Why climate data matters to climate-related financial risk assessments
Aug 01 2022
Climate-related financial risk assessment is a relatively new field that has quickly become a central focus of climate policy. For instance, under the proposed Securities and Exchange Commission (SEC) climate risk rule, all publicly traded companies in the United States would be required to disclose how they are impacted by the physical risks of climate change, including slow-moving chronic risks such as drought and acute event-driven risks like flooding. In practice, this would require thousands of companies to pay a lot more attention to climate change — a fact not lost on the climate analysis and accounting services industry. Financial service providers are already reportedly spending billions to hire staff to help clients comply with the new rule. Climate science is now big business.
The proposed SEC rule reflects a broader change in how stakeholders across society relate to climate data — an interest area once limited primarily to researchers and policymakers. Financial institutions, publicly traded companies, and individual consumers are increasingly interested in understanding how physical climate risks will affect them. In turn, climate information is being incorporated into the decision to purchase a home or assess the creditworthiness of national governments. Practically, this means that broad swathes of global economy are impacted by how key financial system actors generate and interpret future-facing climate information, such as how projections of anthropogenic carbon emissions are transformed into the estimated economic risk of future water stress.11F Dolan et al. (2021) Evaluating the economic impact of water scarcity in a changing world Nature Communications In other words, the technicalities of climate data matter to more people than ever before.
We recently released an open dataset of globally downscaled climate models using the latest data from the Coupled Model Intercomparison Project (CMIP6). This dataset allows users to access and compare multiple downscaling methods across different future climate scenarios. Given the technical nature of our initial release, we anticipate that it will primarily be used by other climate scientists. Nevertheless, the analysis that went into our data release is also the foundation of the climate-related financial risk assessments undertaken across the public and private sector today.
This commentary focuses on climate data and analytics and the new wave of practitioners interested in using that data for physical financial risk assessments. We describe the basics of climate risk modeling and provide an overview of today’s climate analytics ecosystem. Finally, we identify why we think the financial risk community should focus more on the underlying communication, accessibility, and transparency of climate analytics.
Where does physical climate data come from?
Quantifying the future impacts of climate change requires using computer models to assess how the accumulation of greenhouse gasses in the atmosphere will cause damage to human and ecological systems. This process begins with global climate models (GCMs). Among other things, these models allow scientists to simulate the effects of different greenhouse gas concentrations on earth systems like the atmosphere and oceans — projecting changes to temperature, precipitation, and ocean chemistry.
To explore what might happen in future decades, scientists all over the world run GCMs under a range of potential emissions scenarios, coordinated within a global initiative called the Coupled Model Intercomparison Project (CMIP). In earlier CMIP iterations, these emissions scenarios were referred to as “Representative Concentration Pathways” (RCPs).22K E Taylor et al. (2012) An overview of CMIP5 and the experiment design Bulletin of the American Meteorological Society The most recent CMIP iteration increased the complexity and variety of scenarios by developing “Shared Socioeconomic Pathways” (SSPs).33V Eyring et al. (2016) Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization Geoscientific Model Development The SSPs explore the ways that emissions scenarios could develop under different plausible future worlds, incorporating societal factors like economic development, land use shifts, and varying levels of climate policy (e.g., negative emissions strategies). These RCPs and SSPs can be thought of as “stories” or “scenarios” describing different ways the world might look in the future. Scientists then feed these emissions scenarios into GCMs to evaluate the climate system response to each scenario.
Running and comparing multiple scenarios is important for at least two reasons. First, having a range of scenarios helps modelers project how different emission trajectories affect future climate outcomes, which helps policymakers and other stakeholders understand how efforts to reduce global emissions can avoid damages to human and ecological systems. Second, understanding the range of possible scenarios helps illustrate the uncertainty in future climate outcomes. Because of their central role translating between emissions scenarios and climate outcomes, GCMs are the foundational inputs for future-facing climate projections that inform data-driven public debate and policy making.
Location-specific climate data requires downscaling
Although GCMs form a foundational input to any downstream analysis, they are only a first step because GCMs operate at a coarse geographic scale. For instance, a future projection from a given GCM might describe precipitation across the entire state of New Jersey with a single data point. But average precipitation at the level of an entire state does not tell a homeowner useful information about where to move to avoid a flood. Instead, useful climate information should facilitate insights at a much finer level — at the scale of a house, an oil rig, or a highway.
To obtain more spatially granular data, we rely on a technique called “downscaling,” which refers to a family of analytical methods for transforming global climate model results into more localized projections — i.e., transforming statewide temperatures on a representative day in 2050 to the expected temperature of an individual county. Scientists run downscaling methods on GCM outputs, including temperature and precipitation, to produce spatially-refined results — the downscaled data from our recent release is at a ~25-kilometer spatial scale. The resulting data reflect localized geographic characteristics like mountain ranges, which are critical when assessing climate risks for a specific location or asset, but are generally absent from raw GCM outputs.
