Ecology

BioassaysSalmon-louse bioassays were performed by the BC Centre for Aquatic Health Sciences (CAHS) as described in Saksida et al.10. Briefly, motile (i.e., pre-adult and adult) L. salmonis were collected from 11 salmon farms in the Broughton Archipelago (BA) between 2010 and 2021 and transported to CAHS in Campbell River, BC. Within 18 h of collection, healthy lice were separated by sex and randomly placed into petri dishes each containing approximately 10 lice (mean ± SD = 9.6 ± 1.1) and subjected to one of six EMB concentrations (either 0, 31.3, 62.5, 125, 250, and 500 ppb or 0, 62.5, 125, 250, 500, and 1000 ppb, depending on suspected variation in EMB sensitivity11). Each collection corresponded to one bioassay, and each bioassay contained roughly four replicates for each sex (4.0 ± 1.3 for females and 3.6 ± 0.9 for males). After 24 h of EMB exposure, lice were classified as alive if they could swim and attach to the petri dish, or moribund/dead otherwise. Lice were kept at 10 °C throughout the process. In total, 34 bioassays were conducted from 11 farms between October 2010 and November 2021.We analysed the proportion of lice that survived exposure to EMB, using standard statistical descriptions that accounted for within-assay dependencies (generalized linear mixed models (GLMMs) with logit link functions, fitted separately to the data from each bioassay). The models included fixed effects for EMB concentration, sex, and the interaction between the two, as well as a random intercept for petri dish. For each analysis, we centered concentration values and scaled them by one standard deviation. We used the GLMM fits to calculate the effective concentrations at which 50% of the lice survived (EC50) in each bioassay. The GLMM for one bioassay produced a singular fit because there was not enough variation in the female survival data to warrant the random-effects structure. We retained the EC50 values resulting from this singular fit because re-fitting without the random intercept yielded identical EC50 values, and removing the entire bioassay from the overall dataset did not qualitatively affect the subsequent analysis.To assess whether the sensitivity of salmon lice to EMB has decreased over time, we fitted a set of five standard GLMs with gamma error distributions and log link functions to the maximum-likelihood EC50 estimates. Each of these five models included binary effects for sex and for whether the farm’s stock had previously been treated, since both affect EMB sensitivity in lice10. The first model included only these two effects and served as a null model that assumed lice did not evolve EMB resistance over time. The second model added a fixed effect for time (i.e., the number of days since January 1, 2010), while the third model included an interaction between time and sex. The fourth and fifth models were identical to the second and third, but with a quadratic effect for time, to account for possible first-order nonlinearity. We were unable to add an effect for farm due to small sample sizes. We performed model selection using the Akaike Information Criterion penalized for small sample sizes AICc25, treating AICc differences of less than two as being indistinguishable in terms of statistical support and selecting the least complex model when that was the case26. The ΔAICc values for the EC50 models were 48.1, 6.1, 4.9, 0, 1.75, respectively.Field efficacyWe used relative salmon-louse counts after EMB treatment (i.e., the post-treatment count divided by the pre-treatment count) as our measure of EMB field resistance between 2010 and 2021 (higher relative counts imply lower treatment efficacy). We defined “pre-treatment” as one month prior to treatment and “post-treatment” as three months after treatment (roughly when one would expect to find the lowest counts in louse populations previously unexposed to EMB), as in Saksida et al.10. We excluded EMB treatments for which an additional, non-EMB treatment was performed within the following three months. In total, there were 73 EMB treatments for which we were able to calculate relative post-treatment counts.Salmon-louse counts were performed by farm staff as described by Godwin et al.27. In short, salmon-louse counts were usually performed at least one per month by capturing 20 stocked fish in each of three net pens using a box seine net, then placing the fish in an anesthetic bath of tricaine methanesulfonate (TMS, or MS-222) and assessing the fish for motile (i.e., pre-adult and adult) L. salmonis by eye.The treatment dataset included the date and type of every treatment that has been performed on a BA farm (i.e., not just the 11 farms with bioassay data). In total, 88 EMB treatments were conducted between 2010 and 2021, of which we were able to calculate relative post-treatment counts for 73 because some months lacked counts or had a non-EMB treatment performed within the following three months. An additional 22 non-EMB treatments (e.g., freshwater and hydrogen baths) were performed, all since the beginning of 2019, but we excluded these data from our analysis.To determine whether field efficacy of EMB treatments has decreased over time, we used GLM-based “hurdle models”—standard statistical descriptions used to accommodate an over-abundance of zeroes in data being analysed. A hurdle model uses two components—one model for whether a count is nonzero and another for the value of the nonzero count—to predict overall mean count. To this end, we fitted three binomial GLMs paired with three gamma GLMs to the relative-count data, each of the paired models being structurally identical in terms of predictors. All of these submodels included a binary fixed effect for previous treatment, as in the EC50 models. The null pair of submodels included no additional terms, the second pair of submodels included a fixed effect for time (i.e., the number of days since January 1, 2010), and the third pair of submodels included a quadratic effect of time (again, to account for possible first-order deviations nonlinearity). We were unable to add an effect for farm due to small sample sizes. We performed model selection of the hurdle models, again using the Akaike Information Criterion penalized for small sample sizes. The ΔAICc values for the three hurdle models were 39.6, 18.3, and 0, respectively. We performed our analyses in R 3.6.028, using the lme4 package29. More

Base run simulationFigure 6 shows the results of the base run simulation. In this scenario, strong vegetation declines over time, while the empty area and flammable vegetation have increasing trends. As such, more fuel would be available for burning, and the wildfire can burn broader areas. Panel (a) shows an oscillatory trend for the burn rate with an average upward trend (To make sure the oscillatory behavior of the model does not fade, Appendix 4 shows the simulation result for 100 years). The observed pattern in the burn rate can be traced back to the patterns of human ignition (Panel b), and the growing trend of vulnerable properties (Panel c). In addition, the results show the long-term declining trend of strong vegetation in our base line simulation (Panel d); over time, stronger vegetation is replaced by flammable vegetation which can lead to more fire. This change in vegetation composition effectively increases the average burn rate. Over time, with more flammable vegetation and with the expansion of vulnerable properties, the likelihood of human-made ignition increases.Figure 6Base run simulation for a 20-year run of the model.Full size imageCoupling effectsFigure 7 shows how the relation between perceived fire risk and the burn rate influences the system. The black line is the base run simulation for comparison. The blue dashed line depicts the condition in which risk perception changes extremely slowly, and the human system is almost disconnected from the natural system. In this situation, if humans underestimate the fire potential, the system burns down nature, resulting in a catastrophic environmental outcome as depicted in panel (a). Panel (a) shows that the burn rate overshoots in the short term but relatively declines due to less remaining natural resources to burn.Figure 7Coupling effect analysis for 20 years. Human ignition unit is Ignition/year, and vulnerable property unit is a million hectares. Strong vegetation and flammable vegetation are provided as the ratio that each occupied the forest area.Full size imagePanel (b) displays the total burn rate throughout the study time to cast further insight into the burn rate sensitivity to perceived risk. The overall burn rate does not significantly change when the risk perception changes from 0.5 to 2, indicating the difference among burn rates in panel (a) is more about the fluctuation timing, but not the size. However, an additional rise in the sense of risk greatly raises the overall burn rate, as seen in panel (a).In the case of prolonged change in risk perception, human ignition continues to increase (panel c) as the perceived risk changes slowly. Furthermore, vulnerable properties are being built faster than their demolition (panel d). A slighter delay in perception leads to a higher frequency of oscillation as depicted in the graphs by the red dashed lines and a longer delay in a lower frequency oscillation, as shown by the purple graphs. Overall, the results are not much different from the base run. We are losing forests (panel e) and have periodic burn rates of increasing magnitude over time.Policy experimentsHere we examine the impact of implementing four proposed policies introduced in Table 2. To prevent the initial condition and transition periods affecting our comparison of proposed policies, we imposed each policy at the fifth year and compared the total burn rates between 10 and 20 years. Figure 8 shows the effect of these policies on different variables. Figure 8Policy implementation. Note: P1: limits vulnerable property development; P2: prescribed burning; P3: effective firefighting; and P4: Clear cutting. Human ignition unit is Ignition/year, and vulnerable property unit is a million hectares. Strong vegetation and flammable vegetation are provided as the ratio that each occupied the forest area.Full size imagePanels (a) and (b) show the burn rate over time and cumulative, respectively. All four policies reduce the burn-rate magnitude compared to the base run. P3 is more effective in early burning-rate reduction compared to other policies, but they ultimately result in similar behavior. It is worth noticing that P1 has the most effect on long-run fluctuation reduction, although its total effect in the time span is less than P3. It seems that firefighting is more effective in the short run, but it fails to dampen the fluctuation and instead limits its growth. This is partly because of the increase in human ignition and settlement due to the success of firefighting in the short run. As a result, people perceive less fire danger and continue to engage in high-risk activities and expand housing in the WUI. The result is further fluctuation in the burn rate even when P3 is implemented. On the other hand, the WUI expansion limitation policy can effectively reduce the burn-rate fluctuation in a timely manner. Implementing P4 causes a reduction in strong vegetation, which leads to flammable vegetation increase. As flammable vegetation is the main fuel for wildfire, this policy cause increase in fuel availability and an increase in the burning rate.Change in human ignition is provided in panel (c). Different levels of human-made ignition are observable, and the reason is that people adjust their high-risk behavior with burn rate, and not with the number of fires. In the firefighting policy, as for a given level of ignition, the burn rate declines, we observe more risky behavior and more human-made ignition. It is interesting to note that, as panel (c) shows, we end up with more WUI under policies 2, 3, and 4. In fact, the reason is that the firefighting, prescribed burning and clear cutting only affect natural sector of the model, decrease burn rate, which decreases risk perception and in turn result in more WUI development. On the other hand, P1 directly targets WUIs.Panel (e) displays the change in strong vegetation, which shows that P4 causes the most reduction in forest tree cover as it directly removes strong vegetation. P2 also causes a decrease in strong vegetation compared to the base run. The reason is that burning flammable vegetation damages young trees and prevents them from developing into solid vegetation. On the other hand, P3 has the least effect on strong vegetation by slowing the damage to young trees and confining the fire. Panel (f) shows the flammable vegetation dynamic after imposing each policy. P3 and P2 reduce flammable vegetation more than P1. However, there is an important difference in how these policies cause the reduction in flammable vegetation. In comparing panels (a) and (b), we see that while P3 causes further increases in the strong vegetation, P2 causes an increase in the empty area. P4 is the only policy that increases flammable vegetation by removing the strong vegetation and providing an empty area to be filled with young vegetation.Overall, it looks like each policy has some marginal effect on containing wildfire, though the magnitudes of effect are not considerable.Replication of United States dataFor model validation, we investigate its ability to fit a single case, United States’ wildfires from 1996 to 2015. We utilize the United States Department of Agriculture’s wildfire database for the conterminous United States (Short, 2017). The results are shown in Fig. 9. In this figure, simulation of burning rate and human ignition (continuous lines, in black) closely follows the real-world data (dotted lines, in red), and the model fairly replicates the historical trends.Figure 9Burning rate and human ignition per unit of forest area. The black line represents the model result, and the red dotted line represents the historical wildfire activity in the conterminous United States.Full size imageCombination policy implementation analysisTo better understand the impacts of our policies, we run different pairs of policies simultaneously. The results illustrate the nonlinear incremental impacts between policies. Simply put, it appears that the impact of several policies is enforced when combined synergistically. In other words, applying several policies might have a greater overall impact than the sum of the policies’ individual effects and suggests that policymakers should avoid searching for a panacea and adopt a broad range of approaches thoughtfully.The results of multiple policy implementations along with single ones are presented in Fig. 10. For example, P1 and P2 each reduce the total burn rate by 4.9% and 4.5%, respectively. While the summation of these effects is 9.4%, simultaneously implementing P1 and P2 lead to a 13.6% burn-rate reduction—P1 controls the human ignition, and P2 reduces the flammable vegetation stock—together, the burn rate is more affected than if implemented separately. The case is more interesting when P1 and P3 are imposed together. The result is a 38% burn-rate reduction compared to 13.9%, which is the sum of solely implementing each policy. The synergic effect happens because P3 lets the flammable vegetation (mainly young trees) age and become strong vegetation. Furthermore, the P1 also prevents human ignition from growing as fast as a single P3 implementation.Figure 10The nonlinear effect of policies. The benefits of implementing multiple policies differ from the sum of the effect of policies. The figure shows the percent of burn rate reduction. Note: P1: limit vulnerable property development; P2: prescribed burning; P3: effective firefighting; and P4: Clear cutting.Full size imageAn interesting case happens when P2 and P3 are implemented together. The synergic effect is less than the sum of separate implementation, mainly because both policies affect the vegetation dynamic and not the human factor in the wildfire. P2 and P3 both cause a lower initial burn rate, but due to the reduction in perceived risk of wildfire and expansion of WUI, this effect quickly disappears. This is another evidence for the importance of considering the problem as an interconnected natural and human system, where effective policies should address both sides.Finally, an interesting result emerges when all policies impose together. Surprisingly, imposing all policies together does not have the most impact on the total burn rate (32.5%), which is less than the P1 and P3 effect (38.0%). The reason relates mainly to the fact P2 and P4 both cause increase in flammable vegetation after empty area filled, which lead to more burning rate after a delay.Sensitivity analysisWe conducted a series of sensitivity analysis to check the model’s robustness to our assumptions. Specifically, we conducted a Monte-Carlo analysis and changed several parameter values to determine the range of outcomes. The results are reported in Appendix 2. In summary, the focus was on parameters that can take on substantially different values from those assumed in the model, including parameters used for risk perception formulation, its effect on human behavior, such as time to perceive risk and time to change behavior, in addition to fractional burning rate per ignition, average s burning, initial flammable vegetation, initial strong vegetation, human ignition multiplier, and initial vulnerable property. As described in the Appendix, for most of these variables, we changed the corresponding variable up to double its base run value. Moreover, we test different values for initial strong vegetation and initial flammable vegetation changing them between zero and their base run values. Each sensitivity test is the outcome of 2000 simulation runs using a uniformly distributed random distribution of the parameters within the specified intervals. The results are qualitatively robust, and their variability is within reasonable limits (See Figure A1). More

