Urban heat statistics show that cities can be warmer than nearby rural areas, but the size and meaning of the difference depend on the measurement. The evidence below covers surface and air-temperature contrasts, health outcomes, neighborhood variation, mitigation scenarios and the observation systems used to compare places.
Key Urban Heat Statistics
The most useful headline figures span temperature, health, exposure and mitigation:
- 1.2 °C to 4.1 °C was the observed summer urban-minus-nonurban minimum-temperature difference across five Spanish cities from 2014–2018.
- 10–15 K separated global city surface-temperature hotspots from relatively cool parts of the same city in a 2003–2019 mapping study.
- 0.50 ± 0.20 K per decade was the mean urban-core surface-warming trend across more than 2,000 global city clusters from 2002–2021.
- 29% greater warming occurred in urban cores than in rural background areas in the same 2002–2021 global comparison.
- More than 0.23 K per decade of mean warming was attributed to urban expansion in Chinese and Indian city clusters from 2002–2021.
- 45% higher mortality risk was estimated for the warmest 2% of days in 85 European cities from 2015–2017 compared with surrounding rural areas.
- 67% higher mortality risk was estimated in the most-built-up European areas than in the least-built-up areas during heat-extreme days from 2015–2017.
- 0.25 additional deaths per 100,000 people per day was the median UHI addition during European heat extremes from 2015–2017.
- 7% lower mortality risk was estimated from UHI during European cold extremes from 2015–2017 compared with rural areas.
- Relative risk of 1.12 was estimated for non-external mortality among Hong Kong residents aged 75 and older in high-UHIdh areas at the 99th temperature percentile during 2010–2019 hot seasons.
- 12%–44% of postal codes in each of seven U.S. cities showed significant mortality increases on extreme-heat days in the study period reported by the researchers.
- An odds ratio of 1.039 per 1 °C temperature increase was estimated for Greater London during May–September 2007–2016.
- 0.42% of total heat mortality was estimated as attributable to the UHI in Ho Chi Minh City from 2010–2013.
- 7.4 heat deaths prevented per additional 1 km² of green space per 1,000 people was estimated for Ho Chi Minh City from 2010–2013.
- 31% higher mortality risk was estimated between 22.5 °C and 35 °C in eight Swiss cities during May–September 2003–2016.
- 0.14 °C of air-temperature cooling per percentage-point increase in tree canopy was estimated in a Phoenix residential-neighborhood model.
- Up to 1 °C of daytime ambient-air reduction was modeled for Athens across combinations of cool roofs, roads, green roofs and trees.
- Up to 11.5 °C of daytime surface-temperature reduction was modeled for Athens on a heatwave day.
Contents
- Urban heat island temperature differences
- Global urban heat trends and exposure
- Urban heat mortality during hot and cold extremes
- Neighborhood and population health-risk statistics
- Urban heat mitigation statistics
- Urban heat measurement data and city indicators
Urban Heat Island Temperature Differences
Urban heat island intensity is not one universal metric. Studies distinguish air or canopy temperature, land-surface temperature, urban-rural contrasts and differences between hotspots and cooler neighborhoods.
In Spain, the 2014–2018 observational analysis of five cities found the following summer differences in mean minimum temperature:
| City | Urban-minus-nonurban summer mean Tmin |
|---|---|
| Murcia | 1.2 °C |
| Valencia | 4.1 °C |
The study covered Murcia and Valencia among five Spanish cities, and its result concerns minimum temperature rather than maximum temperature. Impact of urban heat islands on morbidity and mortality in heat waves: Observational time series analysis of Spain’s five cities reports the city-specific observatory comparison.
A global surface-temperature mapping study covering 1,711 cities from 2003–2019 found that city hotspots could be 10–15 K hotter at the surface than relatively cool parts of the same city. This is surface urban heat island intensity, or SUHI, rather than directly measured near-surface air temperature. Global long-term mapping of surface temperature shows intensified intra-city urban heat island extremes provides the global mapping result.