Downscaling is a key step, but translating downscaled climate data into projected hazards and then into climate-related financial risk almost always requires further analysis, which we turn to below.
From location-specific data to climate-related financial risk
The downscaled climate data in our CMIP6 webmap includes minimum daily temperatures, maximum daily temperatures, and precipitation. These inputs can be used to estimate the impacts of specific climate hazards — a term used by the Intergovernmental Panel on Climate Change to describe potential damaging physical events or trends.44V Möller et al. (eds.) (2022) Annex II: Glossary Climate Change 2022: Impacts, Adaptation, and Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change
As an example, the fire risks dataset we released last year used downscaled CMIP6 data as an input into a fire impacts model. That model in turn projects the probability that at least one moderate- or high-severity fire will occur within a 20-year period, providing important information about which geographic areas are riskier than others.
To understand climate-related financial risk, however, we need to further translate projected physical impacts of climate change into dollars. For example, projections of a future physical hazard like wildfire must be linked with information about the built environment or the real economy to produce insights into potential financial impacts. This linkage can be accomplished in a variety of ways, including simply overlaying a map of future climate-driven damages with investments in the same location.
In other cases, financial-risk analysis can extend all the way from geographic exposure to potential changes in financial assessment metrics, like when central banks evaluate economic resilience through stress testing. For example, in its 2021 economy-wide climate stress test the European Central Bank used physical climate data from a consultancy called 427 (now Moody’s Climate Solutions) to produce estimates of the probability of loan default under future projections of climate hazards. The 2021 ECB stress test did not affect bank activities, but subsequent climate stress tests could result in changes to bank capital requirements. Efforts like the ECB stress test, which link climate projections to financial impacts and potential regulatory actions, are rapidly emerging. They also represent a fundamental evolution in the kinds of analysis for which climate models have typically been used.
The rise of climate analytics providers
Meeting the needs of central banks, businesses, and consumers is a new role for climate change data. Over the past few decades, the provision of climate data and information has instead been broadly responsive to the needs of policy professionals and research scientists. Climate scientists are both producers of climate data — often through global efforts like CMIP6 — and also users of climate data in research applications. Likewise, policy professionals rely on climate models when developing policy proposals, such as efforts to calculate the social cost of carbon (SCC) or inform international climate negotiations.
Governments also provide data that is often used in economic analysis or for adaptation planning and resilience efforts. In the United States, federal agencies like the Environmental Protection Agency (EPA) and the Federal Emergency Management Agency (FEMA) collect and provide multiple climate datasets. For example, EPA publishes climate indicator dashboards and FEMA reports datasets of public disaster expenditures.
But today, increasingly, many individuals and firms rely on climate analytics service providers to help translate climate data — whether from GCMs or observations — into financial risks like the probability of mortgage loan default.55T Fiedler et al. (2021) Business risk and the emergence of climate analytics Nature Climate Change Climate analytics service providers target the gap between the existing climate data ecosystem and the newer needs of financial-sector and real-economy actors. They provide bespoke, location-specific analytics to understand where, and to what degree, investments, operations, and supply chains are at risk from climate impacts.
Despite the large number of raw datasets produced by researchers and government agencies, climate analytics providers exist because, in many cases, these datasets do not meet the “last-mile” needs of companies and individuals translating those data into location or asset-specific risks. There are several reasons for that disconnect.
First, as we describe in our downscaling explainer article, there are few open source downscaled datasets that make it easy to access and compare multiple downscaling methods.
Second, it is expensive to analyze climate models, both in terms of direct computational costs and the kind of expertise required to translate and produce risk analysis. Few firms have the necessary expertise in-house. Interviews with companies undertaking climate-related financial risk assessments confirm that these barriers lead most to rely on pre-processed data or third-party providers.
Finally, climate datasets provided by government agencies are often inappropriate for use in financial risk applications because the information is insufficient, whether in spatiotemporal scale, coverage, or data formats. Recently announced efforts to better integrate climate and financial data from different government sources are underway but still nascent.
These core challenges have led private companies to dominate the climate analytics market, where a recent wave of consolidations and mergers has concentrated several providers in legacy financial industry firms like Moody’s and S&P. While the majority of these companies contract directly with clients, major real estate sites like Redfin now also allow prospective homebuyers to see climate risk “scores” provided by third-party applications.
