Ross Creek Particulate Matter Speciation Report
A Report from the Fort Air Partnership
Originally published Jan 3, 2022. Last updated: February 25, 2022
1 Introduction
The Fort Air Partnership (FAP) is a not-for-profit organization dedicated to monitoring the air people breathe within a 4,500 square kilometer Airshed located to the northeast of Edmonton. The FAP performed a three year monitoring study of speciated Particulate Matter (PM) at the Ross Creek station (53.71622° N and 113.19994° W) from March 21, 2018 to April 10, 2021.
This report provides information on the characteristics and potential sources of the speciated PM collected over the three-year monitoring project and how this compares to other speciated PM monitoring sites in Canada and the United States.
1.1 Background
1.1.1 Particulate Matter
Particulate Matter (PM) is a term used to describe particles found in the air. PM is a complex mixture of solid and liquid phases that can vary in shape, size, and chemical composition. PM is typically described using the aerodynamic diameter (size fraction). PM2.5 , also known as fine particulate matter and the focus of this study, is the term for particles smaller than 2.5 micrometers. PM10 (course particulate matter) is the term for particles smaller than 10 micrometers. In Alberta there are both provincial objectives and federal standards for PM2.5. The Canadian Ambient Air Quality Standards (CAAQS) are national standards used to identify and manage issues in air quality and consist of both daily (29 \(\mu\)g/m3) and annual (8.8 \(\mu\)g/m3) standards. A 24-hour Alberta Ambient Air Quality Objective (AAAQO) is also in place for PM2.5 (29 \(\mu\)g/m3). Furthermore, PM2.5 is included in the determination of the Air Quality Health Index (AQHI); a tool used to relate ambient air quality (including fine particulate matter) to health in Canada.
Some particulates are directly emitted into the atmosphere such as windblown dust, road dust (such as brake and tire wear), diesel soot, wildfire smoke, and marine aerosols; these are referred to as primary PM. Other particle components referred to as secondary PM are formed in the atmosphere through atmospheric reactions of gaseous precursors like sulphur dioxide, nitrogen dioxide, ammonia, and volatile organic compounds. Key examples of secondary PM include ammonium sulphate, ammonium nitrate, and organic matter. Organic matter, is a very complex set of partially oxidized organic (carbon, hydrogen, oxygen) compounds which condense into the particle phase. Examining the chemical composition of particulate matter can aid in understanding the relative contributions of local and regional sources of pollution that are contributing to local PM concentrations. The most important components in terms of total PM mass are usually organic matter, sulphate, nitrate, ammonium, elemental carbon, and windblown soil or dust (in this report collectively referred to as crustal matter).
1.1.2 Project background
A 2011 assessment of the FAP air monitoring network found that a lack of speciated measurements of PM2.5 posed “a weakness of the network to help characterize emissions sources, provide validation information for models, and provide information for others to evaluate exposure.” FAP undertook this sampling project following a key recommendation from the assessment, that FAP “consider adding a speciated measurements program of PM2.5 at a single site….”
In addition, FAP is often asked by the public or municipalities if the network monitors for metals and toxic substances. Some toxic substances and many metals are collected in speciated monitoring of PM2.5. Thus, the additional information that this speciation project and report provides is considered of interest to FAP stakeholders and the public.
The PM speciation data from this project will also inform work under the Capital Region Particulate Matter Response Plan, in which FAP is actively involved. It can also be compared to other National Air Pollution Surveillance (NAPS) sites in Alberta to better understand PM composition in the province.
1.1.3 Study objectives
The PM speciation project sought to meet the following objectives:
- Assess PM2.5 mass components and gaseous precursor concentrations.
- Quantify major components of PM2.5 including ammonium nitrate, black carbon and organic carbon.
- Quantify important gas phase species including ammonia and nitric acid, and
- Quantify important tracers for specific PM2.5 sources.
These four primary objectives are meant to improve understanding of:
- PM2.5 composition at the monitoring location in Fort Saskatchewan.
- The major primary and secondary sources of PM2.5 measured in the city of Fort Saskatchewan.
- The composition of PM2.5 in comparison to other monitoring sites.
- The source of anthropogenic PM2.5 in the city of Fort Saskatchewan and their proximity to the Ross Creek monitoring site.
- The relative contributors to high 24-hr averages and annual averages of PM2.5 at the Ross Creek site.
1.1.4 Study area
Figure 1 shows the location of the Ross Creek station and the nearest NAPS speciated PM site at Edmonton-McIntyre. The Ross Creek station is situated between the city of Fort Saskatchewan to the southwest and the petrochemical and industrial facilities of the Industrial Heartland of Alberta to the northeast. There is also a separate heavy industry area to the southwest referred to as the Strathcona Industrial Area. Fort Saskatchewan is a city of approximately 27,000 residents and is part of the greater Edmonton Metropolitan area (population ~1.3 million). The Industrial Heartland has multiple petrochemical facilities along the North Saskatchewan River. The land use surrounding the Ross Creek monitoring site also includes agricultural land. The primary FAP monitoring objective for the Ross Creek station is to measure the influence of local industrial emissions on air quality.
Figure 1. Map of Ross Creek and Edmonton-McIntyre stations
Images of the Ross Creek monitoring station and its speciated PM2.5 sampling equipment are shown in Figure 2 and Figure 3, respectively.
