The instruments at the heart of the network are known as MAX-DOAS, short for Multi-Axis Differential Optical Absorption Spectroscopy. The technique is elegantly simple in concept: a spectrometer points at the sky at several slightly different elevation angles, capturing sunlight that has scattered down through the lowest layers of the atmosphere. Because each trace gas absorbs light at its own characteristic set of wavelengths, the depth of those absorption fingerprints in the recorded spectra reveals how much of each pollutant lies in the column of air between the instrument and the sun. Fitting the measurements across multiple viewing angles allows scientists to separate the pollution concentrated near the surface from material higher aloft, which is why the approach is often described as hyperspectral vertical remote sensing from the ground. It is, in effect, the terrestrial counterpart to the satellites it is designed to check.

Why does the atmosphere over China warrant such an elaborate ground-based effort? The country has deployed a formidable fleet of satellite instruments that map atmospheric composition from orbit, and global datasets from missions such as those carrying UV-visible spectrometers now deliver daily pollution maps at ever-finer resolution. But satellites do not speak for themselves. Their retrievals depend on assumptions about clouds, aerosols, surface reflectivity, and the vertical distribution of gases, and those assumptions can introduce biases that vary by region and season. Long-term, independent, standardized ground-based column measurements are the gold standard for validating satellite products, diagnosing where retrieval algorithms go wrong, and fusing multiple data sources into coherent pictures of atmospheric change. Until now, the authors note, China lacked a long-term ground-based column dataset covering multiple atmospheric constituents and processed within a unified framework.

The geographic reach of the new network is one of its most striking features. The 53 sites stretch from roughly 22.5 degrees north to 44.1 degrees north in latitude, and from 80.1 to 122.7 degrees east in longitude, an area that encompasses dense urban agglomerations, suburban transition zones, industrial corridors, coastal cities and islands, landlocked basins prone to trapping haze, arid and semi-arid interior regions, and the Tibetan Plateau. This diversity is not incidental. Different environments stress satellite retrievals in different ways—bright deserts, dark water, high-altitude terrain with unusual pressure and aerosol profiles—and a validation network that samples all of them can reveal systematic errors that a handful of urban stations never would.

The coverage statistics tell the story of a mature and hard-working observing system. Aerosol optical depth boasts the broadest footprint, with all 53 sites contributing 1,073 valid site-months of data. Nitrogen dioxide and formaldehyde, two of the most closely watched pollutants in atmospheric chemistry, each cover 52 sites with more than 1,000 valid site-months apiece. Sulfur dioxide and nitrous acid, harder targets spectroscopically, cover 45 and 44 sites respectively, while ozone columns span 29 sites and 588 valid site-months. For the months in which data are valid, the median monthly completeness reaches 83.9 percent for ozone, nitrogen dioxide, formaldehyde, and sulfur dioxide, and 86.7 percent for nitrous acid and aerosol optical depth. In practical terms, that means the instruments were capturing usable spectra on the vast majority of possible days, even through the disruptions of weather, instrument maintenance, and the operational realities of running dozens of remote stations for half a decade.

A dataset is only as trustworthy as its comparisons, and the team subjected their record to a battery of independent checks. Against satellite products, the ground-based MAX-DOAS columns of nitrogen dioxide achieved a correlation coefficient of 0.899, an impressively tight agreement for a reactive trace gas whose concentrations can change by orders of magnitude within a single day. Formaldehyde columns correlated at 0.734 with satellite retrievals, and ozone at 0.551, a weaker but still meaningful relationship given the very different vertical sensitivities of ground spectrometers and orbiting instruments. For aerosol optical depth, the benchmark was AERONET, the worldwide network of sun photometers that has long served as the reference standard for satellite aerosol validation; the correlation there was 0.879. The researchers also compared their column measurements with surface concentrations reported by the China National Environmental Monitoring Centre, finding correlations of 0.720 for nitrogen dioxide, 0.624 for sulfur dioxide, and 0.678 for ozone—evidence that the column measurements track what people are actually breathing at street level, even though columns integrate the entire atmosphere above.

Each of the six measured quantities carries its own scientific weight. Nitrogen dioxide is a combustion tracer and a precursor of both ozone and particulate nitrate, making it a central player in photochemical smog. Formaldehyde serves as a proxy for volatile organic compounds, the other essential ingredient of ozone formation, so paired nitrogen dioxide and formaldehyde columns allow scientists to diagnose whether ozone production in a given region is limited by nitrogen oxides or by organics—a distinction that determines which emission controls will actually work. Nitrous acid, measurable only in small concentrations and notoriously difficult to observe, photolyzes in morning sunlight to launch the radical chemistry that drives smog formation. Sulfur dioxide tracks coal burning and industrial activity, while ozone itself is both a respiratory hazard and a greenhouse gas. Aerosol optical depth ties the gas-phase chemistry to the particle pollution that dominates visibility and health impacts across much of eastern China.

The applications the authors envision extend well beyond traditional validation. They highlight the dataset’s role in diagnosing retrieval biases in satellite products, supporting the integration of satellite observations with artificial intelligence methods and bias correction, and evaluating chemical transport models—the numerical simulations that forecast air quality and attribute pollution to its sources. As machine learning increasingly enters atmospheric science, training and testing those models demands exactly the kind of independent, consistently processed ground truth this network provides. A model or algorithm calibrated against one satellite can silently inherit that satellite’s errors; a reference built from the ground, with a unified framework across sites and species, offers a way to break that circularity.

The decision to release the entire record freely on Zenodo, under the Creative Commons Attribution 4.0 license, reflects a broader shift in how atmospheric science handles data. Six years of multi-species column observations from a nationally distributed network would once have been the private asset of a single group; here it is a community resource that any researcher, from Beijing to Boston, can download, scrutinize, and build upon. The work is currently a preprint under open review, with the discussion open until mid-November 2026, meaning the scientific community can weigh in on the methods before final publication in Earth System Science Data.

For anyone who has watched China’s air quality story unfold over the past decade—from the haze crises that made international headlines to the aggressive emission controls that followed—this dataset offers something rare: a continuous, standardized, ground-level optical record of how the atmospheric column itself has changed through a period of rapid transition. It captures the chemistry of megacities and the pristine baseline of the plateau, the combustion plumes of industry and the biogenic whiffs of forests, all quantified by the same instruments, reduced by the same algorithms, and validated against the same independent references. In atmospheric science, where every measurement carries hidden assumptions, that kind of consistency is not a luxury. It is the foundation on which the next generation of satellites, models, and pollution policies will stand.

Subject of Research: A ground-based hyperspectral remote sensing dataset of atmospheric trace gas vertical column densities and aerosol optical depth across China from 2020 to 2025

Article Title: A dataset of ground-based VCDs (O3, NO2, HCHO, HONO, SO2) and AOD observations from the hyperspectral vertical remote sensing network in China (2020–2025)

Article References: Li, Y., Xing, C., Jiao, P., & Liu, C. (2026). A dataset of ground-based VCDs (O 3 , NO 2 , HCHO, HONO, SO 2 ) and AOD observations from the hyperspectral vertical remote sensing network in China (2020–2025). https://doi.org/10.5194/essd-2026-743

Image Credits: AI Generated

DOI: 10.5194/essd-2026-743

Keywords: MAX-DOAS, air pollution, satellite validation, aerosol optical depth, nitrogen dioxide, ozone, formaldehyde, nitrous acid, sulfur dioxide, China, atmospheric composition, Earth System Science Data