Stratospheric aerosol injection, or SAI, is the most discussed form of solar geoengineering. The idea is to mimic the cooling effect of large volcanic eruptions by deliberately injecting sulphur dioxide into the stratosphere, where it converts into tiny, long-lived sulphate aerosol particles. These particles scatter incoming sunlight back to space, offsetting some of the warming caused by greenhouse gases. The precedent is real: the 1991 eruption of Mount Pinatubo in the Philippines injected roughly 15 to 20 teragrams of sulphur dioxide into the stratosphere and temporarily lowered global mean surface temperatures by more than one degree Celsius. But deliberate SAI remains deeply controversial, with potential side effects including shifts in regional precipitation, altered temperature gradients between hemispheres and between poles and equator, and delayed recovery of the stratospheric ozone layer. A sudden halt to a deployment could even trigger a catastrophic rapid warming known as the termination shock.

The governance problem is what makes detection so urgent. No global framework currently exists to regulate SAI, despite repeated calls from the United Nations Environment Programme, the European Parliament, and national scientific academies. According to studies under the EU-funded Co-Create project, small-scale outdoor SAI experiments could be carried out today with existing technology at a cost of between 10 and 100 million US dollars, using stratospheric balloons to release anywhere from a few kilograms to a few hundreds of tonnes of sulphur dioxide. Such near-term and mid-term experiments are widely seen as a likely entry point to larger, uncoordinated deployments. A typical moderate volcanic eruption injects a few teragrams of sulphur dioxide, tens of thousands to billions of times more than these exploratory experiments, which is precisely why existing satellites, designed to monitor volcanic plumes, would miss them entirely.

The study tested whether a future satellite instrument could close this surveillance gap. The researchers focused on the Changing-Atmosphere Infra-Red Tomography explorer, or CAIRT, a mission concept that was one of four candidates down-selected in the European Space Agency’s Earth Explorer 11 call. CAIRT would have been the first limb sounder in space to use imaging Fourier-transform thermal infrared technology, observing the limb of the atmosphere at tangent heights from 5 to 115 kilometres with a spectral resolution of 0.25 wavenumbers or better across the mid-infrared range from 718 to 2200 wavenumbers. This design would deliver three-dimensional profiles of trace gases and aerosols with vertical resolution between 1 and 3 kilometres and horizontal resolution between 100 and 300 kilometres, along with temperature profiles resolved to 2 kilometres or finer. Although CAIRT was ultimately not selected for implementation, its scientific goals were recognised as timely, and the concept remains unique in its potential to observe the middle atmosphere.

To evaluate the concept rigorously, the team built what they call a pseudo-reality from state-of-the-art climate model simulations. They used the Community Earth System Model version 2 with the Whole Atmosphere Community Climate Model version 6, run under the Geoengineering Model Intercomparison Project within CMIP6. Specifically, they took the Geo SSP5-34-OS 2.0 scenario, in which sulphur dioxide is injected at four latitudes between roughly 20 and 25 kilometres altitude to hold global warming to 2 degrees Celsius. In the first simulated deployment year, 2034, the injection rates are deliberately tiny: about 0.035 teragrams per year at 15 degrees north, equivalent to roughly 100 tonnes per day, and about 0.005 teragrams per year at 15 degrees south, around 15 tonnes per day. These amounts are characteristic of the near-term experiments that could realistically be performed unilaterally.

From this pseudo-reality, the researchers generated realistic synthetic observations, or pseudo-observations, of what CAIRT would actually measure, using a dedicated Fast Level 2 Simulator that propagates the instrument’s expected random and systematic errors and spatial smoothing characteristics, derived from the mission’s Phase 0 and Phase A feasibility studies. The technique exploits the distinctive rotational-vibrational absorption and emission bands of sulphur dioxide near 8.7 and 7.3 micrometres, and the spectral signatures of sulphate aerosol droplets peaking around 12.5 and 8.3 micrometres, which arise from the undissociated sulphuric acid in the highly acidic stratospheric droplets. Typical total uncertainties for individual retrievals were 0.15 to 0.25 parts per billion in molar fraction for sulphur dioxide and 1.0 to 1.5 parts per trillion in mass fraction for sulphate aerosol, remarkably close to background stratospheric levels.

