Introduction

Further protection of heritage sites is made explicit in Sustainable Development Goal11.41. World heritage sites (WHSs) are considered of Outstanding Universal Value to humanity2, representing an irreplaceable legacy. Unprecedented climate challenges have emerged as one of the perils confronting natural world heritage sites (NWHSs), which may jeopardize their value3,4,5. The Outstanding Universal Value of the NWHS is reflected in its natural beauty or outstanding biodiversity, ecosystem and geological value6. Once the stress tolerance threshold of the NWHS is exceeded, its value suffers an irreversible loss.

NWHSs are becoming increasingly vulnerable in facing the adverse environmental impacts of global climate change7. Compared with preindustrial times, global average temperatures have increased by 1 °C, exacerbating habitat loss and degradation, competition or predation by invasive species8,9. In scenarios where emissions remain unregulated, we can expect future temperatures to increase greatly, leading to more frequent and intense occurrences of high temperatures, extreme precipitation, and droughts10,11,12,13,14,15. Climate change-related stresses can directly threaten the foundational elements that make up the outstanding value of the NWHS. For example, global warming has led to an increase in the probability of extreme rainfall and catastrophic flooding16,17,18. Species at risk from dangerous climates surge with temperatures exceeding 2 °C. Especially in the tropics, more than 80% of species are affected by 4 °C warming19,20,21. Although we have recognized the desirability of the WHS to engage with climate change4,22,23, the future climate challenges for the global NWHS remain unclear. The global network of the NWHS helps raise awareness of the impacts of climate change on human societies, biodiversity and ecosystem services. Due to the lack of an evaluation of climate change for the NWHS, it is difficult to identify the NWHS most vulnerable to climate change and, by extension, to develop appropriate strategies to manage present and future heritage adaptation from climate change.

Here, we integrated all existing natural world heritage site boundary data and organised a spatial database of 250 natural world heritage sites. We defined extreme events by climate indicators provided by Expert Team on Climate Change Detection and Indices (ETCCDI)24. For those extreme climate events that occur in NWHSs, we define the exposure of NWHSs to extreme climate events by integrating their spatial boundaries. We analyse the exposure of NWHS to climate extremes using projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6). This analysis examines how impacts vary under the Shared Socioeconomic Pathway (SSP) scenarios through to the year 2100, and considers the compound effects of multiple extreme events such as heat waves, heavy rainfall, and droughts. Given the landscape attributes of the heritage site and the conservation of heritage site biodiversity, We specifically focus on forest NWHSs due to their high biodiversity, and on those located in the tropics due to their greater susceptibility to climate impacts25. In this study, we emphasise that global climate change will place additional climate pressure on global NHWS protection, further increasing the challenge of protecting and sustaining them. Accelerating the deployment of carbon neutrality can co-benefit NHWS protection and the realisation of Sustainable Development Goals11.4.

Results

Exposure of natural world heritage sites to extreme climate events

Over the historical period (2000–2015), our observations show that 44.8% NWHSs experienced extreme heat events (Supplementary Fig. 1). Under the high emission SSP scenario of SSP370 and SSP585, more NWHSs would experience extreme heat events than historical period. (Fig. 1c, d). By 2100, nearly all NWHSs will experience extreme heat exposure, with 98.4% under SSP370 and 99.2% under SSP585. For the SSP245 scenario, more than half (69.2%) of the NWHSs will experience extreme heat exposure (Fig. 1b). Only the SSP126 scenario has a decrease in extreme climate events for NWHSs, with only 2% of NWHSs are expected to face extreme heat exposure (Fig. 1a). In all future scenarios, extreme heat is the dominant extreme climate event in the NWHSs. Other extreme climate events such as heavy rain and drought affect 14.8% and 26.8% of NWHSs under the SSP585 scenario, respectively. Under SSP370, heavy rainfall affects comparable NWHSs as in SSP585, but drought affects 5.6% fewer NWHSs. The SSP126 scenario affects fewer NWHSs than any of the other scenarios, with heavy rain affecting 3.6% and drought affecting 7.6% of the NWHSs.

Fig. 1: Exposure of natural world heritage sites to extreme climate events under different SSP scenarios.
figure 1

There are spatial and scenario differences in the exposure of NWHSs to extreme climate events. a–d Show the exposure of the NWHS under SSP126, SSP245, SSP370 and SSP585, respectively. e Shows the transformation of the exposure of NWHSs under different SSP scenarios. As emission intensity increases, NWHSs that were previously unaffected by extreme climate events may face also extreme climate. The coloured circles show different extreme climate events. The grey linking lines in the legend show the composition of composite extreme climate events.