Marine Ecology Research Group and Centre for Integrative Ecology, School of Biological Sciences, University of Canterbury, Christchurch, New ZealandMads S. Thomsen, Luca Mondardini, David R. Schiel & Alfonso SicilianoDepartment of Bioscience, Aarhus University, 4000, Roskilde, DenmarkMads S. ThomsenSmithsonian Tropical Research Institute, Apartado, 0843-03092, Balboa, Ancon, Republic of PanamaAndrew H. Altieri, Viktoria M. M. Frühling, Seamus B. Harrison & Gerhard ZotzEnvironmental Engineering Sciences, University of Florida, Gainesville, FL, USAAndrew H. Altieri & Christine AngeliniDepartment of Biological Sciences, Macquarie University, Sydney, NSW, AustraliaMelanie J. Bishop & Semonn OleksynDipartimento di Biologia, Università di Pisa, CoNISMa, Via Derna 1, 56126, Pisa, ItalyFabio Bulleri & Joachim LangeneckMarine Sciences, University of Georgia, Athens, GA, USARoxanne FarhanCentre for Marine Science and Innovation, School of Biological, Earth and Environmental Sciences, University of New South Wales, Sydney, NSW, AustraliaPaul E. Gribben & Brendan S. LanhamSydney Institute of Marine Science, Chowder Bay Road, Mosman, 2088, Sydney, NSW, AustraliaPaul E. Gribben & Brendan S. LanhamCoastal Ecology Lab, MOE Key Laboratory for Biodiversity Science and Ecological Engineering, School of Life Sciences, Fudan University, 2005 Songhu Road, 200438, Shanghai, ChinaQiang HeInstitute for Biology and Environmental Sciences, Carl von Ossietzky University Oldenburg, Oldenburg, GermanyMoritz Klinghardt, Tristan Schneider & Gerhard ZotzSchool of Biological Sciences and UWA Oceans Institute, University of Western Australia, Perth, WA, AustraliaYannick Mulders & Thomas WernbergDepartment of Biology and Marine Biology, University of North Carolina Wilmington, Wilmington, NC, USAAaron P. RamusNicholas School of the Environment, Duke University, 135 Duke Marine Lab Road, Beaufort, NC, USABrian R. Silliman & Stacy ZhangMarine Biological Association of the United Kingdom, The Laboratory, Citadel Hill, Plymouth, PL1 2PB, UKDan A. SmaleCawthron Institute, Nelson, New ZealandPaul M. South More

