Data application at city scale for urban heat island diagnosis.
Next, we demonstrate the potential of crowdsourced climate data to support urban heat resilience and adaptation policies for the city of London. The analysis of hourly UHI intensity and implications is summarised Fig. 5. The UHI intensity was measured as the temperature difference between each CWS and the mean temperature of the surrounding rural area. Different percentiles across the CWS sample per hour were used to characterise the UHI magnitude in the city in terms of duration, intensity, seasonality, intra-day patterns, and extreme heat situations.
The results show that the UHI intensity exceeded 4ºC (90th percentile) during 731 hours in 2021 (see Supplementary Note 6). This intense urban overheating, predominantly nocturnal, was found to be almost constant throughout the year, with similar UHI intensities even exceeding 4ºC at nighttime during the coolest winter days (Fig. 5a). During the daytime (Fig. 5b), a slight decrease in UHI intensity was observed with higher temperatures, with a considerable reduction in the 25th percentile temperature values, illustrating the urban cool island (UCI) phenomenon in the diurnal hours. These UCIs are associated with cooler urban areas than surrounding rural areas38,39. The mechanisms that determine the UCI intensity seem linked with the degree of compactness of urban areas as well as with other specific urban microclimatic variations (e.g. courtyards10) in the city. They can be observed in Fig. 5f-h and Fig. 5j-l. These figures map the UHI across the city on days with higher night and daytime UHI intensities, respectively, showing negative values in some urban areas in the morning.
The UHI intensity variations in London are more related to meteorological factors, particularly solar radiation, wind, cloudiness, and precipitation, than annual seasonality, as supported by the literature12,40,41. However, this extreme urban heat exposure has higher implications on summer days. Figure 5c illustrates the boxplot distribution of 90th, 75th and 25th percentiles of hourly UHI intensity during the warm season (from June to September 2021), highlighting the hours with higher UHI intensities, predominantly at night, with peak values up to 6.0ºC (90th percentile). These observations are slightly higher than previous studies using data from observations11,41 and simulations21.
Despite the highest UHI intensity prevalence being frequently at night, the results also show some daytime hours with UHI intensity above 4ºC (90th percentile) during the hottest days. Figure 5d shows the overlapping air temperature data of CWS, OWS, and average rural temperature during July and August. In this warm period, the hourly magnitude of the UHI phenomenon is shown in Fig. 5e. These two figures confirm previous findings showing how UHI intensity is higher during extreme heat temperatures and heat waves7. The highest nighttime UHI intensity was found on 18/07 at 21:00 (5.6ºC, 90th percentile); and the highest daytime UHI intensity was found on 20/07 at 17:00 (5.0ºC, 90th percentile), both days with maximum temperatures over 30ºC. These days are highlighted in purple in Fig. 5d,e, and their mapping sequence at different times is shown in Fig. 5f-i and Fig. 5j-m, respectively.
This UHI visualisation helps to understand the implications of aerodynamics and imperviousness of the city affecting this problem, with a not constant pattern every day, and highly influenced by other meteorological factors, particularly precipitation, solar radiation, wind, and cloudiness.
The day 18/07, with the highest night UHI intensity registered, illustrates an example of the urban temperature oscillation of a typical summer day (like the statistical values shown in Fig. 5c). In this typical summer day, the highest UHI intensities are found at night, and with daytime overheating below 2ºC. The phenomenon of UCI at daytime hours can also be noted, where some urban areas registered lower temperatures than surrounding rural areas.
On the other hand, the day 20/07 registered the highest UHI intensity in daytime. This daytime phenomenon with UHI intensities higher than 4ºC (90th percentile) only happened three days during the summer of 2021. This particular daytime overheating can be appreciated in Fig. 5k, which shows an east-west gradient associated with localised summer rain during that timeframe, dropping temperatures quickly from around 15:00 (Fig. 5m).
The analysis demonstrates the high spatio-temporal resolution of urban climate achievable with citizen data, showing how urban climate patterns are not constant, and urban climate experiences a range, especially during the daytime. This differs from existing studies showing that the mechanisms affecting urban overheating (particularly aerodynamics and imperviousness of the city42) are consistent. These results advance the existing literature, on the basis of which localised diagnosis and urban climate adaptation interventions can be efficiently assessed.
Prioritising interventions in persistently overheated urban areas
The first step towards effective heat adaptation interventions in cities is to characterise persistently overheated urban areas to efficiently identify climate risks as a function of hazard, exposure and vulnerability (Fig. 1). We propose using Cooling Degree Hours (CDHs) as a heat stress indicator for UHI diagnosis43,44. CDHs are calculated across London for the whole year (total CDHs) and at specific timeframes to consider only night and daytime hours (daytime CDHs and night CDHs). The results are also statistically compared with London’s Local Climate Zones (LCZs) classification (Fig. 6a), derived from the European LCZ map proposed by Demuzere et al.45, and recently extended worldwide46. The LCZs scheme connects London’s urban regions with uniform characteristics related to urban temperatures (e.g. building morphology, surface cover, material and structure, among others).