Climate service provision matters to financial risk assessment
As the impacts of climate change become more visible, there is growing demand among financial decision makers for physical climate risk assessments. More and more stakeholders are looking for rigorously developed climate risk products that are readily integrated with different financial and economic use cases. Right now, these data products and services are often provided by private firms, which fill gaps in government and academic data. But even as the climate data ecosystem has evolved in response to the demand for analytics, the rapid integration of climate data into business and financial risk assessments raises new considerations for providers and users.
Based on the work we did in our CMIP6 project and previous policy comments, we see three key areas where the broader business and financial risk community should engage in more depth with the underlying climate science.
01 — Investing in communication
The first key area for engagement is communication. The disciplines of physical climate science and finance have distinct technical language and skills. The acceptance of climate change as an important component of financial analysis and regulation has brought these disciplines into close contact with each other. Yet reviews of leading private sector financial disclosure standards, like the Taskforce on Climate-related Financial Disclosures (TCFD), show that these frameworks don’t align with canonical climate science terminology or hazard coverage from the IPCC. Efforts to collect climate risk assessment methods also note a lack of consistent terminology across academic, public, and private sectors.
Inconsistencies and conceptual discrepancies risk misunderstanding and misinterpretation of core scientific concepts. For instance, the scientific community is more confident in projected changes in extreme temperatures66S I Seneviratne et al. (eds.) (2021) Weather and Climate Extreme Events in a Changing Climate Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change than projected changes in hail,77T H Raupach et al. (2021) The effects of climate change on hailstorms Nature Reviews Earth & Environment which is a costly impact of growing interest to insurers. These hazard-specific considerations might seem esoteric to some, but they are of fundamental importance to physical risk assessments. Without a consistent understanding of what we know about each specific climate hazard, risk assessments could treat all climate hazard information as having the same degree of certainty. Better communication frameworks for crucial technical considerations across climate science and finance can help reduce the risk of misapplication of climate information.
02 — Supporting data availability and accessibility
The second area for engagement is data availability and accessibility. For some hazards and sectors like real estate, non-profit data providers are building products that can be accessed by the public. The risk scores provided by sites like Redfin, and other efforts like the Climate Impact Lab (CIL), take steps to fill gaps in publicly available information. However, many of these offerings provide only a limited amount of data for free download, reserving full data access to paying customers. Fully publicly available tools and datasets at a sufficiently high spatial granularity remain hard to access, and most require intensive storage and computational resources to leverage, though we note that some efforts to improve data accessability are underway.
Paywalls on climate data — especially the most useful climate data — by definition restrict access to those that can afford it, and in practice exclude many of those that need it. For instance, while hurricane-exposed municipalities in the U.S. would benefit from bespoke analytics, these services can be out of financial reach for local governments. Depending on the degree of provider transparency, paywalls also limit the ability of researchers, NGOs, and regulators to perform oversight and assess the integrity of providers.
03 — Improving methodological transparency
The third key area for more engagement is transparency. Today, climate analytics providers mostly don’t disclose how they do their analysis, or what they tell their clients about their data. This lack of transparency is often referred to as a problem of “black boxes”, where the “how” of what analytics providers do is not publicly available.
Black boxes in the climate analytics market are concerning because the information provided to clients can be significantly impacted by the assumptions and choice of methods used — for instance, the method used to downscale GCM data can influence the final downscaled results.88H Tabari et al. (2021) Comparison of statistical downscaling methods for climate change impact analysis on precipitation-driven drought Hydrology and Earth System Sciences ,99M A Sunyer et al. (2015) Inter-comparison of statistical downscaling methods for projection of extreme precipitation in Europe Hydrology and Earth System Sciences ,1010J T Abatzoglou et al. (2011) A comparison of statistical downscaling methods suited for wildfire applications International Journal of Climatology Our CMIP6 data release contributes to ongoing research efforts evaluating how different downscaling methods influence global climate change projections. More work is especially needed to understand which downscaling algorithm (or set of algorithms) are most appropriate for any given specific application.
It is also important to know how analytics providers are communicating their results to clients. For instance, are providers bracketing their findings with uncertainty estimates? To fully assess the robustness of an analytics provider, we must be able to understand not just which methods were used but also how the results are described to decision makers.
Next steps
The surge in demand for climate risk analytics reflects a growing universe of stakeholders that are preparing to use climate data to make decisions. The technical details of how we understand future climate risk matters more than ever. As regulators around the world begin to mandate physical risk disclosure, it’s clear that individuals, businesses, and regulators need high-quality climate data that can inform financial and economic planning. The question is how to provide that information within an analytics ecosystem that is equitable and technically robust.
Addressing the core issues of communication, data access, and transparency will require significant work to improve not only the underlying tools, but also to develop governance strategies that support a pipeline of high-integrity, decision-useful climate information in business planning and the financial sector. We look forward to working more on these issues in the future.