Foreword

The note highlights the increasing need for reliable physical climate risk data as climate impacts intensify and affect economic and financial stability. It emphasises persistent gaps in data availability, granularity, and comparability across hazards, regions, and sectors. The Foreword stresses the need for improved data-sharing, enhanced in-house expertise, and support for collaborative initiatives such as CLIMADA and Digital Twin projects.

Introduction

Central banks and supervisors use physical climate risk data for micro-prudential, macro-prudential, monetary policy, and research purposes. Survey results from 21 NGFS members show widespread use of such data, especially for macro-prudential analysis and statistical indicators. Major barriers include data availability, internal technical capacity, and the development of robust climate-risk databases. The note positions itself as a stocktake of leading practices and options to strengthen assessments.

Indicators of financial risks from physical climate risks

Physical risk arises from interactions between hazards, exposure, and vulnerability. Metrics fall into three categories: proxy metrics (e.g., aggregate historical losses), exposure-at-risk metrics (e.g., ECB’s Potential Exposure at Risk), and risk metrics incorporating vulnerability (e.g., Normalised Exposure at Risk and Collateralised Exposure at Risk).

Central banks are increasingly assessing financial impacts such as observed and expected GDP and asset losses, expected credit risk changes, and risk levels based on exposure. The report notes wide variation across commercial physical-risk scores due to different methodologies. Integrating vulnerability remains challenging and data intensive, particularly for granular damage functions.

Data needs to assess climate hazards, exposure, and vulnerability

Climate-related hazard metrics and data sources

Hazards span heat, floods, droughts, wildfires, storms, and sea-level rise. Surveyed authorities assess a broad range of hazards, with relevance varying by country context (e.g., droughts in Spain). Tools such as the Copernicus Climate Data Store and Google Earth Engine support hazard analysis. However, lack of granular and validated spatial data—especially in emerging economies—remains a major limitation.

Climate-related exposure data

Accurate exposure assessment requires granular asset-level data on location, value, characteristics, and ownership. Relying on headquarters locations can miss up to ~70% of expected losses. Privacy constraints restrict access to detailed asset data. Emerging databases (e.g., ETS facility data, Global Energy Monitor, Orbis, RIAD) support analysis but remain incomplete across countries and sectors. Satellite-based proxies (e.g., litpop) provide global coverage but cannot link assets to specific debtors.

Climate-related vulnerability metrics and data sources

Vulnerability depends on asset characteristics, function, and adaptive capacity. Granular damage functions link hazard intensity to monetary loss. Flood damage functions (e.g., Huizinga et al.) are the most developed; wind functions often rely on historical insurance claims (e.g., Emanuel). Heat-related damage remains complex due to varied labour and productivity impacts. Tools such as CLIMADA support event-based probabilistic modelling across hazards.

Physical climate risk from value chains

Indirect risks through supply chains can be significant but are difficult to quantify. Data gaps arise from limited disclosure of supplier networks and cross-border dependencies. Some studies integrate micro-level transaction data (e.g., Belgian administrative datasets). Macro-level models like MRIO and IOTA provide industry-average approximations but lack asset specificity.

Data considerations on climate change adaptation and resilience

Adaptation includes incremental and transformational measures. Central banks rarely incorporate adaptation comprehensively; current efforts focus primarily on flood defences. Maturity pathways recommend moving from baseline risk assessments to input and output metrics for adaptation.

Types and sources of data on adaptation actions and resilience strategies

Corporate surveys (e.g., CDP) collect information on adaptation investments, though consistency varies. LLMs and climate-specific NLP models (e.g., ClimateBERT) enable automated extraction of adaptation disclosures. Pilot frameworks, such as the Oxford-Zurich Adaptation Alignment Assessment, assess governance, risk processes, and targets using AI-supported retrieval. Nature-based solutions play an important role in reducing physical risks, though data availability—especially in emerging economies—remains limited.

The role of insurance data

Insurance coverage data is essential for understanding financial resilience but is often unavailable or highly aggregated. Supervisors may conduct ad hoc data collections, and proxy methods can estimate coverage levels where insurer data is incomplete.

Data challenges with estimating climate-related physical risk to the financial system

Bottom-up assessments require highly granular exposure and vulnerability data, which is costly and difficult to obtain. Top-down approaches face macroeconomic modelling challenges rather than data gaps. Combining both approaches provides complementary insights.

Lessons learned and potential next steps

Key actions include expanding capacity building (training, workshops), enhancing data-sharing across institutions, investing in robust multi-source data systems, and increasing funding for climate-risk research. Improved collaboration between central banks, statistical agencies, financial institutions, and researchers is essential to reduce data gaps and strengthen physical climate risk analysis.