A database-integrated framework to support EU taxonomy aligned physical climate risk assessment in the energy sector

Climate change increases physical risks to energy infrastructure, and sustainability regulations such as the EU Taxonomy consequently require that companies conduct physical climate risk and vulnerability assessments consistent with IPCC concepts and ISO standards. However, existing guidance is often constrained by high data requirements. This study proposes an energy sector database-integrated framework that identifies climate hazards relevant to a specified activity at a given location and then assigns corresponding climate indicators to extract scenario-based climate projections, using a systematic literature review and open access climate data from the Copernicus Climate Data Store. The review of 23 Climate-ADAPT publications identified 4,461 hazard–activity associations across 31 EU Taxonomy energy activities. These associations were linked to 79 relevant Copernicus indicators, resulting in 143 documented hazard–indicator matchings with information on spatial resolution, scenario coverage, and temporal range. The framework requires three readily available inputs: economic activity, technological specification, and geographic location, and incorporates additional geospatial datasets to filter relevant hazards and assign suitable indicators. A case study on electricity generation from solar photovoltaic technology in Aachen, Germany, demonstrates that the framework streamlines hazard identification and enables retrieval of high-resolution, scenario-based climate projections. The framework thereby operationalizes the hazard identification and quantification stages of EU Taxonomy-aligned climate risk and vulnerability assessments, whereas the subsequent evaluation of vulnerability and materiality remains the responsibility of the assessing practitioner. Overall, the framework reduces data burden, improves transparency, and supports scalable application across energy technologies, while facilitating the integration of climate adaptation into corporate planning and decision making.

Climate change (CC) is reshaping the global risk landscape. Extreme weather events rank among the most severe long-term global risks and are becoming more frequent and intense due to CC (Seneviratne et al., 2021). Economic losses from damage to critical infrastructure could increase by more than 1,500% by the end of the century (Forzieri et al., 2015). In the EU energy sector, natural hazards already caused about 37% of power outages between 2010 and 2016 (Tavares da Costa et al., 2023), underlining the need for systematic climate risk assessments (CRAs) to strengthen resilience and reduce losses.

CRAs are increasingly embedded in climate adaptation policy and regulation (European Parliament, 2020; IPCC, 2022, 2023). Under the EU Taxonomy Regulation, climate change adaptation (CCA) is one of six environmental objectives, and both substantial contribution and do no significant harm assessments require a robust climate risk and vulnerability assessment for material physical climate risks (EU Commission, 2021; European Parliament, 2020), following standardized methods. These assessments are expected to follow IPCC concepts and standardized approaches such as ISO 14090:2019 and ISO 14091:2021, which define climate risk as the interaction of hazard, exposure and vulnerability, with vulnerability reflecting sensitivity and adaptive capacity (EN ISO, 2019, 2021; IPCC, 2022, 2023).

However, operational implementation remains difficult. While ISO 14090 and ISO 14091 provide an important conceptual basis, they offer limited guidance on translating these concepts into asset-level assessments and EU Taxonomy-compliant workflows (EN ISO, 2019, 2021).

This challenge is particularly relevant in the energy sector, which contains 31 of the 148 EU Taxonomy economic activities and is the sector with the highest number of activities eligible to make a substantial contribution to CCA (EU Commission, 2021, 2023a, 2023b). The sector was prioritized for its high climate relevance (Technical expert group on sustainable finance, 2020). Indeed, the energy sector is highly relevant, accounting for 27% of the EU-27’s total greenhouse gas (GHG) emissions in 2022, while its energy infrastructure is itself vulnerable to climate-related hazards, requiring targeted adaptation measures (European Environment Agency, 2024b; IPCC, 2023). Accordingly, the European Environment Agency (EEA) calls for CCA integration into infrastructure planning, consistent with the EU Taxonomy requirements (European Environment Agency, 2024a; European Parliament, 2020).

In practice, Taxonomy-aligned CRA is often hindered by high data requirements and a lack of standardized, accessible information (Ostojic et al., 2024). Existing institutional, academic and commercial approaches often lack transparency, asset-level specificity, scalability or full alignment with the IPCC risk framing and EU Taxonomy requirements (Bank of England, 2019; Black Rock, 2019; Luo et al., 2023; Task Force on Climate-related Financial Disclosure, 2020). Commercial software tools such as Moody’s Analytics ESG and climate risk tool (Moody’s Analytics, 2025), Repath (Repath, 2025), or Aladdin Climate (Black Rock, 2025) enable scalable assessments but rely on proprietary, undisclosed methods, reducing transparency and asset-level accuracy.

Public guidance remains limited. Among public guidance, Dorsch et al. (2022) provide a stepwise framework for non-financial firms aligned with ISO 14091 and the EU Taxonomy, but their approach focuses on the manufacturing sector and does not establish explicit activity-specific hazard linkages or assign documented climate indicators with metadata on spatial resolution, scenario coverage and temporal range.

More specialized academic methods exist for selected energy technologies and assets, but they often require extensive asset-level data, remain limited in technology scope, or rely on threshold-based approaches that do not fully reflect IPCC-recommended risk concepts (EN ISO, 2019; Karagiannakis et al., 2025; Luo et al., 2023; Sousa et al., 2022).

At the same time, the EU Taxonomy requires CRAs to screen all relevant hazards listed in Appendix A of the Climate Delegated Act, use state-of-the-art climate projections at the highest available spatial resolution across multiple future scenarios, consider activity lifetimes of at least 10–30 years, and evaluate hazard, exposure and vulnerability consistently with IPCC and ISO concepts (Dorsch et al., 2022; EN ISO, 2021; EU Commission, 2021).

Database-integrated concepts show that linking activities and locations to climate datasets can minimize inputs (Luo et al., 2023), but existing models cover only selected technologies and lack transparent mapping between activities, hazards, and measurable indicators. Thus, a comprehensive open-data workflow translating Taxonomy requirements into practice for all energy activities is still missing.

Methodologically, the study connects several strands of literature: procedural guidance for operationalizing regulatory climate risk requirements (Dorsch et al., 2022; Luo et al., 2023), research on the methodological implementation and data requirements of the EU Taxonomy (Achterfeldt et al., 2026; Ostojic et al., 2024), and emerging scholarship on the policy and governance dimensions of technology-enabled risk mitigation tools (Putera et al., 2026). This study addresses a gap by developing a database-integrated framework supporting CRA, tailored to the energy sector. Building on Taxonomy-oriented procedural guidance (Dorsch et al., 2022) and database-integrated concepts discussed in the literature (Luo et al., 2023), the framework advances existing work in four ways: Covering all 31 EU Taxonomy energy-sector activities.

Developing a documented database that links these activities to relevant physical climate-related hazards based on evidence from Climate-ADAPT, the European Climate Adaptation Platform jointly managed by the European Commission and the EEA (Climate-ADAPT, 2025a).

Translating identified hazards into quantifiable indicators derived from the Copernicus Climate Data Store (CDS) and documents relevant metadata such as spatial resolution, temporal coverage, and climate scenarios.

