How Startups Are Using AI To Analyze Climate Change Risk

Losses tied to extreme weather events have generated losses of over $1 billion in 2021 so far. 

This has created an opportunity for startups using AI to help analyze and hopefully prevent potential threats. JL 

John McCormick reports in the Wall Street Journal:

There have been eight extreme events in 2021 with losses exceeding $1 billion as of July 9. Historic climate data is less effective in predicting future events. The increased frequency and severity of natural catastrophes has sparked a growing industry of climate-analytics companies that help businesses gauge short- and long-term threats. These climate-analytics companies combine historical and current data with artificial-technology, including neural networks, to help account for the increasing velocity of climate change. They use machine learning to analyze data, recognize patterns and make forecasts about specific places

A cold snap in Texas earlier this year shut down refineries and chemical plants. Drought in Taiwan continues to put pressure on semiconductor manufacturing capacity. And in the U.S. Pacific Northwest, businesses are still recovering from losses as a result of recent record high temperatures.

The increased frequency and severity of natural catastrophes has sparked a growing industry of climate-analytics companies that aim to help businesses gauge both short- and long-term threats.

Weather and climate forecasting services have existed for decades, but historic climate data is proving to be less effective in helping to predict future events. There have been eight extreme events in 2021 with losses exceeding $1 billion as of July 9. In all of 1980 there were three, adjusting for inflation, according to the National Oceanic and Atmospheric Administration.

More than 8,000 suppliers of goods and services to large corporations reported that $1.26 trillion of their revenue is likely at risk over the next five years because of climate change, deforestation and water security, according to a report by CDP, a nonprofit platform for corporate environmental disclosures.

These new climate-analytics companies combine troves of historical and current data with sophisticated artificial-technology techniques and systems, including neural networks, to help account for the increasing velocity of events linked to climate change.

Westchester Group Investment Management, an investment arm of asset manager Nuveen LLC, was looking to invest in farmland growing a particular crop across Spain and Portugal and worked with San Francisco-based ClimateAi Inc.

---

The startup, which was founded in 2017 and trains its climate-risk system on the long-range global climate models, found that climate over the next few years wouldn’t be optimal for the crop’s success, said Martin Davies, WGIM’s president and chief executive.

“It is essential for climate hazard assessment and adaptation to become the ‘new normal’ in corporate, investor and government decision-making,” said Mekala Krishnan, a partner at the McKinsey Global Institute. “Added to this is the increasing call from stakeholders—including the public, regulators and investors—for transparency from companies on their climate risk exposure.”

Investors are paying attention to such companies. Japanese insurance holdings company Sompo Holdings Inc. in June invested $45 million in Menlo Park, Calif.-based One Concern Inc. The same month, private-equity firm TPG’s Rise Fund invested $100 million in Louisville, Ky.-based Climavision Operating LLC. More recently, ClimateAi said on July 21 it raised $12 million in venture-capital funding.

Climate intelligence startups differ in their approaches, but, for the most part, use machine learning to analyze hordes of data, recognize patterns and make local forecasts about specific places or events, according to Himanshu Gupta, ClimateAi’s chief executive.

Many also use neural networks, systems composed of various layers of artificial neurons designed to mimic the way a human brain works.

Neural networks have the advantage of being able to tackle unstructured information, such as weather and climate maps or graphs of historical temperature data, more easily, contributing to more refined and accurate forecasts, said Chris Goode, Climavision’s CEO.

Neural networks also can learn to weigh the latest data more heavily in their calculations, said Sara Menker, CEO of climate analytics firm Gro Intelligence Inc.

Neural networks factor in time, allowing older information to decay in importance, said Stephen Messer, co-founder and vice chairman of AI company Collective[i] and a neural network expert. The world’s ice, for example, is melting at an accelerated pace. A neural network will recognize that and factor out older ice-melt rates.

But the technology does have limitations. Companies need adequate data to train their models and there isn’t always enough data. One example is hail, where limited observations make it hard to train AI models, said Mr. Gupta of ClimateAi.

The technology isn’t only helping businesses identify risks but also opportunities.

Pacific Seeds, the Australian unit of Advanta Seeds, worked with ClimateAi to determine last year that there would be a higher demand for sorghum seed in northern New South Wales, said Barry Croker, managing director of the seed business.

Analyzing the state of the oceanic patterns that influence precipitation, ClimateAi forecast the region would see a significant rainfall event during the latter half of the spring and summer growing season. Mr. Croker said the company moved sorghum inventory into the area in advance. Mr. Croker said the company captured an additional 5%-10% in sales.

Pacific Seeds uses ClimateAi not only to assist with sales forecasting but in other areas across its business, including with research and development.

“By having a better understanding of both shorter- and longer-term climatic conditions, we can improve our own operations but just as importantly the operations of Australian growers, our customers,” Mr. Croker said.