As environmental challenges and industrial emissions intensify, researchers and innovators are exploring advanced atmospheric interventions to combat widespread pollution. While traditional pollution-control methods remain widely used, increasing complex environmental conditions are creating demand for smarter interventions that can operate across dynamic atmospheric conditions.
Multispectral, hyperspectral, microwave imaging, and other advanced sensing technologies allow researchers to look beyond visible changes on the ground and extract detailed information about atmospheric and environmental conditions. These sensors capture data across a broad range of wavelengths, with different gases and particles absorbing or reflecting light in characteristic patterns. This makes it possible to identify and quantify specific atmospheric constituents, such as harmful gases. Column-average concentrations derived from these measurements can reveal the three-dimensional distribution of pollutants throughout the atmosphere, offering a more detailed picture than surface-level observation alone. When combined with land and other environmental data, these measurements can help identify areas requiring different environmental interventions.
Artificial intelligence (AI) and machine learning (ML) are further expanding the capabilities of modern climate engineering and environmental monitoring. A recent study highlights how these technologies enhance vulnerability mapping. AI is used in analyzing large and diverse datasets for assessing exposure to climate-related hazards. Machine learning models integrate information such as topography, land use, population, satellite imagery, and climate projections to produce detailed risk and vulnerability maps. Furthermore, advanced computer vision algorithms applied to satellite observations can detect characteristic emission signatures, track greenhouse gases, and pinpoint industrial super-emitters such as methane leaks from pipelines or power plants in near-real-time.
Together, these technologies are enabling a shift from simply observing atmospheric conditions to using real-time environmental data to guide targeted interventions. One example we identified in our previous patent roundup is an AI-assisted approach that uses atmospheric sensing to identify areas affected by acidic gases and inform targeted cloud seeding operations.
From atmospheric monitoring to targeted acid rain mitigation
Researchers and inventors have pivoted from observational monitoring to addressing specific chemical hazards like acid rain. Acid rain is a major environmental issue driven by sulfur dioxide (SO₂) and nitrogen oxides (NOₓ) emissions. While traditional approaches to reducing acid rain have focused on broad emission reductions such as the use of catalytic converters in vehicles or transitioning to renewable energy sources, these methods do not account for real-time atmospheric gas concentrations or the specific environmental vulnerabilities of the land where precipitation falls.
U.S. Patent No. 12,673,884 describes a computer implemented method that enables the selective diffusion of acid rain by analyzing atmospheric gas concentrations and tailoring cloud seeding to specific land zones. The system is designed to identify hazardous acidic gases, locate environmental hot spots, and classify land regions to ensure that resulting precipitation provides ecological benefits rather than harm.
The system uses satellites equipped with sensors such as spectrometers or hyperspectral imagers to produce satellite images. These satellite observations can identify concentrations of acidic gases and generate three-dimensional column-average measurements representing the vertical distribution of gases throughout the atmosphere. By combining these measurements with land data, terrain topography, and land usage, the system can locate pollution hot spots and classify surrounding areas into vulnerable zones such as forest zones, wetlands and paddy fields, barren lands, and other sensitive areas.
The system then determines the appropriate amounts of cloud seeding materials and alkaline compounds for each zone. For instance, in forest zones, it can increase nitrogen to accelerate tree carbon sequestration. At the same time, the system can reduce sulfur dioxide to lower methane emissions in wetlands. Once determined, cloud-seeding units are deployed to execute targeted precipitation.
This approach illustrates how advances in satellite sensing, AI, and environmental data analysis could move atmospheric management from passive observation toward targeted intervention, with environmental conditions informing not only where action is needed but also how that intervention should be carried out.
This patent entitled “Acid rain diffusion based on vulnerable zone classifications” was filed on July 23, 2023 and granted on July 7, 2026. Manikandan Padmanaban, Sarbajit Rakshit, and Jagabondhu Hazra are listed as inventors. Legal representation was provided by Yee & Associates.
To gain deeper insight into the developments in this automated environmental management field, we connected with one of the patent’s inventors.
Conversations
We spoke with Manikandan Padmanaban, Senior Research Engineer at IBM Research and one of the inventors behind U.S. Patent No. 12,618,991. In our conversation, Manikandan shared how emerging technologies advance environmental research today and where the field is headed in the coming years.
Q: Satellite imagery has become increasingly accessible in recent years. How has this changed the way researchers approach environmental monitoring and analysis?
A: Satellite imagery has fundamentally changed environmental monitoring by allowing us to move from sparse, localized measurements toward frequent, large scale and spatially consistent observations of the Earth.