In the first step, the total CDHs for London wards, illustrated in Fig. 6b, show the annual implications of the UHI in the city, increasing CDH exposure by more than 60% (from 3133 to 5064 CDHs) for some urban areas. Analysing the data by comparing total CDHs with night/daytime CDHs, the results demonstrate how the delimitation of hot spots (dash lines) is not convergent between daytime and night. This is derived from urban areas that can operate very well or badly during specific spatial and temporal regions. So, CDHs distinguishing between these timeframes help to identify urban problems more efficiently.
The daytime CDHs map (Fig. 6c) shows a delimited hotspot area with a CDH increase of more than 50% (up to 1432 CDHs). Daytime hotspots are found not to be clearly associated with LCZs, with overheating predominantly located in some open mid-rise (LCZ6) and open low-rise areas (LCZ5), but with large variability, presenting the highest interquartile and maximum-to-minimum difference. Areas with an open arrangement and mostly paved surfaces (LCZ8, LCZ10), typical of shopping centres and industrial areas, also presented slightly high daytime CDD values. This daytime overheating delimitation seems associated with areas with higher surface temperature obtained in previous studies in London using data from Landsat 847 and AVHRR instrument48. Imperviousness and the lack of evapotranspiration from water surfaces or green infrastructure have been discussed as the primary focus influencing this daytime overheating22,42. Here, solutions based on evapotranspiration infrastructures, solar protection and albedo may help to overcome this problem.
In the night CDHs map (Fig. 6d), areas with an increase higher than 240% (up to 766 CDHs) during the night are delimited. Night hotspot fits very well with compact urban zones (LCZ1–3; Fig. 6e), a statement consolidated in the literature38,39,49,50. Here, the influence on urban morphology, or aerodynamic roughness, seems to be the primary source influencing this nighttime overheating4. However, it was found that large paved surfaces (LCZ8, LCZ10) also presented slightly high night CDDs. Here, night overheating mainly seems to be an urban structural problem (reducing heat dissipation capacity) but is also affected somewhat by surface cover properties (increasing heat absorption), requiring higher adaptation efforts. These results demonstrate the potential of citizen data to identify accurate and localised climate risks – based on identified urban hazards and characteristics – and select the best available interventions for cost-effective climate resilience and adaptation.
Toward effective city climate action plans
This study demonstrates the benefits of citizen weather data from open-access weather platforms to measure urban climate at the highest spatio-temporal resolution levels and perform diagnoses to prioritise cost-effective localised adaptation. Citizen weather stations involve more than 550,000 observations globally, providing a network 12 times larger than the professionally operated worldwide. We identify cities with more significant data availability per continent, highlighting broad applicability in North America, Europe, Oceania, and some regions of Asia and South America.
A data pre-processing and analytic procedure is proposed and validated, achieving high accuracy results with a mean temperature deviation of 0.48ºC, inside the precision range of best available sensors (from ± 0.3 to ± 0.5ºC). It’s application in London demonstrates the potential of such urban weather observations for diagnosing and building heat resilience in cities. The results show how the London urban heat island phenomenon is almost constant throughout the year, with major implications during the hot season, achieving a mean urban overeating close to 4ºC at night and below 2ºC in the daytime. This urban overheating increased outdoor heat exposure by more than 60% in some urban areas, with Cooling Degree Hours (CDHs) rising from 3133 in the surrounding areas to 5064, having serious implications for human health and the economy. The intra-day analysis also demonstrates how the delimitation of hotspots is not convergent between daytime and night, with no association between daytime overheating and local climate zones classification. While local climate zones classification showed a strong relationship with night overheating, daytime hotspots were more associated with higher surface temperatures. Aerodynamics or imperviousness affected urban overheating at daytime and night differently, requiring different climate adaptation solutions. This novel approach can help quickly measure extreme urban heat exposure situations in cities, map the persistent urban hotspot areas, identify risks to support climate action plans, and even monitor the adaptation progress in real-time.
Such a new citizen data source would allow not only better support the implementation and monitoring of cost-effective climate adaptation strategies in cities but also to generate more accurate climate information for climate studies and architectural design. For example, in the case of London, buildings in night hotpots might require additional heat dissipation strategies for night ventilation to avoid overheating if no measures are taken to improve the microclimatic response of their surroundings. This consideration can be an important intervention since a dramatic variation of night CDH heat exposure, higher than 240% (up to 766 CDHs), was found. Including and scaling citizen science data, with the proposed open source procedure for analysis, can be an important step in delivering timely, accurate and localised urban responses to the unprecedented rise in temperatures and extreme heat.
Limitations and scalability
The following limitations should be considered when using citizen weather data. The data availability is limited to developed countries, with a significant data scarcity in emerging and developing regions. It should be regarded that although CWS may be located mainly in the built environment, there might be a potential bias in green areas, which may be overlooked. Also, future standardisation of sensor location and protection may help reduce the number of outliers in the raw data.
These data sources and procedures are highly scalable since they can be combined with additional data sources, such as other meteorological networks, private monitoring, or even additional citizen science networks. The data used in this study were primarily used as a cost-efficient solution for high-resolution monitoring of urban climates through crowdsourcing. However, the proposed approach is not limited to Netatmo and Wunderground, and it can be combined or integrated with other data to support urban climate diagnosis, especially in those locations more affected by rising temperatures. We argue that this validated citizen science approach could drive the deployment and expansion of these low-cost CWS networks globally to support climate adaptation.