6.6 Nonpoint Quality Assurance

For the nonpoint data category, the dedicated quality assurance (QA) team focused on six key aspects of QA for nonpoint data submissions listed in Table 6.7.

Table 6.7: Key Nonpoint QA issues, causes, and steps taken to address issues.
Type of QA Issue  Causes  Steps taken by QA Team to Address
Impossible Sums. Examples include HAPVOC > VOC, PM PRI ≠ PM-CON + PM-FIL, PM10 < PM2.5 (1)  Emission factors are inconsistent with each other. (2) Incomplete suite of HAPs is provided, and when incongruent datasets (EPA and SLT) are added, they add up to more than VOC. (3) AP aug itself is generating impossible sums (some oil/gas profiles slightly violate this QA check). (1)  Checked HAP augmentation to ensure impossible sums were not generated. (2) Ran iterative QA report on SLT submissions during the window opening to find these errors. Reported back to SLTs. (3) Ran a script on the final selection to check these sums.
Unexpected Pollutants or Missing Pollutants. Reasons include them not on expected pollutants list and EPA will backfill with tool, not on expected pollutants list and EPA can augment HAPs, not on EPA and EPA will not backfill, in EPA datasets but not in SLT datasets. SLT submitted to wrong SCC, or applied an incorrect emission factor, or has additional information that EPA lacks (1)  Created an expected pollutants list for comparison. (2) Ran the iterative QA report on SLT submissions during the window opening to find these errors and report back to SLTs. (3) Ran a script on the final selection to check for unexpected or missing pollutants.
Missing Data, such as missing county or missing SCC.  SLT inadvertently left out data (1)  Ran an iterative report and provided feedback to SLTs when counties or SCCs appeared to be missing. (2) Ran a script on the final selection to check for missing data.
Outliers, such as values too high or too low.
  1. SLT gives data that is outside of acceptable QA limits. (2) EPA estimates are outside of acceptable QA limits. (3) Possibly wrong data units of measure?
Created rankings on a state SCC pollutant basis and compared. Also looked at magnitudes, mainly looking at large orders of magnitudes of difference.
Zeroes clog up our Emissions inventory system, and should only be included if the Nonpoint Survey doesn’t cover the SCC While zeroes are not always a problem, sometimes their submittal changes the way our option group/option set selection process works, so these should be submitted with caution. See Section 6.2.4.3. Reviewed the option group/option set to see if zeroes blocked out data from coming in, inadvertently
Using Old Data SLT submits data using old WW or 2017 default data. Checked the “emissions comment” field for references to old tools or data.

The following subsections discuss how these QA checks were analyzed and identified issues were resolved prior to finalizing the final nonpoint selection for the 2023 NEI.

6.6.1 The Iterative QA Report

Introduced for the 2020 NEI cycle, iterative QA reports, performed on S/L/T-submitted emissions, were not sent out during the 2023 development process due to staffing constraints. Instead, we ran periodic EIS reports on S/L/T-submitted emissions and contacted submitters as we identified possible quality assurance issues such as unexpected or missing pollutants and outliers. EPA hopes to resume these reports for the 2026 NEI.

6.6.2 Expected Pollutants List

To determine whether S/L/T submissions were correct, EPA maintains a list of accurate expected pollutants list, or EPL. The NEI Team updated the existing list of EPA SCCs and EPA non-estimated SCCs that are often submitted by SLTs, and the corresponding pollutants that EPA expects to be emitted from each process. The purpose of the list is twofold: first, to guide the SLTs in providing their submissions for the NEI, and second, to cull out any pollutants that do not belong in the NEI. The expected pollutants list helps everyone to understand what each SCC is supposed to represent, as far as the suite of pollutants, and ultimately leads to a more consistent and cohesive NEI.

New for 2023, this list was added to the Emissions QA Threshold Values Table in the EIS Gateway for access by S/L/Ts. The nonpoint EPL used for 2023 NEI “NonPointSCCs_ExpectedPollutantsList_2023NEI.csv” is available on the “2023 NEI Supplemental Nonpoint data FTP site”. Note that this EPL includes pollutants that EPA does not have the emission factors or methodology to estimate itself. If the “EPA Estimate Category” field is populated, it means there is an existing EPA tool for this SCC, and the EPA tool name is given.

6.6.3 Completeness Reports

A preview of completeness was sent to the S/L/Ts ahead of time, with time for them to correct mistakes and incomplete submissions. In early May 2025 we sent a final report to Air Directors on EPA letterhead. We do not include an example of the report as several S/Ls engaged with EPA after the reports were sent to resolve QA issues prior to finalizing the 2023 nonpoint NEI selection. Additionally, we provided a draft review via the NEI SharePoint site in mid-March 2026 with maps, reports, survey information, and tag out reports alerting S/L/Ts to a final opportunity to review the 2023 NEI while still in draft form.

For agencies that did not submit a complete Nonpoint Survey, we locked in the default response “Yes – Supplement my data with EPA estimates” to ensure EPA data would backfill any potential gaps in data submitted by the agency.

