Not all biodiversity richspots are climate refugia

Anthropogenic climate change is increasingly threatening biodiversity on a global scale. Richspots of biodiversity, regions with exceptionally high endemism and/or number of species, are a top priority for nature conservation. Terrestrial studies have hypothesised that richspots occur in places where long-term climate change was dampened relative to other regions. Here we tested whether biodiversity richspots are likely to provide refugia for organisms during anthropogenic climate 10 change. We assess the spatial distribution of both historic (absolute temperature change and climate change velocities) and projected climate change in terrestrial, freshwater, and marine richspots. The results suggest that although terrestrial and freshwater richspots have been and will be somewhat less affected than other areas, they are not excluded from the impacts of global warming. Their characteristic biota is expected to witness similar forcing as other areas, including range shifts and elevated risk of extinction. Marine richspots have warmed even more, have higher climate velocities and are projected to 15 experience higher future warming than non-richspot areas. Our findings emphasise the urgency of protecting a comprehensive and representative network of biodiversity-rich areas that accommodate species range shifts under climate change. https://doi.org/10.5194/bg-2021-179 Preprint. Discussion started: 12 July 2021 c © Author(s) 2021. CC BY 4.0 License.


Introduction
It has been suggested that the reason that some geographic areas are exceptionally rich in biodiversity (sometimes called 20 "biodiversity hotspots") is because they have had little climate change over geological timescales. This long-term stability has led to high numbers of species that are unique to these "climate refugia" (i.e., endemic species; Dynesius and Jansson, 2000;Jansson, 2003;Harrison and Noss, 2017;Senior et al., 2018;Brown et al., 2020). Here, following Manes et al. (2021), we call these areas biodiversity "richspots" to emphasize the distinction from areas of increased temperature, invasive species, pollution and habitat destruction. If this climate refugium hypothesis is true, then richspots may continue to provide safe 25 harbours (refugia) for species under anthropogenic climate change. Thus, conserving these areas would not only protect species against current human impacts (Halpern et al., 2015;Díaz et al., 2019;Tedesco et al., 2013), such as hunting, fishing and habitat loss, but also limit the effects of climate change on global biodiversity (García molinos et al., 2016).

Biodiversity and climate change
The effects of climate change are now detectable on biodiversity trends since the 1950s (e.g., Chaudhary et al., 2021), and are 30 projected to accelerate in coming decades (e.g. Manes et al., 2021). However, existing human impacts are already devasting biodiversity in all environments. While most confirmed extinctions and threatened species are terrestrial, a higher proportion of freshwater species are threatened, which is reflected in the higher proportion of freshwater richspots impacted by human impacts (Collen et al., 2014;Costello, 2015;Harrison et al., 2018). The rate of species endemism is exceptionally high in freshwater biogeographic realms, at 89-96% for fish in all but one realm, compared to 11-98% for terrestrial vertebrate groups 35 and 17-84% for marine realms (Leroy et al., 2019). Based on species ranges and conservation status, >25% of IUCN-assessed marine species are threatened in 83% of the oceans (O'hara et al., 2019). The lower thermal safety margins of marine ectotherm species renders them more vulnerable to climate change than terrestrial ectotherms (Pinsky et al., 2019).

Biodiversity richspots
Biodiversity richspots have been proposed in many studies based on different criteria, taxa and geographic contexts (e.g. Myers 40 et al., 2000;Mittermeier et al., 2004;Mittermeier et al., 2011;Asaad et al., 2017;Noss et al., 2015). Eighteen different classifications of marine biodiversity richspots alone have been proposed (Jefferson and Costello, 2019). The most comprehensive scheme of richspots is the so-called "WWF Global 200" (Olson and Dinerstein, 2002;G200), which covers terrestrial, freshwater and marine environments, and has been used in previous climate risk assessments (e.g. Warren et al., 2018;Manes et al., 2021). In all cases, the delimitation of richspots was based on expert opinion and limited to a few well-45 known taxa, such as flowering plants and vertebrates. Thus, it is possible that these selections of biodiversity richspots may have taxonomic and/or expert knowledge biases. An objective approach to mapping biodiversity richspots has been applied for the world oceans, using globally standardised data-driven measures of species richness, endemism, habitat, biome and ecosystem distributions (Zhao et al., 2020). This objective designation of representative biodiversity areas (RBAs) indicated https://doi.org/10.5194/bg-2021-179 Preprint. Discussion started: 12 July 2021 c Author(s) 2021. CC BY 4.0 License. that the 30% most biodiversity rich areas of the ocean would contain 68% of all species, 94% of coral reefs and mangrove 50 forests, and 86% of kelp forests and seagrass meadows.