Integrating these elements into a standardized workflow requiring only three inputs related to activity and location, complemented by geospatial filtering, thus reducing data burdens while maintaining asset relevance.

Positioned within the risk concept of ISO 14091 and the IPCC, in which risk results from the interaction of hazard, exposure, and vulnerability, the framework fully operationalizes the hazard dimension through systematic identification and indicator-based quantification, partially captures the exposure dimension through locational screening, and defers the evaluation of vulnerability and materiality to subsequent stages of the assessment process. By combining literature evidence with open climate data, the framework provides a transparent and scalable approach to support consistent, reproducible and regulatory-relevant CRA across the energy sector.

The remainder of this paper is structured as follows. Section 2 describes the four methodological steps of framework development and demonstration. Section 3 presents the results in two parts, the developed databases and the integrated framework (Section 3.1) and their illustrative demonstration in a case study (Section 3.2). Section 4 discusses the findings, their regulatory implications, and the limitations of the approach, and Section 5 concludes.

This study develops a hazard- and indicator-integrated framework to support physical CRA for energy-sector economic activities under the EU Taxonomy. It builds on the practical step-by-step guidance of Dorsch et al. (2022) and the requirement to conduct a robust physical climate risk and vulnerability assessment for Taxonomy-alignment (Dorsch et al., 2022; EU Commission, 2021). Whereas Dorsch et al. (2022) provide high-level procedural guidance, this study delivers its operational implementation for the energy sector through an integrated hazard–activity and hazard–indicator database linking all 31 EU Taxonomy energy activities to quantifiable climate indicators with documented metadata.

The methodological approach comprises four steps (Figure 1): (i) building a literature-based hazard–activity mapping database using a structured review guided by PRISMA, (ii) compiling and characterizing relevant CDS datasets and matching climate indicators to EU Taxonomy hazards, (iii) synthesizing both databases into a location- and technology-sensitive workflow, and (iv) demonstrating applicability through a case study.

The case study applies the framework to the activity “Electricity generation from solar photovoltaic technology” for Aachen, Germany, to illustrate the framework’s workflow, comprising hazard identification, locational filtering, indicator assignment, and exemplary data retrieval and visualization. The case study is intentionally limited to demonstration: it does not conduct a full CRA and does not assess the materiality of identified hazards, but illustrates how the framework can support CRA practitioners in line with EU Taxonomy requirements.

Step 1: mapping climate-related hazards to economic activities

Relevant climate-related hazards for EU Taxonomy energy activities were identified through a structured PRISMA based (Moher et al., 2009) review of scientific and grey literature.

The resulting Hazard-Activity Database maps Taxonomy hazards to energy-sector activities while recording technological and locational characteristics and affected value-chain segments. This aligns with the Dorsch et al. (2022)’s hazard identification procedure, which filters geographically or technologically irrelevant hazards for a given activity.

Following Dorsch et al. (2022), who argue that for economic activities generally the going concern principle applies, the framework assumes that EU Taxonomy activities generally have a lifespan exceeding 10 years. Therefore, long-term climate projections were used. The determination of system components [“investigation objects” in Dorsch et al. (2022)] is not the focus of this study; hazards are assigned at the level of economic activities rather than individual components. To retain a system-oriented perspective and support practical application, documented results include illustrative impact chains and indicate whether a hazard is relevant to the core activity, the upstream/downstream value chain, or only under specific technologies.

The methodological steps of the literature review and database development are detailed in the following:

Data source and search strategy

The literature search (on 05.05.2025) used the EU Climate-ADAPT platform, a European knowledge-sharing portal supporting CCA planning. Climate-ADAPT was deliberately selected as the sole evidence source for four reasons. First, as the official European adaptation knowledge platform, jointly maintained by the European Commission and the EEA, it provides an institutionally curated, quality-assured evidence base directly aligned with the policy context of the EU Taxonomy. Second, its catalogue combines peer-reviewed research with authoritative technical and governmental reports, including assessments by the International Energy Agency, the International Atomic Energy Agency, the EEA, the Joint Research Centre, and the World Bank, therefore, extending well beyond academic journal literature (see Appendix Table A. 1) (Climate-ADAPT, 2025a). Third, several of the retrieved documents are themselves comprehensive reviews or national assessments synthesizing large bodies of primary research, so the effective evidence base is considerably broader than the document count suggests. Fourth, restricting the search to a single curated platform with a documented PRISMA-guided procedure ensures a transparent and reproducible selection process. The implications of this choice for comprehensiveness are discussed in Section 4. To focus on energy-relevant evidence while remaining inclusive, the resource catalogue was filtered by item type (“Publications and Reports” and “Case Studies”), adaptation sector (“Energy”), and language (“English”), with no additional restrictions (e.g., hazards or geographic scope) (Climate-ADAPT, 2025b).

Screening and selection process

Document selection followed the phases of identification, screening, and inclusion documented in Figure 2. During identification, 120 records were retrieved from the resource catalogue and one duplicate was removed, leaving 119 unique records.

Screening proceeded in two consecutive steps. First, titles were screened, and records were excluded only where the title clearly indicated that a document lies outside the scope of the review, for example because it evidently addressed adaptation sectors other than energy or exclusively concerned climate change mitigation. Because titles alone rarely permit definitive judgements on content, this step was applied conservatively and records of uncertain relevance were retained, resulting in the exclusion of 78 records. Second, the abstracts of the remaining 41 records were assessed for indications of a substantive focus on climate-related hazards, risks, or impacts of the energy sector or of specific energy activities. Records were excluded when their abstracts gave no indication of such a focus and the documents were therefore not expected to contain hazard-related information suitable for extraction, for instance because they centred on mitigation measures, policy instruments, or general adaptation strategies. Records without openly accessible full texts were likewise excluded, as data extraction requires the complete document. In this step, 11 records were excluded for insufficient topical fit and 7 for lacking access.

The remaining 23 documents were included as the evidence base for the subsequent analysis. Screening was conducted by a single reviewer applying the predefined criteria described above. The conservative retention of uncertain records at title level, the explicit exclusion criteria, and the full documentation of the selection (see Figure 2; Appendix Table A. 1) mitigate the risk of selection errors and ensure the traceability of the process.

Literature analysis and database structure

Each included publication was analysed to identify every instance in which a climate-related hazard listed in the EU Taxonomy is discussed in relation to a specific energy activity, a group of activities, or the energy sector as a whole. Each such instance, hereafter referred to as a hazard mention, was recorded with the following attributes: (i) the climate-related hazard, (ii) the referenced activity, activity group, or the sector as a whole, (iii) the affected value-chain segment, with core operations assigned by default where the source provided no specification, and (iv) technological or locational conditions stated in the source. Coding relied on closed classification lists comprising the climate-related hazards of the EU Taxonomy, applied on the basis of the definitions compiled in Appendix Table A. 3, and the energy-sector activities of Appendix Table A. 2. Only mentions that could be attributed to one of the listed hazards on the basis of these definitions were included in the database. All hazard mentions were recorded irrespective of whether the source judged the hazard as material, because materiality depends on context-specific exposure and vulnerability and is therefore assessed within the practitioner’s subsequent evaluation rather than encoded in the database. For each entry, the associated risk narrative was captured to document why a hazard is discussed as relevant. The complete entry-level documentation, comprising source, assignment level, qualifiers, and risk narrative, is provided in Supplementary Table 1, and the corresponding documentation of each hazard–indicator matching in Supplementary Table 2. Data extraction was performed by a single reviewer applying these predefined coding rules. Mention frequencies serve exclusively as descriptive statistics of the evidence base and play no role in the identification of relevant hazards within the framework. The data were organized in a Microsoft Excel database using table and pivot functions, supporting future digital integration.

