Weird science

OxEO is raising its first external investment round. This is the fifth in a series of five short posts written to support that effort. If you are a qualified investor and would like to see our deck, please get in touch.

Climate is hot. Every politician, philanthropist and pundit has things to say about the climate emergency, and financial market service providers — always attentive to the zeitgeist — have been quick to respond. In recent years, an industry has developed around ‘climate-related financial risk’ — the practice of integrating what is known about climate change into financial decision-making and disclosure. This is the era of the Climate Service Provider (CSP).

Broadly, CSPs create products using data generated by General Circulation Models (GCMs) to measure financial risk. Most access the wide array of projections data available for free through the Coupled Model Intercomparison Project (CMIP), which provides scenario-based distributions of forecasts. CSPs typically combine predictive climate analytics with other data to assess the risk of an asset or company. Their approaches for doing so are proprietary, and the detailed methodology — the recipe for their secret sauce — is not normally published.

However, most CSPs apply a common ‘Hazard — Exposure — Vulnerability’ framework to their risk assessments. Hazards are potentially destructive physical phenomena, such as floods and wildfires. Exposure relates to the location and attributes of physical assets that might be affected by the hazards. And vulnerability is the likelihood of assets being affected by hazard, based on their exposure.

CSPs use downscaled outputs from the CMIP models to identify hazards and predict vulnerability under different scenarios. Nearly every CSP uses variations on the same set of hazards — because these are the ones that feature in the open access models. These hazards are flood, drought, extreme wind, heat and cold; wildfires, and earthquakes.

Here’s the thing, though. For hazards to present material risk, they almost always need to be associated with extreme weather events. Climate models can tell us many important things. But they cannot predict the weather. As Tanya Feidler and colleagues explain in a recent, excellent paper:

“Enabling GCMs to simulate weather events reliably requires a revolution in climate model design because GCMs do not represent or capture weather-scale phenomenon (defined here as 1–10 km). Given GCMs cannot currently resolve the interactions of the changing climate envelope with weather phenomena, the use of these models to inform business decisions influenced by these phenomena is a misuse.”

In other words: climate is what you expect, weather is what you get. Conflating the two is a problem, no matter the access a CSP has to the best science. At OxEO, in common with CSPs, we use climate models to inform our insights by working at the intersection of earth observation science and climate science. However unlike most CSPs, climate model outputs are not the primary source of our value added. We use earth observation to assess the sustainability footprint (GHG emissions, land use, water use, biodiversity) of real assets. GCMs support our capabilities by informing longer term projections on the size and shape of an asset’s sustainability footprint. As CGMs continue to improve, so will our understanding of the relationships between climate, weather and the variables that we observe.