Refining Risk Assessments with High-Resolution Climate Simulations and Advanced Econometric Modeling

This interview with Nicolas Schneider, Senior Research Engineer - Macroeconomist at EDHEC-Risk Climate Impact Institute, was originally published in the December 2024 newsletter of the Institute. To subscribe to this complimentary newsletter, please contact: [email protected].

In this interview, Nicolas Schneider, Senior Research Engineer and Macroeconomist at EDHEC-Risk Climate Impact Institute, explores the advancement of high-resolution climate-economic modeling to assess climate risks at regional and subnational levels. He explains the necessity of moving beyond global averages to uncover localized vulnerabilities, threshold and non-linear effects, highlighting the innovative methods and data improvements behind this research. Nicolas discusses how this work builds on recent studies, addresses critical gaps in climate models, and delivers actionable insights for policymakers, investors, and financial institutions. He also shares key findings on regional economic impacts, the importance of adaptation strategies, and the role of robust climate projections in enhancing both our temporal and spatial understanding of future climate risks for investors.

Over the last couple of years, the Institute has focused its research on extending integrated climate economics modeling to incorporate the advances of climate science and make them suitable for financial economics and asset pricing applications. Our modeling has focused on the global macroeconomic effects of climate change. Your work represents a significant shift by examining economic impacts at a regional and subnational level. Could you explain why this more granular approach is necessary, what specific methods you have used to achieve this finer resolution, and how it changes our understanding of climate risks?

Nicolas Schneider: The boundaries between global and high-resolution estimates are actually quite porous. A bit of history might help here. In the early days of climate-economic modeling, Integrated Assessment Models (IAMs), such as William Nordhaus's DICE model, emerged as the first systematic attempts to quantify climate change damages in economic terms. The DICE model, formalized in the 1990s, integrates and extends Nordhaus's earlier work from the 1970s, which explored the economic trade-offs of rising CO₂ emissions and global warming risks. These models relied on global aggregate damage functions, often calibrated using cross-sectional studies that linked temperature increases to economic losses, particularly in agriculture and energy. The damage functions typically exhibited a quadratic form: impacts increased non-linearly with warming, but assumed a relatively smooth and gradual acceleration. 

While innovative, this approach had significant limitations. DICE’s quadratic functions smoothed out regional differences and often underestimated the severity of damages, particularly in extreme warming scenarios. They also relied heavily on advanced-economy data (e.g., the U.S. and Europe), failing to capture how disproportionate impacts emerge in regions or sectors with higher vulnerabilities. 

A pivotal insight came from agriculture-sector studies[1], which revealed a more pronounced non-linear relationship between temperature and productivity losses. For example, crop yields decline gradually up to a certain temperature threshold, but beyond that point, losses amplify disproportionately, leading to collapses in agricultural output. These studies emphasized that real-world impacts are not just non-linear but also threshold-driven—a key nuance missed by early IAMs. 

Building on these findings, Burke et al. (2015) moved beyond theoretical IAM assumptions by adopting an empirical approach to link observed country-level GDP data with temperature exposure. They demonstrated that temperature-driven economic impacts scale disproportionately as warming intensifies, mirroring the threshold effects seen in agriculture. Crucially, Burke et al. showed that aggregating GDP damages bottom-up—from grid-cell-level climate data—produced global loss estimates that were substantially higher than those predicted by the smooth, quadratic damage functions of IAMs like DICE. 

The next significant advancement came with Kotz et al. (2024), who increased the spatial resolution from 166 countries to 1,660 subnational regions. This revealed even sharper regional heterogeneities in climate exposure and economic losses, which further amplified global damage estimates. Kotz et al. also incorporated extreme weather components—like daily temperature variability and heavy rainfall—showing that impacts were not only non-linear but also driven by the frequency and intensity of climate shocks. 

In sum, this evolution—from early IAMs to high-resolution empirical studies—marked a major shift in understanding climate risks. Moving from aggregate, global models to more granular approaches has uncovered threshold effects and regional vulnerabilities that were previously overlooked, highlighting the need for even finer-scale analyses to capture the true scale of economic damages.

You have just finalised a major paper that models the economic impact of climate change impacts at an unprecedented granular level in terms of geography. This study comes on the heels of the paper by Kotz et al. (2024). Could you explain how our study builds on and extends these findings?