Grouping of economic activities

Because EU Taxonomy activities are often more specific than the scope used in the reviewed literature, activities were clustered into technology groups (e.g., thermal power plants) to enable consistent hazard assignment across comparable activities. The grouping is documented in Figure 3, and hazards were subsequently reassigned to the original disaggregated EU Taxonomy activities to obtain an activity-level database.

*Numbers are taken from Figure 3.

Step 2: linking climate-related hazards to relevant climate variables

The second methodological step identifies and assigns climate indicators from publicly available datasets to the climate-related hazards captured by the EU Taxonomy.

It corresponds to the third step in Dorsch et al.(2022)’s practical CRA guidance and addresses the EU Taxonomy requirement to use the latest and highest-resolution climate projections (Dorsch et al., 2022; EU Commission, 2021).

In line with the EU Taxonomy’s recommendation of the CDS as a suitable data source (EU Commission, 2021), CDS datasets (C3S, 2025a) were systematically screened and matched to the climate-related hazards of the EU Taxonomy applying the following steps:

Data collection

To ensure relevance for EU-based applications and the use of climate data about the present as well as future prediction projections, the CDS catalogue was filtered for datasets with (i) spatial coverage “Europe” and (ii) temporal coverage “Present” and “Future.” This filter returned 15 datasets; three were excluded because one did not provide future projections despite the filter setting, and two focused on highly specific variables (winter sports; tiger mosquitoes) that were outside the scope of this study. A detailed list of all datasets and the rationale for exclusion can be found in Table 1.

For each included dataset, all available climate indicators were documented in a structured Excel table to support traceability and future digital integration.

Recorded metadata includes unit, variable documentation, spatial resolution, temporal coverage, smallest temporal aggregation, availability of historical records, and availability of Representative Concentration Pathways (RCP) scenarios.

Hazard assignment

Each variable was matched to the EU Taxonomy hazard it best represents by comparing dataset documentation with hazard definitions (see Appendix Table A. 3).

Variables without clear correspondence were noted as non-relevant.

Step 3: framework synthesis

Findings from Steps (1) and (2) are combined into a structured framework for systematic hazard identification and evaluation of EU Taxonomy activities. Following Dorsch et al. (2022) it connects the underlying databases in a way that supports consistent hazard selection and indicator assignment for a specified activity.

To incorporate location-specific exposure conditions, the framework integrates geospatial datasets [CORINE Land Cover (CLC) and GlobPermafrost] capturing locational features affecting hazard exposure. Finally, the framework is operationalized as a process workflow that links (i) user inputs (economic activity, technology specification, and geographic location) with (ii) the hazard–activity and hazard–indicator databases and (iii) geospatial layers, enabling users to identify relevant hazards for a given activity and location and retrieve suitable, pre-processed climate data to support subsequent risk assessment.

Step 4: case study: practical application of the framework

A case study illustrates the framework’s application using real climate data. The selected activity is Electricity generation using solar photovoltaic technology (EU Commission, 2021), in Aachen, Germany.

Germany was selected because it is the EU’s largest industrial economy and hosts a large number of companies affected by EU Taxonomy reporting requirements (Eurostat, 2025).

Solar PV was chosen for its high asset volume in Germany: 76,663 PV installations with capacity >100 kW were registered (31.07.2025), compared with 30,522 wind plants (on−/offshore) (German Federal Network Agency, 2025; Appendix Table A. 4). Given the high number of distributed PV assets, a scalable and systematic approach to support CRA is particularly relevant for this activity.

Procedure

The case study applies the framework to (i) identify relevant climate-related hazards for the selected activity and location and (ii) retrieve corresponding climate indicators. Only the most frequently cited hazards in the reviewed literature were further analyzed, selected based on their frequency of occurrence across the included sources. Climate data for these hazards were downloaded from the CDS and processed in Python (Visual Studio Code) to generate location-specific outputs for hazard evaluation.

The results are presented in two parts: 3.1 framework development and 3.2 case study. Framework development comprises step 1–3 of the presented methodology. The second part illustrates the framework’s practical applicability through a case study (step 4 of the methodological framework).

Framework development

The three steps: (i) literature-based mapping of climate-related hazards to EU Taxonomy activities, (ii) linkage of climate-related hazards to climate indicators, and (iii) inclusion of both resulting databases into a comprehensive CRA framework are presented in the following.

(i) Literature based mapping of climate-related hazards to EU Taxonomy energy activities

The review covers 23 publications identified via the EU Climate-ADAPT platform. From these sources, 703 energy-related hazard mentions (as defined in Section 2.1) were extracted. The number of hazard mentions varies substantially across sources (Figure 4). Each mention was assigned to the corresponding EU Taxonomy activity or activities referenced in the literature. For each entry, the affected value-chain segment was recorded (core, upstream, downstream). Technical and locational specifications reported in the sources were also captured.

Assignment logic and activity groups

Hazard–activity relationships were assigned on three levels: to a specific EU Taxonomy activity, to an activity group (see Figure 3), or to all energy activities. This approach supports consistent mapping because the reviewed literature typically discusses technologies or systems at a more aggregated level than the EU Taxonomy activity list. For example, the EU Taxonomy distinguishes several nuclear-related activities (Appendix Table A. 2), while the reviewed sources (Appendix Table A. 1) do not differentiate at that level. These activities are therefore grouped as “nuclear power plant” for the purpose of the database. Other groups, such as “thermal power plant,” are adopted directly from the literature where they are commonly used in CRAs, for example in Sieber (2012) and Alvater et al. (2011). Of the 703 hazard mentions, 324 are linked directly to a specific activity, 298 to an activity group, and 81 to the entire energy sector. After disaggregation to the EU Taxonomy activity level, the database contains 4,461 hazard–activity associations (database entries, referred to as datapoints in the context of the case study). This increase reflects the difference between aggregated inputs and fully disaggregated outputs. Sector-wide hazard mentions account for 56.29% of all disaggregated entries, although they represent 11.52% of the input. Group-level assignments generate 36.45% of the disaggregated entries, while activity-specific mentions account for 7.26%. Each database entry is labelled with its assignment level, that is, whether the association originates from activity-specific, group-level, or sector-wide evidence (Supplementary Table 1). Users can therefore filter the database to associations based on activity-specific evidence only, or assign lower weights to group-level and sector-wide entries in their analyses. The identification, assignment, and disaggregation process is illustrated in Appendix Figure A. 1.

