How Can We Control Climate Change?

Venice is sinking, California is on fire. We may believe in climate change or not but violent, and in many cases unpredictable, weather has a tremendous impact on our lives, not only destroying our homes but changing the way we grow crops and livestock, what we eat, wear or build.

Though machine learning techniques are well-known in climate change research as methods of big data handling, and weather events prediction, there still remain large uncertainties concerning the global climate changes and the human role in them.

According to a World Economic Forum report, Harnessing Artificial Intelligence for the Earth, the term AI refers to computer systems that “can sense their environment, think, learn, and act in response to what they sense and their programmed objectives.”

Such IT giants as Microsoft believe that artificial intelligence can become a “game changer” in controlling climate change and fighting environmental issues.

So far, AI applications in this sphere fall into two of the broadest categories: predicting and handling of extreme weather conditions and predicting and drawing scenarios of the climate change.

The first use of a deep-learning system to identify tropical cyclones, atmospheric rivers and weather fronts was reported in 2016. By now it has achieved 89% to 99% accuracy in identifying these events and the next steps as seen by the team would be to extrapolate the techniques to study all kinds of extreme events — including ones not yet identified. At present, machines are also able to predict how long a storm will last and its severity. By improving weather forecasts, these types of programs can help keep people safe.

However, a recent boost of data collection, including data provided by satellites, has greatly expanded the possibilities of AI’s application to climate science, turning the latter essentially into a data problem.

For instance, climate prediction models are learning how to utilize accurate representations of clouds and their atmospheric heating and moistening — a critical factor taking into account the debates around the impact of the increased greenhouse gas concentrations on global warming. A paper recently published online in the American Geophysical Union demonstrates how machine learning techniques can represent clouds in coarse resolution (~100 km) climate models and show how they respond to a change in greenhouse gases. This study has the potential to better predict the climate’s response to rising greenhouse gas concentrations both globally and regionally.

Another direction of research is monitoring forests and assessing deforestation from different perspectives. One of the largest-scale project dealt with the determination of rainforest tree species capable of withstanding hurricanes based on high-resolution flyover images. It is only AI, not a human that can figure out what the various species of trees look like from above in the flyover images. At the same time, it is crucial to understand how hurricanes change the distribution and composition of forests because when forests are damaged, vegetation decomposes and emits more CO2 into the atmosphere. Even when trees grow back, they are smaller and store less carbon. If climate change results in more extreme storms, some forests will not recover, less carbon will be stored, and more carbon will remain in the atmosphere, exacerbating global warming.

Looking at climate change-related projects as a whole, it becomes obvious that several dozens of models exist, each having benefits yet unable to cover all issues. To extract intelligence from them, machine learning algorithms were developed by Monteleoni to create weighted averages of the roughly 30 climate models used by the Intergovernmental Panel on Climate Change. By learning the models’ strengths and weaknesses, these algorithms generate better predictions than conventional approaches that treat all models equally.

Even though AI proves to be effective in climate and weather prediction, it is still treated with suspicion among meteorologists. Since machines can’t tell you “how” they arrive at certain predictions or decisions, most climate professionals don’t feel comfortable relying on only what the machines suggest will happen, but use machine insight along with their own professional analysis to complement one another.

Nevertheless, in the future, machine learning and deep learning techniques coupled with computer vision can revolutionize environmental studies, helping protect oceans and forests, wildlife and livestock. The findings and predictions made by AI can be applied to agriculture, making it smart, or to urban development in the new IoT networks.

In India, AI has already helped farmers get 30 percent higher groundnut yields per hectare by providing information on preparing the land, applying fertilizer and choosing sowing dates. In Norway, AI helped create a flexible and autonomous electric grid, integrating more renewable energy. And it is only the beginning of the world where understanding and ultimately controlling global and regional climates can help us not to survive, but to thrive.