Erebia pronoe exhibits highly structured and strongly differentiated mitochondrial lineages, which are consistent with the distribution of previously described morphotaxa and analyses of Dincă et al.10 These genetic lineages are also reflected to varying degrees in the nuclear markers. The observed mito-nuclear discordances can be explained by different evolutionary rates of genetic markers, the effects of Wolbachia infections, and introgression. These aspects are discussed in more detail in the following sections on the phylogeographic history of this species complex.Mito-nuclear discordance and the systematic status of Erebia melas
Based on genital morphology and nuclear markers, E. melas represents a distinct group to E. pronoe. The common area of origin of both species was probably located in the eastern Alps, which is supported by a RASP analysis based on the nuclear markers. However, E. melas acts as an ingroup of E. pronoe based on the mitochondrial markers, and a RASP analysis indicates a common origin for both taxa in the Carpathian region. Since most Erebia species in Europe have at least parts of their distribution in the Alps21 and are adapted to Alpine environments and habitats22,23, we consider an eastern Alpine origin of the ancestor of E. pronoe and E. melas more likely. This hypothesis subsumes the assumption that the genetic proximity on the mitochondrial level was probably caused by hybridisation and introgression events, which could have occurred as a result of several eastward advances of E. pronoe to the Balkan Peninsula (see below). This seems plausible, because the ability and tendency of E. pronoe to hybridise with other Erebia species have been demonstrated repeatedly12,24,25.The existence of Wolbachia strain 2 in both species, and its distribution from the Pyrenees (in E. pronoe) to the Balkan Peninsula (in E. melas) also speaks for a common origin of both species. Thus, Wolbachia strain 2 might represent the ancient strain present in the common ancestor of this species group, surviving today at the geographic margins (i.e. Pyrenees, western Alps, Balkan Peninsula), but which at some time was replaced in the centre of the butterfly’s range (i.e. the eastern and central Alps) by strain 1. The link between co-occurrence in a common area and prevalence of one Wolbachia strain was also recently demonstrated in other Erebia species26 and might facilitate mitochondrial introgression27.Intraspecific differentiation and glacial refugia of Erebia pronoe
The Pyrenean region is inhabited by one of the oldest and most differentiated intraspecific lineages of E. pronoe. The high genetic diversity in the Pyrenees speaks for large effective population sizes throughout time, enabled by mostly altitudinal shifts in response to climatic cycles, and a lack of major genetic bottlenecks. Compared to the Pyrenean group, the genetic diversity of the western Alpine populations, also well differentiated from all other groups, is lower. This lower diversity was probably the result of repeated cold stage retreat to a geographically more restricted refugium at the foot of the south-western Alps, a well-known refugial area for numerous species28.We cannot say conclusively whether the populations in the Pyrenean region or in the western Alps differentiated first, due to the contradictory genetical markers. The higher evolutionary rate of the mitochondrial markers, the allopatric distribution, and the hybridisation with diverse Erebia species may have led to a greater differentiation of the Pyrenees and/or a loss of the genetic link between the western Alps and the Pyrenees. Since a link between the western Alps and the Pyrenees is still well reflected in the nuclear data set and by the shared Wolbachia strain 2, we consider the most likely scenario to be an early Pleistocene or even Pliocene expansion from the western Alps to the Pyrenees, with subsequent isolation and differentiation. Thus, the Pyrenees-western Alps populations might first have separated as one group from an eastern Alps group s.l., as suggested by nuclear information, and not in two independent events, as suggested by mitochondrial genes.Simultaneously to the split between western Alps and Pyrenees, a separation of the eastern Alpine group s.l. into a southern Alpine subgroup and an eastern Alpine subgroup should have occurred. The southern Alpine subgroup displays a high genetic diversity in their nuclear markers, but a significantly lower diversity in the mtDNA. This might be explained by the existence of a cold-stage refugial area in the southern Alps or their margin, supporting the constant survival of large populations, but also a reshaping of the mtDNA patterns through introgression from the eastern Alpine subgroup during secondary contact when both subgroups expanded into formerly glaciated east-central Alpine areas. The isolated occurrence of Wolbachia strain 1 and mitochondrial haplotypes H29 and H30 (shared with the eastern Alps subgroup) in the southern Alps further support the hypothesis of gene flow from the eastern Alpine region into the southern Alpine populations and vice versa.The eastern Alpine subgroup probably survived glacial periods in a large, cohesive refugium at the eastern edge of the Alps, as has been demonstrated for numerous other species28. This area is also seen as a potential centre of origin of the entire taxon. From there, a recent (most likely postglacial) dispersal must have taken place, which should be responsible at least partly for the star-like pattern of this group in both mitochondrial and nuclear haplotype networks. However, further dispersal events out of the eastern Alps during previous interglacials and maybe even going back to the Pliocene have to be postulated to explain the entire range dynamics in E. pronoe.Apparently, multiple advances out of the eastern Alps into the Balkan mountain systems have taken place from several independent glacial refugia in the region, as indicated by the different mtDNA lineages in Slovenia, western Balkan mountains, and eastern Balkan mountains. A separation between the eastern and western Balkans, and hence also separate glacial refugia in both areas, was frequently observed for mountain taxa28,31. This pattern may have resulted from a succession of independent dispersal events from the eastern Alps throughout the younger Pleistocene, with subsequent regional extinction events and/or independent dispersal events across the Carpathians, as has been demonstrated for numerous other species29.A similar pattern of two independent colonisation events also applies to the Carpathians. Thus, the highly isolated populations in the south-eastern Carpathians must go back to an older expansion out of the eastern Alps. This probably took place during one of the last interglacial phases. The route most likely followed the Carpathian arc, but only a few populations survived at their south-eastern edge. This underlines the phylogeographic independence of this part of the Romanian Carpathians, which is also supported by studies on numerous other mountain species30,31,32. On the other hand, the Tatra mountains, as the northernmost part of the Carpathians, were colonised very recently, most likely postglacially, out of the eastern Alpine area. The strong and rather recent link between these two areas is also supported by phylogeographic studies on many taxa30,33,34.Because of the slower evolutionary rate of nuclear DNA and the resulting incomplete lineage sorting, nuclear markers can contribute little to the reconstruction of these presumably recent events. In line with that, the Valais lineage also has little nuclear differentiation but is clearly distinguished from the western and eastern Alpine lineages by the exclusive mtDNA haplotype H17 and Wolbachia strain 3. The presence of a single, highly differentiated mtDNA haplotype and an exclusive Wolbachia strain indicates a selective sweep. This lineage most likely represents a chronological relict of an interglacial expansion of the eastern Alpine subgroup to the western-central Alps surviving since then in this area, finding glacial refugia in nearby unglaciated areas and becoming infested by a Wolbachia strain not present in any other E. pronoe lineage, hence accelerating its differentiation.Another selective sweep was probably the cause of the mito-nuclear unconformity in the southern Alps lineage. The occurrence of the mtDNA haplotypes H29 and H30 and the Wolbachia strain 1 indicate mitochondrial hybridisation between the eastern and southern Alpine lineages during an expansive interglacial phase. As a result, Wolbachia infection probably occurred, which might have impoverished the mitochondrial diversity of the southern Alps lineage.Consequences for subspecific differentiation in Erebia pronoe
In general, the support given by our data for the so-far described subspecies decreases from west to east. Erebia pronoe glottis Fruhstorfer, 1920, distributed in the Pyrenees, represents the best-supported subspecies. Fixed mitochondrial amino acid changes emphasize the distinctness of this taxon, which might be well advanced in the process of speciation; we cannot even exclude the possibility that it has already reached full species rank. The genetic separation of the western Alps from the Valais, geographically separated along the main Alpine ridge, justifies the recognition of the taxa E. pronoe vergy (Ochsenheimer, 1807) and E. pronoe psathura Fruhstorfer, 1920, respectively, and is supported by both marker sets as well as by the existence of two different Wolbachia strains. The eastern Alpine subgroup resembles the nominotypical E. pronoe pronoe. The existence of at least one lineage in the southern Alpine area is supported by both marker sets. A finer separation based on the mitochondrial markers is not possible, because of recent introgression events affecting east Alpine haplotypes, as also indicated by the existence of Wolbachia strain 1. This population group could be assigned to the taxon E. pronoe gardeina Schawerda, 1924, or to E. pronoe tarcenta Fruhstorfer, 1920, considering their ranges. Nevertheless, a final decision requires further regional studies. Erebia pronoe fruhstorferi Warren, 1933 was accepted to be widely distributed in the Balkan mountain systems. However, our data suggest independent lineages in the western and eastern Balkan mountain systems of which only the eastern populations can be assigned to this taxon. The lineage of the Slovenian Alps is primarily based on mitochondrial markers and morphological characteristics7. The existence of an independent lineage for the highly isolated populations in the southern Carpathians, justifies the subspecies status of E. pronoe regalis Hormuzachi, 1937. Both marker sets display a differentiation, which was more pronounced in the nuclear than in the mitochondrial DNA. More