Hazard distribution

Hazards are categorized according to the four EU Taxonomy hazard groups: temperature-related, wind-related, water-related, and solid mass-related (Supplementary Table 1). Water-related hazards represent the largest share of database entries at 37.21%, followed by temperature-related hazards at 34.48%, wind-related hazards at 23.78%, and solid mass-related hazards at 4.53%. Within water-related hazards, flooding is most frequent (477), including coastal, fluvial, pluvial, and groundwater floods. Droughts (397), sea-level rise (333), and extreme precipitation (179) follow. Glacial lake outburst is not mentioned in relation to energy activities. Water stress (4), saline intrusion (6), and ocean acidification (31) appear only rarely (Figure 5). Wind-related hazards are dominated by storms (671). Temperature-related hazards are mainly heat waves (511) and changing temperature levels (496). Heat stress (7) and temperature variability (16) are rarely mentioned. Solid mass-related hazards are least represented. Coastal erosion (99) and subsidence (62) are most frequent within this category. Soil degradation and soil erosion are not mentioned in relation to energy activities (Figure 5).

The database contains more acute than chronic hazards (3,137 vs. 1,324). Acute hazards, short-duration extreme events such as heat waves, storms, and floods (Pörtner et al., 2022), occur more than twice as often as chronic hazards, which are gradual, long-term shifts such as changing temperature levels, sea-level rise, and ocean acidification (Pörtner et al., 2022). Among the 12 hazards referenced more than 100 times, 9 are classified as acute.

Distribution across energy activities

The database supports identification of potentially relevant hazards for each EU Taxonomy energy activity. However, the number of hazard datapoints differs substantially across activities. Some activities are linked only via sector-wide hazard mentions, while others also include group-level and activity-specific mentions (Appendix Figure A. 1). For interpretation, activities are grouped by hazard count. High hazard counts (180–230 datapoints) occur for several major power generation technologies, including nuclear, gas-based activities (fossil and non-fossil), and bioenergy. Concentrated solar power is the only renewable energy source in this segment. Electricity transmission and distribution also fall into this category of high hazard counts. Moderate hazard counts (130–179) cover most renewable generation activities, including geothermal, solar, hydro, wind, and photovoltaic. Heat and cooling from bioenergy and gas-based district heating are also within this segment. Lower hazard counts (below 130) occur for smaller-scale or less established activities such as heat pump operation, waste-heat utilization, ocean energy generation, energy storage, heat and cooling from renewable sources, and biofuel manufacturing for transport. Overall, large-scale generation and grid operation dominate the dataset, while specialized or emerging activities are less represented (Appendix Figure A. 1).

A binary hazard–activity matrix summarises which hazards are potentially relevant to each energy activity (see Figure 6). Acute temperature-related hazards apply across all activities, creating horizontal clusters. Other hazards apply only to specific activities, or to none. In total, 8 of 28 hazards are linked only to specific activities. The structure of Appendix Figure A. 3 reflects the assignment logic. Sector-wide assignments lead to uniform rows, group-level assignments create clustered patterns, and activity-specific assignments create isolated entries. Figure 6 does not consider technical or locational conditions, and it does not differentiate value-chain segments. It therefore represents potential hazards prior to filtering. Including these factors is essential in CRAs and would reduce relevance for some hazards.

Locational and technological conditions and value-chain segmentation

Locational conditions are explicitly reported in 29.75% of database entries. Five location types are captured from the literature: coastal areas, flood-prone areas, mountain areas, permafrost regions, and structures on or in water. Coastal locations (498) and structures on or in water (382) are most frequent. Floodplains (241) and permafrost regions (205) follow. Figure 7 shows which hazards are specifically relevant to the locations covered by the literature review. Systematic recording of locational information supports filtering for geographically conditioned hazards, as recommended by Dorsch et al. (2022). Most location-specific assignments originate from hazards mapped to the entire energy sector (93.44%).

Only one technological condition is identified in the review: the use of water cooling. Water cooling is common in thermal power plants such as liquid fuel-fired plants, nuclear plants, and concentrating solar power plants. For activities using water cooling, the database indicates increased relevance of specific temperature-related and water-related hazards due to impacts on cooling efficiency, water availability, and water intake clogging.

Value-chain segmentation is recorded for each hazard–activity datapoint. Entries are assigned to core operations, upstream, downstream, or both upstream and downstream. If sources do not specify the affected segment, and the context does not allow a conclusion, the entry is assigned to core operations. Most climate-related hazards relate to core operations: Out of 4,461 datapoints, 4,013 (89.96%) are assigned to core operations. Upstream climate-related hazards account for 309 entries (6.93%) and mainly affect fuel supply. Hazards affecting both upstream and downstream account for 136 entries (3.05%). Downstream is assigned to only three entries (0.07%). This is consistent with the scope of the study, since downstream effects often relate to changes in energy demand. Several sources assess demand changes as a CC effect, but these changes are driven by regulatory, behavioral, or economic factors. They are therefore considered transitional risks and are not included because the database focuses on physical climate risks only.

(ii) Linking climate-related hazards to relevant climate variables and indicators

This phase links each EU Taxonomy climate-related hazard to climate indicators from the CDS. The purpose is to enable quantification of hazards within CRAs.

Hazard–indicator mapping approach and coverage. Twelve CDS datasets were identified as relevant (see Table 1). Across these datasets, 148 distinct climate indicators were screened. Each indicator definition was reviewed and assigned to one or more EU Taxonomy climate-related hazards where a clear correspondence existed. Indicators without correspondence to any EU Taxonomy hazard were recorded as not assigned.

This process yields 143 hazard–indicator matchings (Supplementary Table 2). In total, 79 of 148 indicators (53.02%) are matched to at least one EU Taxonomy hazard. The remaining 69 indicators contain climate information that is not aligned with the EU Taxonomy hazard list. Figure 8 shows that the number of indicators differs strongly across datasets. It also shows large differences in the share of indicators that can be mapped to EU Taxonomy hazards. Some datasets, for example, (1), (6), (7) and (11) (see Table 1) include activity-specific indicators, such as renewable power generation potential. These are relevant for energy sector analysis but are not mapped because they do not represent EU Taxonomy hazards. Other datasets, e.g., (49), (5), (12) (see Table 1), include highly specific indicators, such as marine biochemistry parameters, which are also not relevant to EU Taxonomy hazards.

Distribution across hazard categories and gaps

Water-related hazards are covered by the largest number of indicators. The mapping identifies 35 distinct indicators linked to water-related hazards. These include indicators for changing precipitation patterns and precipitation types, precipitation and hydrological variability, sea level rise, water stress, drought conditions, heavy precipitation, and flood events. Flood-related indicators are represented through maximum river discharge metrics for different return periods.

