Ecology

Sites overview and characteristicsThis study focused on three regions located in subantarctic, arctic, and subarctic latitudes. The respective latitudinal and longitudinal ranges covered in this study were: 54.95 to 52.08 °S, and 72.03 to 67.34 °W in Patagonia; 67.44 to 67.54 °N, and 86.59 to 86.71 °E in Siberia; 63.21 to 68.63 °N, and −150.79 to −145.98 °W in Alaska (Figs. 1 and 2). The exact coordinates for each sample were included in the submitted dataset. The field campaigns were conducted in 2016, during the summer for each respective region: January-February in Chilean Patagonia, June-July in Alaska and July-August in Siberia.Fig. 1Location of the three areas included in this study (panel a). The permafrost state and the number of sites and samples per region is indicated for each area. General views of 5 sites are provided as examples (b–f). Panel B provides a large view of the ecosystem surrounding the wetland ALP2 (Alaska, exact location indicated by the white circle). Lake PCL1 (panel c) is representative of the lakes on Navarino island (Chilean Patagonia). The glacial lake SIL2 is shown in panel d. At site SIP5, the hollow at first plan is surrounded by palsa (hummock, second plan), characterized by dark organic matter and lichen vegetation (panel e). The PPP3 peatland shown in panel f is dominated by Sphagnum magellanicum, like most peatlands in the area.Full size imageFig. 2Maps of sampling sites in Patagonia, Alaska and Siberia, indicating the ecosystem type (lake, wetland, soil). The tables show the complete- (in white) and the partial- (in grey) characterization sites. The exact coordinates of each sample are provided in the data record (See data records section).Full size imageFor every site included in the present study, a set of nine qualitative environmental and/or ecological site-scale descriptors was selected and adapted from ENVO Environment Ontology40, which included for example permafrost state, biome, environmental feature and vegetation type (Table 1, Fig. 3). Permafrost state was obtained from the NSIDC permafrost map41. The biome, large-scale descriptor based on climate and vegetation criteria, was derived from Olson et al.42. Temperate forest, boreal forest, and tundra biomes were included. The environmental features that were representative for the three regions were considered: lakes, wetlands, broadleaf/coniferous/mixed forest soils, grassland, tundra, and palsa. All the metadata was included in the submitted dataset. Table 2 summarizes the main types of sampled ecosystems and their main characteristics in the three regions, while Supplementary Table S1 provides the details of each sampling site.Table 1 Overview of the dataset contained in Mimarks sheet.Full size tableFig. 3Description of the qualitative environmental/ecological descriptors used to describe every sample, derived from ENVO Environment Ontology40.Full size imageTable 2 Main types of sampled ecosystems in the three studied regions.Full size tableIn Alaska, the studied area ranged from the Alaska Range and Fairbanks area (interior, continental climate, 63–65°N, discontinuous permafrost) up to Toolik Field Station (North Slope, arctic climate, 66–69°N, continuous permafrost; Fig. 2). The physiochemistry and CH4 emissions of lakes ALL1 (Killarney lake), ALL2 (Otto lake), ALL3 (Nutella lake), and ALL4 (Goldstream lake) were previously characterized35. A number of heterogeneous soil and wetland samples were collected around the studied Alaskan lakes and/or from monitored sites, as detailed in Supplementary Table S1. In the Alaska Range and Fairbanks area, soils were mostly covered by mixed or taiga forests, alpine tundra, and bogs or fens wetlands. In the norther Brooks Ranges mountain system, the landscape was piedmont hills with a predominant soil of porous organic peat underlain by silt and glacial till, all in a permafrost state, characterized mainly by Sphagnum and Eriophorum vegetation, as well as dwarf shrubs.In Siberia, the studied area was located in the discontinuous permafrost region surrounding Igarka, on the eastern bank of the Yenisei River (Fig. 2). This region was mainly covered by forest, dominated by larch (Larix Siberica), birch (Betula Pendula), and Siberian pine (Pinus Siberica), and palsa landscapes (frozen peat mounts), the latter being dominated by moss, lichens, Labrador tea and dwarf birch. In degraded areas, thermokarst bogs were dominated by Sphagnum spp. and Eriophorum spp. Land cover was an indicator of permafrost status, since forested areas reflected a deep permafrost table ( >2 m) associated with Pleistocene permafrost, while palsa-dominated landscapes were indicative of the presence of near-surface ( More

