 from 1999 to 2009 [11]. e objective of this project is to develop a methodology to identify residential wood burning as a major contributor to the higher ambient concentrations of PM 10 , PM 2.5 , and CO in Escondido during winter months (November, December, January) and on holidays. In addition, this study's focus is on the types of re burned in domestic replaces, and the weather patterns in the area of study. e types of wood burned, the physical condition of people in the a ected areas, and any weather phenomena can ultimately change the overall air quality in the wintertime, and can amplify the e ects of poor air quality on the nearby population.

Literature Review
Wood burning used for heating and cooking is the most common type of biomass burning around the world. When wood burns, it generates gases and particles including PM 10 and PM 2.5 . ese emissions also contain high arsenic concentrations, CO, polycyclic aromatic hydrocarbons, benzopyrene, as well as a mixture of condensable organic products that conglomerate and solidify [12][13][14]. During the coldest times of the year, wood consumption increases worldwide, including most parts of the U.S. [15]. e e ect of wood burning emissions on human health has become a key concern in recent years. A long-time exposure to wood smoke has been associated with the dwar ng of children, wheezing in children, and cardiovascular and respiratory illnesses [14,16]. e emissions from wood burning are in uenced by the size and the moisture level inside the wood.
icker and moister logs burn less e ciently. As a result, burning such logs can create higher concentrations of particulate matter in the air than other gases [17]. A replace furnace is not a very e cient heat source because most of the heat exits up the chimney stack with the smoke [14]. When wood is burned, the gaseous products are released when pyrolysis temperatures are reached [15]. Organic materials are transformed into gaseous components and solid residue which are composed of xed carbon and ash [18]. e denser, cold air near the surface from the evening temperature inversion, combined with the average residential chimney stack's closeness to the ground is cause for the wood burning emissions to remain in the lower atmosphere. Inversions trap cold air under a layer of warmer air. e cold and dense air then limits the vertical dispersion of the hot chimney fumes that contain the emissions, keeping them at lower altitudes. ese particles linger close to the ground and increase the ambient concentration of emissions in the a ected area. e most common types of wood burned around San Diego County are oak, eucalyptus, avocado, and pine. Kleeman, et al. [6] studied the chemical composition of the particles emitted during oak, pine, and eucalyptus wood burning. e samples measured were 17.2 kg of pine (burned for 189 min), 15.4 kg of oak (burned for 165 min), and 18.9 kg of eucalyptus (burned for 218 min). e particles emitted by the wood samples burning were primarily composed of organic compounds with a mass distribution that peaked between 0.1-0.2 µm in diameter, which meant they were all categorized as PM 2.5 [6]. e weather also a ects the air quality in southern California. California weather patterns are in uenced by the Paci c High (PH) and the Aleutian Low (AL) pressure systems. e interaction of these two circulation systems determines the air exchange along the west coast of the U.S. e air exchange is also in uenced by the peninsular mountain chains that run along the California coast, impeding the coastal air's movement. ese pressure systems and the inland mountains allow for a marine layer over the coast. e local marine layer is restricted to a depth of 3,000 feet or less. A subsidence inversion occurs, and traps the marine air which also captures pollutants and smog.
Additionally, wind from a speci c weather event could either disperse the marine layer or move air that is higher in emissions to other locations in the county, or move air inland faster than would have happened without the wind event. Southern California is not prone to severe wind weather events such as tornadoes. San Diego county, and Escondido in particular, saw zero wind events of 74 miles per hour or higher (the de nition of a Category 1 weather event) during the study period [19].