Figure 2. Picture of the Ross Creek station
Figure 3. Ross Creek station instruments
1.1.5 Comparison sites
Ross Creek is a relatively unique location for speciated PM2.5 monitoring, as the combination of being adjacent to a suburban town and a heavy industrial zone is not similarly duplicated at any other Canadian or American sites. Comparison sites included in this report provide a representative cross section of monitoring site characteristics.
Six speciated PM2.5 monitoring sites were selected from Canada and the United States for comparison to the Ross Creek site. The complete set of seven sites are shown in Figure 4. Selected Canadian sites are intended to provide a cross-section of Canada. The three Canadian sites selected are Edmonton-McIntyre, Ottawa Downtown, and Burnaby South in Vancouver. The Edmonton-McIntyre site is located in South Edmonton just north of a major roadway, Whitemud Drive NW. The Burnaby South site is in a residential area and is relatively far from any industrial or roadway sources. The Ottawa Downtown site is also in a residential area near the Rideau River.
To complement the Canadian comparison sites, three speciated PM2.5 monitoring sites from the United States were also selected for comparison to the Ross Creek site. These include Granite City, Illinois; Deer Park, Texas; and Teddy Roosevelt National Park, North Dakota. The Granite City site is in East St. Louis, Illinois, near a petrochemical upgrader and heavy industry including a major steel facility. The Deer Park site is located near the Houston ship channel in Texas, another major North American petrochemical refinery area. Finally, the Teddy Roosevelt National Park site is located near upstream shale oil production in rural North Dakota; this site is relatively similar in terms of geography and climate to Fort Saskatchewan although it does not have major industrial facilities or population centers within 10 km.
Figure 4. Map of selected comparison sites
1.2 Overview
The Methods section describes the data collection and processing steps used in this study. The Monitoring protocols section describes data collection and analysis methods. The R and packages section describes the software tools used, while the Data processing section lists the data acquisition, cleaning, and processing for use in the analysis.
The Results section describes the data analysis findings. The Summary statistics section provides a summary of the sample counts, averages, and detection limits for every parameter measured at the Ross Creek site. The Monthly mass comparison section shows the PM2.5 time series of mass concentrations at Ross Creek and six comparison sites. A section on Reconstructed fine mass describes the patterns in speciated PM2.5 components and their relative contributions. The Meteorology section provides a description of wind patterns at the Ross Creek site through use of wind roses. Finally, the Factor analysis section describes source apportionment methods, results, seasonal patterns, and association with wind patterns using polar plots.
The Discussion section provides a broader comparison of results from the paper within the context of other scientific studies of PM in North America. It also provides a description of the limitations of the study and a brief conclusion.
The Acknowledgements section lists authors of this report, organizations that performed monitoring of speciated PM2.5, and other helpful contributors. Finally, the References lists the citations and references for peer-reviewed articles and software packages used.
2 Methods
2.1 Monitoring protocols
Monitoring of speciated PM2.5 at Ross Creek followed the current NAPS sampling protocols as documented in Dabek-Zlotorzynska et al. (2011). NAPS was established in 1969 to monitor and assess the quality of ambient (outdoor) air in the populated regions of Canada. The goal of the NAPS program is to provide accurate and long-term air quality data of a uniform standard across Canada.
Ross Creek speciated PM2.5 sampling began on March 21, 2018 and ended on April 10, 2021. Samples were collected for 24-hours midnight to midnight, every six days following the NAPS schedule for intermittent sampling. In total, 187 integrated PM2.5 samples were collected during the study, including one field blank. No samples were reported as invalid by the analytical laboratory or field technicians. Some samples were flagged as slightly damaged or discolored by the analytical laboratory; these were used as reported.
Samples were taken using two intermittent samplers:
A Met One SuperSASS model speciation sampler with 3 cartridges: “A” “B” & “C.”
Module A - Anodized aluminum inlet, pre-fired quartz filter; measurements of organic carbon and black carbon using IMPROVE_A temperature protocol.
Module B - Anodized aluminum inlet, Teflon and pre-fired quartz filter; measurements of PM2.5 mass and organic carbon artifact
Module C - Teflon coated inlet, Teflon and nylon filter, sodium carbonate and citric acid denuders, ions, sugars metals, nitrate, sulphate, and volatile components of sulphate, nitrate, and ammonia.
A Thermo Scientific model 2000i Partisol sampler equipped with a 47mm Teflon filter
Mass flow controllers maintained the flow rates of the fine and coarse particle streams at 15 liters per minute. The samplers were calibrated every three months during the sampling period.
Integrated (24-hr) measurements include:
- Gravimetric PM2.5 mass
- Ions measured by ion chromatography (IC). Key species include sulphate, nitrate, ammonium, chloride, calcium, sodium, and potassium
- Carbonaceous species by IMPROVE_A temperature protocol thermal optical reflectance (TOR). Key species include organic carbon (OC), black carbon (BC), and carbon fractions (OC1-OC4, EC1-EC2, POC) that volatilize at different temperatures.
- Metals measured by inductively coupled plasma mass spectrometry (ICPMS). Key species include iron, titanium, manganese, vanadium, nickel, chromium, cadmium, arsenic, and lead.
- Light elements measured by x-ray fluorescence (XRF). Key species include silicon, chlorine, bromine, calcium, and sulfur.
- Volatile PM2.5 components were measured on backup citric acid filters; these include sulphur dioxide (SO2), ammonia (NH3), and nitric acid (HNO3).
