How artificial intelligence can help in environmental studies

How artificial intelligence can help in environmental studies

The potential of artificial intelligence (AI) to analyze large volumes of data has been explored in the most diverse areas — and in the environmental sciences it is no different. A new report from US government research departments details advances and possibilities in this area following a workshop with hundreds of experts.

  • 10 artificial intelligence collaborations with space exploration
  • Scientists use AI to create eco-friendly concrete

The Artificial Intelligence for Earth System Predictability (AI4ESP) was held in late 2021, with 700 participants from 112 institutions around the world. Scientists and researchers discussed the challenges and the development of an infrastructure, computational and human, to provide accurate, fast and well-informed data about what happens on the planet.

As for the themes, nine focal points related to climate phenomena and natural environments were discussed. They were: hydrology, watersheds, coastal dynamics, soil, ocean, ice, climate variations and extreme events. For each topic, participants discussed the application of techniques such as neural networks, co-design and AI architectures, which can help in decision-making.

“We need new AI methodologies that incorporate process understanding and respect the laws of physics to make reliable and applicable predictions of the behavior of Earth systems,” says Charu Varadharajan, researcher at Lawrence Berkeley National Laboratory. He adds that the work carried out at the workshop, and now published in the report, sets 2-, 5- and 10-year targets for developments in each of the nine focal points.

In addition to the development of techniques for AI, scientists point to the need for adaptations in government agencies and research institutions in order to implement the expected results. This includes, for example, skilled labor capable of operating advanced technology.

Source: US Department of Energy Via: Phys.org