Full size image

NWHSs are not only exposed to single climate extreme, but also to compound extreme climate events of extreme heat, heavy rain, and drought. By the end of this century, the exposure of NWHSs to compound extreme climate events will be closely linked to greenhouse gas emission and other human activities (Fig. 1e). If the temperature targets set by the Paris Agreement are achieved, such as SSP126 (Fig. 1a), our results suggest that NWHSs will face reduced exposure to high temperatures and compound extreme climate events are avoided. Under the SSP245 scenario, 17.2% of NWHSs would be exposed to compound extreme climate events. This exposure increases to 31.2% under the SSP370 scenario. In the uncontrolled scenario (SSP585), 36.4% of NWHSs suffer from compound extreme climate exposure. The 78.4% of these exposures occur in NWHSs located at mid-high latitudes and in tropical regions.

There is observational evidence indicate that NWHSs are already experiencing extreme weather events, including heat and droughts, which pose challenges to their preservation (Supplementary Fig. 1). For example, extreme heat caused causes severe bleaching of coral reefs26. Future model projections suggest that uncontrolled emissions under SSP585 would expose 36.4% of NWHSs to face compound extreme climate events. Compared to SSP245, the proportion of NWHSs exposed to extreme heat in SSP585 increases by 40%, and exposure to compound events rises by 19.2%.

Forest natural heritage sites may face more extreme climate events without control

The dominant landcover type of NWHSs reflect the natural characteristics of NWHSs and their ecological functions. Different dominant landcover types of natural heritage sites have different sensitivities to climate change and the natural environment. Applying global landcover data, we identify the dominant landcover types of NWHSs and assign each NWHS a different landscape type (Supplementary Fig. 2). Forest NWHSs constitute the largest proportion, at 32.8%, of all NWHSs and forest ecosystems are highly sensitive to climate change, especially to increases in temperature and changes in precipitation patterns27,28. There are differences in the exposure of forest NWHSs to extreme climate events under different future scenarios (Fig. 2). As scenarios escalated, the exposure to singular extreme events was gradually shifting towards exposures involving compounded extreme climatic conditions. Under the SSP245 scenario, 19 forest NWHSs are exposed to compound extreme climate events. This number increases to 33 under the SSP585 scenario, which is nearly half of the forest NWHSs. In addition, a study suggests that forest NWHSs face greater challenges in recovering from compounded extreme events, particularly under scenarios of increasing temperatures29.

Fig. 2: Exposure of forest natural world heritage sites to extreme climate events under different SSP scenarios.
figure 2

The number of exposure of forest NHWSs to extreme climate events varied across scenarios. The compound extremes is represented using an overlay of colours. Stars indicate single SSP scenarios. The lines indicate the changes in the corresponding amounts under the SSP scenarios.

Full size image

Forest NWHSs are expected to extreme heat exposure and drought stress. The models indicated differences between scenarios in high temperature days (TG90P) and 90th temperature, and yet negligible differences in high rain days(R95P) and in Standardized Precipitation Evapotranspiration Index (SPEI) (Fig. 3a–d). The high-emission scenario exhibited more intense extreme exposures than other scenarios. Under the SSP585 scenario, by the end of the century, the number of high-temperature days was projected to reach 78 ± 4(Fig. 3a), with peak temperatures reaching 31.5 ± 2.2 °C(Fig. 3b). Decreases in SPEI indicate that drought is an increased challenge to forest NWHSs (Fig. 3d), forest ecosystems often have difficulty recovering from over drought conditions30. This scenario presented a more critical situation regarding additional extreme climate exposures. Compared to SSP585, there was a 64% reduction in high-temperature days and 3.7 °C lower maximum temperatures in the SSP126 scenario. Under the SSP126 scenario, the pressure from heavy rainfall and drought did not increase (Fig. 3c).

Fig. 3: Forest natural world heritage sites to face increased exposure to climate extremes under SSP Scenarios.
figure 3

The forest NWHS varies widely across SSP scenarios, with different increments and trends. a Change in the number of high-temperature days in the forest NWHS(TG90P) (b), change in the 95th percentile temperature (c), change in the number of heavy rainfall days(R95P) (d), and change in the SPEI of the forest NWHS. a–d Figures with differently coloured solid lines indicate different SSP scenarios, and shaded areas indicate confidence intervals.