Nicolas Schneider: The influential work of Kotz et al. (2024), adopted into the latest scenarios of the Network for Greening the Financial System (NGFS), advanced our understanding of climate-driven economic risks by demonstrating the importance of regional detail and extreme weather events. Our study builds on this foundation by expanding its scope and ensuring greater consistency across methods and data.

We contribute to this important body of work in three key areas:

• 1. Higher Spatial Resolution:

We extrapolated the analysis to 3,672 subnational provinces, covering regions responsible for 95% of global economic production. This finer resolution reveals localized climatic exposure and economic heterogeneity with greater precision, uncovering impacts that are smoothed over at coarser scales. As a result, we observe a more severe global damage function due to the heightened sensitivity of certain regions.[2] 

• 2. Validation Across Data and Methods:

To strengthen confidence in our findings, we tested both Kotz et al.’s results and our extensions using alternative climate datasets and multiple statistical methods, including parametric, non-parametric, and semi-parametric approaches. This validation demonstrates the robustness of our extensions, while aligning with Kotz et al.’s core findings on regional disparities and climate-driven economic damages.

• 3. Refining Climate Model Inputs:

We addressed known biases in a small subset of the climate models underpinning the work of the Intergovernmental Panel on Climate Change (IPCC) which overestimate warming due to cloud feedback biases. Following Hausfather et al. (2022), we excluded these models to ensure our damage projections are based on plausible and reliable climate scenarios, reducing the risk of upward bias.[3].

By improving the spatial resolution, validating results across multiple data sources and methods, and ensuring reliable climate inputs, our study provides a more detailed and consistent assessment of climate-driven economic risks.

This analysis highlights two critical insights:

• 1. Localised risks matter:

Small-scale variations in climate exposure and economic activity can produce disproportionately large impacts, particularly in vulnerable regions and sectors.

• 2. Global implications:

Accurately capturing these local effects increases aggregate global damage estimates, underscoring the urgency of both adaptation and mitigation strategies.

In short, we complement and extend Kotz et al.’s work by offering greater spatial detail and ensuring methodological robustness, providing policymakers, central banks, and investors with reliable insights into climate risks.

Fig. 1: Spatially distributed province-level projections of climatically-driven temperature shift impacts (%) on gross regional output per capita, epoch 2099 compared to historical baseline. Color gradient shows the multi-model median impacts of 15 ’likely’ CMIP6 global climate models (GCMs) simulated under a SSP5-8.5 vigorous warming scenario.

Source: EDHEC-Risk Climate Impact Institute

Your research includes catastrophic economic projections, with GDP losses of up to 85% in certain regions under extreme warming scenarios. However, our team’s probabilistic work suggests that more moderate warming environments are much more likely by the end of the century. Why report results for extreme scenarios, and what do your projections reveal for the more likely scenarios? Which regions or sectors stand out as particularly vulnerable or resilient based on your high-resolution approach?

Nicolas Schneider: We report results for both severe (RCP8.5) and moderate (RCP4.5) warming scenarios to provide a balanced perspective on potential outcomes. What we can say is that the probabilistic analysis we are conducting in parallel suggests that RCP8.5 is far from being a central scenario but also that it is not a complete outlier. There is therefore prudential value in taking it seriously, as it allows us to explore tail risks that are extreme but not impossible, particularly in the absence of strong mitigation. 

In short, our high-resolution approach identifies agriculture and energy as critical focal points, given their sensitivity to weather fluctuations and their central role in the macroeconomic transmission of climate risks. Regions like West Africa, Central America, and parts of South Asia emerge as particularly vulnerable under both moderate and extreme warming scenarios, while sectors tied to food security and energy systems demand the greatest attention. 

Under moderate warming (RCP4.5) by the end of the century, GDP losses remain significant but are more geographically concentrated and relatively manageable, with declines of up to 25% in vulnerable regions. These include parts of West Africa, Central America, and South Asia, where persistent high temperatures strain agricultural systems and energy infrastructure. Parts of southern Europe and the south-western United States also stand out due to rising heat stress, which undermines infrastructure performance and reduces labour productivity. 

By contrast, under severe warming (RCP8.5) by the end of the century, GDP losses could reach up to 85% in the most exposed regions. Areas such as northern Nigeria, western Mali, and parts of tropical Asia are particularly vulnerable due to compounding effects, including agricultural collapse under sustained heat extremes and limited adaptive capacity. Coastal areas, meanwhile, face additional risks such as erosion, sea-level rise, and intensifying extreme weather events, which amplify damages to infrastructure and economic stability. 