Methodology
To estimate the impact of the pollutants released by wood burning in the area of Escondido, a baseline ambient concentration level was established [20]. e data recorded at the SDAPCD's Escondido collection station were sorted and analyzed in terms of the monthly averages for the 11 years studied (2000 to 2010). Daily values were available for PM 10 and PM 2.5 , and hourly data were available for CO from 2000 to 2010, and for PM 2.5 from 2008 to 2010 [20]. Before each test was completed, the entire data set was sorted by time and parameter, and the non-applicable data was eliminated. Statistical tests compared winter months and summer months, holiday times to the rest of the winter, weekends and weekdays, evenings and daytime, early evening hours and night, and burning to smoldering res for each of the parameters, namely PM 10 , PM 2.5 , and CO.
Excel was a useful tool in sorting the large data sets and comparing data e ciently. Both Student's t-tests and Mann-Whitney U-Tests (MWU) were used in Excel to compare the data sets and determine if they were statistically di erent from each other. e t-test is used for smaller sample sizes (ranging from n=2 to n=30) and computes, then compares the means using a p-value (value that indicates level of signi cance) to determine if the samples are signi cantly di erent from one another. e MWU test ranks the data in ascending order and uses rank sums to generate a z-value (value that indicates signi cance based on percent con dence and the critical value) that is then used on a standard distribution curve to determine signi cance. is test is more reliable than a t-test for very large sample sizes, such as those in the thousands, or where sample sizes in the two groups are very di erent from each other. Each of these tests were appropriate for this study because the study compared only two groups and each comparison used only one independent variable. Each test was done with a signi cance level of 0.05 (95% con dence).
For the summer versus winter tests, all data were separated by parameter (PM 10 , PM 2.5 , or CO), ordered by date, and then separated by month. Each month of data were categorized as summer or winter (summer de ned as February-October and winter de ned as  (Table 1). Of course, without any available data revealing the actual replace usage of the residents in and around Escondido, the times of the parameters were assumed and chosen based on the temperatures in the area and the normal tra c patterns.
To assist in attributing high concentrations of all these pollutants to wood burning, the data were analyzed over the New Year's holiday to determine if there was a peak in ambient concentrations speci cally on this night of the year over all years studied. For daily data points (PM 10 and PM 2.5 ), averages of all data collected December 30 th through January 2 nd were calculated and graphed. For hourly data points (CO and PM 2.5 ), averages of all data collected were graphed from 0:00 on December 30 th to 23:00 on January 1 st to show hourly changes. Assuming that there would be a signi cant increase in emission concentrations over the New Year's holiday, data for July 4 th were also graphed for each parameter in order to rule out reworks that happen on both July 4 th and New Year's as the source of higher ambient concentrations.
To eliminate tra c as a major source of these particulate emissions, the daily and hourly tra c for January 2012 in the area were analyzed with tra c volumes along with ambient pollutant concentrations. In addition, all emissions data available from 2000 to 2010 were averaged by each day or hour, depending on whether the sample was collected daily or hourly, and then graphed against the matching tra c data sets.
To further assess whether PM 2.5 ambient concentrations were due to replace emissions, the speciated carbon data from 2001 to 2007 were separated into elemental and organic carbon totals, and then plotted against the other collected lter data over time to look for correlations in changing emissions levels.

Results and Discussion
e ndings are presented in time series plots, separated by air pollutant type over the entire period studied. Table 2 shows a summary of all results, including signi cance and corresponding p-values and z-values.

PM 10
As expected, emissions in winter months are signi cantly higher than summer months (p=0.0305, z-value outside z-critical) ( Table  2). e two holiday weeks (December 22 nd to January 4 th ) showed signi cantly higher PM 10 concentrations than the rest of the winter months (p=0.0198, z was outside z-critical) ( Table 2). e results indicated these weeks had the most elevated PM 10 concentrations for the entire year. No test was conducted to determine if the traditional holiday days, Christmas and New Year's, were marked by signi cantly higher ambient PM 10 concentrations than the rest of their corresponding weeks because of the lack of a su cient number of daily samples on these holidays (samples were taken every six days). Table 2, weekend PM 10 concentrations are not signi cantly higher than weekday concentrations (p=0.6957, z was not outside z-critical). However, the average PM 10 concentration of the weekend group was higher in value than the average concentration of the weekday group (Table 2). Figure 1 shows the average PM 10 monthly levels over the 11 years of data that were collected. e bold line on the graph at 50 µg/m 3 shows the California state standard and the trend line shows that the average concentrations have slightly decreased over the decade. e peak at October 2003 and again at October of 2007 can be attributed to the county-wide wild res during these months [17]. ese values were not excluded from this analysis.