Spatial and temporal distribution of ecological vulnerabilityBased on the SPCA model, the temporal and spatial distribution of ecological vulnerability in Shennongjia is obtained, as shown in Fig. 3. From 1996 to 2018, the area of micro vulnerability areas continued to increase and occupied a dominant position. Moreover, their distribution pattern tended to be gradually integrated, indicating that the structure and function of the ecosystem in most areas of Shennongjia were relatively complete, and in a healthy and stable state. However, the ecological environment of the severely vulnerable areas in the northeast, south and southwest of Shennongjia is in a trend of continuous deterioration, and the risk of extreme vulnerability is gradually emerging. From the spatial distribution of ecological vulnerability in 2018, it can be seen that the extremely vulnerable areas have increased significantly, and exhibit a dense and continuous distribution trend in some areas, accompanied by the development of rapid urbanization and highway traffic construction. There are also high-risk ecological vulnerable zones and the extremely vulnerability areas.Figure 3Spatial and temporal distribution of ecological vulnerability in Shennongjia. Spatial and temporal distribution of ecological vulnerability for (a) 1996, (b) 2007, (c) 2018 in Shennongjia, China.Full size imageIt can be seen from the area proportion of different levels of vulnerable areas (Fig. 4) that the area proportion of micro and extremely vulnerable areas increased significantly. Specifically, the area proportion of micro vulnerable areas increased from 59.98% in 1996 to 71.02% in 2018, while the area proportion of extremely vulnerable areas increased from 1.23% in 1996 to 7.32% in 2018. This shows that the ecological vulnerability of Shennongjia exhibits a significant two-level differentiation trend.Figure 4Proportion of the area of vulnerable districts at all levels in Shennongjia.Full size imageDynamic change of ecological vulnerabilityDuring the study period, the areas with a positive fitting slope account for more than 90% of the total area of the study area, which indicates that the overall vulnerability of Shennongjia presents a downward trend. According to the natural discontinuity point method, the dynamic change results of ecological vulnerability in Shennongjia are divided into five levels (Fig. 5), in order to discern the spatial angle more intuitively and clearly. It can be seen that the ecological vulnerability of most regions exhibits a decreasing trend, while the ecological vulnerability of certain regions increases.Figure 5Dynamic changes of ecological vulnerability in Shennongjia. Changes in the ecological vulnerability of Shennongjia in different periods: (a) 1996–2007, (b) 2007–2018, (c) 1996–2018.Full size imageFrom 1996 to 2007, whether the spatial distribution trend of ecological vulnerability increased or decreased is not obvious. However, from 2007 to 2018, the areas with significantly increased ecological vulnerability were concentrated in Yangri and Songbai in the northeast and near the Hongping airport in Shennongjia in the midwest. During this same time period, in the areas around the main urban areas and along the roads that were seriously disturbed by human activities, ecological vulnerability also exhibited a decreasing trend.Change trend of comprehensive ecological vulnerability indexAnnual change of the comprehensive ecological vulnerability indexThe results of the comprehensive ecological vulnerability index of 1996, 2007, and 2018 are 2.77, 2.71, and 2.51, respectively. From the annual change of the ecological vulnerability index in Shennongjia (Fig. 6), it can be seen that the ecological vulnerability of Shennongjia showed a downward trend from 1996 to 2018, and the stability and health of the ecosystem were improved overall.Figure 6Annual change of the comprehensive ecological vulnerability index. CEVI, comprehensive ecological vulnerability index.Full size imageAmong them, the decline of ecological vulnerability is relatively small from 1996 to 2007, which may be ascribed to the preliminary implementation of restrictive policies, such as banning logging and returning farmland to forest, which reduced ecological exposure factors, such as illegal logging and deforestation. From 2007 to 2018, the comprehensive index of ecological vulnerability in Shennongjia decreased significantly, which is mainly due to the designation of national nature reserves and the implementation of various ecological protection projects36. While reducing the exposed ecological disturbance, it simultaneously markedly improved the adaptability of the ecosystem, and further reduced the overall ecological vulnerability of the region.Changes of the comprehensive ecological vulnerability Index in different townsAccording to the comprehensive index of ecological vulnerability of eight towns in the Shennongjia (Table 5, Fig. 7), the ecological vulnerability difference of each town is obvious. In 2018, the comprehensive index of ecological vulnerability of each town is lower than that in 1996 and 2007. The results show that the average value of CEVI is, from high to low, Yangri, Xiaguping, Songbai, Xinhua, Jiuhu, Hongping, Muyu, and Songluo. The maximum value of the CEVI appeared in Yangri in 1996, and the minimum value occurred in Songluo in 2018.Table 5 Comprehensive ecological vulnerability index of towns.Full size tableFigure 7Radar chart of the comprehensive ecological vulnerability index of towns.Full size imageDriving factors of spatial and temporal evolution of ecological vulnerabilityThe formation and evolution of ecological vulnerability in Shennongjia constitutes a dynamic process, which is the result of interactions of human and natural factors. Based on the principle of SPCA of ecological vulnerability, the transformed principal components are extracted, and the rotated factor load matrix is obtained to reflect the different effects of various factors on the evaluation results. Each principal component possesses a different ability to explain the original index factors, but it has similar rules in the first four principal components (Table 6). The cumulative contribution rate of the first four principal components in the three groups of data reached more than 80%, which can reflect the information of most factors, and thus it has good representativeness.Table 6 Principal component loading and score.Full size tableAmong the first principal component and the third principal component, the contribution of land-use type index (C9) is higher; in the second principal component, the contribution of population density (C1) is higher; among the fourth principal components, the contribution of vegetation coverage (C13) is higher. Moreover, the contribution of other factors in different years and main components is dissimilar.The influence of land-use type on ecological vulnerabilityWhether due to natural or human factors, the original properties of the ecosystem are altered by changing the surface cover. Therefore, land-use type is an important factor affecting regional ecological vulnerability. The difference of surface cover leads to the difference of ecological community, and then produces varied ecological environmental benefits. Forest land is the most important land-use type in the study area, and the ecological vulnerability of the distribution area is mainly micro degree and light. However, consider the important ecological value of the forest ecosystem, attention should be given to its vulnerability. The ecological vulnerability of the construction land is mainly severe and extreme, which is largely due to the expansion of construction land, which destroys the original ecological structure and ecological community. Furthermore, a large number of manmade patches replace natural patches in the construction land, and biodiversity decreases, leading to the decline of the stability of ecological structures and the increase of vulnerability.The influence of population density on ecological vulnerabilityPopulation density is one of the most direct exposure factors in the vulnerability of ecological environments. Population density is generally higher than that in high area, and it is also a region with a developed economy and high urbanization. In these areas, human activities are frequent, which usually impart a negative disturbance to the natural environment, including the rapid expansion of cultivated land and construction land area, as well as high discharge of production and domestic wastewater waste, which has caused great pressure on the ecological environment, leading to a significant increase in ecological vulnerability.The influence of vegetation cover on ecological vulnerabilityFrom 1996 to 2018, the vegetation coverage of the Shennongjia exhibited an overall upward trend, which is of positive significance to the reduction of the vulnerability of the ecosystem. Vegetation, as the main body of the land ecosystem, maintains the balance of ecological environment through interactions with climate, landform, and soil37. Extant literature shows that the change of vegetation coverage is an major factor of regional ecological environment change, and has a clear indication function for the change of regional ecological environment38. The spatial distribution trend of ecological vulnerability in the Shennongjia is markedly similar to that of vegetation coverage. The ecological vulnerability of regions with higher vegetation coverage is lower, exhibiting a significant negative correlation. In the Shennongjia, the change of vegetation coverage is also obviously influenced by human factors.Contribution of landscape pattern index to ecological vulnerabilityThe spatial distribution of each index in Shennongjia have been obtained from previous studies47. From the unary linear regression analysis, in the years of 1996, 2007 and 2018, the NP, LPI, AI, DIVISION and SHDI are all significantly correlated with the ecological vulnerability index (Fig. 8).Figure 8Scatter plot of linear regression of landscape pattern index and ecological vulnerability index. EVI, ecological vulnerability index.Full size imageIn the case of different independent variable combinations in 1996, 2007 and 2018, the multiple regression relationship between the independent variable and the dependent variable of each group is significantly correlated, and the multiple linear regression equation of the full model is obtained as follows:$$1996{:};;{text{ Y}} = 6.443 + 0.014{text{X}}_{1} + 0.006{text{X}}_{2} – 0.038{text{X}}_{3} – 0.066{text{X}}_{4} + 0.058{text{X}}_{5}$$$$2007{:};;{text{ Y}} = 4.497 + 0.016{text{X}}_{1} + 0.007{text{X}}_{2} + 0.793{text{X}}_{3} – 0.047{text{X}}_{4} – 0.305{text{X}}_{5}$$$$2018{:};;{text{ Y}} = – 1.980 + 0.037{text{X}}_{1} + 0.006{text{X}}_{2} + 0.703{text{X}}_{3} + 0.019{text{X}}_{4} – 0.123{text{X}}_{5}$$The contribution rate of landscape pattern index to ecological vulnerability in different years of 1996, 2007, and 2018 is shown in Table 7. The contribution of AI and NP to ecological vulnerability in 1996 was high; the contribution of NP and AI to ecological vulnerability was higher in 2007; and the NP in 2018 had the highest contribution to ecological vulnerability, reaching 95.77%.Table 7 Contribution of the landscape pattern index to the ecological vulnerability index.Full size tableBased on the analysis results from 1996 to 2018, the contribution of NP and AI to ecological vulnerability is relatively high. The main reason for this is that the forest coverage rate of Shennongjia is as high as 91%. Specifically, with the forest as the landscape matrix, the NP is small and the connectivity between patches is high, showing a trend of aggregation. The degree of landscape fragmentation is relatively low and decreases annually, and ecological vulnerability decreases with the decrease of the degree of landscape fragmentation, Therefore, the impact of NP and AI on ecological vulnerability is highly significant.The AI and ecological vulnerability index always exhibit a significant negative correlation in the study period. In the 1996 research results, the contribution of AI to ecological vulnerability is the most obvious. Combined with the spatial distribution of ecological vulnerability, it can be seen that most of the severe and extremely vulnerable areas are distributed in areas with low AI. Most of them are the distribution areas of artificial patches, such as rural living areas, airports, tourism centers, etc., which are obviously disturbed by human activities, resulting in low connectivity among various landscape types, which greatly reduces the aggregation degree of landscape and increases regional vulnerability.There is also a significant positive correlation between the NP and the ecological vulnerability index. This is especially the case in 2018, when the contribution of the NP to ecological vulnerability is as high as 95.77%, which is mainly attributable to the urbanization construction of Songbai town in Shennongjia. Combined with the land-use structure map, it can be seen that the number of construction land patches in the northeast region increased sharply. In this process, the renewal of patches aggravates the degree of landscape fragmentation and plays a key role in the aggravation of regional vulnerability risk.Although the impact of LPI, SHDI and DIVISION on ecological vulnerability always exists, the contribution is not very significant. Among them, SHDI contributed 10.38% in 2007, which was more sensitive to the unbalanced distribution of each patch type. In areas with high SHDI, landscape heterogeneity is high, the ecological pattern is unstable, and ecological vulnerability increases. More

Quicke, D. L. Parasitic Wasps (Chapman & Hall Ltd., 1997).
Google Scholar 
Godfray, H. C. J. Parasitoids: Behavioral and Evolutionary Ecology (Princeton University Press, 1994).
Google Scholar 
Mayhew, P. J. & van Alphen, J. J. M. Gregarious development in alysiine parasitoids evolved through a reduction in larval aggression. Anim. Behav. 58 , 131–141 (1999).Mayhew, P. J. & Hardy, I. C. W. Nonsiblicidal behavior and the evolution of clutch size in bethylid wasps. Am. Nat. 151, 409–424 (1998).ADS 
CAS 
PubMed 

Google Scholar 
Schmidt, J. M. & Smith, J. J. B. Correlations between body angles and substrate curvature in the parasitoid wasp Trichogramma minutum: A possible mechanism of host radius measurement. J. Exp. Biol. 125, 271–285 (1986).
Google Scholar 
Boivin, G. & Baaren, J. The role of larval aggression and mobility in the transition between solitary and gregarious development in parasitoid wasps. Ecol. Lett. 3, 469–474 (2000).
Google Scholar 
Rosenheim, J. A., Wilhoit, L. R. & Armer, C. A. Influence of intraguild predation among generalist insect predators on the suppression of an herbivore population. Oecologia 96, 439–449 (1993).ADS 
PubMed 

Google Scholar 
Mayhew, P. J. The evolution of gregariousness in parasitoid wasps. Proc. R. Soc. Lond. B Biol. 265, 383–389 (1998).
Google Scholar 
Harvey, P. H. & Partridge, L. Murderous mandibles and black holes in hymenopteran wasps. Nature 326, 128–129 (1987).ADS 

Google Scholar 
Pexton, J. J. & Mayhew, P. J. Competitive interactions between parasitoid larvae and the evolution of gregarious development. Oecologia 141, 179–190 (2004).ADS 
PubMed 

Google Scholar 
Pexton, J. J. & Mayhew, P. J. Immobility: The key to family harmony? Trends Ecol. Evol. 16, 7–9 (2001).CAS 
PubMed 