Temperature-related hazards are the second best-covered category. The database includes 31 distinct indicators linked to acute and chronic temperature hazards. This includes indicators for changing air, freshwater, and marine temperatures. It also includes indicators for heat stress, temperature variability, heat waves, cold-wave or frost conditions, and wildfire-related metrics. Heat stress indicators include composite indices such as the Universal Thermal Climate Index, which combines temperature, humidity, wind, and radiation variables.

Wind-related hazards are represented by 13 distinct indicators. All are assigned to storm-related hazards. Most are also linked to changing wind patterns. These indicators mainly capture wind speed at different heights. One indicator captures extreme wind speed days.

No reviewed CDS dataset provides indicators for solid mass-related hazards as defined by the EU Taxonomy.

A comparison with the EU Taxonomy hazard list shows missing CDS indicator coverage for the following hazards: glacial lake outburst, saline intrusion, cyclones or hurricanes or typhoons, tornados, coastal erosion, soil degradation, soil erosion, solifluction, avalanche, landslide, and subsidence. The causes and implications of these gaps are discussed in Section 4.

Scenario availability, temporal coverage, and spatial resolution

EU Taxonomy requirements for CRAs refer to using state-of-the-art climate projections across the existing range of future scenarios consistent with the activity lifetime, and the delegated act explicitly lists RCP2.6, RCP4.5, RCP6.0, and RCP8.5 as future scenarios.

Among the 79 indicators matched to hazards, RCP4.5 is available for all except two sea-level indicators. These two sea-level indicators have no RCP scenario. They are modelled under SSP5-8.5, which builds on RCP8.5 and adds socioeconomic assumptions (Moss et al., 2010). RCP8.5 is available for 73 of 79 indicators. RCP2.6 is available for 40 indicators. This limited availability of optimistic scenarios does not prevent use of the mapped indicators for EU Taxonomy-aligned assessments. As argued by Dorsch et al., 2022, an assessment does not need to cover the full set of scenarios. It should cover a range from optimistic to pessimistic futures. This can be implemented using RCP4.5 or RCP2.6 as optimistic scenarios and RCP8.5 as a pessimistic scenario. RCP6.0 is not covered by any mapped indicator.

In addition to projections, 53 indicators include historical data derived from climate reanalysis. All indicators also include retrospective coverage, with the shortest time span reaching back to 2018. This historical information can serve as a baseline to contextualize projections (Dorsch et al., 2022).

Temporal coverage is assessed against the going concern principle and EU Taxonomy expectations on activity lifetimes (Dorsch et al., 2022; EU Commission, 2021). The analysis shows that 65 of 79 indicators provide projections until 2,100. Among the remaining indicators, five reach until 2098, five until 2085, and four until 2065. The temporal scope therefore aligns with the requirement to support assessments over at least the next 10 to 30 years (Dorsch et al., 2022; EU Commission, 2021).

The EU Taxonomy also requires the use of the highest available spatial resolution (EU Commission, 2021). In the mapped indicator set, spatial resolution ranges from 30 km 2 for 44.3% of indicators to 5 km 2 for 25.3% of indicators in the highest-resolution datasets. The remaining 30.4% have intermediate resolutions between 11km 2 and 13.2km 2. Documenting scenario availability, temporal coverage, and spatial resolution enables filtering of indicators to those best suited for EU Taxonomy-aligned CRAs.

(iii) Framework synthesis - An integrated workflow supporting EU Taxonomy-aligned CRA

This step consolidates the two previously developed databases into an integrated framework supporting EU-Taxonomy-aligned CRA. It merges the hazard mapping results and the hazard–indicator linkage into a structured workflow that supports systematic identification and quantification of relevant climate-related hazards for any EU-Taxonomy economic activity.

The framework incorporates locational specificity by integrating geospatial datasets, including CLC and GlobPermafrost, to capture local environmental characteristics that influence hazard exposure. The overall logic is illustrated in Figure 9.

The workflow uses three main inputs: economic activity, technological specifications, and locational coordinates. Potentially relevant climate-related hazards are first identified via the Hazard–Activity Mapping Database. The hazard list is then filtered in two steps. Technological dependencies are checked first. Locational dependencies are checked second. Hazards identified as irrelevant are excluded at each step. For each remaining hazard, the framework assigns the most suitable indicator from the hazard–indicator database. Indicator selection follows EU Taxonomy requirements and prioritizes indicators with the highest available spatial resolution. At least two IPCC RCP scenarios are considered.

The selected indicators are then used to quantify the remaining hazards by visualizing the corresponding indicator data.

Demonstration of the framework’s applicability: case study solar

The case study applies the framework (see Figure 9) to the EU Taxonomy activity “Electricity generation using solar photovoltaic technology” for Aachen, Germany.

Hazard identification

The first step of the framework is the identification of relevant hazards. All potentially relevant climate-related hazards for the selected activity are retrieved from the Hazard–Activity Mapping Database (Supplementary Table 1) and then screened for technological and locational dependencies. The database assigns 130 hazard–activity datapoints to activity 4.1 (“Electricity generation using solar photovoltaic technology”; see Appendix Figure A. 2), covering 19 of the 28 EU Taxonomy hazards, meaning 9 hazards are not linked to the activity in the database. Of the 130 datapoints, 90 are not location-dependent and 40 are linked to coastal areas, flood-prone areas, permafrost, or on/in water.

The location Aachen, Germany (50.7751°N, 6.0840°E) is characterised using the CLC dataset (Copernicus Land Monitoring Service, 2025). The land cover classification indicates primarily urban fabric and industrial and commercial units, surrounded by agricultural land and forest (see Figure 10).

The broader regional context indicates no proximity to large inland water bodies, coastal areas, marshes, or intertidal zones. The nearest waterbody recorded in CLC is the Wehebachtalsperre, approximately 17 km west of the city centre, and the nearest coastline is over 190 km away. As a result, the PV activity in Aachen is not located in a coastal zone, on/in water, or within a flood-prone area.

Since CLC does not include permafrost, the GlobPermafrost Map Portal (GlobPermafrost, 2021) is used to assess permafrost relevance. The closest permafrost regions are located in the Alps (Alfred-Wegener-Institute and Helmholtz Centre for Polar and Marine Research, 2025), so permafrost-related hazards are not considered directly relevant for Aachen. Based on these locational checks, all 40 location-specific datapoints are excluded. This makes coastal erosion, ocean acidification, permafrost thawing, and sea level rise irrelevant for the case study. The influence of the site location is confined to the hazard identification step. At coastal or marine sites, such as an offshore wind installation, location-dependent hazards including sea level rise, coastal erosion, ocean acidification, and storms would instead be retained as relevant (44 of the 132 datapoints assigned to this activity carry locational conditions). All subsequently described steps, from indicator assignment to data retrieval and visualisation, proceed identically, as the corresponding hazard–indicator matchings are documented for all hazards in Supplementary Table 2. Furthermore, no technological dependencies are identified for PV within the 130 datapoints, so no hazards are excluded for technological reasons.