Full size image

The vulnerable biodiversity of natural world heritage sites will suffer more from high-temperature stress in the tropics

Unique biodiversity is an core element of the outstanding universal value of NWHSs. Climate change has spatially heterogeneous impacts on the biodiversity of NWHSs globally. In the tropics, biodiversity would be severely challenged by high temperatures under a high emissions scenario21. We specifically focused on the biodiversity of 106 NWHSs in the tropics.

There were differences in the intensity and frequency of extreme heat exposure for tropical-distributed NWHSs under different future pathways (Fig. 4a, b, d). Tropical NWHSs in the SSP585 scenario experienced a higher mean warming, very high temperatures (90th percentile temperature) and higher number of high-temperature days (TG90P) than the other scenarios.

Fig. 4: Differences in temperature of natural world heritages at different biodiversity.
figure 4

Given the conservation of biodiversity in the NWHSs,14 NWHSs in tropical areas are worthy of attention. a, b, d Show the changes in the 90th temperature, the number of high temperature days (TG90P), and the average temperature suffered by tropic NWHSs in different scenarios for different levels of biodiversity, respectively. The box plot visually represents the transition from the SSP126 to SSP585 scenarios through a change in colour, from green to red. Boxes indicate the first and third quartiles, and the line in the centre indicates the median. The whiskers extend to the smallest and largest values within 1.5 times the interquartile range from the lower and upper quartiles, respectively. The red (Biodiversity Intactness Index (BII) ≤ 0.8) to grey (BII > 0.8) panels on the x-axis indicate different biodiversity levels. The colours of the pie charts correspond to the categories of the biodiversity index. c Spatial distribution of the tropic NWHSs (BII ≤ 0.8). The red circles identify the selected NWHSs, while the grey circles represent the other tropical NWHSs. The area of the circle reflects the size of the NWHS. The histogram shows the distribution of the area of these sites, with the red portion representing the 14 selected NWHSs and the grey portion representing the other tropical NWHSs. The regions are classified as L1-L5, representing the five area levels of the NWSs. The height of the bar indicates the number of tropic NWHSs in levels L1-L5. The tropical region is defined as the zone near the equator, approximately between 23.5°N and 23.5°S.

Full size image

Notably, regardless of the SSP scenarios, NWHSs with vulnerable biodiversity status (9.3% of NWHSs with Biodiversity Intactness Index (BII) < 0.7 and 5.2% of NWHSs with BII of 0.7–0.8) have similar high temperature days as other NWHSs with better biodiversity quality, however the former will experience higher extreme heat. Abnormal high temperatures can put the biodiversity of NWHSs at risk, and even amplify the vulnerability of NWHSs31, especially on NWHSs that already have fragile quality of biodiversity (Fig. 4a, Suppelementary Fig. 4).

Therefore, we identified tropical NWHSs with fragile biodiversity based on the criterion of BII less than 0.8. These 14 biodiversity-vulnerable NWHSs need more attention (Fig. 4c, detail see Supplementary Table 1). These 14 NWHSs are located in South America, Africa’s mainland, and on various coasts and small islands. 64.3% of the 14 NWHSs are less than 5000 Km2. They are often parks and protected areas, which means that they have been recognised as areas in need of protection.

