The extreme heat risk measures how many days per year the maximum temperature exceeds 35 °C near the asset. It is an absolute count, the same for every site: the class does not depend on the site's past climate but on the number of days, because that number is what weighs on people, plants and power grids. More days, higher risk.
Indicator. Days per year with daily maximum 2 m temperature above 35 °C (TX35). Unit: days.
Data. "Monthly extreme hot days" index of the Copernicus climate atlas (C3S Atlas): CMIP6 models on a 1° grid, bias-adjusted with the ISIMIP method against WFDE5 observations for 1980–2005. Models per scenario: SSP1-2.6 16, SSP2-4.5 17, SSP3-7.0 16, SSP5-8.5 19. The site value is the one of the nearest land grid point.
Calculation.
(1) For each model the 12 months are summed to obtain the days of the year.
(2) For each model the 5-year mean centred on the year is computed; at the ends of the period the nearest full window is used.
(3) The value is the median across models. The 10th and 90th percentiles are kept as an uncertainty range.
Thresholds → level A–F. Absolute classes, the same for every site, derived from the global 1995–2014 distribution over land cells (area-weighted quintiles of the cells with at least 1 day per year):
• A: fewer than 1 day
• B: 1 to 6.5
• C: 6.5 to 24.5
• D: 24.5 to 65.1
• E: 65.1 to 137.3
• F: more than 137.3 days per year.
Validation. The 1995–2014 values were compared with MeteoSwiss and Hong Kong Observatory stations. A 1° cell averages coast, hills and plains, so it can underestimate the local heat of an urban or inland site.
From physical to impact. The physical level (A–F) is combined with the production process's sensitivity through the risk matrix (6 physical levels by 5 sensitivity bands) → impact level. Version 2025 also turned the impact level into a value at risk with a damage factor per class, a xNatura estimate without a published source: it is no longer used.
Economic damage. Version 2026, published source: the production lost for the extra days with a maximum temperature above 35 °C, with the effect estimated by the OECD on 2.7 million firms in 23 advanced economies, 2000-2021 (Costa et al. 2024, Table C.2): each day above 35 °C lowers the value added per employee by 0.0339%.
• Days: change of the days above 35 °C between the 20 years around the year and 1995-2014, for each of the ten NASA NEX-GDDP-CMIP6 models, then the mean of the models; fewer hot days count as 0.
• Annual loss = employees × value added per employee × (1 − e^(−0.000339 × extra days)); value added per employee: Eurostat by country and sector, World Bank and ILO outside Europe.
• It measures the same effect as the Heat stress risk (labour productivity): the site total counts the larger of the two, never the sum. The OECD sample is of advanced economies: elsewhere it is applied the same, stated.
Economic damage. The production damage of heat (working hours lost) is counted only once, in the heat stress risk, so that the same hours are not counted three times.
Sources.
• [Copernicus Interactive Climate Atlas (C3S Atlas)](Copernicus Interactive Climate Atlas (C3S Atlas))
• [ETCCDI climate extremes indices](ETCCDI climate extremes indices)
• [Lange 2019, ISIMIP3BASD bias adjustment](Lange 2019, ISIMIP3BASD bias adjustment)
• [Climate Change 2021: The Physical Science Basis](Climate Change 2021: The Physical Science Basis)
• [Costa et al. 2024, The heat is on: heat stress, productivity and adaptation among firms (OECD)](Costa et al. 2024, The heat is on: heat stress, productivity and adaptation among firms (OECD)).
Indicator. Days per year with daily maximum 2 m temperature above 35 °C (TX35). Unit: days.
Data. "Monthly extreme hot days" index of the Copernicus climate atlas (C3S Atlas): CMIP6 models on a 1° grid, bias-adjusted with the ISIMIP method against WFDE5 observations for 1980–2005. Models per scenario: SSP1-2.6 16, SSP2-4.5 17, SSP3-7.0 16, SSP5-8.5 19. The site value is the one of the nearest land grid point.
Calculation.
(1) For each model the 12 months are summed to obtain the days of the year.
(2) For each model the 5-year mean centred on the year is computed; at the ends of the period the nearest full window is used.
(3) The value is the median across models. The 10th and 90th percentiles are kept as an uncertainty range.
Thresholds → level A–F. Absolute classes, the same for every site, derived from the global 1995–2014 distribution over land cells (area-weighted quintiles of the cells with at least 1 day per year):
• A: fewer than 1 day
• B: 1 to 6.5
• C: 6.5 to 24.5
• D: 24.5 to 65.1
• E: 65.1 to 137.3
• F: more than 137.3 days per year.
Validation. The 1995–2014 values were compared with MeteoSwiss and Hong Kong Observatory stations. A 1° cell averages coast, hills and plains, so it can underestimate the local heat of an urban or inland site.
From physical to impact. The physical level (A–F) is combined with the production process's sensitivity through the risk matrix (6 physical levels by 5 sensitivity bands) → impact level. Version 2025 also turned the impact level into a value at risk with a damage factor per class, a xNatura estimate without a published source: it is no longer used.
Economic damage. Version 2026, published source: the production lost for the extra days with a maximum temperature above 35 °C, with the effect estimated by the OECD on 2.7 million firms in 23 advanced economies, 2000-2021 (Costa et al. 2024, Table C.2): each day above 35 °C lowers the value added per employee by 0.0339%.
• Days: change of the days above 35 °C between the 20 years around the year and 1995-2014, for each of the ten NASA NEX-GDDP-CMIP6 models, then the mean of the models; fewer hot days count as 0.
• Annual loss = employees × value added per employee × (1 − e^(−0.000339 × extra days)); value added per employee: Eurostat by country and sector, World Bank and ILO outside Europe.
• It measures the same effect as the Heat stress risk (labour productivity): the site total counts the larger of the two, never the sum. The OECD sample is of advanced economies: elsewhere it is applied the same, stated.
Economic damage. The production damage of heat (working hours lost) is counted only once, in the heat stress risk, so that the same hours are not counted three times.
Sources.
• [Copernicus Interactive Climate Atlas (C3S Atlas)](Copernicus Interactive Climate Atlas (C3S Atlas))
• [ETCCDI climate extremes indices](ETCCDI climate extremes indices)
• [Lange 2019, ISIMIP3BASD bias adjustment](Lange 2019, ISIMIP3BASD bias adjustment)
• [Climate Change 2021: The Physical Science Basis](Climate Change 2021: The Physical Science Basis)
• [Costa et al. 2024, The heat is on: heat stress, productivity and adaptation among firms (OECD)](Costa et al. 2024, The heat is on: heat stress, productivity and adaptation among firms (OECD)).