The results were striking. In the simulated February 2034 scene, CAIRT-like pseudo-observations clearly detected both injection sites, pinpointed their altitude at 25 to 30 kilometres, and quantified the intercepted sulphur dioxide mass at 3.1 tonnes against a true value of 3.4 tonnes at the northern site, and 0.4 tonnes against 0.5 tonnes at the southern site, biases of roughly 10 and 25 percent. Even the weaker Southern Hemisphere injection, cumulatively only a few hundred tonnes by the end of February, stood well above the detection limit. The plume’s descent, rapid zonal dispersion, and poleward transport were all faithfully captured. In practical terms, this means a CAIRT-like instrument could detect and quantitatively characterise a unilateral geoengineering test within hours to a few days of the first injection, something no current satellite can do.

The technique also proved capable of tracking the aftermath. Over the simulated year 2034, the pseudo-observations followed the conversion of sulphur dioxide into sulphate aerosol as the new layer formed at the injection longitudes, spread zonally to encircle the globe, and then dispersed toward higher latitudes in both hemispheres. Differences between the synthetic observations and the underlying model truth generally did not exceed 10 percent within the plume. The instrument even reproduced the self-lofting of the aerosol plume, a subtle process in which the particles absorb solar radiation, heat the surrounding air diabatically, and lift the layer to higher altitudes over time, a phenomenon recently observed after volcanic eruptions such as Raikoke in 2019.

By contrast, the study’s audit of existing capabilities reads as a catalogue of gaps. Solar occultation instruments like SAGE III/ISS achieve excellent precision but have sparse coverage and would only distinguish SAI perturbations at injection rates of 1 to 2 teragrams of sulphur per year, far above exploratory levels. Limb-scattering instruments such as OMPS-LP have worse signal-to-noise ratios, and both operate in the visible range, so they cannot chemically identify sulphate aerosol. Space LiDARs like EarthCARE’s ATLID detect weak aerosol layers but cannot determine composition, while nadir infrared sounders like IASI can identify sulphate but with little or no vertical resolution. The infrared occultation instrument ACE-FTS is expected to be discontinued within a few years, and several other middle-atmosphere missions are nearing retirement, threatening what researchers have called a data desert in stratospheric composition monitoring.

The message of the study is therefore twofold. First, the world is currently blind to the earliest and most plausible forms of geoengineering, the small tests that could precede larger unilateral deployments. Second, high-spectral-resolution mid-infrared limb emission sounding offers a proven-in-simulation path to closing that gap, with benefits extending well beyond geoengineering oversight to wildfire and space-debris aerosol sources, gravity wave characterisation, and understanding of stratospheric circulation in a changing climate. Whether a CAIRT-like mission flies in the coming decades may determine not just what science can see, but whether the international community can verify, and thus govern, one of the most consequential climate interventions humanity might ever attempt.

Subject of Research: Satellite detection of stratospheric aerosol injection geoengineering using mid-infrared limb emission sounding

Article Title: The enhanced capabilities of mid-infrared limb emission sounding to observe stratospheric aerosol injection geoengineering

Article References: Sellitto, P., Kosary, M., Höpfner, M., Funke, B., Hoffmann, A., Ungermann, J., Errera, Q., Tilmes, S., & Sinnhuber, B.-M. (2026). The enhanced capabilities of mid-infrared limb emission sounding to observe stratospheric aerosol injection geoengineering. Atmospheric Measurement Techniques, 19(19), 6379-6393. https://doi.org/10.5194/amt-19-6379-2026

Image Credits: AI Generated

DOI: 10.5194/amt-19-6379-2026

Keywords: stratospheric aerosol injection, geoengineering, CAIRT, mid-infrared limb sounding, sulphur dioxide, sulphate aerosols, satellite remote sensing, solar radiation management, ESA Earth Explorer, climate monitoring, stratosphere, unilateral deployment

The research team, led by Neelam Chaudhary and Sunita Verma of Banaras Hindu University, together with colleagues from the Space Applications Centre, Birla Institute of Technology Mesra, and Poornima College of Engineering, examined five key pollutants: particulate matter with diameters of 10 micrometres or less (PM10) and 2.5 micrometres or less (PM2.5), ground-level ozone (O3), nitrogen dioxide (NO2), and sulphur dioxide (SO2). These were paired with six meteorological variables: planetary boundary layer height, relative humidity, temperature, precipitation, wind patterns, and surface pressure. The goal was to disentangle how much of India’s air quality story is written by emissions and how much by the atmosphere’s capacity to dilute or concentrate them.