To keep the application illustrative, the case study focuses on the three most frequently cited hazards for PV in the hazard mapping database. These are changing temperature (air, freshwater, marine water), heat waves, and storms (including blizzards, dust and sandstorms). This restriction applies solely to the data retrieval and visualization step and serves the illustrative purpose of the case study. The complete framework output, comprising the assigned indicators for all 15 hazards remaining after locational screening, is provided in Appendix Table A. 7. An EU Taxonomy-aligned assessment would require the quantification of all identified relevant hazards (see Section 4).

Climate indicator selection

In the second step, for each hazard the highest resolution indicators need to be identified. Indicators are selected from the Hazard–Indicator Mapping Database (Supplementary Table 2) with the aim of meeting EU Taxonomy requirements. Selection follows a two-step filter. First, indicators with the highest spatial resolution are prioritised. Second, indicators are screened for availability of optimistic and pessimistic climate scenarios, with preference for wider scenario coverage such as RCP2.6 and RCP8.5 compared to RCP4.5 and RCP8.5.

For the three selected hazards, the database provides 18 indicators for changing temperature, 15 indicators for heat waves, and 13 indicators for storms. For changing temperature, three indicators at 5 km 2 resolution are available under RCP2.6, RCP4.5, and RCP8.5. These are: 2 m air temperature (monthly mean), water temperature in catchments (monthly mean), and water temperature in local streams (monthly mean), each provided for 2011–2040, 2041–2070, and 2071 (Berg et al., 2021b; Berg et al., 2021c). For heat waves, two suitable indicators at 12 km 2 resolution are selected and are available under RCP4.5 and RCP8.5. These are maximum temperature (X th percentile) and heat wave days (C3S Climate Data Store, 2024b; Hooyberghs et al., 2019a,b,c).

For storms, the two-step selection identifies seven variables at 13.2 km 2 resolution covering RCP2.6, RCP4.5, and RCP8.5. These are, however, raw 3-hourly wind velocity components and pressure fields rather than aggregated hazard indicators, and their transformation into extreme-wind metrics would require processing very large data volumes. For the illustrative purpose of this case study, the aggregated indicator extreme wind speed days at 30 km 2 resolution under RCP4.5 and RCP8.5 is used instead (C3S Climate Data Store, 2024a, 2024b). This selection satisfies the framework’s minimum requirement of two RCP scenarios but deviates from the stated preference for the highest resolution and widest scenario coverage, and it therefore constitutes a practical constraint of the demonstration rather than a recommendation. The higher-resolution variables remain fully documented in Supplementary Table 2 for applications with corresponding processing capacity. Download configurations for the retrieved datasets, including the underlying model configuration, are documented in Appendix Table A. 5). Accordingly, the visualized results reflect the projections of the specific model configurations listed there, while differences between projections from alternative combinations of global and regional climate models are not represented. Given the illustrative purpose of the case study, this inter-model variability is not examined further here.

Visualization

The last step to perform is the visualization of the data. Downloaded CDS data are provided in Network Common Data Format, which supports multidimensional arrays (Unidata, 2018). The datasets are gridded, and each grid cell is defined by its latitude and longitude. For Aachen, the relevant cell is identified by selecting the grid cell with the smallest distance between its center point and the target coordinates (see Figure A3). Indicator values are then extracted for that cell by indexing it using xarray and visualized using the matplotlib.pyplot. The data processing performed in the case study comprises solely the described extraction and visualization of the indicator values. The Python scripts implementing these steps are available from the corresponding author upon request (see Data and Code Availability Statement).

The resulting figures illustrate hazard evolution at the case location. Near-surface air temperature increases moderately under RCP2.6 by about 1 °C by end-of-century, while RCP8.5 shows a stronger increase with more pronounced summer warming (Figure 11).

Water temperatures in catchments and streams show similar warming patterns (Figures 12a,b), with summer temperatures remaining lower than air temperatures but increasing towards end-of-century, exceeding +2 °C under RCP8.5 compared to the current RCP2.6 baseline (Berg et al., 2021b; Berg et al., 2021c).

Extreme heat indicators show increases under both scenarios (Figures 13, 14), with stronger rises under RCP8.5 after mid-century (Hooyberghs et al., 2019a,b,c). Heatwave days increase strongly over the 21st century under both scenarios, based on a heatwave definition of at least three consecutive days above the 99th percentile of historical (1981–2010) daily maximum temperatures (C3S Climate Data Store, 2024b).

Extreme wind speed days show high interannual variability in both RCP4.5 and RCP8.5 without a clear trend (Figure 15), under a definition based on exceeding the 98th percentile of historical (1981–2010) daily wind speeds (C3S Climate Data Store, 2024a, 2024b).

The visualized results also indicate the uncertainty associated with the choice of climate scenario. For each hazard, the distance between the RCP curves reflects how strongly the projected development depends on the assumed emission pathway. This dependence is pronounced for the temperature-related hazards, where the curves diverge markedly towards the end of the century, whereas no systematic difference between the scenarios is apparent for extreme wind speed days.

Further steps for EU Taxonomy-alignment

A full EU Taxonomy-aligned CRA would require visualisation and evaluation of all relevant hazards for the activity. It would then need to assess material physical climate risks by jointly evaluating hazard, exposure, and vulnerability, as proposed by Dorsch et al. (2022). The framework developed here primarily addresses the hazard dimension and includes exposure in a limited way via location-specific filtering. A full risk assessment also requires climate impact chains that describe cause–effect relations between hazards and the activity system. Existing climate impact chains can be used for CRA (Dorsch et al., 2022). To support further assessment, the Hazard - Activity Mapping Database (Supplementary Table 2) includes cause–effect statements extracted during the literature review. Statements relevant to the three hazards in this case study are compiled in Appendix Table A. 6). Finally, once material risks are identified, EU Taxonomy requirements imply that adaptation measures need to be identified, assessed, and implemented.

The literature review underlying the Hazard-Activity Database reveals a structural mismatch between the EU Taxonomy’s activity-level logic and the aggregation level of hazard evidence on the Climate-ADAPT platform. Evidence is typically reported for sectors or technology groups rather than individual activities, complicating direct mapping. To address this, hazard information is captured at three levels (activity, activity group, sector) and disaggregated to the activity level to match the EU Taxonomy structure. This enables a pragmatic integration of an evidence base that is often more aggregated than the Taxonomy classification, while still delivering an activity-resolved dataset for downstream use.

A central pattern is the dominance of temperature- and water-related hazards. This dominance is largely driven by a small number of frequently cited hazards within each category: mainly cited are floods, droughts, sea-level rise, and heat events; others (e.g., water stress, glacial lake outburst) are rarely cited. This distribution reflects both literature focus and database design. Importantly, database frequencies should not be interpreted as direct measures of hazard importance. Hazards assigned at sector or activity-group level are disaggregated across many activities, inflating their statistical weight. By contrast, hazards linked only to specific activities remain less frequent. Frequency thus indicates how widely a hazard is discussed, not its absolute relevance, and is reported for descriptive purposes only, while the identification of relevant hazards within the framework relies on the binary hazard–activity assignment and the subsequent filtering steps.