6.6.4 EPA-estimated emissions QA

EPA requires all data inventory developers, including contractors, to be responsible for reviewing any emissions data they provide, as well as keeping track of and reviewing the Input Templates that they upload into the Wagon Wheel.

Upon providing EPA-generated estimates, each contractor provided a spreadsheet of QA checks they performed on the data, as well as keeping a tally of and reviewing the Input Templates that were uploaded into the Wagon Wheel. We provide a QA Contractor Checklist “QA checklist for contractors.docx”, available on the “2020 NEI Supplemental Nonpoint data FTP site” that outlines all of the QA a contractor must perform when providing emissions data to EPA via a tool.

6.6.5 Input Template Review

Input template review is the responsibility of the contractor and was performed on a rolling basis (i.e., as they were submitted to the EIS Gateway). While Input Templates weren’t incorporated formally until after the submission window closed, getting back to the S/Ls in a timely fashion ensured that mistakes were caught early in the process. Some error checking was implemented via feedback reports in EIS upon upload by S/Ls. More in-depth review was performed by the contractor once templates were downloaded from EIS for input in the Wagon Wheel tool.

6.6.6 Reviewing S/L/T data after the EIS submittal window has closed

After the EIS nonpoint data category submittal window closed, we checked the S/L/T-submitted emissions data for four main categories.

  1. Completeness

    1. Tag out unexpected pollutants. We’ve already given them a heads up during the window opening with the iterative QA reports.

    2. Tag out incomplete HAPs. We also tagged out process records if CAPS were incomplete; for example, missing NH3 from agricultural livestock waste.

  2. Old data/methods

    1. The Wagon Wheel emissions comment field includes the version of the tool; we tagged out data for sources where old tools used noted and activity data had known updates in the latest version of the tool. We also reviewed and tagged out data significantly different from EPA estimates or previous S/L/T submittals where the comment field indicated “engineering judgement” and no other supporting documentation was provided.

  3. Check Nonpoint Survey Responses

    1. If a S/L chose “No -do not supplement” but their submittal had missing CAPs. There should not be missing CAPs, and this would have also been caught as incomplete on Completeness Reports.

    2. Any tagged out S/L emissions data required a “YES” on their Nonpoint Survey; sometimes we had to tag out a S/L Nonpoint Survey response (from “No”) to ensure the NEI would capture EPA estimates when S/L data were tagged out.

    3. We asked states to update the Nonpoint Survey answer themselves if not time-limited; otherwise, EPA tagged out their survey responses in these cases.

  4. Percent change from previous NEIs

    1. This does not work for new or changed SCCs, or for some county changes (e.g., changes for Connecticut in 2023)

    2. Evaluated the minimum, maximum, and mean values from the last 3 NEIs (2014, 2017, 2020) – compared to the 2023 submitted value. We looked more deeply at 2023 values outside the min/max/20% from mean

    3. Graphed 2020 vs 2023 for the values that got flagged.

Each NEI Nonpoint “sector lead” reviewed QA team findings and reported back for team discussion on follow up and reconciliation.

6.6.7 Data Tagging Summary

6.6.7.1 S/L/T emissions tagging

We tagged out 2023 SLT nonpoint emissions for various reasons, including but not limited to the following observations:

• Submittal of VOC HAPs that in sum, exceed submitted VOC
• Submittal of HAP metals that in sum, exceed submitted PM25-PRI
• VOC HAPs submitted with no corresponding VOC, or HAP metals submitted with no corresponding PM (exception for non-combustion mercury sources)
• Apparent submittal of filterable PM as primary PM component.
• Apparent unit of measure issue when comparing to EPA values, or ratio of HAP to associated CAP and EIS HAP Augmentation multiplication factors; for example, benzene being < 0.01% of evaporative VOC
• Double count with point inventory submittal; for example, similar emissions submitted for point and nonpoint railyards
• SLT request with or without EPA solicitation of an identified QA issue
• Submittal to wrong SCCs
• Unexpected pollutants such as metals in commercial cooking, livestock waste, VOC in road dust, mercury in composting, etc.
• GHGs submitted for stationary nonpoint sources
• All S/L/T-submitted biogenic emissions; we prefer to use the consistent and complete dataset from the BEIS model.

In most cases not involving mass balance (e.g., VOC HAPs > VOC), unexpected pollutants, or obvious errors, we collaborated with SLTs on the observed issue and a recommended course of action. In most cases, SLTs agreed with these recommendations and tags were created. A complete list of all tags applied to the 2023 SLT nonpoint emissions submittals is available in the workbook “2023NEI_SLT_Nonpoint_emissions_Tags_21may2026.xlsx” on the “2023 NEI Supporting Data and Summaries” site.

6.6.7.2 S/L Input Template review and Nonpoint Survey tagging

We compared SLT Input Template activity data submittals with EPA default activity and reached out to agencies where we saw significant outliers. In most cases, SLTs were able to resolve the conflict and provide either updated activity data or removed their template to accept EPA default data.