To detect long-term trends in noisy monthly data, the team applied the Mann–Kendall statistical test, a non-parametric method widely used in environmental science because it makes no assumptions about the underlying data distribution and is robust against outliers and gaps. Trend magnitudes were quantified with Sen’s slope, a median-based estimator of the rate of change that resists distortion by individual extreme months. Correlation analysis then linked pollutant behaviour to meteorological conditions, providing a statistical portrait of how weather modulates pollution across India’s diverse climate zones.

The headline finding concerns particulate matter. Across numerous monitoring sites in the Indo-Gangetic Plain—the densely populated arc stretching from Punjab through Delhi and Uttar Pradesh to West Bengal—PM concentrations showed statistically significant declines. Sen’s slope values ranged from −0.893 to −0.097 micrograms per cubic metre per month for PM2.5 and from −1.320 to −0.169 micrograms per cubic metre per month for PM10. Over the full study period, these rates compound into substantial reductions, a signal the authors associate with the tightening of emission controls under initiatives such as the National Clean Air Programme, which targets air quality improvement in 131 Indian cities.

Yet the improvement is far from uniform. Monitoring sites in Punjab bucked the regional trend and showed increasing particulate pollution, which the researchers attribute to localised emission sources—most plausibly the intensive agricultural activity and crop residue burning for which the state is known. Post-monsoon stubble fires in Punjab and Haryana have long been identified as a dominant driver of severe winter smog episodes across northern India, and the new analysis suggests that in some locations these local sources continue to overwhelm any broader cleanup. A few southern sites also displayed region-specific behaviour that diverged from national patterns, underscoring how varied India’s pollution regimes are.

Perhaps the most consequential result is the rise of ozone. Surface ozone exhibited increasing trends in several regions, with Sen’s slope values between 0.041 and 0.529 micrograms per cubic metre per month. Ozone at ground level is not emitted directly; it forms photochemically when sunlight drives reactions between nitrogen oxides and volatile organic compounds. The study highlights two mechanisms behind the increase: enhanced precursor concentrations that feed ozone production, and aerosol inhibition—the counterintuitive process by which particulate pollution actually suppresses ozone formation. Dense aerosol loads scatter and absorb solar radiation, reducing the photolysis rates that drive ozone chemistry, and they also alter the uptake of hydroperyl radicals onto particle surfaces. As particulate levels fall, more sunlight reaches the surface and more radicals survive, accelerating ozone production. Recent modelling work by other Indian researchers has warned that continued PM reductions could drive significant rises in surface ozone, and the observational trends in this study appear consistent with that prediction.

Correlation analysis reinforced this chemical tug-of-war. Particulate matter showed positive associations with NO2 and SO2, with correlation coefficients of roughly 0.5 to 0.6, reflecting their shared combustion origins in traffic, industry, and residential energy use. Ozone, by contrast, was inversely correlated with particulate matter, with coefficients ranging from about −0.2 to −0.8, a spread that captures the interplay of aerosol attenuation of sunlight, photochemical interactions, and region-specific precursor chemistry in regulating surface ozone levels. The strength of the inverse relationship varied by location, indicating that the ozone response to particulate cleanup will differ markedly from one airshed to another.

The meteorological analysis revealed a stark north–south divide in the atmosphere’s ventilating capacity. Over northern and eastern India—Delhi being the notable exception—planetary boundary layer height and wind speed declined over the study period, while relative humidity and surface pressure increased. A shallow boundary layer acts like a lid, trapping emissions within a shallow volume of air near the surface, and weak winds prevent that air from being flushed out. Higher humidity favours the formation of secondary aerosols and the hygroscopic growth of existing particles, while elevated surface pressure is typically associated with stagnant, subsiding air masses. Together, these shifts created conditions favouring pollutant accumulation precisely where much of India’s population and industry are concentrated, helping explain why winter pollution episodes remain so severe even as long-term averages decline.