Google Scholar 
Godfray, H. C. J. The evolution of clutch size in parasitic wasps. Am. Nat. 129, 221–233 (1987).
Google Scholar 
Laing, J. E. & Corrigan, J. E. Intrinsic competition between the gregarious parasite, Cotesia glomeratus and the solitary parasite Cotesia rubecula (Hymenoptera: Braconidae) for their host Artogeia rapae (Lepidoptera: Pieridae). Entomophaga 32, 493–501 (1987).
Google Scholar 
Pexton, J. J. & Mayhew, P. J. Clutch size adjustment, information use and the evolution of gregarious development in parasitoid wasps. Behav. Ecol. Soc. 58, 99–110 (2005).
Google Scholar 
Reitz, S. R. & Adler, P. H. Fecundity and oviposition of Eucelatoria bryani, a gregarious parasitoid of Helicoverpa zea and Heliothis virescens. Entomol. Exp. Appl. 75, 175–181 (1995).
Google Scholar 
Wei, K., Tang, Y. L., Wang, X. Y., Cao, L. M. & Yang, Z. Q. The developmental strategies and related profitability of an idiobiont ectoparasitoid Sclerodermus pupariae vary with host size. Ecol. Entomol. 39, 101–108 (2014).
Google Scholar 
van Alphen, J. J. M. & Visser, M. E. Superparasitism as an adaptive strategy for insect parasitoids. Ann. Rev. Entomol. 35, 59–79 (1990).
Google Scholar 
Mayhew, P. J. & Glaizot, O. Integrating theory of clutch size and body size evolution for parasitoids. Oikos 92, 372–376 (2001).
Google Scholar 
Samková, A., Hadrava, J., Skuhrovec, J. & Janšta, P. Reproductive strategy as a major factor determining female body size and fertility of a gregarious parasitoid. J. Appl. Entomol. 143, 441–450 (2019).
Google Scholar 
Hardy, I. C. W., Griffiths, N. T. & Godfray, H. C. J. Clutch size in a parasitoid wasp: A manipulation experiment. J. Anim. Ecol. 61, 121–129 (1992).
Google Scholar 
Visser, M. E. The importance of being large: The relationship between size and fitness in females of the parasitoid Aphaereta minuta (Hymenoptera: Braconidae). J. Anim. Ecol. 63, 963–978 (1994).
Google Scholar 
Sagarra, L. A., Vincent, C. & Stewart, R. K. Body size as an indicator of parasitoid quality in male and female Anagyrus kamali (Hymenoptera: Encyrtidae). Bull. Entomol. Res. 91, 363–367 (2001).CAS 
PubMed 

Google Scholar 
Bezemer, T. M. & Mills, N. J. Clutch size decisions of a gregarious parasitoid under laboratory and field conditions. Anim. Behav. 66, 1119–1128 (2003).
Google Scholar 
Takagi, M. The reproductive strategy of the gregarious parasitoid, Pteromalus puparum (Hymenoptera: Pteromalidae). Oecologia 68, 1–6 (1985).ADS 
PubMed 

Google Scholar 
Jervis, M. A., Ferns, P. N. & Heimpel, G. E. Body size and the timing of egg production in parasitoid wasps: A comparative analysis. Funct. Ecol. 17, 375–383 (2003).
Google Scholar 
Waage, J. K. & Lane, J. A. The reproductive strategy of a parasitic wasp: II. Sex allocation and local mate competition in Trichogramma evanescens. J. Anim. Ecol. 53, 417–426 (1984).
Google Scholar 
Waage, J. K. & Ming, N. S. The reproductive strategy of a parasitic wasp: I. Optimal progeny and sex allocation in Trichogramma evanescens. J. Anim. Ecol. 53, 401–415 (1984).
Google Scholar 
Rabinovich, J. E., Jorda, M. T. & Bernstein, C. Local mate competition and precise sex ratios in Telenomus fariai (Hymenoptera: Scelionidae), a parasitoid of triatomine eggs. Behav. Ecol. Sociobiol. 48, 308–315 (2000).
Google Scholar 
Goubault, M., Mack, A. F. & Hardy, I. C. W. Encountering competitors reduces clutch size and increases offspring size in a parasitoid with female–female fighting. Proc. R. Soc. B Biol. 274, 2571–2577 (2007).
Google Scholar 
Duval, J. F., Brodeur, J., Doyon, J. & Boivin, G. Impact of superparasitism time intervals on progeny survival and fitness of an egg parasitoid. Ecol. Entomol. 43, 310–317 (2018).
Google Scholar 
Mesterton-Gibbons, M. & Hardy, I. C. W. The influence of contests on optimal clutch size: A game–theoretic model. Proc. R. Soc. Lond. B Biol. 271, 971–978 (2004).
Google Scholar 
Koppik, M., Thiel, A. & Hoffmeister, T. S. Adaptive decision making or differential mortality: What causes offspring emergence in a gregarious parasitoid? Entomol. Exp. Appl. 150, 208–216 (2014).
Google Scholar 
Heimpel, G. E. Host–parasitoid population dynamics. In Parasitoid population biology (eds Hochberg, M. E. & Ives, A. R.) 27–40 (Princeton, 2000).
Google Scholar 
Zaviezo, T. & Mills, M. Factors influencing the evolution of clutch size in a gregarious insect parasitoid. J. Anim. Ecol. 69, 1047–1057 (2000).
Google Scholar 
Kazmer, D. J. & Luck, R. F. Field tests of the size-fitness hypothesis in the egg parasitoid Trichogramma pretiosum. Ecology 76, 412–425 (1995).
Google Scholar 
Segoli, M. & Rosenheim, J. A. The effect of body size on oviposition success of a minute parasitoid in nature. Ecol. Entomol. 40, 483–485 (2015).
Google Scholar 
Gao, S. K., Wei, K., Tang, Z. L., Wang, X. Y. & Yang, Z. Q. Effect of parasitoid density on the timing of parasitism and development duration of progeny in Sclerodermus pupariae (Hymenoptera: Bethylidae). Biol. Control 97, 57–62 (2016).
Google Scholar 
Anderson, R. C. & Paschke, J. D. The biology and ecology of Anaphes flavipes (Hymenoptera: Mymaridae), an exotic egg parasite of the cereal leaf beetle. Ann. Entomol. Soc. Am. 61, 1–5 (1968).
Google Scholar 
Hoffman, G. D. & Rao, S. Oviposition site selection on oats: The effect of plant architecture, plant and leaf age, tissue toughness, and hardness on cereal leaf beetle, Oulema melanopus. Entomol. Exp. Appl. 141, 232–244 (2011).
Google Scholar 
Samková, A., Hadrava, J., Skuhrovec, J. & Janšta, P. Host population density and presence of predators as key factors influencing the number of gregarious parasitoid Anaphes flavipes offspring. Sci. Rep. UK 9, 1–7 (2019).ADS 

Google Scholar 
Hardy, I. C. W. Sex ratio and mating structure in the parasitoid Hymenoptera. Oikos 69, 3–20 (1994).
Google Scholar 
Godfray, H. C. J. Models for clutch size and sex ratio with sibling interaction. Theor. Popul. Biol. 30, 215–231 (1986).MATH 

Google Scholar 
Hardy, I. C. W. Non-binomial sex allocation and brood sex ratio variances in the parasitoid Hymenoptera. Oikos 65, 143–158 (1992).
Google Scholar 
Petersen, G. & Hardy, I. C. W. The importance of being larger: Parasitoid intruder–owner contests and their implications for clutch size. Anim. Behav. 51, 1363–1373 (1996).
Google Scholar 
Klomp, H. & Teerink, B. J. The significance of oviposition rates in the egg parasite, Trichogramma embryophagum Htg. Arch. Neerl. Zool. 17, 350–375 (1967).
Google Scholar 
May, R. M., Hassell, M. P., Anderson, R. M. & Tonkyn, D. W. Density dependence in host–parasitoid models. J. Anim. Ecol. 50, 855–865 (1981).MathSciNet 

Google Scholar 
Hoddle, M. S., Van Driesche, R. G., Elkinton, J. S. & Sanderson, J. P. Discovery and utilization of Bemisia argentifolii patches by Eretmocerus eremicus and Encarsia formosa (Beltsville strain) in greenhouses. Entomol. Exp. Appl. 87, 15–28 (1998).
Google Scholar 
Samková, A., Raška, J., Hadrava, J. & Skuhrovec, J. Scarcity of hosts for gregarious parasitoids indicates an increase of individual offspring fertility by reducing their own fertility. bioRxiv https://doi.org/10.1101/2021.03.05.434037 (2021).Article 

Google Scholar 
van Dijken, M. J. & Waage, J. K. Self and conspecific superparasitism by the egg parasitoid Trichogramma evanescens. Entomol. Exp. Appl. 43, 183–192 (1987).
Google Scholar 
van de Vijver, E. et al. Inter-and intrafield distribution of cereal leaf beetle species (Coleoptera: Chrysomelidae) in Belgian winter wheat. Environ. Entomol. 48, 276–283 (2019).PubMed 

Google Scholar 
Samková, A., Hadrava, J., Skuhrovec, J. & Janšta, P. Host specificity of the parasitic wasp Anaphes flavipes (Hymenoptera: Mymaridae) and a new defence in its hosts (Coleoptera: Chrysomelidae: Oulema spp.). Insects 11, 175 (2020).PubMed Central 

Google Scholar 
Bezděk, J. & Baselga, A. Revision of western Palaearctic species of the Oulema melanopus group, with description of two new species from Europe (Coleoptera: Chrysomelidae: Criocerinae). Acta Entomol. Mus. Nat. Pragae 55, 273–304 (2015).
Google Scholar 
Anderson, R. C. & Paschke, J. D. Additional observations on the biology of Anaphes flavipes (Hymenoptera: Mymaridae), with special reference to the effects of temperature and superparasitism on development. Ann. Entomol. Soc. Am. 62, 1316–1321 (1969).
Google Scholar 
R Core Team. A Language and Environment for Statistical Computing. R Foundation for Statistical Computing (R Core Team, 2020).
Google Scholar 
Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48 (2015). https://CRAN.R-project.org/package=lme4. More

Urban, M. C. Accelerating extinction risk from climate change. Science 348, 571–573 (2015).ADS 
CAS 