Speciated PM2.5 data collected at the Ross Creek station were provided directly by the Desert Research Institute (DRI) analytical laboratory in a Microsoft Excel spreadsheet in September of 2021.
NAPS data for Edmonton-McIntyre, Burnaby South, Ottawa Downtown and all other sites was available for all of 2018 and was partially complete for 2019 when downloaded in August 2021; 2020 data were not yet publicly available as of February 2022.
Speciated PM from the three U.S. sites were selected from the IMPROVE (Interagency Monitoring of Protected Visual Environments) and CSN (Chemical Speciation Network) datasets. These data are available publicly here. Data for 2018-2020 were acquired in February, 2022.
Continuous data collected at the Ross Creek site were obtained from the Alberta data warehouse. Data collected from March 2018 through April 2021 at the Ross Creek site were acquired in August, 2021. This data were used to generate hourly and 24-hr wind roses.
2.2 R and packages
The R statistical programming language package (v4.1.1 “Kick Things”) developed by the R Core Team (2021), was used to create an integrated and interactive report . R and its libraries are free and open source software. The R libraries used to acquire, process, and visualize the datasets include:
- Data import and processing libraries: tidyverse, janitor, readxl, lubridate, and data.table; Wickham et al. (2019), Firke (2021), Wickham and Bryan (2019), Grolemund and Wickham (2011), Dowle and Srinivasan (2021)
- Data visualization libraries: openair, leaflet, DT, scales, grid, htmltools; Carslaw and Ropkins (2012), Cheng, Karambelkar, and Xie (2021), Xie, Cheng, and Tan (2021), Wickham and Seidel (2020), Cheng et al. (2021)
- RMarkdown template library: Rtemps; Zobolas (2020)
R Markdown is a file format for making dynamic documents (HTML, PDF, or MS Word, for example) in R. R Markdown allows the author to embed R code into the document.
2.3 Data processing
2.3.1 Ross Creek integrated data
The final Ross Creek dataset was acquired in September, 2021 directly from the DRI laboratory. All data was imported from four separate MS Excel worksheets. A table of parameter names and methods of analyses were used to convert lab specified identifications to plain language chemical species. Additionally a separate method detection limit (MDL) worksheet was imported to provide laboratory reported MDLs for each species.
Data were combined together from the four separate worksheets into a single data frame. A single blank sample collected on June 20, 2018 was removed. Metadata were added to the data to ensure that each sample included plain languages values for date collected, collection channel, analysis method description, method detection limit (MDL), and a below MDL flag.
Four samples had additional laboratory flags. These were: mineral particles in deposit (1 sampled filter), discoloration on deposit (1 sampled filter) and scratch or scrape on deposit (2 sampled filters).
2.3.2 NAPS data
NAPS data were acquired from the NAPS web data repository zip files. No data was available for 2020 or 2021 as of February, 2022. Similar, to the Ross Creek dataset, the data were provided in MS Excel spreadsheets with individual analysis types in separate worksheets. English language spreadsheets were imported for 2018 and 2019 data worksheets for each measurement type used for later site comparisons. Data from all sites and years were combined into a single data frame. Field and trip blanks were removed from the NAPS dataset and invalid (“M1” flag) samples were removed.
2.3.3 United States comparison data
Daily average samples from the three US comparison sites were acquired from the U.S. Environmental Protection Agency Air Quality System in February 2022. All samples were pre-cleaned to remove NULL concentration samples. Samples are typically invalidated only if something goes wrong during collection or measurement (e.g., sample not collected, bad flow rate, destroyed filter). Data below MDLs were reported as is with no censoring or substitution of below MDL data. Parameters were blank corrected using monthly average trip blanks from the network, but there are no backup filter corrections for volatile species like nitrate and sulphate as done in the NAPS network. The United States sites had 36 months of valid PM2.5 data from 2018 to 2020 available as of February 2022.
2.3.4 Data processing summary
Data were imported for each of the individual data sets. Blanks were removed from the datasets. In creating summary statistics and monthly averages, no treatment of data below reported method detection limits (MDL) was performed. Data below detection were included as reported.
Monthly average concentrations of PM mass and PM components required at least 3 valid samples per month. Months with fewer than 3 valid samples for a given parameter were not included.
Sulphate and nitrate were calculated as corrected values based on subtracting backup filters for the Ross Creek and NAPS data. United States sites do not have backup filters and thus semivolatile corrections are not possible; concentrations of sulphate and nitrate from U.S. sites are likely biased high relative to the Canadian sites.
2.3.4.1 Reconstructed Fine Mass Calculations
Reconstructed fine mass (RCFM) uses sampled concentrations to calculate the major components that make up most of the total PM2.5 mass at North American sites. The major components of PM2.5 mass are organic matter (OM), elemental carbon (EC), soil (crustal matter), ammonium, nitrate, sulphate, and salt (NaCl).
The evolution and assumptions of RCFM calculations are discussed in Chow et al. (2015). There are multiple accepted methods for calculating RCFM for each of the components, which can depend on specific site characteristics and speciation profiles for salt, crustal matter, OM, and even ion components like sulphate and nitrate. For this study, component mass reconstruction is calculated following methods outlined in Dabek-Zlotorzynska et al. (2011), Chow et al. (2015), and Wang et al. (2021), and Malm et al. (1994).