As shown in
November-January based on the time that replaces are active in this area of California). en each larger data set was compared using both types of statistical tests. All dates without data collected were eliminated before testing to avoid false results.
To further test for conditions within winter months, the winter data were again separated by type, ordered by date, eliminated if no data were collected, and then categorized for each test to create a complete data set for each parameter. e tests included the following: Weekend (Saturday and Sunday) versus Weekday (Monday through Friday): is test was done to separate possible high values of these pollutants from tra c patterns and actually attribute possible pollutant contributions to replace emissions instead of another source.
Christmas Day (December 25 th ) versus Christmas Week (December 23 rd to December 27 th ): Many people take vacation to be home with family during this time and school is o en out for the holiday week.
New Year's Day (January 1 st ) versus New Year's Week (December 28 th to January 3 rd ): ese two holiday comparisons were de ned as two days before and two days a er because the data were collected over multiple years where the holidays and school breaks fell on di erent weekdays.
Two holiday weeks (December 22 nd to January 4 th ) versus winter months (November through January except for the two weeks aforementioned): is test was done to determine if higher emission concentrations in the area were attributed to domestic wood burning during the holidays compared to the rest of winter.
CO (2000 to 2010) and some PM 2.5 data (2008 to 2010) were collected hourly instead of once per day.
is data were subjected to the same tests in the same way as the data above, as well as some additional, more speci c tests. To complete these tests, the data from winter months were sorted by date, then by time and separated based on the parameters of each test. We tested for the following: Night (6:00 pm-12:00 am, active re times only) versus Daytime (6:00 am-5:00 pm): e time frame of 6 am to 5 pm was determined to be 'daytime' because this is a window that incorporates the morning and evening tra c rush, as well as most of the commercial tra c during the day. e 'night' time frame was de ned as 6 pm to 12 am because this window incorporates the assumed time that people use their replaces before falling asleep and letting them burn out or continue smoldering.
Early evening hours (6:00 pm-9:00 pm) versus the rest of the night (10:00 pm-6:00 am): is test was to separate tra c as a possible source of higher ambient concentrations of emissions in early evening hours from other possible sources at night.
Burning re times (6:00 pm-12:00 am) versus smoldering re times (1:00 am-7:00 am): is test was to determine if ambient concentrations of emissions were higher when people rst got home in the evenings and maintained a burning re, or if the emission concentrations were higher when the res were not maintained and were le as smoldering coals. e reason for the possible di erence can be attributed to the fact that burning and smoldering res are di erent in their burning e ciency and could possibly be creating chemically unique emissions.
All tests listed above were done by both Student's t-test and the PM 2.5 data were tested to compare summer versus winter months, the two holiday weeks together versus winter months, the holiday days versus their corresponding week, and weekend versus weekday levels ( Table 2). e results showed that the PM 2.5 concentrations for winter months were signi cantly higher than the summer months (p=0.001, z outside z-critical), which was expected. e two holiday weeks (December 22 nd to January 4 th ) showed signi cantly higher concentrations of PM 2.5 than the rest of the winter months (p=0.0002, z outside z-critical) ( Table 2). e PM 2.5 levels for the traditional holiday days, Christmas and New Year's, were higher in average value than the rest of their corresponding weeks (December 23 rd to 27 th and December 28 th -January 3 rd , respectively). PM 2.5 concentrations for Christmas were not signi cantly higher than the rest of the week (p=0.080, z outside z-critical contrary to t-test; t-test values are more reliable for this sample size, so the result is considered signi cant), but they were signi cantly higher for New Year's Day (p=0.049, z outside z-critical) ( Table 2). PM 2.5 concentrations for weekends were signi cantly higher than for weekdays (p=0.032, z outside z-critical).  Table 2, hourly data points for PM 2.5 from 2008 to 2010 were also tested in the same fashion as daily PM 10 and PM 2.5 values, as well as night versus day, early evening versus the rest of the night, and burning versus smoldering tests. Winter PM 2.5 concentrations were signi cantly higher than summer (p=0, z outside z-critical). PM 2.5 concentrations over the two holiday weeks (December 22 nd to January 4 th ) were signi cantly higher than winter months (p=0, z was outside z-critical) ( Table 2). Christmas Day (p=0.001, z was outside z-critical) and New Year's Day (p=0, z was outside z-critical) PM 2.5 concentrations were both signi cantly higher than their corresponding weeks.
Weekend PM 2.5 concentrations were higher than on weekdays (p=0.028, z outside z-critical). Nighttime PM 2.5 concentrations were signi cantly higher than daytime (p=0, z was outside z-critical) ( Table  2). e ambient PM 2.5 concentrations during early evening hours were not signi cantly higher than those collected during the rest of the night (p=0.195, z was not outside z-critical). Finally, data collected during burning times showed PM 2.5 concentrations were signi cantly higher than those during smoldering times (p=0, z was not outside z-critical; contrary to t-test, t-test values are more reliable for this sample size, so the result is considered signi cant) ( Table 2).

