Will artificial intelligence help stop climate change?

Will artificial intelligence help stop climate change?

Can Artificial Intelligence, increasingly trusted to tackle our toughest problems, can help us understand and stop climate change?

Your phone can recognize your face, and your bank can reject any transaction, although it cannot teach you how to spend money wisely. An online supermarket is pushing for vegan products because you once bought oat milk there, and as soon as you watched one TV show last month, the online movie theater started bombarding you with B movie ads.

An increasing number of gadgets and services are using artificial intelligence (AI) technology, which is gradually penetrating all areas of our lives. Scientists, entrepreneurs and the public sector trust AI to solve the most pressing problems of modern society. One of them is understanding the nature of climatic processes and possible scenarios for their development. Already today, technology is helping us organize a large amount of data. Against this background, the question arises: can they be used to mitigate the effects of climate change and adapt to the new reality?

“When we talk about AI, we often mean machine learning (ML), which is a collection of algorithms that can ‘learn’ from data,” said Dr. David Rolnick, assistant professor at the University of Pennsylvania. “In principle, AI will not perform better than humans, but it will be much faster, it will be able to identify patterns by analyzing huge amounts of data.” It is this ability to instantly process large amounts of data, organize information and find relationships that made AI a tool that has radically changed the rules of the game in all sectors of the economy.

All this applies to climatology and climate change monitoring. The amount of climate data collected by satellites has reached unprecedented proportions. Weather forecasts are made with the highest degree of detail. There are still many uncertainties embedded in climate models and scenarios. Scientists are using AI to process and organize large amounts of data and make more accurate predictions that will allow society and the environment to adapt to future reality. “ML allows you to learn complex behaviors from a dataset without understanding physical phenomena,” explains Dr. Peter Duben, ECMWF Research Fellow. – The more data, the better the tools. As more information accumulates, machine learning tools will get better and better. It means,


AI helps scientists analyze satellite images and make predictions

“Machines help us measure and monitor the world around us, which is critical to making better decisions in the face of an uncertain future,” says Natalia Tkachenko, Lead Research Fellow in Data Processing and AI at Oxford University. – AI is not just a collection of data, it is the ability to identify patterns and relationships in a complex world. The end result is always a decision or processed information. ”

Scientists are successfully using AI to obtain images of the Earth at higher spatial resolution. “AI is doing an excellent job of acquiring spatial information, this is one of its superpowers,” says Dr Pierre-Phillippe Mathieu, head of the F-lab of the European Space Agency. This is also the opinion of Dr. Vincent Peuch, Director of the Copernicus Atmosphere Monitoring Service (CAMS). “It compares satellite images very effectively and automatically tracks changes in land cover, so it is suitable for use in areas where accurate ground monitoring is not possible. It also helps to speed up the process of computer modeling, make it less expensive, especially for detailed weather forecasts that require constant updates. ”

The Copernicus Climate Change Service (C3S) and CAMS are testing and using AI to detect changes in vegetation and tree cover, refine urban air quality forecasts and automatically process satellite imagery, Dr Peuch said.

In the Amundsen Sea, off the coast of West Antarctica, scientists at the British Antarctic Research Administration (BAS) at the Turing Institute are using MO to detect and monitor the processes of icebergs breaking into smaller pieces and trained AI to predict future changes in the ice sheet. AI allows them to interpret the results of these predictions and even better understand how climate variables affect each other over space and time.

AI is increasingly being used to solve a wide variety of environmental and social problems. The University of Washington plans to use AI to monitor and improve forecasting of sea heat waves. Tanzania’s Conservation Center will use AI for aerial photography of wildlife and human observation to prevent conflicts between animals and humans. In Boston, experts are testing an AI-powered GreenCityWatch application that will collect information about the number and condition of trees in cities.

Artificial intelligence is successfully used in agriculture. FarmBeats technology, powered by the Microsoft Azure cloud platform, consolidates data from ground sensors, cameras, tractors and drones, and develops MO models to monitor agricultural activities and improve farm resilience to climate change. “Agricultural producers determine the timing of planting, watering, harvesting and other activities based on weather conditions,” said Ranviir Chandra, chief scientist at Microsoft Azure Global. “However, the weather forecasts come from weather stations, not from the farm. One of our AI algorithms combines detailed weather models and weather station data with information from sensors installed on the farm, for forecasting the weather at the location of the farm. By filling in the information gaps with farm data, you can generate forecasts that will enable farmers to make better decisions. ”

Will AI become a powerful new tool for predicting climate change?