The hazard distribution also reflects research priorities. The strong emphasis on a limited set of hazards likely corresponds to vulnerability patterns commonly discussed for energy systems, including thermally sensitive processes, dependence on water availability, and exposure of electrical equipment to ambient conditions. A similar imbalance is visible for hazard temporality: acute hazards are represented more than twice as often as chronic hazards. This reflects research bias, although chronic hazards remain relevant for long-lived energy assets. Evidence coverage varies: established generation and grid technologies dominate, while emerging ones are underrepresented, likely due to fewer scientific studies rather than lower climate relevance.

The dataset structure supports the two-step filtering logic by Dorsch et al. (2022). About 45% of datapoints are activity- or activity-group-specific, indicating technological contextualisation, while around 30% include geographic qualifiers. This suggests that the dataset contains the qualifier types needed to operationalise the filtering logic in practice, as proposed by Dorsch et al. (2022).

Regarding hazard quantification, 53% of reviewed CDS indicators can be linked to at least one EU Taxonomy hazard. Hazards that are frequently mapped in the Hazard - Activity Database are generally covered by multiple indicator options. Remaining gaps affect solid mass-related hazards, specific acute wind events such as cyclones, hurricanes, and typhoons, saline intrusion, permafrost thaw, and glacial lake outburst. The causes and implications of these gaps are examined below.

Integrating hazard–activity and hazard–indicator databases into one workflow reduces data burden for CRA practitioners. Similar to Luo et al. (2023), the framework is driven by three easily available inputs: economic activity, location, and technical characteristics. The case study demonstrates that site-specific hazard filtering and climate-projection visualisation can be implemented efficiently.

With respect to the constituent steps of a complete CRA, the framework’s coverage is differentiated as follows. Hazard identification is fully operationalized through the hazard–activity database and the technological and locational filters. Exposure screening is partially operationalized: the locational filter establishes whether a site lies within hazard-prone location types, but asset-level exposure characterization is not performed. Vulnerability assessment is not operationalized, as it requires asset-specific information on sensitivity and adaptive capacity, and therefore remains with the practitioner. Impact-chain development is supported rather than operationalized: the cause–effect statements documented for each database entry (Supplementary Table 1; Appendix Table A. 6) provide evidence-based building blocks for constructing impact chains in line with Dorsch et al. (2022). Materiality assessment is deliberately deferred, as it depends on the joint, context-specific evaluation of hazard, exposure, and vulnerability. Adaptation-measure selection lies outside the framework’s scope and is addressed only as a potential future extension connecting screening outputs to organizational preparedness and adaptation implementation (Putera et al., 2026). The framework thus concentrates practitioner effort on the steps that genuinely require case-specific judgement, namely vulnerability and materiality, while streamlining the data-intensive steps of hazard identification and quantification.

Key limitations relate to data sources and methodological choices. Hazard identification relies on peer-reviewed studies and institutional technical reports available on Climate-ADAPT, ensuring policy relevance but limiting comprehensiveness compared to multi-database searches in Scopus or Web of Science. Literature gaps and research biases may therefore propagate into the hazard–activity database. This is evident in the activity-level coverage: three activities are only linked at sector level, and 21 only at sector and activity-group level, indicating that many EU Taxonomy activities are not yet addressed individually in the considered literature base. In addition, screening and data extraction were performed by a single reviewer. Although explicit eligibility and coding criteria were applied and records of uncertain relevance were retained during screening, single-reviewer designs carry a residual risk of selection and coding bias, and inter-coder reliability could not be assessed. The entry-level documentation in the Supplementary material mitigates this risk by keeping every coding decision open to scrutiny. Finally, the case study is illustrative in scope. It is confined to a single activity and location, and its quantification step covers three hazards, so it illustrates the framework’s workflow rather than a completed Taxonomy-aligned assessment.

Beyond these limitations, the case-study results are subject to sources of uncertainty that are inherent in state-of-the-art climate projection data and would therefore be present in any EU-Taxonomy-aligned assessment drawing on these sources, irrespective of the framework applied. The first is scenario uncertainty, reflecting that future emission pathways are unknown. The EU Taxonomy addresses this by requiring the use of climate projections across a range of future scenarios, which the framework operationalizes, following Dorsch et al. (2022), by quantifying each hazard under at least two RCP scenarios spanning an optimistic and a pessimistic future, so that every output displays the scenario range rather than a single trajectory. In the case study this range is substantial for the temperature-related hazards—the end-of-century warming signal increases from about 1 °C under RCP2.6 to about 2.5 °C under RCP8.5, and heatwave days diverge markedly after mid-century—whereas the extreme-wind signal shows high interannual variability without a clear trend under either scenario. Scenario choice thus dominates the uncertainty of the thermal hazard signals at the case location, while it hardly affects the storm signal. The second inherent source is model spread: even under the same scenario, projections from different quality-assured combinations of global and regional climate models differ. The demonstration retrieved a single model chain per indicator (Appendix Table A. 7), consistent with its illustrative scope, so this spread is not displayed in the figures. For Central Europe, the direction of the temperature-related signals is consistent across the underlying model ensemble, so that a different model chain would alter the magnitude rather than the direction of the case-study conclusions, whereas projected changes in wind extremes are known to differ more strongly between models — a further reason for the cautious interpretation of the storm indicator. Since the matched CDS datasets provide multiple model chains, documented in Supplementary Table 2, the identical workflow can be repeated across chains to quantify this spread, which is identified as a direction for future applications. Third, the reference periods of the threshold-based indicators (e.g., 1981–2010 for heatwave days and extreme wind speed days) are fixed within the pre-computed CDS products and are disclosed in the indicator documentation and figure captions. A different baseline would shift the absolute exceedance counts but not the direction of the projected change, and varying it would require reprocessing raw model output, which lies outside the scope of a framework deliberately built on quality-assured, pre-aggregated indicators.

In addition to these inherent uncertainties, the results are sensitive to two methodological choices made within the framework, both governed by transparent rules rather than case-by-case judgement. The first concerns spatial resolution: the framework prioritizes the highest available resolution, in line with the EU-Taxonomy requirement, and documents the native resolution of every indicator. The single deviation in the case study—the storm indicator retrieved at 30 km 2 in place of the 13.2 km 2 raw variables—is documented transparently in Section 3.2. Coarser grid cells average local extremes over a larger area, which primarily affects localized phenomena such as wind, so the storm results warrant additional interpretive caution, whereas the smoothly varying temperature fields at the case-study location are considerably less sensitive to resolution. The second concerns the selection among the indicators matched to a hazard. Rather than prescribing a single indicator, the framework documents all matchings with their metadata, rendering this choice transparent and adjustable. The case study illustrates robustness to this choice for heat extremes, where the two complementary indicators—the 95th percentile of daily maximum temperature and heatwave days—yield consistent signals. The subjectivity of the underlying hazard–indicator matching has been acknowledged as a limitation above, and the standardization of indicator selection remains identified as future work. Overall, the uncertainties inherent in the climate information affect the magnitude, but not the direction, of the hazard signals identified in the demonstration, and the framework is designed to make them visible—through mandatory scenario ranges and documented model chains, resolutions, baselines, and indicator metadata—rather than to conceal them behind a single trajectory.