This section lists each of the equations used to calculate the RCFM components . Specifically the equations below list our calculations.
\[OM=OC*1.6\] The laboratory results include organic carbon (OC) concentrations. OC has a stoichiometric relationship to particulate matter organic matter (OM). OM is calculated as 1.6*OC, and is the largest contributor in most months. For example, OM is a multiplier of OC multiplied by X, where X can be as low as 1.2 (freshly emitted OC, high hydrogen:oxygen ratio) and as high as 2.5 (highly aged OC, low hydrogen:oxygen ratio), depending on the age and oxidation of organic carbon. A value of 1.6 was selected for this study, which is representative of moderately aged aerosol that may be typical of a site downwind of a metropolitan area with only moderate amounts of fresh local emissions. However, this average OM multiplier may not be appropriate for individual days where OM could be primarily from long-distance transport.
\[crustal\ matter=2.2*Al+2.49*Si+1.63*Ca+2.42*Fe+1.94*Ti\]Crustal matter mass was calculated as described in Malm et al. (1994), using iron, silicon, titanium, calcium, and aluminum. Ross Creek and NAPS measurements include ICP-MS and XRF measurements of some species. In this analysis, XRF measurements of aluminium, calcium, and silicon and ICP-MS acid digest measurements of iron and titanium were used to calculate crustal matter for Canadian sites. The US sites only measured elements using XRF and so all values of crustal matter for those sites are XRF only.
\[NaCl=Na + Cl\] Soluble sodium and chloride measured with ion chromatography were used to calculate salt mass. Chlorine measured using XRF was not used.
\[NH_{4}NO_{3}~=1.29*NO_{3}\]
The volatile backup corrected nitrate concentrations for Canadian sites and the uncorrectable nitrate values reported for the US sites were used to calculate ammonium nitrate mass (nitrate). In both US and Canadian cases, the concentration of nitrate was multiplied by 1.29 to account for full neutralization by ammonium in NH4NO3 as described in Chow et al. (2015).
\[(NH_{4})_{2}SO_{4}~=1.375*SO_{4}\]
The volatile backup filter corrected sulphate concentrations for Canadian sites and the uncorrectable sulphate values reported for the US sites were used to calculate the ammonium sulphate mass (sulphate). In both US and Canadian cases, the concentration of sulphate is multiplied by 1.375 to account for full neutralization by ammonium in (NH4)2SO4 as described in Chow et al. (2015). An investigation of the ion balance showed that the nitrate and sulphate are fully neutralized and the ammonium calculations are consistent with measured ammonium to within 10% for all months at the Ross Creek site.
For elemental carbon (EC) the IMPROVE_A TOR reported concentration was used directly for all sites.
3 Results
3.1 Summary statistics
Summary statistics for the speciated Ross Creek data are shown in Table 1. All collected species are shown for each collection channel and sampling methodology. For example, PM2.5 mass was collected in both the Super SASS B cartridge and Partisol samplers and measured gravimetrically. Count indicates the number of valid samples. Mean indicates the sample average for all valid samples. The analytical method detection limit (MDL) is reported by the laboratory and indicates a concentration threshold below which the concentrations are not significantly different from a value of zero. The pctBelowMDL column indicates the percentage of samples that were reported at or below the MDL; high percentages indicate that the concentration for a large number of samples were lower than the MDL. As a rule of thumb, mean concentrations are considered reliable with high certainty no more than 25% of samples are below MDL. Summary concentrations for each method and channel are reported as is, blanks and samples flagged as invalid or missing by the analytical laboratory were removed.
3.2 Monthly mass comparison
Table 2 shows the mean and median PM2.5 concentrations at the seven sites for the data available in the time period 2018-2021. Ross Creek, Deer Park, Granite City, and Teddy Roosevelt had comparable data collection periods of 36 months. Data completeness at the Canadian NAPS sites ranged from 20-22 months, ending in mid-2019. The Ross Creek site mean and median concentration is slightly higher than the nearby Edmonton site and significantly higher than Burnaby South, Ottawa, and Teddy Roosevelt. The Ross Creek median and mean is significantly lower than the sites at Deer Park and Granite City.
Table 3 provides a sortable and searchable table with the individual monthly average values of PM2.5 for the seven sites.
Figure 5a shows the time series of monthly average PM2.5 gravimetric mass reported at Ross Creek and the six comparison sites. We note that the Edmonton-McIntyre site is about 31 km to the Southwest of the Ross Creek site; all other sites are at least 500 km away.
The annual Canadian Ambient Air Quality Standard (CAAQS) of 8.8 \(\mu\)g/m3 is shown as a dashed red line in Figure 5. The CAAQS are used to evaluate data for air quality management decisions and do not include wildfire impacted samples. This standard is included in some figures as a general guideline. Monthly averages should not be directly compared to annual standards, but the annual average can be used as a guideline to highlight the typical concentrations at each site. Monthly average PM2.5 concentrations at Ross Creek are quantitatively similar to Edmonton but are higher than the two urban sites of Ottawa and Burnaby South and the Teddy Roosevelt National Park site. The high concentration peak in August 2018 observed at Ross Creek and Edmonton-McIntyre sites coincided with a very active wildfire month in British Columbia with smoke transported throughout Alberta.
Figure 5b shows scatter plot of monthly average PM2.5 at each of the six comparison sites compared to Ross Creek. Ross Creek concentrations are always on the x-axis. The 1:1 line is shown as a dashed black line. Points near the 1:1 line indicate that concentrations between the two sites are similar. Points well below the line indicate higher concentrations at Ross Creek. Points above the line indicate higher concentrations at a comparison site (e.g., some monthly points at Deer Park and Granite City sites).