Discussion

Climate change impacts on the WHS are widespread, and have become one of the primary factors threatening the WHS32,33. Previous studies have already highlighted that extreme heat from climate change is causing irreversible ecological damage to WHS34, manifesting as increased coral bleaching on the Great Barrier Reef26, seagrass mortality in Western Australia35, and disruptions to the Lake Baikal food web36. Building on this, our research identifies similar extreme heat events at Lake Baikal and other WHS locations in Western Australia, and further extends these observations to an additional 248 NWHSs that have previously been underexplored. We have assessed the extreme climate events that these heritage sites are likely to experience. In the future, high temperatures will continue to be a major extreme climate challenge for these heritage sites if we do not take effective measures to control emissions. In Andalusia, Spain, the WHS faces challenges from extreme heat as well as from heavy rain and drought. These extreme weather events pose threats to the landscape structure of the heritage site37. Our study assessed not only the exposure of NWHSs to individual extreme weather events but also gave special attention to their exposure to composite extreme weather events. Due to data limitations, we only considered biodiversity in terrestrial NWHSs and not in the ocean. In the tropics, 14 NHWSs were identified for priority attention based on biodiversity conservation principles. Among the NWHSs studied here, two of the selected 14 are already listed on UNESCO’s list of world heritage in Danger (Lake Turkana National Parks has identified and potentially threatened the impacts of climate change38), five of them have already suggested the impact of climate change, as described in the UNESCO world heritage list. We have expanded our vision to a global scale and have increased our focus on future extreme climate event exposure to NWHSs. This lack of a global perspective hampers effective conservation assessments and future planning. Our study offers a global, multi-scenario climate assessment. Our results show that NWHSs will be exposed to an increasing number of extreme climate events by the end of the century, especially a high-emission scenario almost all NWHSs will be exposed. However, heritage sites face not only climate pressures but also excessive interference from human activities, including mining, tourism, and logging39,40,41.

Extreme weather events are a natural part of heritages and are unavoidable in the future. Despite the uncertainties in simulations of models42,43, a global assessment of the challenges of climate extremes on world heritage sites under future simulations is still necessary. Extreme climate events tend to have stronger impacts on ecosystems, and shocks that exceed thresholds may upset the balance of ecosystem resilience29,30,44,45,46. Increasingly high heat extremes are impacting the ecosystems of NWHSs and threatening biodiversity. Model result suggest that rigorous climate change can help NWHSs mitigate exposure to shocks from extreme weather events in the future. For example, extreme ocean temperatures are causing the Great Barrier Reef in the Australian region to bleach more frequently each year. If no response is taken, the Great Barrier Reef will likely experience bleaching once a year26. Reducing emissions is vital to any solution, but it is beyond the control of local organisations and the window of time to prevent the worst effects of anthropogenic warming is closing47. Although it is still unclear how successful mitigation actions could be, it is clear that mitigation efforts and management of human activities are necessary to protect the world heritage sites.

Methods

Natural world heritage site and landcover data

The World Heritage Sites List provides the coordinates, names, and accompanying text information for 267 nature world heritage sites. World heritage sites are not always described by accurate shape and extent, often intentionally, to protect fragile heritage sites. The World Database on Protected Areas (WDPA) provides data on the boundaries of protected areas36, which also contain the boundaries of natural world heritage sites. By integrating both, the rich text information and the boundary data provided by WDPA, we have aggregated to form a spatial data on natural world heritage sites. This contains the spatial extent coordinates, centre ___location, area, and textual descriptive information for 250 natural world heritage sites. Although there are still 17 natural world heritage sites missing, 94% of the area is covered.

Global land use dataset was obtained from Gong et al.48. The spatial resolution of the land use dataset is 10 m. It distinguishes between attributes such as forest, water body, and bare land, and can provide information on the environment within natural world heritage sites.

Historical reanalysis data

The European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 reanalysis is an Integrated Forecasting System (IFS)-based product within C3S49. The daily mean data used in this study is obtained from the ERA5 dataset with a horizontal resolution of 0.25° × 0.25° from 2000 to 2015, including 2 m surface air temperature and precipitation.

Future scenario simulation and model output

Shared socioeconomic pathways (SSPs) is a set of pathways for modelling future societal developments that are closely related to the impacts of climate change and the formulation of climate policy. These pathways include not only projections of greenhouse gas emissions, but also socioeconomic factors such as population growth, economic development, energy consumption and land use50. SSP126 is a low emission scenario. Mitigation measures are taken worldwide to keep global warming below 2 °C51. SSP245 is a medium emissions scenario that assumes that future development of the world takes into account economic, social, and environmental goals in a balanced way.SSP245 is a baseline scenario commonly used in many integrated assessment models (IAMs)52. SSP370 is a high-emission scenario in which socio-economic development focuses on regionalism and competition. SSP585 This is the most extreme high-emissions scenario, which assumes that fossil fuel use continues to increase, resulting in very high greenhouse gas emissions. Under this scenario, global warming could exceed 4 °C, which would lead to severe climate change impacts and environmental catastrophe52.