Google Scholar 
Goulson, D. The insect apocalypse, and why it matters. Curr. Biol. 29, R967–R971 (2019).CAS 
PubMed 

Google Scholar 
Wagner, D. L. Insect declines in the Anthropocene. Annu. Rev. Entomol. 65, 457–480 (2020).CAS 
PubMed 

Google Scholar 
Heikkinen, R. K. et al. Assessing the vulnerability of European butterflies to climate change using multiple criteria. Biodivers. Conserv. 19, 695–723 (2010).
Google Scholar 
Montgomery, G. A. et al. Is the insect apocalypse upon us? How to find out. Biol. Conserv. 241, 108327 (2020).
Google Scholar 
Hufnagel, L. & Kocsis, M. Impacts of climate change on Lepidoptera species and communities. Appl. Ecol. Environ. Res. 9, 43–72 (2011).
Google Scholar 
Geyle, H. M. et al. Butterflies on the brink: identifying the Australian butterflies (Lepidoptera) most at risk of extinction. Austral Entomol. 60, 98–110 (2021).
Google Scholar 
Merckx, T., Huertas, B., Basset, Y. & Thomas, J. A global perspective on conserving butterflies and moths and their habitats. Key Topics in Conservation Biology 2, 237–257 (2013).
Google Scholar 
New, T. R. Moths (Insecta: Lepidoptera) and conservation: background and perspective. J. Insect Conserv. 8, 79–94 (2004).
Google Scholar 
Wagner, D. L., Fox, R., Salcido, D. M. & Dyer, L. A. A window to the world of global insect declines: Moth biodiversity trends are complex and heterogeneous. Proc. Natl. Acad. Sci. USA 118, e2002549117 (2021).CAS 
PubMed 
PubMed Central 

Google Scholar 
Van Langevelde, F. et al. Declines in moth populations stress the need for conserving dark nights. Glob. Chang. Biol. 24, 925–932 (2018).ADS 
PubMed 

Google Scholar 
Green, K. et al. Australian Bogong moths Agrotis infusa (Lepidoptera: Noctuidae). 1951–2020: decline and crash. Austral Entomol. 60, 66–81 (2021).
Google Scholar 
Sánchez‐Bayo, F. & Wyckhuys, K. A. Further evidence for a global decline of the entomofauna. Austral Entomol. 60, 9–26 (2021).
Google Scholar 
Rönkä, K., Mappes, J., Kaila, L. & Wahlberg, N. Putting Parasemia in its phylogenetic place: a molecular analysis of the subtribe Arctiina (Lepidoptera). Syst. Entomol. 41, 844–853 (2016).
Google Scholar 
Witt, T. J., Speidel, W., Ronkay, G., Ronkay, L. & László, G. M. Subfamilia Arctiinae in Noctuidae Europaeae. Volume 13. Lymantriinae and Arctiinae including phylogeny and check list of the quadrifid Noctuoidea of Europe (eds. Witt, T. J. & Ronkay, L.) 81-216 (Entomological Press, 2011).Dowdy, N. J. et al. A deeper meaning for shallow‐level phylogenomic studies: nested anchored hybrid enrichment offers great promise for resolving the tiger moth tree of life (Lepidoptera: Erebidae: Arctiinae). Syst. Entomol. 45, 874–893 (2020).
Google Scholar 
Zahiri, R. et al. Molecular phylogenetics of Erebidae (Lepidoptera, Noctuoidea). Syst. Entomol. 37, 102–124 (2012).
Google Scholar 
Holloway, J. D. The Moths of Borneo 6: family Arctiidae, subfamilies: Syntominae, Euchromiinae, Arctiinae; Noctuidae misplaced in Arctiidae (Camptoma, Aganinae) (Southdene Sdn. Bhd., 1988).Černý, K. & Pinratana, A. Arctiidae. Moths of Thailand 6, 1–283 (2009).
Google Scholar 
Černý, K. A review of the subfamily Arctiinae (Lepidoptera: Arctiidae) from the Philippines. Entomofauna 32, 29–92 (2011).
Google Scholar 
Bucsek, K. Erebidae, Arctiinae (Lithosiini, Arctiini) of Malay Peninsula – Malaysia (Institut of Zoology SAS, 2012).Bolotov, I. N., Kondakov, A. V. & Spitsyn, V. M. A review of tiger moths (Lepidoptera: Erebidae: Arctiinae: Arctiini) from Flores Island, Lesser Sunda Archipelago, with description of a new species and new subspecies. Ecol. Montenegrina 16, 1–15 (2018).
Google Scholar 
Dubatolov, V. V. New genera and species of Arctiinae from the Afrotropical fauna (Lepidoptera: Arctiidae). Nachr. Entomol. Ver. Apollo 27, 139–152 (2006).
Google Scholar 
Ferro, V. G., Melo, A. S. & Diniz, I. R. Richness of tiger moths (Lepidoptera: Arctiidae) in the Brazilian Cerrado: how much do we know? Zoologia (Curitiba) 27, 725–731 (2010).
Google Scholar 
Schmidt, B. C. A new genus and two new species of arctiine tiger moth (Noctuidae, Arctiinae, Arctiini) from Costa Rica. Zookeys 9, 89–96 (2009).
Google Scholar 
Dubatolov, V. V. Tiger-moths of Eurasia (Lepidoptera, Arctiidae) (Nyctemerini by Rob de Vos and V. V. Dubatolov). Neue Ent. Nachr. 65, 1–106 (2010).
Google Scholar 
Fibiger, M. et al. Lymantriinae and Arctiinae, including phylogeny and check list of the quadrifid Noctuoidea of Europe. Noctuidae Europaeae 13, 1–448 (2011).
Google Scholar 
Koshkin, E. S. Moths (Lepidoptera, Macroheterocera, excluding Geometridae and Noctuidae s.l.) of the Bureinsky State Nature Reserve and adjacent territories (Khabarovsk Krai, Russia) [In Russian]. Amur. Zool. J. 12, 412–435 (2020).
Google Scholar 
Kullberg, J., Filippov, B. Y., Spitsyn, V. M., Zubrij, N. A. & Kozlov, M. V. Moths and butterflies (Insecta: Lepidoptera) of the Russian Arctic islands in the Barents Sea. Polar Biol. 42, 335–346 (2019).
Google Scholar 
Bolotov, I. N. et al. The distribution and biology of Pararctia subnebulosa (Dyar, 1899) (Lepidoptera: Erebidae: Arctiinae), the largest tiger moth species in the High Arctic. Polar Biol. 38, 905–911 (2015).
Google Scholar 
Bolotov, I. N. et al. New occurrences, morphology, and imaginal phenology of the rarest Arctic tiger moth Arctia tundrana (Erebidae: Arctiinae). Ecol. Montenegrina 39, 121–128 (2021).
Google Scholar 
Bolotov, I. N., Gofarov, M. Y., Kolosova, Y. S. & Frolov, A. A. Occurrence of Borearctia menetriesii (Eversmann, 1846) (Erebidae: Arctiinae) in Northern European Russia: a new locality in a disjunct species range. Nota Lepidopterol. 36, 65–75 (2013).
Google Scholar 
Dubatolov, V. V. Borearctia gen. n., a new genus for the tiger moth Callimorpha menetriesi (Ev.) (Lepidoptera, Arctiidae) [In Russian]. Entomol. Rev. 63, 157–161 (1984).
Google Scholar 
Hori, H. An unrecorded species of the Arctiidae [In Japanese]. Kontyu 1, 86 (1926).
Google Scholar 
Eversmann, E. Lepidoptera quaedam nova in Rossia observata. Bulletin de la Société Impériale des Naturalistes de Moscou 19, 83–88 (1846).
Google Scholar 
Koshkin, E. S. Life history of the rare boreal tiger moth Arctia menetriesii (Eversmann, 1846) (Lepidoptera, Erebidae, Arctiinae) in the Russian Far East. Nota Lepidopterol. 44, 141–151 (2021).
Google Scholar 
Krogerus, H. D. Vorkommen von Callimorpha menetriesi Ev. in Fennoskandien, nebst Beschriebungen der verschiedenen Entwicklungsstadien [In German]. Not. Entomol. 24, 79–86 (1944).
Google Scholar 
Saarenmaa, H. Conservation ecology of Borearctia menetriesii [online]. http://www.bormene.myspecies.info/en (2011-2021).Berlov, O. E. & Bolotov, I. N. Record of Borearctia menetriesii (Eversmann, 1846) (Lepidoptera, Erebidae, Arctiinae) larva on Aconitum rubicundum Fischer (Ranunculaceae) in Eastern Siberia. Nota Lepidopterol. 38, 23–27 (2015).
Google Scholar 
Staudinger, O. & Rebel, H. Catalog der Lepidopteren des palaearctischen Faunengebietes. Vol. 1. Th. Famil. Papilionidae-Hepialidae (R. Friedländer & Sohn, 1901).Filipiev, I. Lepidoptera [In Russian]. Russkoe Entomologicheskoe Obozrenie 16, 376–378 (1916).
Google Scholar 
Fabritius, G. R. Anmärkningsvärda fynd av fjärilar, bland dessa den för Europa nya Callimorpha menetriesii Ev. [In Finnish]. Meddeland. Soc. Fauna Fl. Fenn. 40, 47–49 (1914).
Google Scholar 
Carpelan, J. Callimorpha menetriesii Ev. återfunnen [In Finnish]. Meddeland. Soc. Fauna Fl. Fenn. 48, 108–109 (1921).
Google Scholar 
Kurentzov, A. I. Zoogeography of the Amur Region [In Russian] (Nauka Publisher, 1965).Dubatolov, V. V. Tiger moths (Lepidoptera, Arctiidae: Arctiinae) of South Siberian mountains (report 2) [In Russian] in Arthropods and Helminths, Fauna of Siberia Series (ed. Zolotarenko, G. S.) 139–169 (Nauka Publisher, 1990).Klitin, A. K. New record of the tiger moth Borearctia menetriesii on Sakhalin Island [In Russian]. Bulletin of Sakhalin Museum 16, 269–271 (2009).
Google Scholar 
Nupponen, K. & Fibiger, M. Additions to the checklist of Bombycoidea and Noctuoidea of the Volgo-Ural region. Part II. (Lepidoptera: Lasiocampidae, Erebidae, Nolidae, Noctuidae). Nota Lepidopterol. 35, 33–50 (2012).
Google Scholar 
Koshkin, E. S. Preliminary results of the examination of the fauna of Higher Moths (Macroheterocera, excluding Geometridae and Noctuidae) of the upper Bureya River basin (Khabarovsk Region) [In Russian]. Proceedings of Grodekovsky Museum (Nature of the Far East) 24, 65–75 (2010).
Google Scholar 
Marttila, O., Saarinen, K., Haahtela, T. & Pajari, M. Idänsiilikäs Borearctia menetriesi (Eversmann, 1846) [In Finnish] in Suomen kiitäjät ja kehrääjät [Macrolepidoptera of Finland] 265–266 (Kirjayhtymä Oy, 1996).Lappi, E., Mikkola, K. & Ryynänen, J. Idänsiilikäs Borearctia menetriesii, tervetuloa takaisin! [Welcome back Borearctia menetriesii] [In Finnish]. Baptria 29, 28–29 (2004).
Google Scholar 
Silvonen, K. Borearctia Dubatolov, 1985 [online]. Kimmo’s Lepidoptera Site, Finland. http://www.kolumbus.fi/~kr5298/lnel/a/bormenet.htm (2010).Bolotov, I. N. et al. Menetries’ Tiger Moth Range and Ecology Database (1840s-2020). figshare https://doi.org/10.6084/m9.figshare.15000399 (2022).Dirzo, R. et al. Defaunation in the Anthropocene. Science 345, 401–406 (2014).ADS 
CAS 
PubMed 