What is particularly exciting is the growing diversity of satellite observations. Multispectral, hyperspectral, microwave imaging and other sensing technologies allow us to look beyond visible changes on the ground and extract information about atmospheric and environmental conditions. For example, spectral observations can be used to identify signatures associated with pollutants such as NOx and SO2 and help locate areas with elevated concentrations. Our work explores precisely this idea using satellite spectral observations to identify spatial pollution hotspots within a geographic region.
This changes the research question from simply “What can we observe?” to “What can we infer from those observations, what is likely to be affected, and what can we potentially do about it?”
Q: AI is playing a growing role in interpreting vast amounts of Earth observation data. What do you see as the biggest advantages of combining satellite imagery with machine learning?
A: The biggest advantage is the ability to convert massive volumes of Earth observation data into context-aware environmental intelligence.
Satellite data can tell us a great deal about atmospheric and surface conditions, but environmental impacts depend heavily on context. The same level of atmospheric pollution, for example, may have very different implications over a forest, wetland, agricultural region, barren land or densely populated area.
AI can help connect these layers. It can identify spatial patterns and hotspots, classify vulnerable regions, integrate weather and land characteristics, and ultimately estimate potential environmental impact. In our work, for example, satellite-derived pollutant information is combined with terrain and land-use characteristics to distinguish among different vulnerable zones.
So, we see the real advantage as going beyond image recognition. The opportunity is to use AI to understand the relationship between what we observe, where it occurs, the surrounding environmental context, and its potential consequences.
Q: Environmental monitoring often requires integrating information from multiple sources. What are some of the technical challenges in building systems that can reliably analyze and interpret geospatial data?
A: One of the biggest challenges is that environmental data are extremely heterogeneous. Satellite imagery, atmospheric observations, weather forecasts, terrain information, land use maps and ground measurements can all have different spatial resolutions, temporal frequencies, coordinate systems and levels of uncertainty.
A reliable system therefore has to do much more than simply combine datasets. It needs to align them correctly in space and time, account for missing or uncertain observations, and preserve meaningful relationships between environmental variables.
For example, our framework considers satellite spectral observations together with terrain, land use and meteorological variables such as precipitation, humidity, wind speed and direction, temperature and pressure. Wind is particularly important because what a satellite observes at one location does not necessarily mean the pollutant originated there or will remain there.
Another important challenge is ensuring that AI predictions remain consistent with our understanding of the underlying physical and environmental processes. For environmental decision-making, accuracy alone is not enough, we also need reliability, interpretability and awareness of uncertainty.
Q: Your work demonstrates how satellite analytics can move beyond simply observing environmental conditions toward supporting informed decision-making. How do you see this capability evolving in the coming years?
A: We see the evolution happening in three stages: observation –> prediction –> decision support.
Satellite analytics has traditionally been very strong at observation, showing us what is happening and where. AI increasingly enables the second stage: predicting how conditions may evolve and estimating their potential environmental consequences.
The next step is particularly interesting: using that information to evaluate possible actions. In our work, for example, the system does not stop after identifying pollution hotspots. It considers the vulnerability of the underlying region, estimates an appropriate response, and evaluates the potential environmental benefit against the cost of intervention.
In the future, we expect these capabilities to become increasingly dynamic and closed loop: observe environmental conditions, identify emerging risks, predict impacts, evaluate alternative interventions, support an action, and then use subsequent observations to measure the outcome and adapt the strategy.
That could fundamentally change satellite analytics from primarily an environmental monitoring technology into an important component of environmental decision support systems.
Q: Which environmental applications do you believe stand to benefit the most from advances in satellite analytics and AI, such as pollution monitoring, disaster response, ecosystem management, or climate resilience?
A: All of these areas have significant potential, but we think the greatest benefits will come in problems that are both spatially distributed and highly dynamic.
Pollution monitoring is a good example. Air pollutants move across geographic and administrative boundaries, making them difficult to characterize using ground sensors alone. Satellite observations can provide broader spatial coverage, while AI can combine those observations with weather, terrain and land use information to identify hotspots and estimate which regions are most vulnerable.
Disaster response is another important opportunity because satellite data can provide rapid observations over areas where ground infrastructure may be unavailable or damaged. Similarly, ecosystem management can benefit because AI can help distinguish how the same environmental stressor may affect forests, wetlands, agricultural areas and other sensitive ecosystems differently.
Our work illustrates this context dependent approach by considering different environmental objectives for different types of regions rather than applying a uniform response everywhere.
More broadly, climate resilience may ultimately be the largest opportunity because it brings many of these capabilities together, continuous observation, risk identification, forecasting, vulnerability assessment and targeted adaptation.
