In recent years, climate science has dramatically advanced, bolstered by sophisticated computing and satellite imagery. These innovations have led to the rise of predictive models that assess the likelihood of environmental disasters, such as floods and wildfires, at an unprecedented level of detail even down to individual building lots. Startups developing these models have flourished, backed by significant investment from venture capital and private equity.
Promise and Perils of Predictive Models
These climate models are already influencing crucial decisions in the U.S. and beyond. Companies like Fannie Mae, Aon Plc, Allstate Corp., and Zurich Insurance Group AG, alongside major banks, real estate firms, and public agencies, are utilizing these tools to climate-proof their assets. Additionally, major ratings agencies, including Moody’s Corp. and S&P Global Inc., have incorporated climate risk modeling into their operations through strategic acquisitions.
While the need for such predictive models is undeniable especially as natural disasters become more frequent and severe the industry is not without its challenges. One major concern is the lack of transparency in many of these models, which are often considered “black boxes.”
This opacity means that the methodologies behind the predictions are not open to rigorous, independent scrutiny. A report from White House scientific advisers last year highlighted this issue, warning that some climate risk predictions might be of “questionable quality.” The research nonprofit CarbonPlan echoed these concerns, stating that decisions based on opaque models could impact billions of lives and trillions of dollars.
Discrepancies in Flood Risk Models: A Case Study
A Bloomberg Green analysis comparing two flood-risk models in Los Angeles County underscores the uncertainties associated with these tools. The analysis found that the models, designed to predict the same risk, agreed only 21% of the time on a single, relatively simple metric. This inconsistency is troubling, especially given the real-world consequences these models could have on communities. For instance, government agencies or insurers might base decisions on these models, leading to potentially misguided investments in flood prevention or inappropriate increases in insurance premiums.
The two models analyzed were created by New York-based First Street Technology Inc. and researchers at the University of California, Irvine. The Irvine team’s model, which was developed specifically for Los Angeles County, identified areas like Compton and Long Beach as high-risk for flooding.
In contrast, First Street’s model highlighted the San Gabriel Valley and Westside areas like Beverly Hills and Santa Monica as most vulnerable. Such discrepancies could lead to vastly different strategies for flood-defense projects, potentially neglecting disadvantaged communities that might be at higher risk.
Need for Greater Transparency and Validation
The differences in these models raise critical questions about the reliability of climate risk predictions. The Irvine team argues that their model might be more accurate due to its use of high-resolution ground-elevation data and detailed information on local drainage infrastructure. Meanwhile, First Street defends its model by asserting that its predictions align more closely with real-world flood damage claims, such as those made to FEMA’s Individual Assistance program.
Adding to the complexity, a third model by CoreLogic Inc., a significant player in the climate risk modeling space, also shows less than 50% agreement with either First Street’s or Irvine’s model. This further highlights the variability in climate risk assessments depending on the methodologies and data inputs used.
As climate risk modeling becomes increasingly integrated into industries like insurance and real estate, the need for transparency and rigorous validation of these models is more pressing than ever. Experts caution that while these models can provide valuable insights, the lack of clarity at finer scales where the models are less accurate poses significant risks. These tools, when misapplied or misunderstood, could disproportionately harm vulnerable communities, exacerbate inequalities, and lead to misinformed policy decisions.
The debate over climate risk modeling serves as a reminder that while these tools are powerful, they are not infallible. As the world continues to grapple with the impacts of climate change, ensuring that the models guiding our decisions are as accurate and transparent as possible will be crucial in mitigating future risks and safeguarding communities.
Reference: https://www.bloomberg.com/graphics/2024-flood-fire-climate-risk-analytics/?