Monitoring-site mix and PM2.5 trends in Los Angeles County (2015–2024)

Research question

Does the apparent county trend change when the same monitoring sites are followed throughout?

This case uses public EPA monitoring observations to examine how changing monitoring-site membership affects the apparent 2015–2024 PM2.5 trend in Los Angeles County. It compares the observed network with six sites observed in every month, and examines seasonality, between-site differences, sampling coverage and alternative measurement networks. It is descriptive research, not a regulatory compliance assessment, causal estimate or population-weighted exposure analysis.

Data and comparison design

The comparison uses an annual observed-network average with an average over the six fixed sites. The first can change because site values change or because sites enter and leave; the second holds site membership fixed. Endpoint and multi-year-window comparisons also ask different questions.

Recorded findings

Temporal summary	Observed network	Six fixed sites
2015–2024 endpoint change	−0.15 µg/m³	+0.81 µg/m³
2015–2017 versus 2022–2024	Decline	Decline
Meaning	Site values and site membership can both change	Membership held fixed
Estimand	Equal-site monitoring summary	Not population exposure or attainment

This case uses public EPA monitoring observations to examine how changing monitoring-site membership affects the apparent 2015–2024 PM2.5 trend in Los Angeles County. It compares the observed network with six sites observed in every month, and examines seasonality, between-site differences, sampling coverage and alternative measurement networks. It is descriptive research, not a regulatory compliance assessment, causal estimate or population-weighted exposure analysis.

The primary analysis uses parameter 88101 and selects one monitor per site-year. It calculates site-month means, gives sites equal weight within each month, then gives months equal weight within each year. Under this estimator, the observed-network mean changes from 10.952 to 10.802 µg/m³ (−0.149), while the fixed six-site mean changes from 11.254 to 12.064 µg/m³ (+0.811). The endpoint-change contrast is approximately 0.960 µg/m³. Comparing 2015–2017 with 2022–2024 instead gives declines of 0.591 and 0.239 µg/m³, respectively. Site mix therefore affects apparent improvement, but the non-monotonic series does not establish a steady countywide increase or decline.

Missing-day and coverage assumptions affect the magnitude. The separate 88502 network uses different sites and measurement methods and is not pooled with 88101. Cadence-based coverage is an analyst-defined proxy, not verified scheduled-sample completeness. There is no imputation, meteorological adjustment, formal significance test or causal identification. Full findings, sensitivity results and interpretation appear in the saved Session response and bundled tables.

What the comparison shows

The opposite endpoint changes show that monitoring-network composition matters. Declines under the multi-year comparison further show why the chosen temporal summary must accompany any trend claim.

Assumptions and limits

These equal-site averages describe monitoring results, not population exposure or regulatory compliance. Comparing 2015–2017 with 2022–2024 gave declines in both networks; coverage assumptions and the comparison window affect the conclusion.

Sources and research materials

The unchanged research package comes from upstream Case 10. It includes the archived session, analysis materials, results and figures. The poster uses an AI-generated illustrative photograph. This introduction summarizes archived findings; the analysis was not rerun.