Climate velocity and range shifts
Climate velocity (Loarie et al., 2009;Burrows et al., 2011) is a key concept to understand the origin and fate of biodiversity richspots under climate change. The velocity of climate change is the pace and direction at which a specified climate variable changes across geographic space due to changing climate. For example, climate velocity for temperature is the speed at which 55 points of the same temperature (isotherms) move due to changing climate (distance time -1 ). Regions of high climate velocities are those with low topographic relief on land, particularly flooded grasslands and deserts (Loarie et al., 2009), tropical and Arctic regions; as well as offshore tropical and polar regions in the oceans (Burrows et al., 2011;Burrows et al., 2014;García Molinos et al., 2016;Brito-Morales et al., 2018;Brito-Morales et al., 2020).
Some terrestrial areas that have experienced low climate velocities since the last glacial maximum are rich in endemic species 60 and hence more likely to be identified as richspots (Sandel et al., 2011). Climate velocities are also able to predict the direction and pace of past and future species range shifts (Pinsky et al., 2013;Brito-Morales et al., 2018). Marine species tend to follow the physical pathway dictated by climate velocities more closely than terrestrial species probably due to fewer dispersal barriers than on land and the smaller thermal safety margins of marine species (Sunday et al., 2012;Pinsky et al., 2019). Spatial patterns of climate velocities show regions where species are expected to leave, pass through or arrive within a certain period under a 65 particular climate change scenario (Burrows et al., 2014). Elevated climate velocities are expected to be especially problematic for endemic species, which may have limited dispersal ability (Sandel et al., 2011;Brito-Morales et al., 2018), especially when they live in enclosed seas such as the Mediterranean, from where they can be trapped under global warming.

Anthropogenic climate change in richspots
Current policies put the world on track for around 3°C of heating by the end of the of the century (Hausfather and Peters, 70 2020). Manes et al. (2021) suggested (based on studies available for half of the richspots), that at this degree of warming, 92% of land-based endemic species and 95% of marine endemics face negative consequences, such as a reduction in abundance.
With the doubling of global warming from 1.5°C to 3°C, there is at least a 10-times increase in local extinction risk in biodiversity richspots: rising from 2% for all species on land and sea to 20% and 32% at risk. Of endemic species, 34% and 46% in terrestrial and marine ecosystems, and 100% and 84% of island and mountain species were projected to face high 75 extinction risk, respectively. The fact that these species are endemic suggests they cannot disperse to other areas, and thus a local extinction within a richspot would mean global extinction. In contrast, introduced invasive species were projected to be unaffected by climate change or benefit from it, while their expansion will further threaten the survival of native species.
However, if warming rates are lower inside than outside these richspots, then impacts of climate change should be reduced relative to other regions. Until now, there has been no comparison of recent or projected global warming inside and outside 80 biodiversity richspots. Here we assessed the past and future-projected magnitude of climate change in biodiversity richspots https://doi.org/10.5194/bg-2021-179 Preprint. Discussion started: 12 July 2021 c Author(s) 2021. CC BY 4.0 License.

4
and compared those variables with other regions. We tabulated historic differences in temperatures of the past 50 years and climate change velocities, as well as projected temperature and precipitation change. Our findings are of importance for policies to address biodiversity loss and climate change.