Google Scholar 
Young, H. S., McCauley, D. J., Galetti, M. & Dirzo, R. Patterns, causes, and consequences of anthropocene defaunation. Annu. Rev. Ecol. Evol. Syst. 47, 333–358 (2016).
Google Scholar 
Conrad, K. F., Warren, M. S., Fox, R., Parsons, M. S. & Woiwod, I. P. Rapid declines of common, widespread British moths provide evidence of an insect biodiversity crisis. Biol. Conserv. 132, 279–291 (2006).
Google Scholar 
Sánchez-Bayo, F. & Wyckhuys, K. A. G. Worldwide decline of the entomofauna: A review of its drivers. Biol. Conserv. 232, 8–27 (2019).
Google Scholar 
Simmons, B. I. et al. Worldwide insect declines: An important message, but interpret with caution. Ecol. Evol. 9, 3678–3680 (2019).PubMed 
PubMed Central 

Google Scholar 
Didham, R. K. et al. Interpreting insect declines: seven challenges and a way forward. Insect Conserv. Diver. 13, 103–114 (2020).
Google Scholar 
Boyes, D. H., Evans, D. M., Fox, R., Parsons, M. S. & Pocock, M. J. Is light pollution driving moth population declines? A review of causal mechanisms across the life cycle. Insect Conserv. Diver. 14, 167–187 (2021).
Google Scholar 
Raven, P. H. & Wagner, D. L. Agricultural intensification and climate change are rapidly decreasing insect biodiversity. Proc. Natl. Acad. Sci. USA 118, e2002548117 (2021).CAS 
PubMed 
PubMed Central 

Google Scholar 
Wagner, D. L., Grames, E. M., Forister, M. L., Berenbaum, M. R. & Stopak, D. Insect decline in the Anthropocene: Death by a thousand cuts. Proc. Natl. Acad. Sci. USA 118, e2023989118 (2021).CAS 
PubMed 
PubMed Central 

Google Scholar 
Schowalter, T. D., Pandey, M., Presley, S. J., Willig, M. R. & Zimmerman, J. K. Arthropods are not declining but are responsive to disturbance in the Luquillo Experimental Forest, Puerto Rico. Proc. Natl. Acad. Sci. USA 118, e2002556117 (2021).CAS 
PubMed 
PubMed Central 

Google Scholar 
Berry, P. A. M., Smith, R. G. & Benveniste, J. ACE2: the new global digital elevation model in Gravity, Geoid and Earth Observation (ed. Mertikas, S. P.) 231–237 (Springer, 2010).Kurentzov, A. I. My travels [In Russian] (Far Eastern Publishing House, 1973).Dubatolov, V. V. A catalogue of type specimens of Palaearctic tiger moths (Lepidoptera, Arctiidae, Arctiinae) preserved in the collection of the Zoological Institute of Russian Academy of Sciences (St. Petersburg) [In Russian]. Entomol. Rev. 75, 338–356 (1996).
Google Scholar 
Bailey, R. G. Explanatory Supplement to Ecoregions Map of the Continents. Environ. Conserv. 16, 307–309 (1989).
Google Scholar 
Olson, D. M. & Dinerstein, E. The Global 200: Priority ecoregions for global conservation. Ann. Mo. Bot. Gard. 89, 199–224 (2002).
Google Scholar 
Olson, D. M. et al. Terrestrial Ecoregions of the World: A New Map of Life on Earth. BioScience 51, 933–938 (2001).
Google Scholar 
Beaumont, L. J. et al. Impacts of climate change on the world’s most exceptional ecoregions. Proc. Natl. Acad. Sci. USA 108, 2306–2311 (2011).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Smith, J. R. et al. A global test of ecoregions. Nat. Ecol. Evol. 2, 1889–1896 (2018).PubMed 

