01 Refine your plan for a first source apportionment analysis

After collecting an archive of manual filter samplers, it is time to analyze the archived filters and conduct a receptor modeling analysis. Building on Stage 1 guidance (Source Attribution, Steps 2 and 6 and the key category analysis from the Emissions Inventory Stage 1, Step 8) you can develop a conceptual model for the regional sources of air pollution that affect your jurisdiction. Engaging academic institutions can significantly enhance the quality of the source apportionment study. Their expertise in advanced analytical techniques, such as receptor modeling and chemical speciation analysis, ensures accurate data development and interpretation. Collaborating with academics can also provide access to specialized equipment and sophisticated data analysis software, thereby reducing costs and resource burden for your agency.  These institutions may have research grants or funding opportunities that can be leveraged to support the study financially and professors with long academic careers can provide institutional memory, which can be a challenge for government agencies that experience a higher rate of staff turnover. Overall, incorporating academic expertise in planning and execution maximizes the impact of source apportionment efforts and strengthens air quality management planning. Examples such as the USEPA-funded IMPROVE program, which is hosted by Colorado State University, or ACE centers, such as the Carnegie Mellon University example listed or the EU-funded EMME-CARE program  (each linked below) highlight this kind of academic partnership that help to provide an understanding of regional sources of air pollution. . 

02 Train staff

Start by establishing clear objectives for the training program and clear roles for staff who will conduct the analysis (e.g. chemical speciation versus mathematical receptor modeling versus trajectory analysis to identify geographic source regions). Training may take several forms, but can include workshops, seminars, and online courses led by experienced professionals or academics in the field. Incorporate hands-on training sessions where staff can practice using software tools for modeling and data analysis.  

The curriculum should cover the fundamental concepts of receptor modeling, including various methodologies such as Chemical Mass Balance (CMB), Positive Matrix Factorization (PMF), and/or Unmix. Training should also include detailed training with the specific statistical software being used for the receptor modeling exercise.  Use of real-world examples and practical exercises on actual datasets can help to recognize patterns that are typical of ambient air pollution.   

03 Develop a chemical speciation database

Collating and formatting spectated air quality data for input into Positive Matrix Factorization (PMF) or Unmix analysis requires careful preparation to ensure the data is suitable for these statistical modeling techniques. Ensure that air quality samples are collected consistently across designated monitoring stations with specified time intervals (e.g., every 3rd day or every 6th day, etc.). Include all relevant species, such as all the measured components of particulate matter (PM2.5, PM10), metal species, ions, total volatile organic compounds (VOCs), and other key gaseous pollutants that may influence air quality. In general, more species can lead to more accurate identification but adds to the analytical burden. See Srivastava et al for a recent Source Apportionment analysis of Beijing that includes a wide variety of measured species in their analysis.

After the data has been collected, organize your data in a matrix format where rows represent sampling events and columns represent the measured pollutants/species. This structure supports the mathematical requirements of PMF and Unmix. Address any missing data points. If specific values are not available, use appropriate methods (like interpolation) or replace them with values derived from surrounding data, ensuring transparency in reporting. Ensure that all data is in consistent units (e.g., µg/m³ for particulate matter, ppb for gases).  

Perform quality control checks to validate the integrity and accuracy of the data. Remove any outliers or erroneous data entries that could skew the results. Add flags to indicate data quality (e.g., good, questionable, poor) to provide context during analysis, as well as metadata that provide context such as site identifiers, dates/times, sampler type or calibration details.  

Finally, export the dataset in a format compatible with PMF or Unmix software (e.g., CSV or Excel). Ensure that header rows accurately describe each data column. 

04 Develop a source profile database

To prepare a locally specific database of source profiles for air quality receptor modeling, start by reviewing globally available datasets like AP-42 and EU datasets to identify relevant source profiles for your region. Customize the profiles by collecting local emissions data, including industrial activities, traffic patterns, and specific land uses. While it may be premature to conduct your own field sampling to capture local PM chemical compositions, ensure that you cover various important conditions and seasons by checking with local academics or authorities who may have measured emissions from important sources. Normalize data that you may add into a database to ensure consistency, and categorize profiles based on source types (e.g., traffic, industrial, natural sources). Finally, document metadata for transparency, including methodology, sampling dates, and location specifics, ensuring usability in receptor modeling studies. 

