Not All Climate Risk Models Are Created Equal

Climate risk modelling is rapidly becoming a core tool in financial decision-making. Banks are stress-testing portfolios, insurers are reassessing exposures, and asset managers are increasingly expected to quantify physical climate risks across their investments.

Yet a fundamental challenge remains: climate risk models can produce very different answers for the same asset in the same location.

This does not necessarily mean one model is wrong. Rather, it reflects the inherent complexity of translating climate science into financial risk metrics.

1. Good governance matters more than a single "correct" number

One of the most important — and often misunderstood — points about climate risk modelling is that there is no single "correct" climate risk number.

Different models may produce materially different results for the same hazard because they rely on different assumptions, datasets, and modelling techniques. This is particularly true for forward-looking risks where historical data provides only limited guidance.

Confidence therefore does not come from a single output but from strong model governance.

Financial institutions should evaluate climate models using frameworks similar to those applied to credit or market risk models. This includes:

• Transparency around methodologies
• Clear documentation of assumptions
• Peer review and scientific validation
• Ongoing model testing and benchmarking

Where possible, institutions should also consider a multi-model approach, comparing outputs across credible models to better understand uncertainty.

In climate risk modelling, governance, not precision, creates confidence.

2. Start with science, not marketing claims

The climate analytics industry is expanding quickly, and vendors often emphasize headline metrics such as spatial resolution or simplified climate "scores."

But these metrics can be misleading.

Institutions should begin with science-first questions when evaluating climate models:

• Are the underlying models peer-reviewed?
• Are assumptions clearly documented?
• Are uncertainty ranges provided rather than single-point estimates?
• Are the limitations of the model explicitly acknowledged?

Simplistic climate adjustments can also be a warning sign. For example, models that assume uniform warming impacts across regions, or that automatically translate increased rainfall into higher river flood risk, may overlook the complex physical processes that govern climate hazards.

Robust climate models should be grounded in credible science, not simplified shortcuts.

3. Resolution does not equal accuracy

Climate risk vendors often promote extremely high spatial resolution outputs, sometimes down to tens of meters.

But resolution alone does not guarantee accuracy.

The quality of the underlying data often matters more than the apparent precision of the output. For example, flood modelling depends heavily on the accuracy of terrain data, particularly digital elevation models (DEMs). Poor or inconsistent terrain data can smooth out risk patterns and produce misleading results across a portfolio.

Similarly, projections of future climate change rely on global and regional climate models that typically operate at 10–100 km resolution. Downscaling techniques can refine spatial detail, but they cannot fully reconstruct the physical processes occurring at smaller scales.

In other words, high-resolution outputs may sometimes give a false sense of precision.

Understanding the limitations of the underlying climate data is essential when interpreting model results.

4. Climate modelling is a chain — from science to financial loss

Translating climate data into financial risk metrics involves a long modelling chain.

A typical process includes:

• Raw climate projections
• Hazard modelling (e.g., flood depth or wind speed)
• Exposure modelling (asset characteristics and location)
• Vulnerability functions (how assets respond to hazards)
• Loss estimation and financial impact
• Portfolio aggregation and stress testing

Each step introduces modelling choices and assumptions that can materially influence the final results.

For example, two models may produce similar flood depths but very different loss estimates depending on how they model building vulnerability or insurance coverage.

Because of this complexity, institutions should ask vendors to clearly explain and document each stage of the modelling chain.

Understanding how climate science becomes a financial number is essential to interpreting the output correctly.

5. Forward-looking modelling and scenario thinking are essential

Perhaps the biggest challenge in climate risk modelling is that the most relevant risks lie in the future.

Backtesting models against historical events is important, but it is not sufficient. Climate change means that the future will likely include hazard combinations and intensities that have not yet been observed.

Models therefore need to be forward-looking, incorporating climate projections from sources such as the IPCC and testing results against multiple climate scenarios.

Flood modelling illustrates this challenge particularly well. Unlike heat or wind, flood risk requires detailed terrain data, complex water-routing physics, and the interaction of multiple climate drivers such as rainfall, river flow, and storm surge.

Historical flood maps, such as those produced for regulatory purposes, are therefore often insufficient for long-dated financial decisions like mortgages or infrastructure investments.

In addition, climate risks increasingly involve compound and cascading events: floods combined with storm surge, drought followed by wildfire, or infrastructure failures triggered by extreme weather.

These interactions remain at the frontier of climate science. Because they are difficult to quantify probabilistically, the most effective approach today is scenario analysis — exploring plausible impact chains and stress-testing portfolios against them.

In climate risk, the future is not an extrapolation of the past, and models must reflect that.

From models to decisions

Ultimately, the purpose of climate modelling is not simply to generate data.

It is to support better decisions about risk, resilience, and capital allocation.

For financial institutions, this means moving beyond the search for a single "correct" climate risk number and focusing instead on model governance, scientific credibility, and transparent uncertainty.

Climate models will never eliminate uncertainty.

But when built and used correctly, they can make that uncertainty visible — and manageable.