Observed climatic change and climate velocities
To assess past changes of temperature, we used monthly interpolated data from the CRU TS 4.05 (Harris et al., 2020) and the HadISST1.1 (Rayner et al., 2003) compilations for the near-surface air and ocean surface temperatures (hereafter air and ocean temperatures). We used measurements compiled from the last 50 years to assess the magnitude of change from averages of the 90 1971-1980 until the 2011-2020 interval. The same timespan (1971-2020) was used to calculate the velocities of climate change for temperature. Climate change velocities were calculated separately for air and ocean temperatures using the "VoCC" R package (Molinos et al., 2019). Original data layers of historic air temperature had 0.5 × 0.5; ocean temperature had 1 × 1, latitude-longitude degrees of resolution. Antarctica was not represented in the air temperature data.

Future climate projections 95
Future projections of changes of air and ocean temperatures, as well as aggregated precipitation were downloaded from the IPCC Atlas of the AR6 report (Iturbide et al., 2020). These are the results of 6th Phase of the Coupled Model Intercomparison Project (CMIP6, CMIP6, Eyring et al., 2016), and represent multi-model averages at different stages of warming when global warming reaches the +1.5, +2, and +3°C thresholds compared to the simulated pre-industrial baseline . Data layers that represented the same stage of global warming were averaged across four different scenarios (SSP126, SSP245, SSP370 100 and SSP585) to obtain a single expectation for the three stages of warming. The data also show comparisons to the preindustrial baseline. Changes of precipitation were rescaled to mm year -1 . The downloaded future-projected climate data had 1 × 1 degree resolution.

Richspot schemes
The impact of climate change was assessed using three richspot schemes: (i) The "WWF Global 200" (G200) scheme of Olson 105 and Dinerstein (2002) is the most comprehensive and was designed to represent areas prioritized for conservation on land, freshwaters and ocean; (ii) the partly overlapping group of terrestrial "hotspots" (hereafter called Myers) proposed by Myers et al. (2000) with the modifications of Mittermeier et al. (2011) and Noss et al. (2015), which is based on species endemism and habitat loss; and (iii) the 30% highest marine biodiversity areas of Zhao et al. (2020). Results based on air temperature and precipitation were used to assess terrestrial and freshwater, and ocean temperature data were used to assess marine richspots. 110 https://doi.org/10.5194/bg-2021-179 Preprint. Discussion started: 12 July 2021 c Author(s) 2021. CC BY 4.0 License.