Global Urban Heat Trends and Exposure
The global city-cluster analysis covered more than 2,000 city clusters and measured satellite-derived land-surface temperature trends from 2002–2021. Mean urban-core surface warming was 0.50 ± 0.20 K per decade, with the uncertainty reported as one standard deviation.
Urban-core warming was 29% greater than warming in rural background areas. In city clusters in China and India, the study attributed more than 0.23 K per decade of the mean warming trend to urban expansion, a statistical decomposition rather than a thermometer-only observation. Surface warming in global cities is substantially more rapid than in rural background areas reports these estimates.
The European mortality study evaluated 85 cities at 500 m spatial resolution during 2015–2017. Its annual aggregate illustrates why annual and extreme-event measures must be kept separate: 70 of 85 cities had a weak protective net annual UHI effect, with a median 2.8 fewer deaths per 100,000 people per year.
The 90% confidence interval for that European median was 1.7–3.5 fewer deaths per 100,000 per year, while the between-city interquartile range was 0.7–4.4 fewer deaths per 100,000 per year. The annual result includes cold-weather protection and does not mean UHI is protective during heat extremes. Economic valuation of temperature-related mortality attributed to urban heat islands in European cities explains the city-level estimates.
Urban Heat Mortality During Hot and Cold Extremes
During the warmest 2% of days in the 85-city European study, UHI added a median 0.25 deaths per 100,000 people per day. The 90% confidence interval was 0.21–0.27, and the interquartile range across cities was 0.18–0.29 deaths per 100,000 per day.
The corresponding relative result was a median 45% higher mortality risk than in rural surroundings, with an interquartile range of 30%–61%. These are modeled estimates for 2015–2017, and relative risk is not an absolute death count.
Cold extremes produced a different result. UHI reduced mortality by a median 0.05 deaths per 100,000 people per day, with a 90% confidence interval of 0.04–0.07 and a median relative-risk reduction of 7%.
Selected annual net estimates show the importance of local climate and population structure:
| City | Estimated annual net UHI-attributable temperature-related effect, 2015–2017 |
|---|---|
| Glasgow | 16 fewer deaths per 100,000 people per year |
| Porto | 12 fewer deaths per 100,000 people per year |
| London | 11 fewer deaths per 100,000 people per year |
| Trieste | 5 additional deaths per 100,000 people per year |
| Genoa | 4 additional deaths per 100,000 people per year |
| Bologna | 3 additional deaths per 100,000 people per year |
These estimates combine hot-day harm and cold-day protection, so they should not be substituted for heatwave effects. The source is Economic valuation of temperature-related mortality attributed to urban heat islands in European cities.
Hong Kong used local definitions for 2010–2019 hot seasons: a very hot day had a maximum temperature of at least 33 °C, and a hot night had a minimum temperature of at least 28 °C. At the 99th percentile, non-external mortality relative risk was 1.08 in high-UHIdh areas and 1.05 in moderate-UHIdh areas; for non-cancer mortality, the corresponding estimates were 1.11 and 1.06.
Among Hong Kong residents aged 75 and older, the 99th-percentile non-external mortality relative risk was 1.12 in high-UHIdh areas and 1.07 in moderate-UHIdh areas. These estimates were compared with study-specific minimum-mortality temperatures of 27.7 °C and 28.4 °C for non-external mortality, and 28.1 °C and 28.3 °C for non-cancer mortality. Urban heat island effect-related mortality under extreme heat and non-extreme heat scenarios: A 2010-2019 case study in Hong Kong provides the study definitions and estimates.
Neighborhood and Population Health-Risk Statistics
Urban heat risk varies inside cities. A seven-city U.S. study analyzed 2.1 million daily all-cause mortality records, and significant mortality increases on extreme-heat days appeared in 12%–44% of postal codes within each city.
The range identifies spatial heterogeneity; it is not a national share or a count of heat-caused deaths. Geographic dimensions of heat-related mortality in seven U.S. cities reports the postal-code analysis.