Figure 5a. PM mass time series at Ross Creek and comparison sites
Figure 5b. Comparison of monthly average PM concentrations
3.3 Reconstructed fine mass
The primary mass components of the three year average PM2.5 at the Ross creek site over the study period are OM, sulphate, nitrate, and ammonium. The annual mean concentration at Ross Creek was 7.66 \(\mu\)g/m3 over the three year monitoring period. This three-year mean concentration includes all valid monthly means and is including exceptional events like the wildfires of August 2018 and September 2020.
Table 4 displays PM2.5 reconstructed fine particulate mass concentrations by month and year for the Ross Creek and Edmonton-McIntyre sites.
OM concentrations at Ross Creek had a study average of 2.45 \(\mu\)g/m3, which is 33% of the total mass. This was lower than the 2.8 \(\mu\)g/m3 and 42% average contribution at the Edmonton-McIntyre site. Sulphate at Ross Creek was 1.4 \(\mu\)g/m3 and is 19% of the mass, both higher than the Edmonton values of 1.11 \(\mu\)g/m3 and 17% of the mass. Nitrate at Ross Creek was 1.15 \(\mu\)g/m3 and is 16% of the mass, which is considerably lower than the 1.97 \(\mu\)g/m3 and 29% contribution at Edmonton.
The shorter sampling period at Edmonton (~20 months of data) results in skewed mean concentrations from the very high outlier months of August 2018 for OM and March 2018 for nitrate. Those two high concentrations with a shorter monitoring period have almost double the average weight compared to the 36 month sampling period at Ross Creek. The relative contribution of OM is generally highest in the warmer months; sulphate concentrations are relatively constant throughout the year and nitrate contributions are elevated during the colder months.
The three smaller components of reconstructed fine particulate mass (RCFM) are crustal matter, EC, and salt. Crustal matter was calculated at 0.45 \(\mu\)g/m3 (6% of mass), EC was measured as 0.27 \(\mu\)g/m3, and salt was measured at less than 0.05 \(\mu\)g/m3.
Figure 6 shows the RCFM at the Ross Creek (6a) and Edmonton-McIntyre (6b) sites. Both figures show stacked bar plots indicating the monthly average components contributing to the PM2.5 mass. The red dashed line indicates the annual average CAAQS at 8.8 \(\mu\)g/m3. Both the Ross Creek and Edmonton-McIntyre sites monthly averages are typically below the annual standard value; with the notable exception of the month of August 2018. The highest monthly value at both sites is the August 2018 wildfire smoke event in British Columbia, as documented in the Alberta Air Zones Report 2016-18. The very high organic matter (OM) in August 2018 is consistent with particulates from smoke. It is also noteworthy that the March 2018 Edmonton-McIntyre episode dominated by high nitrate concentrations occurred just prior to the deployment of speciated PM2.5 monitoring at the Ross Creek site. A smaller nitrate episode in the February and March 2019 time period is visible at both Ross Creek and Edmonton-McIntyre.
August 6, 2018, August 18, 2018, and September, 18, 2020 are removed by Alberta Environment and Parks as wildfire smoke days at Fort Saskatchewan. Alberta Environment and Parks calculates CAAQS metrics and excludes exceptional events that cannot be directly controlled by management actions, such as smoke from wildfires.
Figure 6a. Ross Creek Reconstructed fine mass by month
Figure 6b. PM mass time series at Edmonton-McIntyre
Figure 7 shows a scatter plot of monthly average RCFM components at the Ross Creek and Edmonton-McIntyre sites. The black dashed line indicates the 1:1 line; note both axes are on a logarithmic scale. Symbols above the line indicate higher component concentrations at the Edmonton-McIntyre site; symbols below the line indicate indicate higher component concentrations at the Ross Creek site. Overall, most symbols are clustered around the 1:1 line, indicating comparable concentrations at the two sites.
The components at the two sites with the most notable concentrations are OM, nitrate, sulphate, and ammonium. OM and nitrate concentrations were comparable at the two sites, with no notable differences in mean concentrations. Sulphate and ammonium concentrations were slightly higher at the Ross Creek site.
Of the minor components, salt, EC, and crustal matter appear to be consistently higher at Edmonton-McIntyre. This is consistent with its proximity to major roadways (Whitemud Drive NW and 91st Street) and the increased influence of vehicular emissions, road salt, and road dust.
Figure 7. Comparison of monthly average RCFM concentrations
Figure 8 shows the monthly average RCFM at comparison sites. Canadian sites are arranged on the left and U.S. sites are arranged on the right. Ross Creek concentrations are typically higher than those observed at the Burnaby South and Ottawa Downtown sites during the comparable time periods. Burnaby South is in Vancouver, this area experiences more rainfall than Edmonton or Ottawa and has clean Pacific air as background Vedal et al. (2003); it is expected to have lower particulate concentrations. Ottawa has somewhat high OM and wintertime nitrate but does not have the sulphate or nitrate levels seen at the Edmonton and Ross Creek sites. Sulphate levels in the eastern half of Canada and the U.S. have rapidly declined from the early 2000s as shown in Chan, Gantt, and McDow (2018). Concentrations in eastern Canada were 4-6 \(\mu\)g/m3 at sites in Windsor, Simcoe, Toronto, and Montreal in 2003-2008 as shown in Dabek-Zlotorzynska et al. (2011). Concentrations at Ottawa in 2018-19 were below 1 \(\mu\)g/m3. The decline in the U.S. is attributed to reductions of >70% in SO2 emitted from coal-fired power plants.