Most records of N-tiger cats were from savanna environments, and it was not surprising that this vegetative formation has a key influence on the N-tiger cat range in the Amazon. The bulk of the L. t. tigrinus distribution lies in the savannas, dry forests and shrublands of the Cerrado and Caatinga biomes. These are also the areas with the vast majority of records for this lowland subspecies (Supplemental Fig. S5). Hence, L. t. tigrinus is more associated with savannas and savanna-like environments than with rainforests. In fact, more than 80% of the records in the Amazon were within 100 km of a savanna patch. Colonization of the northern savanna formations of the Amazon by the N-tiger cat likely occurred during the forest-savanna shifts of the glacial period18, and the cat currently shows a patchy distribution. Strong evidence of established biogeographic corridor connections between the savannas of the Cerrado and those of the Amazon exists, suggesting northward expansion of the former during glacial periods, perhaps predating the Last Glacial Maximum19,20,21. Further corroborating this evidence, tiger cat ‘gene flow’ niche modelling showed prior connectivity between the Guiana population and that of Central Brazil and no connectivity with the Andean population22. Additionally, Guianan tiger cat skin patterns are found in savanna and transitional savanna/Amazon areas and in the semiarid shrub-woodland of Brazil and are very distinct from the patterns of the tiger cats from the Andes of northwestern South America and Central America (Supplementary Information Fig. S6).The bioclimatic variables in the best model also supported the cat’s preference for savanna areas. The best model indicated a positive effect of precipitation in the driest month on the probability of the presence of the N-tiger cat, likely indicating the Aw/As climates of tropical savannas23. These climates are marked by seasonal variation in rainfall, with a pronounced dry season. Higher rainfall during the dry season favors the growth of vegetation, which results in some tree cover within the savannas. Thus, our results agree with previous research suggesting that tiger cats avoid open savanna formations24. Similarly, the species had a significant negative response to net primary productivity. This also supports the species’ avoidance of dense lowland rainforests, which are the most productive habitats. In the Amazon biome, the least productive areas are found in more open landscapes25.The N-tiger cat’s range considered from an ecoregion perspective12 could biogeographically explain its distribution in the Amazon. All records but 2 fell within Guiana savannas, Guiana highland forest, Guiana rainforest, part of the Uatumã-Trombetas rainforest bordering the Guianas or all of it connecting to Gurupá and Monte Alegre varzea forests, as well as Marajó varzeas, the interfluve Tocantins-Araguaia/Maranhão, and the southern block of the interfluve Xingu/Tocantins-Araguaia. There were two records from the Negro-Branco moist forest, which also includes savanna-like “campinarana” formations. The range also reaches the transitional babaçu palm forests of Maranhão and the Mato Grosso seasonal forests (Supplementary Information Fig. S7, Table S3). The N-tiger cat’s range in the Amazon was determined by combining records with species distribution modeling, also matching the ecoregion perspective.Outside the Guiana Shield and likely the savanna patches of the region of the Upper Negro River, in other parts of the Amazon, the N-tiger cat seems to be restricted to the forests of the eastern Amazon, along the arc of deforestation and to transitional areas with savanna formations. The presence and absence points at camera-trapping sites could explain the N-tiger cat’s range in the Amazon and define its distribution range in the biome. Absence points, for instance, were usually located in dense rainforest habitats throughout the Amazon biome.The species may occasionally occupy rainforests, such as those of the Guianas, where it tends to be very rare. At a site in central Suriname, after an enormous trapping effort of  > 20,000 trap days in four years by cat specialists, over an area  > 1100 km2, no records of the N-tiger cat were found (Supplementary Information Table S2), although its presence is expected in that area26. This finding attests to the inherent rarity of this felid in its limited range within the Amazon. However, could its association with the arc of deforestation be related to the replacement of forest by bushy