Monoxide
Carbon monoxide (CO) data used for this study were collected on an hourly basis, while most PM data provided were collected daily. CO was tested for summer versus winter months, the two holiday weeks together versus winter months, the holiday days versus the other days in their corresponding week, weekends versus weekdays, night versus daytime, early evening versus night, and burning versus smoldering, which is essentially the rst half of the night versus the latter half of the night ( Table 2). e results showed that CO concentrations in winter months were signi cantly higher than during the summer months (p=0.001, z was outside z-critical). e two holiday weeks (December 22 nd to January 4 th ) showed signi cantly higher ambient concentrations of CO than the rest of the winter months (p=0, z was outside z-critical) ( Table 2). e traditional holiday days, Christmas and New Year's, showed a higher average concentrations of CO than the rest of their corresponding weeks, but were not signi cantly higher (Christmas Day p=0.086, z was not outside z-critical, New Year's Day p=0.061, z was not outside z-critical). is makes the combined two holiday weeks the highest in average concentrations of CO in the year.
As shown in Table 2, the weekend CO concentrations were signi cantly higher than weekdays (p=0.0002, z was outside z-critical). Early evening hour concentrations were signi cantly higher than the rest of the night (p=0, z was outside z-critical). Nights were marked by signi cantly higher concentrations of CO than daytimes (p=0, z was outside z-critical), and the same occurred for burning versus smoldering times (p=0, z outside z-critical) ( Table 2). Figure 3 shows average monthly ambient concentrations of CO over 11 years. e error bars on each monthly average are the minimums and maximums in the month. e high values in October 2003 can also be attributed to the county-wide wild res [17]; however, these high concentrations were not seen during the October 2007 wild res due to the location of the res relative to the monitoring station. Statistical calculations did not exclude these values. Figure 4 shows the average daily concentrations of PM 2.5 and PM 10 from December 30 th to January 2 nd plotted as daily values over all years collected. ese data show the peak in values over the New Year's holiday where people burn res not only for heat, but also, because they stay up until midnight, as is customary to the holiday. Actively burning res emit more gases than smoldering res, so if they are active for a longer period of time over the holiday, as opposed to smoldering for many hours, the concentrations of gaseous emissions in the coldest hours of the night are higher. Higher ambient concentrations of particulate matter may also be expected because the holiday is in the wintertime during a period when the temperature inversion traps emissions in the area.   J Environ Anal Toxicol speci c pollutant showed that there were higher ambient concentrations over the time period when replaces are used all night. Another source of these higher ambient concentrations to consider are reworks used to celebrate the holiday. In order to determine if this was a likely source contributing to the higher ambient concentrations found on New

Additional holiday analysis
Year's, the July 4 th period was plotted for each pollutant. For daily PM 10 and PM 2.5 values, the daily average concentrations were plotted from July 2 nd through July 6 th . e hourly data for PM 2.5 and CO were plotted as hourly concentration averages from midnight on July 4 th to 11:00 p.m. on July 5 th in order to observe any hourly changes across the July 4 th holiday.

