Artificial intelligence can be used to analyze massive amounts of data from climate simulations, but more training data are needed.
By Javier Barbuzano
Extreme weather events, whether scorching temperatures that ruin crops or killer storms that drown coastal towns, are likely to be more frequent and more powerful with climate change. Quantifying the increase in these extreme events (and their economic and public health costs) requires combing through thousands of gigabytes of data that climate models generate every day.
Scientists can’t just look at the results of their climate models and count hurricanes or droughts. Instead, they are turning to machine learning to find such extreme weather events in their models’ data.
For decades, modelers have relied on heuristics—mathematical definitions of an object of interest—to pinpoint extreme weather events. But heuristics cannot capture the complexity and variability of weather, which are very difficult to condense into a set of values and threshold conditions. Although many algorithms based on heuristics can detect atmospheric rivers (air currents that transport water from the tropics to the poles), for instance, their results tend to be unreliable, with large discrepancies between algorithms. Researchers hope to replace them with a new generation of artificial intelligence (AI) algorithms.
Computer Visionaries
To build these better algorithms, climate scientists have turned to a particular subset of machine learning techniques known as deep learning. Deep learning systems do not require a set of human-defined rules and values to guide their output. Instead, researchers “train” a system with hundreds or thousands of solved examples that the system then analyzes to create its own relevant rules.
“In the [19]80s, [19]90s, and 2000s, people kept coming up with heuristics for defining what makes a pedestrian, what makes a car, what makes a face, and so on and so forth,” said Prabhat, a computer scientist who leads big-data initiatives at the Lawrence Berkeley National Laboratory in California. “In the last 10 years it has been conclusively proved that AI and, in particular, deep learning techniques are truly well suited for solving [the computer vision] problem. We felt that we could apply the same idea to finding extreme weather patterns.”
To test this approach, Prabhat and his colleagues attempted to train a deep learning network to recognize and draw labels around two types of extreme weather in high-resolution simulation data: tropical cyclones and atmospheric rivers, both of which are associated with heavy rainfall. After being trained with over 500 labeled examples, their system can detect most atmospheric rivers and tropical cyclones in simulation data it has never seen before. Their work is described in a preprint article in the open-access journal Geoscientific Model Development.
.
Experienced climate scientists have been “very convinced that the deep learning network is identifying the right patterns,” says Karthik Kashinath, a machine learning scientist leading the project at the National Energy Research Scientific Computing Center. That’s not to say the project is perfect—the system still yields a number of false positives and false negatives—but the researchers hope it can be improved with more training data.
Generating Training Data
Obtaining reliable training data is one of the main challenges for deep learning applications. For extreme weather patterns, the labeling has to be done by experts in the field. “We’re going directly to the climate experts, the meteorologists, who have been looking at these patterns for many years, so they have a good sense of what they look like,” Kashinath said. But labeling these images is a tedious and time-consuming process.
To speed things up, Kashinath and Prabhat, with many collaborators, created ClimateNet, a crowdsourced extreme weather training database for deep learning networks. The goal is to share the labeling effort by opening it to other researchers and institutions around the world. The team also created an online labeling tool called ClimateContours, which allows other scientists to upload their own data sets and label them.
“If a dozen institutions around the world can provide labels corresponding to maybe a dozen [extreme weather] patterns, then we can just solve these problems once and for all, and we can move on to the next set of problems,” Prabhat said.
“The unique aspect to ClimateNet is that instead of relying on one definition of what an atmospheric river or tropical cyclone is, it crowdsources the information using experts, all with subtly different ideas on how these features should be defined in data sets,” said Christine Shields, a climate scientist at the National Center for Atmospheric Research in Boulder, Colo. “All of those viewpoints are then used as a way to ‘train’ the deep learning algorithm, thus incorporating, implicitly, these different definitions.”
Better data sets will lead to better AI models and could unlock untapped results from climate simulations, said Claire Monteleoni, an associate professor in computer science at the University of Colorado Boulder who has applied AI to hurricane track prediction but wasn’t involved in this project. “Much of the data output by climate models hasn’t been analyzed, so AI is really the cheapest way, without having to run new climate model simulations, to gain more insights to help with predicting climate and extreme weather events.”
Enabling High-Precision Analysis
The most basic thing researchers will be able to do with a successfully trained deep learning network will be to count how many extreme events a model calculates will happen in the future and compare the result with present-day numbers. Even so, Prabhat said, that would barely scratch the surface of what these systems will allow.
Researchers can look at underlying data for each event in the model data and extract information such as how much precipitation is expected to fall in a specific area. This information can be used in many ways, including estimating insurance costs in high-risk areas and managing future drought. In places like California, which depend on rain delivered by atmospheric rivers, authorities could be alerted that they have to make contingency plans if those rivers are going to be diverted, Prabhat said.
“It’s not that you could have never answered these questions” without resorting to AI, Prabhat explained, but the new approach will likely make it easier and faster—and possibly better than with heuristics.
“We are able to do precision analytics and get these very high-quality results,” he said, which heightens the credibility of climate models’ results and the plans that rely on them for preparing for the extreme weather events of the future.