Mumbai and Bangalore tell the opposite story. These coastal and peninsular cities experienced higher boundary layer heights, stronger winds, and greater precipitation, all of which aid pollutant dispersion. Rainfall scavenges particles directly from the atmosphere, while a tall, well-mixed boundary layer dilutes emissions through a much larger volume of air. The contrast illustrates a fundamental asymmetry in Indian air quality management: cities with favourable ventilation meteorology can achieve cleaner air with comparatively modest emission reductions, while cities in the stagnant Indo-Gangetic Plain face a much steeper challenge, because even unchanged emissions translate into worsening concentrations when the atmosphere’s mixing depth shrinks.

NO2 and SO2 showed heterogeneous trends across the country, with notable decreases in some regions—likely reflecting stricter emission regulations on power plants and industry—and relative stability elsewhere. The authors emphasise that their findings carry a clear policy message: because meteorological modulation can amplify or mask emission trends, mitigation strategies must be tailored to regional conditions rather than applied uniformly. They also stress the need for continuous monitoring, since ozone’s stealthy rise in a cleaning atmosphere could offset health gains from particulate reductions. Ground-level ozone is a respiratory irritant linked to cardiovascular injury and premature mortality, and it damages crops, meaning that India’s air quality success story may be incomplete until the photochemical side of the ledger is addressed as rigorously as the particulate one.

Subject of Research: Meteorological modulation of ground-level air pollutant trends across India from 2017 to 2024

Article Title: Meteorological modulation of ground-level air quality in India based on in situ measurements

Article References: Chaudhary, N., Mishra, M. K., Payra, S., Prakash, D., & Verma, S. (2026). Meteorological modulation of ground-level air quality in India based on in situ measurements. Environmental Monitoring and Assessment, 198(11), Article 1176. https://doi.org/10.1007/s10661-026-15915-9

Image Credits: AI Generated

DOI: 10.1007/s10661-026-15915-9

Keywords: air pollution, India, PM2.5, PM10, surface ozone, nitrogen dioxide, sulphur dioxide, planetary boundary layer, meteorology, Mann-Kendall test, Sen's slope, Indo-Gangetic Plain

The research team, led by Mahesh Bankar of Dr. Babasaheb Ambedkar Technological University in Lonere, together with Vinayak Patki, Sachin Pore, and Mithun B. Patil, focused on the three pollutants that dominate the air quality picture at the Solapur Municipal Corporation monitoring station: sulphur dioxide, nitrogen oxides, and respirable suspended particulate matter. These pollutants are the primary inputs used to compute the Air Quality Index in India, a single number that condenses complex atmospheric chemistry into a scale that officials and the public can act upon. Rather than treating the index as a purely statistical artifact, the team set out to learn the underlying nonlinear relationships between pollutant concentrations and the resulting index, relationships that simpler models routinely miss.

The technical pipeline behind the model reflects the messy reality of real-world environmental data. Monitoring stations in developing cities often suffer from gaps caused by instrument downtime, calibration issues, and communication failures, and the Solapur dataset was no exception. The researchers applied interpolation to fill in missing values, a technique that estimates absent readings from the surrounding data points. They then eliminated outliers, the anomalous readings that can arise from sensor malfunctions or transient local events, and normalized the features so that all input variables operated on comparable scales. Each of these preprocessing steps matters: unhandled gaps and outliers can destabilize neural network training, while unscaled features can cause the optimizer to weight certain pollutants disproportionately simply because their units produce larger numbers.

At the heart of the study is a three-layer deep neural network trained with the Adam optimizer, a widely used adaptive gradient descent algorithm that adjusts learning rates individually for each parameter. The network was trained to minimize the mean squared error between its predictions and the observed Air Quality Index values, a standard loss function that penalizes large deviations heavily and thereby pushes the model toward accuracy on the most consequential errors. Three-layer architectures of this kind strike a practical balance: they possess enough depth to capture nonlinear interactions among pollutants, such as the way particulate levels respond differently to precursor gases depending on season and meteorology, while remaining small enough to train efficiently on the modest datasets typical of a single monitoring station.