Analysis of climate-change variables
Prior to the analyses, the climatic data layers were resampled to 0.25 × 0.25-degree resolution using the bilinear method, which was necessary to ensure that adequate (albeit smoothed) information is passed to small richspots. All spatial data items (climate variables and richspot schemes) were projected to the Mollweide equal-area projection, which was used throughout the analyses. This step ensured that every pixel represents an equal area, so pixel counts translate to cumulated area and global 115 means are not biased by the unequal spatial sampling along latitudes. For air temperature and precipitation only land-based values were included in this assessment. The coordinates of richspot centroids were tabulated to assess the latitudinal patterns of their distribution and those of their characteristic impacts.
We separated our impact variables (historic temperature difference, climate change velocity, projected warming and precipitation) to values that fall within and outside a richspot-scheme. Within and outside richspot-scheme areas were 120 compared with their respective mean values. We also tabulated the impacts for every individual richspot of all schemes, except that of Zhao et al. (2020) which represents a single area covering 30% of the ocean. Every richspot was characterized with one mean value of the equal-area pixels that fell within its boundaries. To express the uncertainty of within-richspot climate change due to variability among individual richspots in a scheme, we executed bootstrap simulations of richspot-means and tabulated their mean in every simulation trial using the areas of richspots as weights. Errors are reported as the standard deviations of 125 the bootstrap distributions, based on 10 000 trials.
For estimates of historic temperature change, we tabulated the proportion of pixels in a richspot that have been warming in the past 50 years. Richspots that had more than 95% of their pixels above +0°C were considered to have been significantly affected by climate change. We also tabulated the 2.5 and 97.5% percentiles of the distributions of pixels in every richspot and contrasted these with the global and latitudinal means of the respective variables. Richspots where the global mean was above 130 the richspot's 97.5% percentile were considered global refugia, those with the global mean below the 2.5% percentile were considered critically warming. Refugia and critically warming richspots within latitudinal bands were tabulated the same way, and only compared to the variable's value at the latitude of the richspot's centroid (Fig. 4, Table 1).
Despite that 10 of the 43 marine richspots (23%) did not witness significant overall increases of ocean temperature (Okhotsk 145 Sea, Galápagos, Humboldt Current, Fiji Barrier Reef, Benguela Current, Agulhas Current, Rapa Nui, Patagonian Southwest Atlantic, New Zealand Marine, Antarctic Peninsula and Weddell Sea, Fig. 1a), marine richspots on average have been affected more by climate change than terrestrial or freshwater richspots. Ocean temperature in the G200 marine richspots has increased 41% more than outside (0.53 ± 0.06 vs. 0.38, with global average of 0.39 °C) and climate velocities were 69% higher than areas outside (11.24 ± 1.86 vs. 6.64 km decade -1 , Fig. 2a). This difference is less pronounced when the RBA of Zhao et al. 150 (2020) is considered: this area faced 4% more warming (0.41 vs. 0.39 °C) and climate velocities have been 33% larger than outside (8.86 vs. 6.65 km decade -1 ).

Projected changes
Near-surface air temperature is projected to warm considerably faster over land than over the seas (+2.03, +2.64 and +3.93 °C with +1.5, +2 and +3°C of warming). This means air temperatures above land warm more than over the ocean and the global 155 average. However, when compared to other land areas (Antarctica included), freshwater richspots will be less affected by temperature changes, and will see 16, 15 and 14% less warming than areas outside at the +1.5, +2 and +3°C warming stages, respectively (Fig. 3). Terrestrial richspots defined using the Myers scheme are projected to experience similar patterns (20% less within than outside with all three stages), whereas areas using the terrestrial G200 are projected to be about as much affected by temperature changes as areas outside them (5% less than outside). 160 Marine richspots of the G200 will continue to be more affected (12, 13 and 13%) than outside areas, with highest projected warming in the northern, and lowest in the southern, polar regions. The 30% RBA of Zhao et al. (2020), on the other hand, is expected to be only 1% more affected by global warming as other areas.

Variation across richspots
Compared to the global mean temperature changes (both observed and future), most terrestrial and freshwater richspots represent climate refugia (Table 1a, Fig. 4) and only a minority of these (< 20%) are expected to warm critically (Table 1b).
In contrast to terrestrial and freshwater richspots, most marine ones of the G200 are not climate refugia, with the notable 170 exception of the Antarctic richspot (Figure 1), and a considerable number of marine richspots are positioned in high-velocity areas. Almost half of the marine G200 are expected to face higher-than-global warming in this century.

7
The latitudinal distribution of richspots is similar in all three environments (see Supplementary Information). Northern high latitude richspots will warm most, whereas the Southern Ocean and the upwelling on the Atlantic coast of southern Africa will cool (Figs. 1, 4). Following the latitudinal patterns of warming, richspots in the northern hemisphere are disproportionally 175 more affected by the magnitude of temperature increase than those in the southern hemisphere (Fig. 4). Terrestrial and freshwater richspots tend to occur in places where climate velocities are comparatively lower than those suggested by the latitudinal average (Fig. 4b).