Greater London provides a temperature-response estimate for May–September 2007–2016. The analysis linked 185,397 death records to high-resolution daily temperature data and estimated an odds ratio of 1.039 for each 1 °C temperature increase, with a 95% confidence interval of 1.036–1.043.
In Ho Chi Minh City, the 2010–2013 analysis estimated the UHI-attributable fraction of total heat mortality at 0.42%. The city had more than 7 million inhabitants in 2012, about 8.4% of Vietnam’s population, and about 2,660 people per km² in the study setting.
The same study estimated that each additional 1 km² of green space per 1,000 people could prevent 7.4 heat deaths. This is a modeled ecological association, not a randomized intervention result. Green Space and Deaths Attributable to the Urban Heat Island Effect in Ho Chi Minh City gives the population and modeled relationships.
In eight Swiss cities, 53,593 deaths occurred during the May–September 2003–2016 warm-season analysis. Mortality risk rose 31% from the 22.5 °C minimum-mortality temperature to 35 °C, the 99th percentile of warm-season maximum temperature; the reported relative risk was 1.31, with a 95% confidence interval of 1.20–1.42. Socio-environmental modifiers of heat-related mortality in eight Swiss cities: A case time series analysis reports this estimate.
Urban Heat Mitigation Statistics for Trees, Roofs and Surfaces
Mitigation estimates are scenario-specific and should not be treated as guaranteed results for every city. They also distinguish air-temperature changes from surface-temperature changes.
In a pre-monsoon summer-day model for Phoenix residential neighborhoods, each 1 percentage-point increase in tree canopy was associated with 0.14 °C of air-temperature cooling. A modeled 10 percentage-point canopy increase therefore corresponded to 1.4 °C in that study’s linear neighborhood model, not to a measured citywide intervention outcome. Urban forestry and cool roofs: Assessment of heat mitigation strategies in Phoenix residential neighborhoods reports the Phoenix relationship.
The Athens scenario study modeled combined cool roofs, roads, green roofs and trees. It found up to 1 °C of daytime ambient-air reduction, up to 9.5 °C of daytime surface-temperature reduction on a sea-breeze day and up to 11.5 °C of surface-temperature reduction on a heatwave day.
Surface temperature differs from air temperature, and the Athens figures are modeled scenarios rather than a single intervention field trial. On the cooling potential of urban heating mitigation technologies in a coastal temperate city provides the scenario estimates.
Urban Heat Measurement Data and Comparable City Indicators
The U.S. EPA describes several measurement choices relevant to urban heat comparisons. NASA MODIS surface-temperature pixels have 1 km resolution, while Landsat Thematic Mapper thermal data have 120 m native resolution and Landsat Enhanced Thematic Mapper Plus thermal data have 90 m native resolution for the cited thermal bands.
The EPA notes that the cited Landsat uses resampled the data to 30 m. Landsat’s stated repeat cycle is 16 days, but usable observations depend on clear weather and data quality; post-2003 Enhanced Thematic Mapper Plus Scan Line Corrector gaps also apply. Measuring Heat Islands describes these sensor specifications and limitations.
For longer time series, the NCEP/NCAR reanalysis listed by the EPA provides four-times-daily, daily and monthly temperature values from 1984 onward. NOAA Climate Data Online includes 30-year climate normals among its historical temperature products, although station availability and urban representativeness vary.
Hong Kong’s 2016 average population density was 6,777 people per km², rising to 57,530 people per km² in Kwun Tong. Its mortality analysis used 12 general air-quality monitoring stations averaged for each day during 2010–2019; that station count describes pollutant adjustment, not a direct heat-exposure measurement. Urban heat island effect-related mortality under extreme heat and non-extreme heat scenarios: A 2010-2019 case study in Hong Kong supplies the local density and monitoring context.
These definitions matter when comparing city indicators: a satellite surface-temperature hotspot, an urban-rural air-temperature difference, a relative mortality risk and an attributable fraction use different denominators and should not be combined as if they were the same metric.