Q: What does it mean to move from “Earth observation” toward “environmental intelligence”?
A: To us, Earth observation is primarily about measuring and understanding what is happening in the environment. Environmental intelligence goes one step further: it combines those observations with AI, physical understanding and contextual information to determine what those observations mean and what decisions they can support.
For example, detecting an atmospheric pollutant is Earth observation. Identifying where its concentration is unusually high, understanding how weather may transport it, determining which ecosystems or communities are vulnerable, estimating its potential impact, and evaluating possible responses is environmental intelligence.
That is the direction we find particularly promising: not replacing environmental scientists or decision makers, but providing them with a much richer, more timely and spatially detailed understanding of environmental risks and possible responses.
To better understand how innovation in this area is evolving, we examined the US patenting activity within G06V 20/13, a patent classification covering satellite imagery.
Satellite Imagery: Patenting Activity
U.S. patent filings for G06V 20/13 show a variable filing pattern over the past decade, with filings rising steadily from 2016 to a peak in 2019. This growth comes as demand for satellite-based technologies continues to expand across applications such as disaster response, agriculture, defense, infrastructure monitoring, and weather observation.
According to the Business Research Company, the atmospheric satellite segment is expected to grow from $10.9 billion in 2025 to $17.01 billion by 2030, representing a compound annual growth rate of 9.4%. This is attributed to the development of stratospheric platform-based satellites, miniaturized high-resolution sensor payloads, real-time atmospheric data services, AI-powered predictive climate analytics, and commercial satellite weather monitoring.
Satellite Imagery: Top Assignees
The leading assignees for patent filings under G06V 20/13 reveal a patent landscape shaped less by traditional aerospace and imaging giants and more by a mix of drone makers, tech platforms, and companies applying satellite-derived imagery to their core businesses.
Skydio leads the filings with a portfolio centered on autonomous drone navigation and image capture. A notable patent by Skydio that moves beyond passive monitoring to active predictive analysis is U.S. Patent No. 12,330,784, which projects planned 3D flight trajectories directly into captured image space to dynamically evaluate risk. The system optimizes an autonomous vehicle’s path using a visual cost map that assigns higher risk penalties to hazardous or low-confidence regions.
SZ DJI Technology’s portfolio focuses on image capture, stabilization, and presentation across visible and infrared spectrums. A notable example is U.S. Patent No. 11,006,033, which utilizes deep machine learning to recognize, classify, and track multiple moving targets within complex visual scenes in real time. This works in tandem with laser radar ranging to accurately predict target boundaries and dynamically drive autofocusing.
IBM is also listed as one of the top assignees with focus on the analytics and use of AI in analyzing satellite and aerial imagery. This is consistent with its broader enterprise push into geospatial data processing using artificial intelligence and cloud-based image analysis to address climate challenges.
These patents affirm the rise of artificial intelligence in processing satellite data. As Manikandan mentioned, AI helps connect layers of contextual data to better reveal the relationships between satellite observations, their environmental contexts, and their potential impacts.
The presence of other top assignees such as Amazon Technologies, State Farm Insurance, Maxar Space, Boeing, and Toyota Motor further highlights the broad adoption of satellite and aerial imaging technologies across logistics, insurance, aerospace, and automotive industries.
Satellite Imagery: Top Law Firms
Galvin Patent Law is the leading law firm for G06V 20/13 patent filings in the U.S., followed by Lowe Graham Jones and Anova Law Group. Other prominent firms include Armstrong Teasdale, Fish & Richardson, and Athorus. Together, these specialized intellectual property firms manage a steady pipeline of patent work in the satellite imagery industry.
The future of satellite imagery and analytics
The future of satellite imagery and analytics is defined by a fundamental shift from passive environmental observation to active, intelligent decision support. According to Manikandan, this evolution is happening in three stages: from “observation to prediction to decision support.” While traditional satellite technology has long excelled at showing us current ground conditions, the integration of emerging technologies like AI now allows systems to forecast how these environments might change.
Looking ahead, these technologies will function as continuous, automated feedback cycles. Manikandan expects future capabilities to become “increasingly dynamic and closed loop.” That is, systems will continuously monitor risks, recommend interventions, and then use new satellite data to measure the effectiveness of those actions so strategies can be adapted. He identified opportunities in disaster response, ecosystem management, and the broader climate resilience space.
Ultimately, Manikandan acknowledges the vital challenge of ensuring AI predictions remain consistent with fundamental knowledge of respective fields. In making decisions, he posited that accuracy alone is insufficient. Systems must also demonstrate high reliability, clear interpretability, and an awareness of their own uncertainties.