savanna-like vegetation that succeeds abandoned pastures? The other currently recognized subspecies, L. t. pardinoides (the Andean tiger cat) and L. t. oncilla (the oncilla), and the recently split southern tiger cat L. guttulus are all associated with forested areas. Conversely, L. t. tigrinus has higher abundance and is mostly found in the nonforested habitats of the Cerrado and Caatinga domains of Brazil and only rarely in rainforests. Thus, L. t. tigrinus may be an open-habitat (sub)species. However, within savannas, N-tiger cats are restricted to denser savanna formations, with open savannas deemed unsuitable24. In the semiarid Caatinga, the N-tiger cat also prefers denser formations27,28.One of the most interesting findings was the clear relationship between the ranges of the dominant mesopredator and subordinate species. The ranges of ocelots and N-tiger cats in the Amazon were diametrically opposite (Fig. 1), a finding never recorded for felids. The reported ocelot densities and relative abundance indexes (RAIs) in the Amazon range from 0.29 to 0.95 ind/km2 and 0.07–13.2 ind/100 trap-days, respectively7,29. Thus, the expected ocelot density found using modeling that allows for N-tiger cat presence is very low (Fig. 2A). In the Rupununi, the ocelot:N-tiger cat RAI ratio was roughly 10:1, with a very low RAI and expected density for N-tiger cats (see Supplementary Material). The only other relative abundance estimate of tiger cats presented for the Amazon30 was not confirmed as an estimate of tiger cats following inspection of the original records by the authors but as an estimate of margays or ocelots. This antagonistic relationship between ocelots and all other small cat species in their area of sympatry is quite impressive. It is density-dependent, as it seems to take effect only above an ocelot density threshold of 0.12 ind./km231. The influence can range from patterns of density, distribution, and occupancy to spatial and temporal use. Conversely, such an impact was not detected when either the small cats or ocelots were compared to the larger cats31,32,33,34,35.In view of the Red List assessments and applying the limited estimates presented, the expected total population size for N-tiger cats in the Amazon would be approximately 150 and 1622 individuals, considering their AOO or EOO, respectively. Applying the IUCN’s formula for mature individuals8, these numbers would be 45 and 487 individuals for the AOO and EOO, respectively.The ocelot’s preference for very dense rainforests may explain the low probability of N-tiger cat occurrence within the Amazon biome. Notably, most tiger cat records from rainforests and all those from premontane forests came from the Guiana Shield, a region where tropical grasslands and savannas dot more forested landscapes. The Guiana Highlands and Pantepui ecoregions, which make up a considerable portion of the shield, tend to have low ocelot densities (below 0.30 ind/km2), although they do contain some rainforest. Ocelot densities reach some of their lowest values in the Guianan savanna ecoregion (mean ocelot density of 0.029 in the savanna formations), where the N-tiger cat probability of occurrence was highest. At the Karanambu site in the Rupununi, all ocelot records came from either gallery forests or forest patches embedded in the savanna. Although the data did not allow us to test further hypotheses, it is likely that spatial partitioning occurs in the Guiana Shield, with N-tiger cats favoring habitats that are more open. Conversely, areas farther west in the Amazon biome, other than the predicted area, do not have any major savanna patches and are covered mostly by lowland tropical rainforest formations, where ocelots can potentially reach densities in excess of 0.7 ind/km2. Of all Amazonian records of N-tiger cats, only one came from west of the 68th meridian: a preserved specimen from Puerto Leguizamo on the Putumayo River in Colombia. The specimen was identified as L. t. pardinoides by its collector, so it most likely represents an individual that came down from the foothills of the Andes. Alternatively, it could have been caught in the Andean foothills but labeled generally as from Puerto Leguizamo, as museum records do not always present precise locations, like most of those from our dataset; thus, they could represent a broader region, not a single collection location.The records of L. t. tigrinus in the Monte-Alegre Várzea ecoregion and Tapajós-Xingu Moist Forest ecoregion (which shares a border with the Amazon River) are actually