Industries / Energy, water and waste / Environmental services
Monitoring air, water and soil
Monitoring measures what is in the air, water and soil, using fixed stations, instruments in smokestacks, low-cost sensors and satellites, so that regulators and the public know what people are exposed to.
7
Milestones achieved, from 1956 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
In depth
The same ground is charted in more depth, as an industry of its own.
The road
AchievedWatch forOpen
Achieved
1956
Clean air after killer smog
Smog covered London from 5 to 9 December 1952, and the Met Office says about 4,000 people were known to have died, possibly many more. The Clean Air Act 1956 banned black smoke and required homes and factories in cities to switch to smokeless fuels.
LessonFogs stayed smoky for some time after the act, because people needed time to adapt.
Credit
UK Parliament; account by the Met Office
1995
Power plants measured hourly
A continuous emissions monitoring system (CEMS) is a set of instruments in a smokestack that measures pollution all the time. EPA's Part 75 rule, first published in January 1993 for the Acid Rain Program that began in 1995, requires fossil-fuel power units over 25 megawatts to report sulfur dioxide, carbon dioxide, nitrogen oxides and heat input for every hour they run, and EPA publishes the data; it found that only 0.6% of operating hours from 2015 to 2019 had missing or invalid data.
LessonEPA calls monitoring "the backbone" of cap-and-trade, a system in which plants trade a fixed total of pollution allowances, because a ton must be measured the same way everywhere before it can be traded.
Credit
US Congress (Title IV of the 1990 Clean Air Act Amendments); EPA Clean Air Markets Division
2008
An embassy publishes its readings
In 2008 the US Embassy in Beijing began posting hourly readings of fine particles (PM2.5, particles 2.5 micrometers across or smaller) from its own monitor, at a time when official Chinese readings were based on the broader PM10 measure. In 2012 China's State Council required PM2.5 and ozone monitoring, starting that year in 4 municipalities, 27 provincial capitals and 3 key regions, and the government ordered 74 cities to publish PM2.5 results.
LessonA 2022 study by the economists Jha and La Nauze, in PNAS, estimated that US diplomatic monitors, now more than 40 worldwide, lowered fine-particle levels in their host cities by about 2 to 4 micrograms per cubic meter.
Credit
US Department of State; China's State Council and Ministry of Environmental Protection
2014
China publishes soil survey
On 17 April 2014 China's environment and land ministries published a national soil survey, run from 2005 to 2013 over about 6.3 million square kilometers, which China Water Risk calls the first nationwide survey. It found 16.1% of the soil sampled over national standards, including 19.4% of farmland, and more than 80% of the problem came from inorganic pollutants such as cadmium, mercury, arsenic, chromium and lead.
LessonThe ministry had earlier classified the results as a state secret.
Credit
Ministry of Environmental Protection; Ministry of Land and Resources
2017
Satellites map pollution daily
The European Space Agency's Sentinel-5P, launched on 13 October 2017 for the EU's Copernicus program, carries the TROPOMI spectrometer, which measures nitrogen dioxide, methane, ozone, carbon monoxide, sulfur dioxide and formaldehyde over almost the whole planet every day. In April 2023 NASA's TEMPO went up on the Intelsat 40e communications satellite; its team calls it the first space instrument to measure trace gases across North America every daylight hour, and NASA released its data in May 2024.
Credit
ESA and the European Commission (Copernicus); NASA (TEMPO)
2021
Low-cost sensors get corrected
Raw readings from PurpleAir, a brand of low-cost particle sensor, ran about 40% too high in most of the US. EPA scientists used nearly 12,000 day-long readings from sensors placed beside regulatory monitors in 16 states to build one US-wide correction, which cut the typical error from 8 to 3 micrograms per cubic meter and is applied on EPA's AirNow Fire and Smoke Map.
LessonAdding a term for humidity cut the error further than a plain straight-line adjustment.
Credit
Karoline Barkjohn, Brett Gantt and Andrea Clements, EPA Office of Research and Development
2025
Methane satellite goes dark
MethaneSAT was launched in March 2024 by the Environmental Defense Fund (EDF) to measure methane from oil and gas regions worldwide. It lost contact on 20 June 2025, and on 1 July EDF said it had lost power and was likely not recoverable; E&E News reported that it cost $88 million and was the first satellite developed by an environmental nonprofit.