Google Scholar  More

Direct habitat lossAccording to the global projections of urban expansion under five SSPs17 (Supplementary Note 3 and Supplementary Fig. 1), 36–74 million hectares (Mha) of land areas will be urbanized by 2100, representing a 54–111% increase compared with the baseline year of 2015. Among these, 11–33 Mha natural habitats (Supplementary Table 1) will become urban areas by 2100. Across SSP scenarios, the patterns of change in losses of total habitat, forest, shrubland, and grassland are consistent with the global projections of urban expansion (Fig. 1). In terms of urban encroachment on wetlands, wetland will undergo the largest loss under scenario SSP4 than under other scenarios. However, if the sustainable pathway of scenario SSP1 is properly implemented, this will enable us to conserve the global wetland. The greatest loss of other habitat will occur under scenario SSP3, but the minimal loss of other habitat will occur under scenario SSP1. Under the five different SSP scenarios, the United States, Nigeria, Australia, Germany, and the UK are consistently predicted to have greater habitat loss due to urban expansion (Supplementary Table 2).Fig. 1: Future direct habitat loss due to urban expansion under SSP scenarios.a The habitat loss by 2100 for each habitat type. Bars indicate the mean habitat loss area (five scenarios) for each habitat type. Error bars represent mean values ± 1 SEM for the loss of each habitat type under five scenarios, n = 5 scenarios. Points represent data in five scenarios. b The losses in total area, forest, shrubland, grassland, wetland, and other land.Full size imageThere are obvious disparities in the hot spots and cold spots of habitat loss under the five SSP scenarios (Fig. 2 and Supplementary Figs. 2–6). Potential hot spots of habitat loss are concentrated in regions such as the northeastern, southern, and western coasts of the United States, the Gulf of Guinea coastal areas, Sub-Saharan Africa, and the Persian Gulf coastal areas. Under scenario SSP5, parts of central and western Europe will also become hot spots. However, under other scenarios, the cold spots will be particularly concentrated in eastern and southern Europe. East Asia and South Asia, which are represented by China, India, and Japan, are dominated by cold spots (Supplementary Figs. 2–6), because these regions may experience a decline in urban land demand from 2050 to 2100 (for examples in China, see Supplementary Figs. 7–11), although they are currently the most populous regions in the world.Fig. 2: Future hot spots and cold spots of habitat loss due to urban expansion under SSP scenarios by 2100.Figures for the United States (a), Europe (b), Africa (c), and China (d) are presented separately. The Gi_Bin identifies statistically significant hot spots and cold spots. Statistical significance was based on the p-value and z-score (two-sided), and no adjustments were made for multiple comparisons.Full size imageOur scenario projections show that the largest natural habitat loss is expected to occur in the temperate broadleaf and mixed forests biome (except for scenario SSP3). In addition, many biomes will experience proportionate loss of natural habitat. These biomes include the tropical and subtropical coniferous forests biome, the temperate coniferous forests biome, the flooded grasslands and savannas biome, the Mediterranean forests, woodlands, and scrub biome, and the mangroves biome (Supplementary Table 3). Although the rate of future habitat loss is small at the global scale, it can be large in some areas. For example, the habitat in the temperate broadleaf and mixed forests may decrease by 1.4% under scenario SSP5. At the ecoregion scale, about 9% of 867 terrestrial ecoregions will lose more than 1% of habitat due to urban expansion (Supplementary Fig. 12). In the future, four ecoregions—the Atlantic coastal pine barrens, the coastal forests of the northeastern United States, and the Puerto Rican moist and dry forests—will experience more than 20% of habitat loss.Urban expansion threatens biodiversity prioritization schemesTo reflect the potential impact of urban expansion on protected areas (Supplementary Note 4), the analyses presented here were based on the assumption that urban expansion within protected areas is not strictly restricted and can even occur in the currently gazetted protected areas (Supplementary Note 5, Supplementary Figs. 13 and 14). In 2015, urban areas with a total area of 30,594 km2 were distributed in 28,152 protected areas, accounting for 12.6% of global protected areas (Supplementary Figs. 15 and 16). Moreover, 38% of the urban land-use changes within protected areas were due to the conversion of natural habitats into urban land between 1992 and 2015. If urban expansion continues without strict restrictions, 13.2–19.8% of the protected areas will be affected by urban land by 2100, and urban land will occur in 29,563–44,400 protected areas with a total urban land area of up to 46,705–89,901 km2 across the five SSP scenarios (the lowest and highest proportions of urban land in each protected area by 2100 under SSP3 and SSP5 scenarios are presented in Supplementary Figs. 17 and 18).We also found that 0.90% of all terrestrial biodiversity hotspots (Supplementary Note 6), which are the world’s most biologically rich yet threatened terrestrial regions24, were urbanized in 2015. And this proportion (0.90%) is higher than that located in the rest of the Earth’s surface (0.51%) in 2015. By 2100, the new urban expansion will additionally occupy 1.5–1.8% of hotspot areas under the five SSP scenarios (Supplementary Table 4). Five biodiversity hotspots are projected to suffer the largest proportion of urban land conversion: the California Floristic Province (6–11%), Japan (6–8%), the North American Coastal Plain (4–8%), the Guinean Forests of West Africa (4–8%), and the Forests of East Australia (2–6%). In contrast, the East Melanesian Islands and the New Caledonia are almost unaffected by urban expansion. Biodiversity hotspots (e.g., the Guinean Forests of West Africa, the Coastal Forests of Eastern Africa, Eastern Afromontane, and the Polynesia-Micronesia) with few human disturbances in 2015 are projected to experience the highest percentage of future urban growth. Compared with the urban areas in 2015, by 2100, the urban areas in these four biodiversity hotspots will experience a disproportionate increase of 281–708, 294–535, 169–305, and 33–337%, respectively.The World Wildlife Fund (WWF) selected the ecoregions that are most crucial to the conservation of global biodiversity as Global 20025 (Supplementary Note 7). However, about 93% of the Global 200 ecoregions will be affected by future urban expansion. Although the proportion of urban land in each ecoregion will be less than 1% in 2100, the urban area located in these ecoregions will experience an increase of 74–160% from 2015 to 2100 across the five SSP scenarios (Supplementary Table 4). Four ecologically vulnerable ecoregions that have the highest urban growth rates are the Sudd-Sahelian Flooded Grasslands and Savannas, the East African Acacia Savannas, the Hawaii Moist Forest, and the Congolian Coastal Forests. By 2100, the urban areas in these four ecoregions will increase by 877–9955, 527–646, 18–902, and 500–1037%, respectively.The five SSP scenarios showed that the urban area is expected to increase by only 73–213 km2 in the Last of the Wild areas26 (see Supplementary Note 8 for descriptions about the Last of the Wild areas) by 2100 (Supplementary Table 4).Impacts of urban expansion on habitat fragmentationThe increasing exposures of natural habitat to urbanized land use may cause long-term changes in the function and structure of the natural habitat that is adjacent to urban areas13. To examine this proximity effect, we investigated the impact of future urban expansion on the nearest distance between urban areas and natural habitat (i.e., the distance from patch edges of urban areas to patch edges of the nearest natural habitats) under different SSP scenarios. Although the global urban area is expected to increase by 36–74 Mha by 2100, the impacts of future urban expansion on adjacent natural habitat are disproportionately large. Future urban expansion will make urban areas much closer to patch edges of 34–40 Mha natural habitat, which will inevitably threaten the natural habitat and increase the risk of biodiversity decline. The effects of urban expansion on adjacent patch edges of natural habitats are remarkably different across different scenarios. Specifically, the area of affected adjacent natural habitat is expected to be 38.45, 34.24, 40.31, 37.84, and 39.42 Mha under SSP1 to SSP5 scenarios by 2100, with the smallest effect under scenario SSP2, and the largest effect under scenario SSP3. Moreover, the scale of urban expansion does not correspond directly with the size of the impact. Several countries, including Mauritania, Algeria, Saudi Arabia, Western Sahara, and the United States, will have a large change in the distance from future urban areas to natural habitats due to urban expansion (Supplementary Table 5). Such effects also varied across different natural habitat types. The distance from the patch edges of urban areas to patch edges of (a) wetland, other land, and forest, (b) grassland, and (c) shrubland will generally be shortened by ~2000, ~1500 and ~900 m, respectively.In addition to the effect on the distance to the habitat edge, urban-caused habitat fragmentation is also reflected in reducing mean patch size (MPS)13, increasing mean edge index (edge density (ED), i.e., edge length on a per-unit area)27, and enlarging isolation (mean Euclidean nearest neighbor distance, ENN_MN)28 (Fig. 3). Taking the global ecoregions as the analysis unit, we found that within a 5 km buffer of urban areas, the median of MPS of natural habitats tends to show an overall decline trend, and the segmentation and subdivision of habitats become more obvious as future urban land expands. The median of MPS is the largest under scenario SSP1, followed by SSP4, SPP2, and SSP3 with some fluctuations in between, and the smallest MPS is found with the most fragmented landscape under scenario SSP5. A smaller patch size indicates that the inner parts of the habitat are subject to higher risk of being influenced by external disturbance. Future urban expansion also tends to cause an increase in the ED of natural habitat, which is often linked with smaller patches or more irregular shapes, and therefore poses a threat to biodiversity that influences many ecological processes (e.g., the spread of dispersal and predation)13,27,28. Scenario SSP1 shows the best performance in maintaining a low habitat ED and a high level of biodiversity conservation. However, under scenario SSP5, ED will experience a rapid increase in the second half of the 21st century. Meanwhile, the ENN_MN will increase substantially in the future, suggesting that areas with the same habitat type will become increasingly isolated, irregular, dispersed, or unevenly distributed due to the barrier of urban land. This will affect the speed of dispersal and patch recolonization. Scenario SSP1 is also most conducive to maintaining the proximity of natural habitats with the same habitat type. Other scenarios show relatively similar performance.Fig. 3: Future urban expansion effects on habitat fragmentation under SSP scenarios.a Mean patch size (MPS), b edge density (ED), c mean Euclidean nearest-neighbor distance (ENN_MN).Full size imageImpacts of urban expansion on terrestrial biodiversityWe focus on biodiversity in three common vertebrate taxa (i.e., amphibians, mammals, and birds) in our analyses. Future land system conversion to urban land will cause an average of 34% loss in the overall relative species richness. Land conversion from dense forest, mosaic grassland and open forest, mosaic grassland, and bare and natural grassland to urban land will cause the highest overall relative biodiversity loss (48%, 95% confidence interval (CI): 34–59% on a 1 km grid). These land systems with a high risk of biodiversity loss are concentrated in the United States, Europe, and Sub-Saharan Africa (Supplementary Fig. 19). Overall, the negative effect of future urban expansion on the total abundance of species will be more pronounced than that on species richness. Urban land changes will result in an average of 52% overall loss in relative total abundance of species. In particular, the losses of dense forest, natural grassland, and mosaic grassland, due to conversion to urban land, will lead to a high risk of species loss (62%, 95% CI: 38–76%).In terms of the number of species (i.e., all amphibians, mammals, and birds), future urban expansion will cause an average loss of 7–9 species and a loss of up to ~197 species per 10 km grid cell by 2100 across the five SSP scenarios (Fig. 4 and Supplementary Fig. 20). Species loss is most likely to be concentrated in Sub-Saharan Africa (particularly the Gulf of Guinea coast), the United States, and Europe. In addition, southeastern Brazil, India, and the eastern coast of Australia are also relatively high-risk areas. However, the specific effects of urban expansion vary substantially across different SSP scenarios. For instance, under scenario SSP5, urban expansion will pose a fatal threat to the global species richness in areas with urban development potential (species richness loss will occur in ~740 Mha land areas), whereas under the divided pathway (SSP4) and regional rivalry pathway (SSP3) scenarios, urban expansion will threaten the richest biodiversity hotspots, such as Sub-Saharan Africa and Latin America (Supplementary Fig. 20).Fig. 4: Potential biodiversity loss due to future urban expansion under SSP scenarios.The biodiversity loss in terms of the number of terrestrial vertebrate species (amphibians, mammals, and birds) lost per 10 km grid cell in the North America (a), Europe (b), the Gulf of Guinea coast (c), and East Asia (d).Full size imageWe also found a loss of up to 12 species of threatened amphibians, mammals, and birds (including vulnerable, endangered, or critically endangered categories defined in the IUCN Red List), and a loss of up to 40 species of small-ranged amphibians, mammals, and birds (small-ranged species are species with a geographic range size smaller than the median range size for that taxon)29 due to future urban expansion by 2100. There are a few scattered areas that will be hotspots for the loss of threatened species, such as West Africa, East Africa, northern India, and the eastern coast of Australia (Supplementary Fig. 21). The loss of small-ranged species will concentrate in fewer areas (Supplementary Fig. 22). We have identified 30 conservation priority ecoregions with high risks of habitat loss and small-ranged species loss due to future urban expansion (Supplementary Table 6). These conservation priority ecoregions are all found in Latin America and Sub-Saharan Africa (Supplementary Fig. 23). However, some hotspots outside of these conservation priority regions, such as tropical Southeast Asia, the west coast of the United States, and northern New Zealand, will also be affected (Supplementary Fig. 23).The top 5% 10 km grid cells with the highest loss in species richness (28–38 species potentially being lost) scatter across adjacent urban areas. However, only 6.4–8.6% of these regions are covered by the current global network of protected areas. These areas are often overlooked, and thus receive relatively low conservation spending. Ecoregions in Sub-Saharan African, Central and South America, Southeast Asia, and Australia will be responsible for the top 43% of average species loss across the SSP scenarios (Fig. 5). Kenya, Swaziland, Brunei, Zambia, Republic of Congo, and Zimbabwe will face the largest potential species richness loss (approximately > 29 species lost per 10 km grid cell) under all five SSP scenarios (Supplementary Fig. 24 and Supplementary Table 7).Fig. 5: Average potential biodiversity loss per 10 km grid cell in ecoregions due to future urban expansion under SSP scenarios.The mean potential biodiversity loss represents the average number of terrestrial vertebrate species (amphibians, mammals, and birds) lost per 10 km grid cell.Full size image More