Adaptation will be crucial but constrained. In regions already struggling with extreme heat, producers may rely on intensive margins of adaptation—such as increasing irrigation, fertiliser use, or adopting alternative crops—to mitigate productivity losses. However, these measures face practical and economic limits, underscoring the importance of early mitigation to reduce warming and avoid catastrophic, irreversible damages. 

By presenting results for both scenarios, our work equips decision-makers with a comprehensive understanding of potential futures: moderate outcomes highlight where adaptation strategies can have the greatest impact, while extreme outcomes underscore the urgency of mitigation.

While much attention often centers on end-of-century outcomes—understandably so, as climate action taken now can still significantly reduce the severity of those long-term consequences—your research also includes projections for mid-century. These mid-century outcomes, however, reflect risks that are largely built-in due to current trajectories, regardless of whether emissions are curbed. What do these mid-century scenarios reveal about regional economic risks, and how can they inform near-term strategies for investors and policymakers?

Nicolas Schneider: Mid-century projections (2040–2050) provide a practical near-term window for understanding climate risks, reflecting damages that are largely locked-in due to historical and current emissions. Unlike end-century simulations, mid-century results carry higher reliability because they involve shorter extrapolation horizons and are underpinned by a vast ensemble of simulations—spanning 30 global climate models, multiple emission pathways, and time horizons. This ensemble-based approach improves the robustness of our findings and allows us to identify regional vulnerabilities with greater precision. 

For a long time, the climate space has largely focused on mitigation targets, particularly reducing fossil fuel use in the energy supply mix. While mitigation remains crucial for long-term stability, the near-certainty of extreme weather events and climate shocks by mid-century underscores the need to place at least as much emphasis on adaptation. It is important to note, however, that mitigation and adaptation objectives may sometimes conflict—for example, in the energy sector, where infrastructure investments for adaptation may rely on high-carbon materials. 

Our work aims to support this shift toward adaptation by answering three key questions that are particularly relevant for investors seeking regional climate risk solutions:

• 1. What is the size of future climate shocks at the local level, and how do they compare to the global average?

High-resolution projections reveal significant heterogeneity in climate risks, with tropical regions and low-lying coastal areas expected to experience above-average damages. For instance, mid-century results indicate severe heat stress and crop failures in parts of West Africa, South Asia, and Central America, while rising sea levels and storm surges threaten coastal infrastructure and habitability.

• 2. In a warmer future, how much will it cost to produce the same output as today?

As physical risks intensify, maintaining economic productivity in exposed regions will require significant additional input factors—such as irrigation, fertiliser, and mechanisation in agriculture, or energy-intensive cooling systems in infrastructure. These adaptations come at a rising cost, particularly in regions with limited economic capacity to respond.

• 3. Who will pay these costs, and what are the asset pricing implications?

The adjustments needed to sustain productivity will not be distributed evenly. Regions and sectors that are heavily exposed to mid-century risks will face mounting climate premiums in asset pricing. Investors and regulators will need to assess how these premiums reflect localised risks and who ultimately bears the economic burden of adaptation.

 While adaptation strategies—like irrigation, mechanisation, and changes to crop varieties—offer pathways to limit economic losses, geography imposes a fixed constraint on their effectiveness. Land is immobile, unlike other production factors, meaning highly exposed regions—particularly those near the equator—will continue to face declining productivity even with adaptation. This raises critical questions about how to shift land usage to more climate-resilient areas to balance global food supply and economic stability. Achieving this will require large-scale processing of high-resolution physical climate data and remote-sensing information to quantify land productivity shocks and their broader macroeconomic implications.

In summary, mid-century projections, grounded in robust ensemble-based approaches, reveal inevitable physical risks that demand immediate attention. While mitigation remains essential to avoid catastrophic end-century outcomes, mid-century results underscore the immediate need to prioritize adaptation strategies and assess their regional costs and financial impacts. By answering these questions, our work helps investors and policymakers identify solutions that address both the scale and distribution of climate risks, while acknowledging the structural limits imposed by geography.

Your study employs high-resolution climate simulations and advanced econometric modelling. Looking ahead, what advancements in data or methods do you think could further improve our understanding of the economic impacts of climate change, particularly at a regional level?

Nicolas Schneider: The results we presented rely on an extensive volume of climate and economic data, processed through high-resolution modelling frameworks. Specifically, we use:

• 30 global climate models spanning multiple scenarios to project regional climate outcomes.
• Methods to improve the regional detail of these projections, achieving resolutions that can identify localised impacts.
• Robust, harmonised datasets that ensure consistency and reliability across regions and sectors.