The Granite City, Illinois site shows high crustal matter contributions but that is actually due to industrial iron emissions that result in very high iron concentrations (multiple micrograms). In contrast, the high crustal matter concentrations at Deer Park, Texas are seasonally consistent with dust events from the Sahara in Africa being transported to Texas Pu and Jin (2021) . Deer Park has sulphate levels that are comparable to Ross Creek, but much higher EC and lower nitrate than Ross Creek and Edmonton. The Teddy Roosevelt site has relatively low PM but does show some seasonal nitrate and evidence of the August 2018 Canadian fires affecting the area.
Figure 8. Reconstructed fine mass at comparison sites
3.3.1 Reconstructed fine particulate mass on high PM days
Figure 9 shows the average contributions of each of the RCFM components to total mass over the three year study period and for the top 10 PM mass days of the three year study period. Individual components are shown as stacked columns; total PM2.5 mass is shown as the black dot. The blue dashed line indicates the 24-hr Alberta ambient air quality objective (AAAQO) of 29 \(\mu\)g/m3. When total PM2.5 mass is higher than the sum of RCFM, that may indicate underestimation of individual components in the calculation, such as organic matter or crustal matter. Higher OC:OM multipliers such as 1.8 or 2.1 would potentially account for much of the underestimate for the high OM days like August 6, 2018, August 18, 2018, and September 9, 2020 where aged wildfire smoke is likely impinging the site. However, it seems less likely that changing the OM multiplier would account for the discrepancy in mass on low OM days like January 28, 2020 or March 11, 2021.
Figure 9. Ross Creek reconstructed fine mass - top 10 days
3.4 Meteorology
Figure 10 shows wind roses at Ross Creek by meteorological season. Wind roses show the direction the wind is blowing from, so if the largest paddle is pointing to the southeast then the wind is most frequently originating from the southeast. For example, in Figure 10 the winter (DJF = December, January, and February) winds are at lower speeds and originate most commonly from the southwest. In contrast, the summer (JJA = June, July, and August) winds have higher speed and are most commonly from the West.
Speciated particulate matter sampling is done on a 24-hr basis at all sites in this study. Hourly winds were vector averaged to generate a 24-hr wind rose for PM sampling days at Ross Creek as shown in Figure 11. With far fewer samples, this wind rose is less granular and more prone to averaging out diurnal shifts in the winds. For the study period, the predominant wind direction was southwesterly and wind speeds were low (0-2 m/s); higher speed westerly and northwesterly winds (2-6 m/s) were also observed. As illustrated in Figure 10, low speed southwesterly winds were dominant during the winter. Higher speed winds from the west and northwest were more frequent during the spring and summer.
Figure 10. Hourly wind roses at Ross Creek by season
Figure 11. Ross Creek sampling days 24-hr wind rose
3.5 Factor analysis
To assess possible sources and origins of PM2.5 at Ross Creek, the EPA PMF 5.0 software package was applied to the Ross Creek speciated data to perform source apportionment. Source apportionment is a statistical technique that uses measurement data with many different samples and species to decompose samples into a linear combination of “factors” that represent covarying species. Applying this technique to PM data allows us to quantitatively identify emissions sources and components of particulate matter that influence concentrations at the Ross Creek site.
Species for analysis were selected based on (1) signal-to-noise ratios > 1.0, such as those in McCarthy et al. (2013) and (2) at least 25% of the species needed to be reported at concentrations above method detection limits. Secondarily, species measured via multiple methods (i.e., ICP-MS vs. XRF metals) were selected based on signal-to-noise ratios for the two methods, with higher values being the basis for selection when multiple measurements of the same species were available.
Source apportionment techniques require the user to determine the correct number of factors. For this analysis, five to eight factor solutions were tested at the Ross Creek and combined Ross Creek and Edmonton-Mcintyre sites. A six factor solution for the Ross Creek data was selected as most reasonable based on examining source apportionment results from the Ross Creek only dataset and combined Ross Creek and Edmonton-McIntyre dataset. The source apportionment results presented are only for the Ross Creek dataset. The six factor solution for the Ross Creek data include factors identified as:
- ammonium nitrate (nitrate)
- ammonium sulphate (sulphate)
- wildfire and residential smoke (smoke)
- salt and road dust metals (salt & metals)
- crustal material and dust (crustal matter)
- organic matter and metals (OC1 & metals)
The PM2.5 in the six-factor solution was comparable to the measured mass and had an R2 of 0.81. It accounted for 92% of the total PM2.5 mass. The analysis did not adequately fit a few high outlier concentration days; the results notably underestimate measured concentrations of the three highest contribution days - two wildfire smoke events and the highest nitrate observation on the first sampling date. Underestimating the highest values is common for statistical fitting techniques like PMF.
Each of the factor contribution allocations of each PM species are shown in Figure 12a. Factor names are based on key chemical species in the factor or based on the cumulative characterization of a factor with an emissions source. For example, the nitrate factor has has the largest portion of the nitrate species allocated to it and the smoke factor contains a large fraction of the organic and elemental carbon thermal fractions (e.g., oc3, oc4, poc, ec1).