One of the ambitious goals that scientists set for AI is to create a “digital twin” of the Earth or a replica of all systems and processes on the planet. “It would be a virtual laboratory for the planet, where we could conduct experiments, develop environmental strategies and evaluate the results,” says Dr. Mathieu. “We already have the building blocks of AI that will go into creating digital twins of environmental objects, and then a digital twin of the Earth,” said Dr. Scott Hosking, environmental data scientist at BAS. “We are not able to track every aspect of our changing planet with the required level of detail. By developing digital twins of the natural environment, we can focus on remote and hard-to-reach areas such as the polar regions, where sometimes it is impossible to even charge the battery. The resulting information can be used in real time to navigate drones and underwater drones to improve the accuracy of future measurements. ”

However, AI is not yet completely reliable. When it comes to predicting climate, scientists warn that there is not enough historical data to “train” algorithms. “AI needs to be trained on historical data,” explains Dr. Judah Cohen, director of seasonal forecasting at Atmospheric and Environmental Research (AER) and climatologist at MIT. “We have been training it using data since 1979, when satellites were widely used, but this information is not enough to develop optimal AI solutions. It may be necessary to resort to creating artificial data through modeling, but it is not yet clear if the simulated data can replace historical data.

Moreover, as Dr. Rolnik noted, AI cannot replace the physical processes responsible for climate formation. “AI has its drawbacks,” adds Dr. Mathieu. “You can always find a correlation between the data, but it doesn’t necessarily mean there is a causal relationship, so experts are needed who can provide explanations based on the laws of physics.”

According to Dr. Dubin of ECMWF, this also applies to weather forecasting models. “There is an opinion that AI and ML are better than traditional technologies to cope with the tasks of operational (several hours ahead) and long-term forecasting. However, it is unlikely that DoD will be able to outperform most other forecasts and thus ‘replace’ weather forecasting models, since the forecasts will not be as accurate in most applications. ”

In addition, AI systems are only good at performing tasks that they have been trained to do. This raises many problems. “You have to make sure that the AI ​​is being used to work with the range of variables it was trained on,” says Dr. Peuch. “Otherwise, you might get false results.” This means that, although the algorithm will correctly analyze the data for which it was created to process, the arrival of new information beyond its competence can lead to inaccurate results. When it comes to climatology, it is not only the data that change, but the climate itself. “When it comes to climate change, algorithms need to be carefully thought out, and as the climate continues to change, we need to make sure that AI is not only using data from the past to predict the future. – adds the director of CAMS “. When it comes to climate change issues, choosing an algorithm becomes a daunting task. “There are many AI solutions out there, and it’s not easy to choose the one that works best for climate prediction,” explains Dr. Cohen. “I believe that choosing and optimizing an AI algorithm that can dramatically improve existing climate predictions will be challenging.”

The use of AI technologies also raises questions related to the acquisition and management of data. “We have no major privacy concerns for data from conventional weather sources,” says Dr. Dyuben. “However, there is so-called“ Internet of Things ”(IoT) data, which today is practically not used for weather forecasts, but in the future can significantly improve it. For example, there are observations from mobile phones and other information obtained through crowdsourcing. There may be questions related to data confidentiality. ” Dr. Tkachenko goes even further in his reasoning. He argues that if the data used in decision formulas have been distorted, then this can lead to negative results. “We indicate the composition of the products on the packages with ready meals, and here we are,

Finally, can climate scientists and environmentalists learn from the use of AI in other industries? “Use AI only when it is necessary to solve a specific problem,” warns Dr. Rolnik. – Easy to be seduced by trendy technology. Every time you use AI, make sure it makes sense. AI applications must be results-oriented and developed in collaboration with all parties who will use and benefit from this technology. It would be a big mistake to think that AI will magically solve all problems. It is a powerful tool, but only one of many that can be used to combat climate change. ”

Leave a Reply

Your email address will not be published. Required fields are marked *