Future improvements should expand the literature base beyond Climate-ADAPT to reduce the risk of missing relevant climate-related hazards and help improve the representation of currently underrepresented activities and include regular updates to maintain the long-term relevance of the database. Such updates would also provide the opportunity to implement dual independent screening and coding with documented inter-rater agreement. Automated approaches using large language models and machine learning could support scalable identification and evaluation of relevant studies. Introducing minimum evidence thresholds per activity could further improve robustness by excluding weakly supported mapping. Activities that only have a small number of research papers focusing specifically on the activity or activity group may then be excluded until sufficient literature is available to ensure a robust basis for hazard identification. Periodic expert review of the hazard–activity assignments would provide an additional validation layer as the database evolves. Applying the framework to real assets in collaboration with practitioners and public authorities, including marine and coastal sites, would provide additional validation of its outputs in operational assessment contexts and would foster the collaborative institutional uptake on which the practical effectiveness of such tools depends beyond technical hazard mapping alone (Valentina et al., 2025a,b).

The CDS climate indicator dataset provides a robust and transparent basis for hazard quantification, but it does not cover all hazards identified in the literature basis. The remaining indicator gaps have different causes. Tornados operate at spatial scales far below the resolution of current regional climate models, so projection-based indicators for them do not exist in any climate data catalogue. Cyclones, hurricanes, and typhoons rarely affect the European domain covered by the screened datasets, and their explicit representation in climate projections remains constrained by model resolution, so European climate services provide no dedicated indicators for them, whereas the extratropical storms relevant to Europe are covered by the matched storm indicators. The solid mass-related hazards, such as landslides, subsidence, and coastal erosion, as well as saline intrusion and glacial lake outbursts, are of a more fundamental nature. They are not climate variables but geo-hydrological processes, triggered by climatic drivers and governed by terrain, soil, and hydrogeological conditions. Climate datasets can therefore only provide their triggers, such as heavy precipitation for landslides or sea level rise and storms for coastal erosion, while quantifying the processes themselves requires additional non-climatic, site-specific data, irrespective of the climate data source used. These gaps consequently persist in any climate data catalogue, including the CDS as the climate data source referenced by the EU Taxonomy itself, and ultimately manifest an operationalization challenge of the EU Taxonomy: the regulation obliges preparers to screen hazards for which climate data alone cannot supply indicators. For practice, three implications follow. First, these hazards cannot be overlooked, because the hazard–activity database identifies their potential relevance independently of indicator availability. Second, for the underlying climatic drivers, the framework assigns quantifiable indicators where these are available in the CDS, thereby enabling a driver-level characterization of the affected hazards. Third, the framework makes explicit for which hazards complementary non-climatic data are required, and the integration of such data is already part of its design, as demonstrated by the CORINE Land Cover and GlobPermafrost datasets used for locational screening. In addition, CDS offers spatial resolutions down to 5 km 2, while locally more detailed datasets may be available from national meteorological services or commercial providers.

The current hazard–indicator mapping is limited to CDS indicators, which does not provide indicators for all EU Taxonomy climate-related hazards. Expanding to additional data sources and integrating high-resolution regional datasets (e.g., flood models or local precipitation extremes) would improve coverage and spatial accuracy.

Future extensions could include activity-specific climate indicators. Although excluded here due to their lack of hazard specificity and thus, a clear cause-effect relationship, they may still support risk assessment by capturing climate impacts. Allowing practitioners to select from multiple EU Taxonomy-aligned indicators could enhance flexibility.

Another limitation is the treatment of hazards as independent. In practice, hazards interact either simultaneously (compounding hazards) or sequentially (cascading hazards), which is not yet captured but could significantly improve CRA realism.

Furthermore, the hazard–indicator matching process currently relies on subjective interpretation. Incorporating expert judgement or standardized matching criteria could improve transparency and reproducibility.

The framework is scalable and suitable for automation and integration in digital tools, including API integration with CDS for real-time data retrieval. Extensions such as activity-specific fragility functions as proposed by Luo et al. (2023), could incorporate vulnerability modelling further reduce data demands and manual effort.

Further development could also integrate adaptation measure databases, enabling decision-support systems linking hazards to mitigation options.

This research developed a climate-related hazard- and indicator-integrated framework supporting CRA in the energy sector, aligned with EU Taxonomy requirements. Building on the step-by-step guidance of Dorsch et al. (2022) the framework targets a central practical barrier to EU Taxonomy-aligned CRAs: high data intensity. It addresses this challenge by integrating two complementary databases: a literature-derived hazard-activity mapping linking EU Taxonomy energy-sector activities to relevant climate-related hazards, and a systematic hazard-indicator mapping connecting these hazards to quantifiable climate indicators from the CDS. By capturing hazards at sector, activity-group, and activity level and disaggregating them to individual activities, the framework bridges the gap between the EU Taxonomy’s granular activity definitions and the broader aggregation level typically used in CRA-oriented literature.

The database analysis indicates that temperature- and water-related hazards dominate the hazard landscape for EU Taxonomy energy activities, with heatwaves, changing temperature, flooding, and drought among the most frequently cited hazards. For these high-priority hazards, suitable CDS indicators are available, including long-term projections under multiple RCP scenarios, enabling hazard quantification that supports EU Taxonomy-aligned assessment needs. At the same time, indicators for less frequently cited hazards remain partially unavailable, which constrains quantification in contexts where such hazards may be material.

The framework reduces the information needs of CRA practitioners to three basic inputs – economic activity, location, and technological characteristics – which are sufficient to identify relevant hazards and retrieve high-resolution climate projection data for hazard evaluation. A case study on solar PV in Aachen, Germany, illustrates the framework’s practical applicability: relevant hazards are identified, geographically irrelevant hazards are excluded, and suitable indicators are assigned to enable location-specific quantification without manual data collection. By streamlining hazard identification and climate-data retrieval, the workflow shifts effort from time-consuming data gathering toward the assessment of material impacts and the development of targeted adaptation measures.

Overall, the framework provides a transparent, systematic, and scalable approach to support hazard identification and quantification in an EU Taxonomy context, thereby facilitating more consistent consideration of CCA within CRA practice and related decision-making. Key strengths include broad hazard coverage, reliance on peer-reviewed literature and institutional open-access data, and a design that is amenable to automation. However, limitations remain: hazard identification currently relies on a single literature platform, which can lead to incomplete hazard profiles for activities that are less represented in the underlying evidence base; some hazards lack corresponding indicators; and the framework primarily addresses the hazard dimension of risk rather than vulnerability or exposure. Expanding literature and data sources and incorporating vulnerability as well as hazard interaction modelling would be important next steps toward a more comprehensive, multi-dimensional CRA tool, consistent with the broader CRA logic.