Elemental abbreviations are used in Figures 12a and 12b to identify elements. The list is arsenic (as), barium (ba), bromine (br), calcium (ca), chlorine (cl), cobalt (co), chromium (cr), copper (cu), iron (fe), potassium (k), manganese (mn), nickel (ni), lead (pb), silicon (si), strontium (sr), titanium (ti), and zinc (zn).
Figure 12a. Ross Creek PMF six factor solution
Figure 12b shows the absolute contributions of each species in each factor; this figure has a logarithmic x-axis which can compress large differences. This is the factor chemical “fingerprint” which identifies the relative contributions of each of the included chemical species. Sometimes a chemical species such as potassium (k) will have no point displayed (e.g., in the nitrate factor). This means that the contribution of that chemical species was less than 10-5 \(\mu\)g/m3 relative to the unit mass. Daily factor concentrations below 0.1 \(\mu\)g/m3 are not displayed in this figure to focus on the higher concentration days.
Figure 12b. Ross Creek PMF six factor solution
Figure 13 displays a time series of factor contributions across the three year monitoring period. Each of the six factors has unique temporal variability. The gray dashed line indicates the 3 \(\mu\)g/m3 concentration line, which is a top 10th percentile concentration for the individual factors. Factors including nitrate, smoke, crustal matter, and salt have significant seasonal variability as illustrated in Figure 14.
Figure 13. Ross Creek PMF factor contribution time series
Figure 14 shows seasonal contributions by factor using box-whisker plots. The line at the waist shows the median concentration contribution and the 95th percentile confidence interval around the median. The orange line shows the median concentration across all individual factor contributions (~ 0.75 \(\mu\)g/m3). If two notches do not overlap, differences between seasons can be considered statistically significant.
Differences in seasonal concentrations indicate that the contributions are seasonal as a result of differences in emissions activity, physical and chemical processes as a function of temperature and solar irradiance, and/or planetary boundary layer height McCarthy et al. (2007). The nitrate contributions are highest in winter and are lowest in summer; nitrate particle formation is very sensitive to cool temperatures as it volatilizes at warmer summer temperatures. The salt and metals factor has a nearly identical seasonal pattern and is likely associated with salt applications to prevent icy roads in the colder months. The sulphate factor is relatively consistent throughout the year; this may be a result of constant emissions activity and consistent frequency of winds originating from sulphur emitting industry (as shown later in Figure 15). The smoke factor contribution is highest in the summer, which is consistent with the large infrequent contributions from wildfires. The smoke factor is also elevated in the winter months, which could be due to residential or agricultural smoke. The OC1 & metals factor is highest in the summer months, which may be due to higher photo-oxidation of biogenic and anthropogenic VOC emissions in the summer months leading to greater secondary OM. Finally, the crustal matter factor is highest in spring (possibly associated with agricultural activities or transported dust) and is lowest in winter when the ground is frozen and predominately snow covered.
Figure 14. Ross Creek PMF factor seasonal variation
Table 6 displays the summary statistics for each of the six factors identified. The ammonium nitrate factor ‘nitrate’ is highest in mean concentration over the entire study period at 1.51 \(\mu\)g/m3, but its median concentration is only 0.30 \(\mu\)g/m3. Concentrations of ammonium nitrate are highly variable seasonally, with very low contributions during summer months. Smoke is the second highest average factor, with higher max and median concentrations compared to nitrate but a slightly lower mean at 1.42 \(\mu\)g/m3. The relatively higher median concentration indicates that the smoke factor, combined with the seasonal patterns in Figure 14 suggest that wintertime smoke, possibly from local/regional residential fireplaces, may be an important contributor. Prescribed agricultural burns in the winter also potentially contribute.
Sulphate and OC1 & metals factors are comparable in mass contribution at ~1.21 \(\mu\)g/m3. Sulphate contributions are consistent seasonally, whereas the OC1 factor is highest in the summer months. The salt & metals factor mean concentration of 0.96 \(\mu\)g/m3 was highest in the winter months. The salt factor contribution is a factor of 10 higher than the reconstructed fine particulate mass estimate of salt in Section 3.3. in both cases the concentrations are highest in the winter months. Finally, the crustal matter factor is the smallest at 0.58 \(\mu\)g/m3; it has highest concentrations in the spring and summer and very low concentrations in the winter.
Figure 15a is a polar plot of contributions for each of the factors. Dots indicate the frequency and magnitude of winds on sample days. Colors indicate the likelihood that absolute contributions from that wind direction and factor are high or low; warmer (orange/red) colors indicate high concentrations that originate from a given direction, while greener and bluer colors indicate low concentrations for a factor on average from that sector. Figure 15a illustrates that the sulphate factor has very high contributions associated with winds from the northeast – this is consistent with emissions from Alberta Industrial Heartland area. In contrast, most other factors have a less dramatic association with winds from any given sector. Figure 15b shows conditional probability factor results at the 75th percentile. This plot allows a user to specify a threshold or range of contributions that are of high likelihood, in this case, 75th percentile. Figure 15b confirms that sulphate contributions are high from the northeast. It also shows that the Salt & metals contribution is highest from the southwest towards Fort Saskatchewan and Edmonton. This could be associated with de-icing operations on roads in and around the city of Fort Saskatchewan.