Discussion
Our results show that although the impacts of climate change have been lower in terrestrial and freshwater richspots, they have 180 been and are projected to be affected by climate change. Marine biodiversity richspots have and are projected to experience greater effects of climate change than other areas. This discrepancy reflects both the spatial distribution of richspots and the latitudinal patterns of climate change. The latitudinal imbalance (i.e. polar amplification) of global warming is expected to further exacerbate the already asymmetric human impact on the marine environment and the biota (Halpern et al., 2015;Sydeman et al., 2021). 185 Although overall warming is expected to affect marine richspots only slightly more based on future projections, the velocity of climate change is extremely high in tropical richspots. Species have already responded to these changes by shifting their latitudinal distributions polewards (Lenoir et al., 2020), which has already led to the loss of thousands of marine species from equatorial latitudes and increases in species richness in the subtropics (Chaudhary et al., 2021).
The high climate velocities in marine richspots seem to contradict the previously suggested relationship between climate and 190 endemism based on long-term climate change velocity (Sandel et al., 2011). In comparison to terrestrial and freshwater areas, the distribution of biodiversity in the ocean is more influenced by environment conditions than geographic isolation; on land old islands and fragmented landscapes have led to higher terrestrial than marine endemism . Also, climate change today is happening on much shorter time scales than what may have influenced the evolutionary origin of richspots and the distribution of endemics. 195 It is also possible that the definition of older richspot schemes is not representative of true biodiversity. The G200 richspots was partly driven by political priorities ("make every nation a stakeholder", Olson and Dinerstein, 2002), and the Myers et al. (2000) richspots were also prioritized based on threat from other human impacts in addition to their rich biodiversity. The systematically lower difference between warming inside and outside the RBA of Zhao et al. (2020) compared to the marine G200 might suggest that the former grasps patterns of endemism and species distributions better than the latter (Fig. 1). 200 The present study did not consider annual variation and additional climatic variables (Fick and Hijmans, 2017) that might influence the distribution of species. Small-scale climate refugia might exist within the individual richspots which are not detected due to the spatial scale of our analyses. Where there is high heterogeneity of climate change velocities (e.g., due to topographic variation) at a spatial resolution finer that the used in our analysis, species may find thermal refugia within https://doi.org/10.5194/bg-2021-179 Preprint. Discussion started: 12 July 2021 c Author(s) 2021. CC BY 4.0 License. 8 terrestrial and freshwater richspots. Thus, projections as used here need to be validated by in situ monitoring of changes in 205 species abundance.

Conclusions
Our findings support the hypothesis that most terrestrial richspots have been climate refugia in a relative sense, but they do not relax concerns regarding the effects of global warming on endemic life. While thousands of species are shifting their geographic ranges rapidly in response to a warming climate, there is a high risk that endemic species will not be able to disperse 210 to more suitable climates and go extinct (Manes et al., 2021). Climate mitigation is thus essential to keeping climate warming to less than 2 o C because this will reduce extinction risk in all richspots (Manes et al., 2021).
Assessment of the impact of climate change on biodiversity richspots is compounded by human-induced losses of species and habitats across all environments. As stated repeatedly in the scientific literature for decades, strict protection of biodiversity from local human impacts within richspots is a most area-effective way to minimize species extinctions (e.g. Mittermeier et 215 al., 2011;Darwall et al., 2018;Zhao et al., 2020). In addition, environmentally sustainable practices inside and outside richspots must facilitate species dispersal between habitats as climate change occurs.

Code and data availability
Past climate data are openly available from the website of the MetOffice Hadley Centre (https://www.metoffice.gov.uk/hadobs/hadisst/) and Climatic Research Unit (University of East Anglia, 220 https://crudata.uea.ac.uk/cru/data/hrg/). Results of the CMIP6 climate data will be made publicly available from the IPCC  (Kocsis et al., 2021).

Author contributions
This paper arose from discussions within IPCC WGII Cross Cutting Chapter on Biodiversity Hotspots. The study was designed by MJC and WK. ÁTK and MJC drafted the first versions of the manuscript. ÁTK carried out all the analyses and finalized the manuscript. All authors contributed to the discussions and revisions of the paper.       Table 1 for the tabulation of refugia and critically warming richspots.