For Scenario Model Intercomparison Project (ScenarioMIP), the part of the international Coupled Model Intercomparison Project 6 (CMIP6) of the World Climate Research Programme (WCRP) provides simulation results for various sets of complex Earth system models under different SSPs. We selected the daily temperature and precipitation variables of 3 models (AWI-CM, TaiESM, BCC-CSM2MR model) under the SSP126, SSP245, SSP370, and SSP585 scenarios in 2015–2100 to calculate the extreme climate indicators individually, and averaged the model-calculated extreme climate indicators.

Extreme climate indicators

Following the ETCCDI suggestion53,54,55 and The Intergovernmental Panel on Climate Change (IPCC) report10, We selected 3 types (extreme heat, heavy rain and extreme drought) of indicators for extreme climate events and calculated 3 types indicators for 4 SSP scenarios (Supplementary Table 2).

For extreme heat indictor, annual maxima of daily maximum temperature (TXx) can be used to indicate an intensity of extreme heat.

$$TX{x}_{j}=\,\max (T{X}_{ij})$$
(1)

\(T{X}_{{ij}}\) be the maximum temperature at i day of period j. We calculate TXx here as the maximum temperature for the year.

Number of days with mean temperature above the 90th percentile (TG90P) is used as a frequency indicator of extreme heat.

$$TG90P={\sum}_{d=1}^{365}\Bigg\{\begin{array}{cc}1 & {{{\rm{if}}}}\,{T}_{{{{\rm{Y}}}}}(d) \, > \, {T}_{{{{\rm{ref90}}}}}(d)\\ 0 & {{{\rm{otherwise}}}}\hfill\end{array}$$
(2)

\({T}_{Y}(d)\) represents dth day in year Y. \({T}_{{ref}90}\left(d\right)\) indicates the 90th percentile value under the reference period. Due to the differences between the models and the differences between the models and the ERA5 data, we set the reference period for each model to be the first 10 years of future simulated data for SSP245.

For heavy rain indictor, Annual maximum 1-day precipitation (Rx1day) can be used to indicate an intensity of heavy rain56. Number of days with precipitation above the 95th percentile (R95P) is used as a frequency indicator of extreme heat.

$${{{\rm{Rx1day}}}}=\,\max \{P(d)|d\in D{{{\rm{a}}}}ys\,of\,Year\}$$
(3)
$${{{\rm{R95P}}}}={\sum}_{d=1}^{365}\Bigg\{\begin{array}{cc}{P}_{Y}(d) & {{{\rm{if}}}}\,{P}_{Y}(d) \, > \, {P}_{{{{\rm{re}}}}f95}(d)\\ 0 & {{{\rm{otherwise}}}}\hfill\end{array}$$
(4)

\(P(d)\) shows the amount of rainfall in the middle of the year. \({P}_{{ref}95}\left(d\right)\) indicates the 95th percentile value under the reference period. We set the reference period for each model to be the first 10 years of future simulated data for SSP245.

For the assessment of drought, we selected SPEI as the evaluation indicator. The difference between Precipitation (P) and Reference Evapotranspiration (ET0) is key for commutating SPEI57.

$$SPEI=Z\left({\sum}_{i=1}^{n}({P}_{i}-E{T}_{i})\right)$$
(5)

Z indicates the Z-score obtained by fitting the cumulative water balance to a standard normal distribution. The i denotes the selected time window, here we choose I = 6 months as the time window.

Exposure for natural world heritage sites

To explore the climate change challenges facing natural world heritage sites, this study focuses on the intensity and frequency of climate extremes. We used specific climate metrics to quantify these impacts. Specifically, we considered extreme heat events, defined as annual maxima of daily maximum temperature (TXx) and Number of days with mean temperatures exceeding the 90th percentile values for the reference period (TG90P) within a 10-year period at the end of the century. Meanwhile, extreme rainfall events were defined as annual maximum 1-day precipitation (Rx1day) and Number of days with precipitation exceeding the 90th percentile values for the reference period (R95P). In addition, drought conditions are assessed by the SPEI, which indicates an increase in the degree of drought when the SPEI is lower than the reference period for the last decade of the century. We considered the NWHS to have experienced exposure to extreme climate events during this period once the calculated extreme climate metrics met the previous requirements. Note that we have chosen here the extreme event exposure on which all models agree. That is, we favour a more conservative estimate.

Consequently, heritage sites where these thresholds are exceeded are considered to be under additional exposure to extreme climate events. NWHSs sometimes experience multiple extreme climate events during some time periods. We identify exposure to compound extreme climate events when two or more of these conditions—extreme heat, extreme rainfall, or drought—are present during the same year of extreme climate exposure.

