As extreme weather events become more frequent around the world due to climate change, the need for further research into our increasingly warm planet grows. Climate research involves not only studying these events but also enabling external researchers to do the same. Developments in artificial intelligence (AI) represent a powerful tool for achieving these goals.
In 2023, NASA collaborated with IBM Research to develop an AI-powered geospatial model. Developed using vast amounts of widely used NASA Harmonized Landsat and Sentinel-2 (HLS) data, the model serves as the basis for various AI-based studies to address environmental challenges. In line with open science principles, the model is freely available to anyone.
Core models serve as a foundation upon which scientists can develop a wide range of applications, providing powerful and effective solutions. “Core models only know what’s represented in the data,” explained Manil Muskie, head of data analytics at NASA’s Office of the Chief Science Data Officer (OCSDO). “They’re like a Swiss Army knife—they can be used for a wide variety of purposes.”
Once a base model has been created, it can be trained on a small amount of data to perform a specific task. To date, the Interagency Implementation and Advanced Concepts Team (IMPACT) and other participants have demonstrated the capabilities of the geospatial base model by refining it to detect fire scars, determine floodwater boundaries, and classify crops and other land-use categories.
This image, taken March 14, 2024, by the Operational Land Imager-2 (OLI-2) on Landsat 9, shows rectangular shrimp ponds along the coast of northern Peru. NASA Earth Observatory/Lauren Dauphin
Given the computational resources required to create the initial baseline model, partnership was essential for success. In this case, NASA provided the data and scientific expertise, while IBM provided computing power and expertise in optimizing AI algorithms. The team’s shared commitment to open science, aimed at making their research accessible, ensures that their model will be useful to as many researchers as possible.
“We recognized from the start that building a large-scale base model wouldn’t be something we could do alone,” Maskey said. “Everything we did with our base models was publicly available—from the pre-training data, code, recommendations, and model weights to fine-tuning the training data and publications. Transparency allows researchers to track why certain data elements or model architectures were used.”
Building on the success of their geospatial baseline model, NASA and IBM Research are continuing their collaboration to develop a new, similar model for weather and climate studies. To bring this model to life, they are collaborating with Oak Ridge National Laboratory (ORNL), NVIDIA, and several universities.
This time, the primary dataset will be the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2)—a comprehensive collection of atmospheric reanalysis data spanning the period from 1980 to the present. Like the geospatial model, the meteorological and climate model is being developed using an open science approach and will be publicly available in the near future.
In this image of Europe, including Poland, taken from the International Space Station, the night sky is illuminated by lights. NASA
Covering all aspects of Earth science will require multiple base models trained on different types of datasets. However, Muskie believes that these future models could one day be combined into a single, comprehensive model, creating a “digital twin” of Earth that will enable unprecedented analysis and prediction of all types of climate and environmental events.
Whatever innovations the future holds, NASA and IBM’s geospatial and climate models will unlock unprecedented advances in Earth science. While powerful artificial intelligence (AI) tools will enhance researchers’ efficiency, the team’s commitment to open science principles expands the possibilities for discovery, enabling anyone to apply these tools practically and pave the way for groundbreaking research that will help better care for the planet.
When it comes to AI, hopes and fears abound. As a machine learning expert, Rochelle Schneider, head of the AI Applications Unit at the European Space Agency’s Φ-lab, can explain why AI plays a vital role in Earth observation. Rochelle has extensive experience extracting valuable insights from Earth observation data for disease prevention and child development.
Rochelle, how would you define the role of AI in society?
— Artificial intelligence uses computers and machines to perform repetitive tasks or functions that can be automated. This allows us to work more efficiently and focus on activities that require essential human skills, such as emotional intelligence, human relationships, and intuition.
See also... New method for observing individual atoms confirms century-old quantum mechanical theory |
How do you feel about the idea that AI will lead to job losses?
— Well, AI will undoubtedly transform society, but it will also create countless opportunities for new careers. Therefore, we should view AI as a tool that facilitates or enhances human abilities and tasks, not as something that will replace us in the labor market. We must inspire the next generation of professionals to take advantage of the benefits of working with AI technologies. For example, as part of the Earth Observation program at ESA, we train young scientists and engineers to use machine learning tools to improve their understanding and processing of data.
Rochelle Schneider. ESA
Can you explain how AI contributes to Earth observation?
— Artificial intelligence (AI) is making a huge contribution to Earth observation. Several satellites are now equipped with AI-powered computers to filter and process data, and there are numerous applications that use AI to generate critical information for end users.
Computer vision clearly illustrates the contribution of AI. We extract many features from satellite images, such as vegetation or coal mines, and while computers have been doing this for a long time, they have traditionally used fixed algorithms or models to detect the desired objects.
Thanks to AI, the model continuously evolves and learns as new data arrives, fundamentally changing our need for automated detection. With vast volumes of Earth observation data generated daily, AI-powered automation plays a vital role in Earth monitoring.
For example, in the Child Connectivity project, implemented by ESA in collaboration with Giga, we assess young people’s access to the internet. As part of this project, we use artificial intelligence algorithms to recognize school buildings based on Copernicus Sentinel-2 data. In Brazil alone, we were able to identify 65,000 schools incorrectly listed in the UNICEF database.
Forecasting is also becoming increasingly important. In another major initiative, implemented jointly with UNICEF, our climate-based AI models were able to predict dengue fever outbreaks a month in advance.
How is the development of artificial intelligence impacting the commercial space sector?
— Commercial Earth observation is growing exponentially, and artificial intelligence also plays a significant role here. Numerous companies have emerged in Europe in recent years, many of which are supported by the European Space Agency’s InCubed program.
These companies use artificial intelligence-powered satellite data processing for their clients, providing real-time information to government departments, disaster response agencies, and industries such as insurance companies.
The Φsat-2 satellite is equipped with an onboard system for processing images from a multispectral camera using artificial intelligence. The satellite already has AI applications installed that will remove clouds from images, create street maps, and detect shipping. Furthermore, Φsat-2 allows commercial companies to download and launch their applications in space.
We expect that Φsat-2 and other satellites with onboard AI will create many new business models and commercial opportunities, even for companies outside the space industry.
One final thought on AI: how do you see its contribution to Earth observation in the future?
— I think we’ll see the interaction of AI and many other emerging technologies in all areas of life. For example, in the field of Earth observation, we’re currently working on the Destination Earth project—an AI-powered digital twin of the planet that will monitor the effects of natural and anthropogenic impacts and predict extreme events.
Taking this a step further, we can imagine a digital twin of Earth in the metaverse, an immersive experience that provides enhanced interactions and simulations through extended reality—a combination of augmented and virtual reality.
More generally, AI tools will increasingly be integrated into ESA’s core operations and Earth observation activities, significantly enhancing our efforts to create a green and sustainable society by providing us with more valuable information for decision-making and the necessary management of the planet.
Comment on Facebook* The owner of the Meta company is recognized as extremist in Russia and is banned.See also... Complete list of HTML/Javascript event handlers with examples List of required documents for a website in the Russian Federation |
