Rank pollution controls by how much exposure they cut.

Model exposure between the monitors

Where monitors are sparse, officials see pollution at a few points and estimate the rest. The UN's 2026 SDG report says the countries most vulnerable to water-quality damage often have the least capacity to measure it, and calls for Earth observation and low-cost sensors to help. The same report puts the population-weighted average level of fine particles, known as PM2.5, in cities at 26 micrograms per cubic meter in 2023. The WHO guideline is 5.

Combine every source of evidence

No single data source shows the whole picture, so the model would combine them. It would cover emissions to air, discharges to water, how pollution moves and mixes, and where people are exposed.

• Air monitor readings gathered through OpenAQ's open API
• Satellite measurements of gases such as nitrogen dioxide and sulfur dioxide from Sentinel-5P
• Traffic, industrial permit and household fuel data, to estimate emissions by source
• River and lake monitoring results and wastewater discharge records
• Soil sampling results and land use history for sites suspected of contamination
• Weather and rainfall data, which drive how pollution spreads

Judge the model on monitors it was not fitted to

We would build an inventory of emissions by source, then run a chemistry and transport model such as CMAQ to estimate concentrations across the area hour by hour. The model is checked against monitors it was not fitted to, and against satellite readings. Water quality would be modeled with SWMM for runoff and sewers and with EPANET for drinking water pipes. For soil, the work would map sampling results against land use history to suggest where to test next. The 2021 FAO and UNEP global assessment of soil pollution covers point-source and diffuse pollution and traces most contaminants to human activity.

Rank controls by exposure avoided per unit of cost

The test for a control is how much exposure it avoids for the money. Environment agencies and city health teams would get evidence they can act on and defend.

• Maps of estimated concentrations between monitors, with uncertainty shown
• The estimated share of pollution at each place that comes from each source type
• Controls ranked by exposure avoided per unit of cost, such as low-emission zones, cleaner household fuels or industrial upgrades
• Suggested sites for new monitors where they would reduce uncertainty most
• A plain-language summary of each scenario, reviewed by a scientist

Check every estimate before it names anyone

An estimate about a named facility goes to the agency for checking before anyone acts on it. Every published estimate states its uncertainty and data date. Exposure results are shown by neighborhood and income group, so the plan shows whether controls reach the most exposed areas. Officials decide on any enforcement, and every run is logged.

Use models that agencies can rerun

An agency should be able to rerun the model behind any control it funds. Each tool below is open source or open data.

• OpenAQ, a nonprofit platform that gathers air quality data from hundreds of sources and shares it through an open API
• Sentinel-5P, the first Copernicus mission dedicated to monitoring the atmosphere, with data through the Copernicus Data Space Ecosystem
• CMAQ, the US EPA's Community Multiscale Air Quality model for ozone, particulates and other pollutants
• SWMM, the EPA's open model for runoff quantity and quality, including pollutant wash-off
• EPANET, for water quality in pipe networks
• QGIS, for mapping samples, sources and exposure