Processing this scale of data requires advanced computational tools, such as high-performance computing systems and shared cloud clusters, which allow us to simulate thousands of climate trajectories. These tools improve the precision of regional risk assessments and help identify how climate change could impact economic activity and asset performance.

Looking ahead, while current models provide valuable insights, gaps remain. Existing projections still face trade-offs between temporal detail (e.g., hourly heatwaves or extreme rainfall) and regional precision. Achieving both requires significant computational resources and data processing capabilities, which current tools cannot yet fully support. Addressing these gaps would allow us to better capture short-lived but intense shocks—like floods and storms—that drive significant localised damages.

In the future, the creation of a “best-in-class” data product combining global coverage at higher spatial precision and detailed, high-frequency variability would significantly improve our ability to assess risks at a regional level. Such advancements will refine localised risk assessments, enhance sector-specific projections, and provide the granularity needed for investors to price risks more effectively.

One final direction involves improving the performance of non-parametric statistical methods to handle the increasingly large climate datasets used in research.[4]

By enhancing the precision of climate-economic insights, these advancements will enable investors to stress-test portfolios, refine asset valuations, and better anticipate how physical risks will affect their holdings.

Footnotes

[1] See inter alia: Schlenker and Roberts, 2009; Burke and Emerick, 2016.

[2] The amplification of global damages reflects the greater heterogeneity revealed at finer scales, which coarser models inherently smooth over.

[3] The exclusion targets models with Equilibrium Climate Sensitivity (ECS) above 4°C or Transient Climate Response (TCR) above 2.2°C. In the Coupled Model Intercomparison Project Phase 6 (CMIP6) ensemble, less than a handful—4 out of 21 models—exceed these thresholds.

[4] While Bayesian Additive Models (BAM) already outperform their Generalized Additive Model (GAM) counterparts through approximated fitting procedures (e.g., restricted maximum likelihood with iterative updates), further advances in computational efficiency and memory optimisation will be critical to scale these methods for global climate-economic studies.

References

• Burke, M., \& Emerick, K. (2016). Adaptation to climate change: Evidence from US agriculture. American Economic Journal: Economic Policy, 8(3), 106-140.
• Burke, M., Hsiang, S. M., \& Miguel, E. (2015). Global non-linear effect of temperature on economic production. Nature, 527(7577), 235-239.
• Hausfather, Z., Marvel, K., Schmidt, G. A., Nielsen-Gammon, J. W., \& Zelinka, M. (2022). Climate simulations: recognize the ‘hot model’ problem. Nature, 605(7908), 26-29.
• Hausfather, Z.,\& Peters, G. P. (2020). RCP8. 5 is a problematic scenario for near-term emissions. Proceedings of the National Academy of Sciences, 117(45), 27791-27792.
• Kotz, M., Levermann, A., \& Wenz, L. (2024). The economic commitment of climate change. Nature, 628(8008), 551-557.
• Schlenker, W., \& Roberts, M. J. (2009). Nonlinear temperature effects indicate severe damages to US crop yields under climate change. Proceedings of the National Academy of sciences, 106(37), 15594-15598.

About Nicolas Schneider, Senior Research Engineer at EDHEC-Risk Climate Impact Institute

Nicolas Schneider is a Senior Research Engineer and Empirical Macroeconomist at EDHEC-Risk Climate Impact Institute. He is working on expanding the Institute’s capabilities in spatial climate econometrics, allowing it to parametrise and therefore project the impact of climate change on the economy. The aim is to enhance the granularity with which the Institute characterises climate change and economic output (considering elements such as regional and sectoral effects. Nicolas holds a PhD obtained in the Geography and Environment Department of Boston University where his doctoral research used high performance computing systems to conduct large-scale processing of high-resolution time- and spatially-downscaled global climate model simulations and generate projections of future climate shift-driven impacts on various economic, energy and agriculture outputs. He holds an MSc in Environmental Economics and Climate Change from the London School of Economics and Political Science and a MSc in Development Economics and Applied Econometrics from Paris-Pantheon Sorbonne University. He has co-authored over a dozen peer-reviewed articles and taught in the Department of General Education at Harvard University and in the Quantitative Methods Seminar of the Oxford Smith School. He has consulted for the New York City's Mayor Office of Climate and as a fellow in the Program on Coupled Human and Earth Systems funded under the US Department of Energy's MultiSector Dynamics Modeling Initiative.