Tra c analysis
Interstate 15 travels through Escondido and carries heavy commercial and other motorized vehicle tra c, whose emissions pollute Escondido's air throughout the year. To identify the impact of other possible major sources of PM 10 , PM 2.5 and CO in the area, the tra c volume against ambient concentrations of each pollutant Once it was found that the air quality di erence between these two seasons was statistically signi cant, the emissions samples taken during the winter month weekdays were compared to the samples taken during the weekends ( Table 2). is di erence was attributed to the higher pollutant emissions emanating from replaces as opposed to tra c emissions. Wood burning stoves were also identi ed as a major emission source because in this area of Southern California, at least outside of the urban section of the city, stoves are used as an e cient way to heat the home. is study's results for weekends versus weekdays and days versus evenings also suggest that tra c is not the primary source of these particulate emissions.

Speciated particulate matter analysis
Speciated PM 2.5 data were available year-round from 2001 to 2007, and were available as separate carbonaceous species that could be sorted and analyzed. Data were broken down into Total Elemental Carbon and Total Organic Carbon and compared to PM 2.5 lter data that were collected on the same day. All three data sets were graphed, with the expectation that the sum of the elemental and organic carbon species would be the same as or less than the PM 2.5 lter collection ( Figure  10). e graph also shows that the carbon species increase and decrease at the same time as the carbon species collected on the lters. It was also shown that the portion of elemental carbon was always less than the portion of organic carbon. Table 3 shows the range and average of percentages of elemental and organic carbon species, separated by summer and winter to show the seasonal uctuation that was also seen in lter tests.

Conclusions
During ideal biomass burning conditions, CO 2 is produced (CH 2 O+O 2 → CO 2 + OH 2 ). CH 2 O represents the composition of the average cellulose in wood [22]. In reality, methane, hydrocarbons, VOC's (Volatile Organic Compounds), and a large list of other          contaminants can be released into the air during wood burning [23] .
Comparison of the results between burning and smoldering tests indicate that the products of wood burning combustion change with the replace temperature. As the re changes from burning to smoldering, ambient concentrations of CO increase. All types of combustion reactions have e ciency values that vary in accordance with the combustion chamber temperature. e temperature inside the replace decreases throughout the time it is used, and the amount of particulate emissions generated increases as re temperature decreases, making smoldering and initial burning the most impactful combustion occurrences [2]. PM 10 , PM 2.5 , and CO data sets collected at the SDAPCD Escondido monitoring station from 2000 to 2010 showed higher ambient concentrations of pollutants in the winter and around the winter holidays. ese higher concentrations were attributed to larger emissions during times of replace usage. is conclusion was drawn by eliminating tra c as a major source of these emissions, and by separating carbonaceous species that showed a relationship with the particulate lter data sets. As expected, the concentration of pollutants present in the air was signi cantly di erent between winter and summer. e results suggest that di erences in measured concentrations may have been due in part to replace wood burning, along with more stable atmospheric conditions during winter months. e hourly CO data collected were more extensive than the daily particulate data collected. is allowed for exibility in the analysis and they also made it easy to isolate measurements based on times of the day. Data could be sorted into various groups since each data point collected was paired with a time of day. Results of the weekday versus weekend tests, as well as the daytime versus nighttime, and early evening versus nighttime tests, show that higher concentrations of CO (Table 2) at these times likely comes from the use of replaces as opposed to other area sources because concentrations are signi cantly higher at speci c times when people are home and using their replaces, furnaces, and water heaters.