In addition to tagging of emissions, we also tagged Nonpoint Survey responses -reverting the Nonpoint Survey to “Yes -Supplement my data with EPA data”- for select source categories at several S/Ls. The reasons for tagging these Nonpoint Survey sources are provided in Table 6.8 but they often correspond to identified issues with SLT-submitted emissions. In cases where SLTs submitted emissions, they often selected “No” in the Nonpoint Survey, so we sometimes needed to also tag out the Nonpoint Survey to allow EPA estimates to supplement their now nonexistent (tagged out) emissions.

There were also several scenarios where a S/L agency submitted an Input Template but also selected “No” in the Nonpoint Survey (nor submitted emissions). S/L Input template activity data submittals are loaded into the EPA Wagon Wheel tool, and the estimates generated are therefore considered “EPA”. This was the primary reason for updating the Nonpoint Survey to include a checkbox for SLTs to indicate whether they submitted an Input Template for the category. Regardless of that check box status, if we discovered an S/L input template and a Nonpoint Survey response of “No’, we reconciled this inconsistency in QA by tagging out the Nonpoint Survey response to allow the S/L activity data-based estimates to make it into the NEI.

Table 6.8: S/L Nonpoint Survey responses tagged with rationale provided.
Agency EPA Tool Estimate Category Reason for Tagging
California Air Resources Board Biogenics Estimates Using BEIS for entire country, only 1 agency (CARB) submitted direct emissions
Memphis and Shelby County Health Department - Pollution Control Gasoline Distribution Tool; Pipeline Gasoline Requested that EPA estimates be used for pipeline emissions.
Minnesota Pollution Control Agency Residential Wood Combustion Tool; all RWC SCCs State ended up not submitting emissions, so by tagging out “Supplement Only At Reported Location – SCCs”, ensured that EPA estimates with MN-supplied input template data, were used.
Minnesota Pollution Control Agency Campfires Tool submitted activity data via input template, but said “No - Do Not Supplement My Data”
New Hampshire Department of Environmental Services ICI Fuel Combustion Tool; 14 out of 18 SCCs submitted activity data via input template, but said “No - Do Not Supplement My Data”
New Hampshire Department of Environmental Services Road Dust Tool; Paved Road Dust submitted activity data via input template, but said “No - Do Not Supplement My Data”
New Hampshire Department of Environmental Services Motor Vehicle and Structure Fires Tool; 2 out of 2 SCCs submitted activity data via input template, but said “No - Do Not Supplement My Data”
Wyoming Department of Environmental Quality Oil and Gas Tool; 15 out of 76 SCCs submitted activity data via input template, but said “No - Do Not Supplement My Data”

6.6.8 Final Review of EPA-generated tool data

  1. Completeness – confirm everything made it into EIS

    1. Review contractor QA checklists

    2. Compare to expected list of EPA Tool SCCs (pulled from Nonpoint Survey and tools into the Expected Pollutants List)

    3. Compare tool pollutant outputs to Expected Pollutants List (run through iterative QA)

    4. Inform contractor of any missing SCCs or pollutants

  2. Check input template submission compared to Nonpoint Survey “Y/N” checkbox

    1. Review from Contractor QA Checklist

    2. Nonpoint Survey detailed report comparison to list of input templates

6.6.9 Final Nonpoint Selection review

A final review of the nonpoint data category includes:

  1. Confirming tagged out data did not make it into the NEI selection

    1. Confirmed “Exclude tagged values” set to “yes” in EIS selection

    2. Run EIS tagging report to confirm all submitted tagged records are included

  2. Pollutant Completeness

    1. Check against Expected Pollutants List

    2. Find explanation for causes of any missing data

    3. Have rationale for any remaining “unexpected” pollutants

  3. SCC/Sector Completeness

    1. Not every county should have a value for every SCC/Sector, but ensure there is an explanation

    2. Compare to previous (2020) NEI, look for:

      1. Whether S/L/T submitted in past vs current NEI: is it SLT vs SLT, EPA vs SLT for example

      2. County or SCC changes since last NEI

  4. Non-EPA SCCs in NEI

    1. What SCCs are S/L/Ts reporting that are not in EPA tools/estimates? We tagged some of these out if they were not reported anywhere else (e.g., human perspiration, motor vehicle fires) to avoid some inconsistencies across states.

    2. Is there any potential overlap with SCCs not included in the Option Group/Option Set assignments (possible double-counting issue)?

    3. Does OGOS unintentionally drop S/L/T emissions?

  5. Accuracy

    1. Final magnitude check comparing relative rankings at sector and state level to previous NEI. Did the relative ranking change significantly for a given state/sector?

    2. Review where there are zero emissions for entire agency/SCC

  6. HAPs

    1. Compare to expected pollutants list

    2. Ensure backfilled via augmentation

    3. Check HAP-VOC vs VOC and correct/tag if necessary.

  7. Ensure all data exclusion rules properly reflected

    1. Nonpoint Survey: SLT data supersedes EPA data where appropriate

    2. Pollutant Groupings: each county/SCC should only have one group level between different datasets

    3. OGOS: review selection SCCs to confirm correct application

  8. PM speciation mass balance: sum of PM species equals PM2.5