LessonA one-satellite mission can end without warning, and EDF said it would keep processing the year of data already received.
Credit
Environmental Defense Fund
Watch for
Hourly air maps over Europe
Sentinel-4 rides on EUMETSAT's MTG-S1 satellite, launched on 1 July 2025 into an orbit that keeps it over Europe. On 8 October 2025 it took its first images of nitrogen dioxide, sulfur dioxide and ozone over Europe and North Africa, and ESA's mission scientist said it "is the first European mission that will deliver hourly air-quality observations." ESA called the images preliminary and the instrument still in commissioning.
Achieved whenESA and EUMETSAT declare it operational and deliver validated hourly data to users and to the Copernicus Atmosphere Monitoring Service (CAMS), Europe's air-quality forecast service.
Claimed by
ESA, October 2025
AI air-quality forecasts
Microsoft Research's Aurora is a large AI model trained on weather and atmosphere data. Its authors reported in Nature that its five-day pollution forecasts matched or beat CAMS on 74% of targets at about 100,000 times lower computing cost, scored against CAMS's own analysis rather than ground monitors. In August 2026 an independent team from Guangdong Ocean University and the Chinese Academy of Sciences tested it against 3,540 ground stations for 2025 and found it competitive, but with higher overall error than CAMS and a tendency to miss high-pollution peaks.
Achieved whenan operational forecaster uses it routinely, or independent tests against ground monitors show it at least as accurate as CAMS.
Claimed by
Microsoft Research and co-authors, May 2025
Open
01
A scorecard for water sensors
A standing public test of low-cost water-quality sensors and test kits (pH, conductivity, turbidity, dissolved oxygen, nitrate and phosphate), each run to a published method against standard solutions and accredited lab results, the way a California air district already tests air sensors.
The briefCredit · unclaimed02
Open up official air data
A machine-readable public feed of one country's official air-quality readings (station, location, pollutant, concentration and time), built from what the government posts only as web pages, images or PDF bulletins, and offered to OpenAQ, the open air-quality database.
The briefCredit · unclaimed
Open 01 · the brief
A scorecard for water sensors
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A standing public test of low-cost water-quality sensors and test kits (pH, conductivity, turbidity, dissolved oxygen, nitrate and phosphate), each run to a published method against standard solutions and accredited lab results, the way a California air district already tests air sensors.
Why it is still open
Researchers testing Arduino-based river sensors wrote in 2024 that "no best-practice guideline for the calibration and validation of low-cost sensor networks exists," and that no truly low-cost nitrogen or phosphorus sensors existed. A 2025 study on the River Wye found that citizen nutrient kits matched lab results between 14% and 62% of the time, depending on the kit. South Coast AQMD's AQ-SPEC center, set up in 2014, tests air sensors only.
Built with what exists
Arduino or Raspberry Pi boards; off-the-shelf probes such as the DFRobot pH, conductivity, turbidity and dissolved-oxygen sensors used in the 2024 study; certified standard solutions; an accredited lab for paired samples; Python and R.
First step
Buy two units each of three DFRobot probes and repeat the 2024 study's one-week bench test with standard solutions.
Done looks like
At least six sensor or kit models tested for repeatability and accuracy by the 2024 study's ISO 21748 method, each with at least 30 river samples paired with lab results. The method, raw data and results are published, and a second team repeats one test and gets results within the stated uncertainty.
Open 02 · the brief
Open up official air data
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A machine-readable public feed of one country's official air-quality readings (station, location, pollutant, concentration and time), built from what the government posts only as web pages, images or PDF bulletins, and offered to OpenAQ, the open air-quality database.
Why it is still open
OpenAQ's 2026 assessment of 198 countries found that 33% share air-quality data fully openly, 28% only partly and 7% not at all, and that another 33% have no government monitoring. Some countries publish only an index, not the concentrations behind it.
Built with what exists
Python; the OpenAQ platform; a model built for extracting from long documents, such as Muse Ember, for PDF bulletins (see the Note "The jobs agents repeat"); a scheduled job on a small server.
First step
Pick a country that posts concentrations only as pages or PDFs, using OpenAQ's report as a starting point, read the site's terms of use, and archive its postings daily for two weeks.
Done looks like
Ninety days of station-level readings for one country that match the official posting in at least 99% of 200 random spot checks, with code and method public, and the feed accepted by OpenAQ or published under an open license.