Assigning landscape type classification

In order to classify landscape types for each natural world heritage site, we undertook a cluster analysis. Firstly, the global land use coverage data allows us to obtain the proportion of each land use type in the natural world heritage sites. Secondly, in cluster analysis, defining the appropriate number of clusters is essential for meaningful categorization. We employed the Davies-Bouldin Index(DBI)58 and Silhouette Coefficient(SC)59 to determine the optimal number of clusters due to their effectiveness in assessing cluster separation and cohesion.

The DBI is also known as the classification suitability index, which calculates the ratio of the sum of in-class distances to out-class distances. SC is a way to evaluate the effectiveness of clustering and combines two factors, namely, cohesion and segregation. while a lower DBI value is indicative of better cluster separation and cohesion, a higher SC value is preferable as it reflects a clearer and more confident clustering assignment

$$DBI=\frac{1}{k}{\sum}_{i=1}^{k}ma{x}_{i\ne j}\left(\frac{{\sigma }_{i}+{\sigma }_{j}}{d({c}_{i},{c}_{j})}\right)$$
(6)

k is the number of clusters, \({\sigma }_{i}\) is the average distance of samples within the ith cluster, \({\sigma }_{j}\) is the ith cluster centre, and d is the distance between two cluster centres Ci and Cj.

$${{{\rm{s}}}}({{{\rm{i}}}})=\Bigg\{\begin{array}{ll}1-{{{\rm{a}}}}({{{\rm{i}}}})/{{{\rm{b}}}}({{{\rm{i}}}}), & {{{\rm{ifa}}}}({{{\rm{i}}}}) \, < \, {{{\rm{b}}}}({{{\rm{i}}}})\\ 0, \hfill & {{{\rm{ifa}}}}({{{\rm{i}}}})={{{\rm{b}}}}({{{\rm{i}}}})\\ {{{\rm{b}}}}({{{\rm{i}}}})/{{{\rm{a}}}}({{{\rm{i}}}})-1, & {{{\rm{ifa}}}}({{{\rm{i}}}}) \, > \, {{{\rm{b}}}}({{{\rm{i}}}})\end{array}$$
(7)

a(i) is the average distance from sample i to other samples in its same cluster (i.e., within-cluster dissimilarity). b(i) is the average distance from sample i to all samples in the nearest other cluster (i.e., between-cluster dissimilarity).

Finally, we identified the dominant land cover type for each World Natural Heritage site based on the category that constitutes the largest proportion of the site’s area. Combining the UNESCO descriptions, 7 landscape types were identified, which were assigned to each natural world heritage site. (See Supplementary Fig. 2 and Table 3 for detailed information).

Biodiversity and correlation analyses

Biodiversity quality is obtained from the Biodiversity Intactness Index (BII) in Newbold et al.60 (Supplementary Fig 3). The BII is useful to reflect the quality of terrestrial biodiversity. Newbold et al.60 offer a precautionary threshold set at a BII value of 0.9 BII. This benchmark serves as a warning sign, suggesting that any loss of biodiversity beyond this limit could disrupt ecosystem functions and services. In this study, we defined different levels of biodiversity vulnerability. NWHSs with a BII < 0.7 are considered severely vulnerable. Those with a BII between 0.7 and 0.8 are deemed vulnerable and those with a BII ≥ 0.9 or higher are classified as relatively safe60.

Rising levels of biodiversity vulnerability are closely linked to changes in temperature change20, both in relation to increases in average temperatures60 and to extreme high temperatures25 (like 90th temperature). We made a correlation between biodiversity and temperature changes by using linear least-squares regression (Supplementary Fig 4). Due to the rich biodiversity of the tropics and its sensitivity to warming25, We selected 106 tropic NWHSs, which accounted for 42.4% of all NWHSs, and calculated the correlation between those BII and temperature variables in these regions We investigated the correlation between biodiversity and 90th percentile temperatures at tropical NWHSs, where the 90th percentile temperatures were calculated based on daily temperature data over the historical period of the model. Our result suggest a negative correlation between biodiversity and extreme high temperatures in tropic NWHSs (Supplementary Fig 4). According to the BII threshold we set (BII < 0.8), we identify 14 NWHSs in the tropics that need to be focussed.