05 Explore global tools and reduced-form chemical transport models for your jurisdiction

At Stage 2, the focus is on performing receptor modeling, but it is not too early to explore reduced form or simplified chemical transport modeling as another effective approach to understanding key sources of air pollution. Tools like GEOS-CF and MERRA2 reanalysis offer global datasets that provide atmospheric conditions and historical pollution data, facilitating a clearer picture of source contributions.  

Consider the CAMS (Copernicus Atmosphere Monitoring Service) air quality forecasts to gain insights into temporal variations in pollutant concentrations and identify trends linked to specific sources. For localized assessments, tools like SIM-Air can simulate traffic-related emissions and their impact on air quality, while Global InMAP (Integrated Modeling and Assessment Platform) allows for risk assessments and source apportionment by integrating emissions and dispersion models.

When using these models, ensure a coherent understanding of input data, known uncertainties, regional characteristics, and validation with local measurements. Continuous calibration and comparison with observational data enhance model reliability, ultimately supporting informed air quality management and regulatory decisions.  

06 Conduct receptor modeling

Run the chosen receptor model, which will allocate pollutant contributions to various sources based on their chemical signatures in your database. The European Guide on Air Pollution Source Apportionment with receptor models has guidance on running this type of model in Chapters 8-11. After running the receptor model, interpret the results by examining the identified sources and their contributions to pollutant levels. Focus on the distribution of contributions across different source categories, such as traffic, industrial emissions, and natural sources.  

Contextualize these findings by comparing them with local activities, such as traffic counts, industrial operations, or seasonal agricultural practices, which may influence the identified emissions.  

To validate the modeling results, conduct a comparative analysis using independent datasets, such as additional monitoring stations or meteorological data. This cross-validation can help assess the robustness of the model outputs. Chapter 5 of the European Guide on Air Pollution Source Apportionment has guidance on some of these QA and validation steps. 

07 Conduct trajectory analysis

Assemble a database of back-trajectories for understanding the geographic locations of the identified sources of air pollution.  Begin by utilizing meteorological models such as the HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model or the FLEXPART model. These tools allow you to calculate the path that an air parcel has followed based on atmospheric and meteorological conditions.

Start with the relevant monitoring sites for your air quality data and determine the timeframes of interest. Generate back-trajectories for various durations (e.g., 72 hours) and multiple start heights (e.g. 100, 500 and 1000 m above ground-level) leading up to each sampling event, focusing on the pollutants included in your receptor modeling.

Store trajectory data in a database, including detailed metadata such as start and end dates, trajectory duration, and atmospheric conditions at different points. Integrate trajectory data with geographic information systems (GIS) to visualize source regions, overlaying this with potential emission sources like industries, traffic corridors, and natural features.  Analysis of the trajectory patterns can identify likely geographic contributors to the observed air pollution, enhancing the interpretation of your receptor modeling results. See Annex E of the CSPAR technical support document or Chapter 10 of The Open Air Book for  examples. 

08  Identify source profiles for improvement

Once you have completed a source apportionment analysis, closely examine the identified pollution sources and their contributions to overall air quality and check for consistency with conceptual model. Identify sources with significant contributions that lack robust or locally specific source profiles, as these represent opportunities for improvement. Refer back to your key category analysis from the Emissions Inventory Stage 1, Step 8 to understand if these findings are consistent with your inventory work. If so, you may need to ensure that the source profiles being used are accurate and locally representative. 

09 Improve the analysis and underlying data

Engage relevant stakeholders, such as regulatory agencies, community groups, and industries, to discuss results and implications for air quality management.  Source apportionment is an ongoing activity.  Identify the suspected greatest uncertainties. Identify the key action steps that will enhance the results next time around.  How can you improve the analysis with more precise chemical speciation, additional species, additional sites or more localized source profiles?  Additionally, consider conducting field studies or targeted sampling to capture emissions directly linked to identified sources.