The performance figures reported by the team are striking. The model achieved a mean absolute error of 3.88 index points and a root mean squared error of 5.92, both small relative to the range of values the Air Quality Index can span. The coefficient of determination of 0.983 indicates that the network’s predictions track the observed values almost perfectly across the evaluation period. For context, an error of under four index points means the model can generally distinguish not just between broad categories such as moderate and poor air quality, but between fine gradations within those categories, which is precisely the level of fidelity needed for meaningful public health advisories.

Equally important is the comparative analysis the researchers conducted against baseline models. Linear regression, the workhorse of classical statistics, assumes that the Air Quality Index responds linearly to pollutant concentrations, an assumption that atmospheric science has repeatedly shown to be false. Support vector regression, a more flexible machine learning method that maps inputs into higher-dimensional spaces, performs better but still struggled to match the deep network. The deep neural network’s superiority over both baselines demonstrates that the relationships embedded in air pollution data are genuinely nonlinear and that architectures capable of representing those nonlinearities deliver tangible gains in predictive skill rather than merely theoretical elegance.

The significance of this work extends well beyond a single city in Maharashtra. Cities across India and the wider developing world operate sparse monitoring networks, often with just one or two reference-grade stations per urban area, and the data those stations produce are frequently incomplete. Sophisticated forecasting systems built on numerical atmospheric models demand extensive computational resources, detailed emissions inventories, and meteorological inputs that many municipalities cannot assemble. A neural network trained on pollutant concentrations from a single station, by contrast, requires only data that municipal corporations already collect and report publicly. The Solapur study used air quality data freely available from the Maharashtra Pollution Control Board, underscoring how accessible such systems have become.

The study also situates itself within a rapidly expanding global literature on machine learning for air quality. Recent years have seen researchers apply recurrent architectures, convolutional networks, hybrid models combining gradient boosting with deep learning, and even quantum-inspired neural networks to pollutant forecasting in cities from Delhi to Kampala. Systematic reviews of the field conclude that artificial intelligence methods now consistently outperform traditional statistical approaches for both monitoring and forecasting tasks. What the Solapur study adds is a demonstration that this revolution reaches down to second-tier cities, not just megacities with dense sensor networks and abundant funding, and that a carefully preprocessed dataset from a single station can support a genuinely high-performing model.

There remain important caveats and directions for future work. A model trained on data from one station captures the pollution dynamics of that location and may not generalize to other neighborhoods or cities without retraining, particularly where emission sources differ, as they do between industrial corridors and residential districts. The study also predicts the index from concurrent pollutant concentrations rather than forecasting it days in advance, and true early-warning systems require lead time. Nonetheless, the framework the researchers describe, from interpolation and outlier removal through normalization, network design, and rigorous evaluation with multiple error metrics, provides a replicable template that other municipalities can adapt to their own monitoring records.

The broader stakes could hardly be higher. Exposure to particulate matter and pollutant gases is linked to respiratory disease, cardiovascular illness, and premature death, and accurate, timely air quality information is a prerequisite for protective measures ranging from school closures to traffic restrictions. If deep learning models of the kind developed for Solapur can be deployed across the hundreds of Indian cities that monitor air quality, they could convert routine regulatory data into actionable forecasts at almost no additional cost. The authors report no funding or conflicts of interest, and they state that the data and code supporting the findings are available from the corresponding author upon reasonable request, an openness that may accelerate exactly that kind of adoption. In a field where the gap between data collection and public benefit has often been wide, this study shows how a modest neural network, trained on freely available measurements, can begin to close it.

Subject of Research: Deep neural network prediction of the Air Quality Index from pollutant concentrations at an urban monitoring station in Solapur, India

Article Title: Deep neural network-based prediction of air quality index (AQI) for municipal corporation station, Solapur, Maharashtra State, India

Article References: Bankar, M., Patki, V., Pore, S., & Patil, M. B. (2026). Deep neural network-based prediction of air quality index (AQI) for municipal corporation station, Solapur, Maharashtra State, India. Theoretical and Applied Climatology, 157(10), Article 651. https://doi.org/10.1007/s00704-026-06588-y

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06588-y

Keywords: air quality index, deep neural network, air pollution, Solapur, Maharashtra, machine learning, particulate matter, sulphur dioxide, nitrogen oxides, environmental monitoring, Adam optimizer, regression baselines