Figure 15a. Polar plot of PMF factors using absolute concentrations
Figure 15b. CPF plots at the 75th percentile (1.6 ug/m3)
4 Discussion
PM2.5 concentrations at the Fort Saskatchewan Ross Creek site (7.7 \(\mu\)g/m3) were slightly higher than those at the Edmonton-McIntyre (6.7 \(\mu\)g/m3) site and much higher than comparison sites in Downtown Ottawa, Burnaby South, BC, and Teddy Roosevelt NP in North Dakota. Ross Creek mean and median PM2.5 concentrations were significantly lower the industrially influenced sites of Deer Park, Texas and Granite City, Illinois. Monthly comparisons of mean PM2.5 concentrations indicated that Ross Creek had higher relative concentration sites in the winter months, and lower relative concentrations in the summer months. The Ross Creek site is more suburban/rural than the Edmonton-McIntyre site but is downwind of the major Edmonton metropolitan area about 40% of the time and can be reasonably expected to be influenced by transport of regional urban emissions from metropolitan Edmonton towards the Fort Saskatchewan area. PM2.5 is often considered a regional-scale issue given the secondary formation of components like nitrate, sulphate, and OM, but the much lower residential density and transportation density in the Ross Creek area does not appear to yield lower concentrations than a central urban site like Edmonton-McIntyre. Six of the 36 months had mean PM2.5 concentrations greater than 10 \(\mu\)g/m3 at the Ross Creek site.
High event days for PM2.5 at Ross Creek were driven by three types of events. Wildfire smoke events were the most influential, driving concentrations well above the 24-hour AAAQO. These events were relatively rare and more likely to happen in the summer or early fall. Observed smoke events include August 6 and 18th of 2018 and September 18, 2019. Wintertime episodes with elevated nitrate concentrations were a secondary event, with concentrations of nitrate greater than 10 \(\mu\)g/m3 often accompanied by elevated sulphate concentrations (as shown in Figure 9). These events were episodic and can result in extremely high regional PM2.5 concentrations during stagnant meteorological conditions. High wintertime PM2.5 episodes during this monitoring study occurred in March, 2018, February and March 2019, and January of 2020. The third type of event observed were high sulphate events. The highest sulphate concentration days were March 21, 2018 and January 28, 2020, both occurred when winter time inversion was observed in the region.
Two key analyses were applied to assess the components of PM2.5: reconstructed fine mass (RCFM) and source apportionment using PMF. The RCFM calculations showed that the biggest contributions to total PM2.5 mass at Ross Creek were OM at 2.45 \(\mu\)g/m3, followed by sulphate (1.40 \(\mu\)g/m3), nitrate (1.15 \(\mu\)g/m3), crustal matter (0.45 \(\mu\)g/m3), EC (0.27 \(\mu\)g/m3), and salt (0.05 \(\mu\)g/m3). In the PMF analysis, the OM is split into two separate source factors; the smoke factor mean concentration was 1.42 \(\mu\)g/m3 while the OC1 & metals factor mean concentration was 1.23 \(\mu\)g/m3. The nitrate factor was the largest contributor at 1.51 \(\mu\)g/m3 and the sulphate factor was 1.21 \(\mu\)g/m3. The salt & metals factor average concentration was 0.96 \(\mu\)g/m3, and is better interpreted as a winter road resuspended particulate factor rather than purely de-icing salt. Finally, the crustal matter factor was the smallest mean contributor at 0.58 \(\mu\)g/m3.
The most interesting local feature at the Ross Creek site evident in the meteorological comparison of source factors was the very clear ammonium sulphate signal that originated from the sector to the northeast. This is consistent with transport of photo-processed emissions from the petrochemical facilities of the Alberta Industrial Heartland. Overall, those concentrations were about 1 to 1.4 \(\mu\)g/m3 on average and constitute between 13-18% of the total PM2.5 mass at the site. Winds were less frequent from the northeast (~10-15% of the time) and therefore the influence of this factor was somewhat muted.
Wintertime nitrate and summertime OM factors do not appear to be particularly strongly associated with local wind directions or wind speeds. Given the similarity of the concentrations of these factors with those observed at Edmonton-McIntyre, it is likely that these are regional factors associated with the Edmonton Metropolitan area and it may not be possible to directly manage these within the Fort Air Partnership region. Similarly, wildfire smoke is most often transported from outside the region and is unlikely to be due to local fires within the Fort Air Partnership boundaries.
The two smaller factors contributing to PM mass may plausibly be due to sources that originate locally within the Fort Air Partnership boundaries. The salt & metals factor was elevated when winds originated from the direction of Fort Saskatchewan and their high concentrations in the winter suggest a possible link with de-icing applications. Crustal matter concentrations were highest in the spring months and may be associated with local agricultural operations generating dust. However, crustal matter was the smallest contributor to total PM2.5 mass and its total contribution was about 5-10% of the total PM2.5 mass on an annual basis.
5 Acknowledgments
The Fort Air Partnership would like to acknowledge the authors of this report, Michael McCarthy, PhD at Radical Research LLC and Sean Raffuse at the University of California at Davis, Air Quality Research Center.
FAP would also like to thank the following project contributors:
- Yayne-abeba Aklilu, PhD at Alberta Environment and Parks for providing a critical review to enhance this work.
- WSP field technicians for collecting the PM samples and operating the Ross Creek Station.
- Desert Research Institute for analyzing the PM samples and providing the data to FAP.