from the small savanna patches of Terra Santa and Alter do Chão, respectively, which are imbedded within the forests of these ecoregions. Similarly, the Negro-Branco Moist Forest ecoregion includes open-canopy white sand forests with savanna-like vegetation, known as ‘campinaranas’36.Although our model predicted a high probability of N-tiger cat presence in the Marajó Várzea ecoregion, the records from the island came from savanna patches and not from flooded forests and mangroves. Hence, we did not include such large areas in the AOO for the subspecies. It is likely that the highly predicted probability of presence there is an artifact of low predicted ocelot density. Nevertheless, the environment there is not suitable for either cat. Our ocelot density model was highly significant and explained almost 50% of the variation in ocelot density. The remaining variation was related to either other variables that could not be measured via satellite imagery (such as prey availability) or the sampling design of the different studies. Nonetheless, ocelot densities predicted from our model across the Amazon were within the expected range for the species29.Why are N-tiger cats absent in camera-trapping studies in Amazonian forests throughout the biome? The most straightforward answer seems to be because they simply are not there (central and western Amazon) or, where present, their numbers are extremely low (Guianas and eastern Amazon). The lack of surveys cannot be cited as a potential reason for their apparent absence because the studies that did not detect the species were conducted throughout the Amazon biome, in all nine Amazonian countries. Some of the areas have been surveyed for several years—or decades in some cases—and have failed to record a single individual (Supplementary Information Table S2). Typically, N-tiger cats appear, even prominently, on cameras in other biomes, such as in the savannas of the Cerrado and semiarid scrub of the Caatinga domain in Brazil, including sites where ocelots are present24,27,37. Clouded tiger cats (L. t. pardinoides) have also been frequently recorded on cameras in the Andes, higher than 1500 m above sea level34,38, but not in lowland Amazonian forests. This finding indicates that the N-tiger cat is not camera-shy. In northern Brazilian savannas, its density can reach 0.25 ind/km2 24. Coincidentally, this highest density estimate of the N-tiger cat is the same as the lowest ocelot density estimate for Amazonian forests24,29.Tiger cats and margays show high similarity, making misidentifications relatively common39. However, the evaluation of  > 3000 camera trap images of small-medium felids in the Amazon revealed that only one mildly resembled a tiger cat, a finding that supports the species being absent there and does not represent a case of mistaken identity with margays or even ocelots7.The Amazonian range of L. tigrinus is very limited, and populations are expected to be very small. With the upcoming split of L. t. tigrinus and L. t. pardinoides into two different species40, this situation would have serious implications for the conservation of the former. Thus, L. t. tigrinus conservation lies outside the “Amazonian safe haven” of most other carnivore species found there7. The Brazilian drylands Cerrado and Caatinga represent such places for L. t. tigrinus populations. Unfortunately, these biomes have had  > 50% of their cover completely removed41. Very importantly, besides being extremely rare in the Amazonian savannas, this rather limited vegetative formation is also considered highly threatened and of conservation priority42. Therefore, the tiger cat could become an emblematic flagship species representing the uniqueness of this vegetative formation in dire need of protection.In short, the picture that emerges is that although the N-tiger cat uses both rainforests and deciduous forests in the Amazon, it seems to be mostly associated with savanna formations and that its distribution in the Amazon is highly influenced by the ocelot, the dominant mesopredator. The N-tiger cat’s inherent rarity, expected population size, and restricted range in the Amazon suggest that this biome does not in fact represent a safe haven for the global conservation of this small felid. In addition to shedding light on and refining the N-tiger cat distribution in the Amazon, this paper highlights the importance of including biological variables, such as the potential impacts of competitors and predators on species presence and distribution, in SDMs. More

Davies EK, Peters AD, Keightley PD (1999) High frequency of cryptic deleterious mutations in Caenorhabditis elegans. Science 285:1748–1751Article 
CAS 
PubMed 

Google Scholar 
Eyre-Walker A, Keightley PD (2007) The distribution of fitness effects of new mutations. Nat Rev Genet 8:610–618Article 
CAS 
PubMed 

Google Scholar 
Eyre-Walker A, Keightley PD (2013) A comparison of models to infer the distribution of fitness effects of new mutations. Genetics 193:1197–1208Article 

Google Scholar 
Fry JD, Keightley PD, Heinsohn SL, Nuzhdi SV (1999) New estimates of the rates and effects of mildly deleterious mutation in Drosophila melanogaster. Proc Natl Acad Sci 96:574–579Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
García-Dorado A (2007) Shortcut predictions for fitness properties at the mutation-selection-drift balance and for its buildup after size reduction under different management strategies. Genetics 176:983–997Article 
PubMed 
PubMed Central 

Google Scholar 
García-Dorado A (2012) Understanding and predicting the fitness decline of shrunk populations: inbreeding, purging, mutation, and standard selection. Genetics 190:1461–1476Article 
PubMed 
PubMed Central 

Google Scholar 
García-Dorado A (2015) On the consequences of ignoring purging on genetic recommendations for minimum viable population rules. Heredity 115:185–187Article 
PubMed 
PubMed Central 

Google Scholar 
García-Dorado A, Caballero A (2021) Neutral genetic diversity as a useful tool for conservation biology. Conserv Genet 22:541–545Article 

Google Scholar 
Garner BA, Hoban S, Luikart G (2020) IUCN Red List and the value of integrating genetics. Conserv Genet 21:795–801Article 

Google Scholar 
Hedrick PW, García-Dorado A (2016) Understanding inbreeding depression, purging, and genetic rescue. Trends Ecol Evol 31:940–952Article 
PubMed 

Google Scholar 
Kardos M, Armstrong EE, Fitzpatrick SW, Hauser S, Hedrick PW, Miller J et al. (2021) The crucial role of genome-wide genetic variation in conservation. Proc Natl Acad Sci USA 118:e2104642118Khan A, Patel A, Shukla H, Viswanathan A, van der Valk T, Borthakur U, … & Ramakrishnan U (2021) Genomic evidence for inbreeding depression and purging of deleterious genetic variation in Indian tigers. Proc. Natl. Acad. Sci. 118Kimura M, Maruyama T, Crow JF (1963) The mutation load in small populations. Genetics 48:1303–1312Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Kimura M (1980) Average time until fixation of a mutant allele in a finite population under continued mutation pressure: Studies by analytical, numerical, and pseudo-sampling methods. Proc Natl Acad Sci 77:522–526Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Morin PA, Archer FI, Avila CD, Balacco JR, Bukhman YV, Chow, W, … & Jarvis ED (2021) Reference genome and demographic history of the most endangered marine mammal, the vaquita. Mol Ecol Resour 21:1008–1020Mukai T (1964) The genetic structure of natural populations of Drosophila melanogaster. I. Spontaneous mutation rate of polygenes controlling viability. Genetics 50:1–19Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Nietlisbach P, Muff S, Reid JM, Whitlock MC, Keller LF (2019) Nonequivalent lethal equivalents: Models and inbreeding metrics for unbiased estimation of inbreeding load. Evol Applic 12:266–279Article 

Google Scholar 
O’Grady JJ, Brook BW, Reed DH, Ballou JD, Tonkyn DW, Frankham R (2006) Realistic levels of inbreeding depression strongly affect extinction risk in wild populations. Biol Conserv 133:42–51Article 

Google Scholar 
Pérez-Pereira N, Caballero A, García-Dorado A (2021) Reviewing the consequences of genetic purging on the success of rescue programs. Conserv Gen 23:1–17Article 

Google Scholar 
Pérez-Pereira N, Wang J, Quesada H, Caballero A (2022). Prediction of the minimum effective size of a population viable in the long term. Biodivers Conserv https://doi.org/10.1007/s10531-022-02456-zRobinson JA, Kyriazis CC, Nigenda-Morales SF, Beichman AC, Rojas-Bracho L, Robertson KM et al. (2022) The critically endangered vaquita is not doomed to extinction by inbreeding depression. Science 376:635–639Article 
CAS 
PubMed 

Google Scholar 
Teixeira JC, Huber CD (2021) The inflated significance of neutral genetic diversity in conservation genetics. Proc Natl Acad Sci USA 118:e2015096118Wade EE, Kyriazis C, Cavassim MIA, Lohmueller KE (2022) Quantifying the fraction of new mutations that are recessive lethal. bioRxiv 1–24, https://www.biorxiv.org/content/10.1101/2022.04.22.489225v1 More

Trisos, C. H., Merow, C. & Pigot, A. L. The projected timing of abrupt ecological disruption from climate change. Nature 580, 496–501 (2020).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Schwörer, C. et al. Holocene climate, fire and vegetation dynamics at the treeline in the Northwestern Swiss Alps. Veg. Hist. Archaeobot. 23, 479–496 (2014).Article 

Google Scholar 
Steinbauer, M. J. et al. Accelerated increase in plant species richness on mountain summits is linked to warming. Nature 556, 231–234 (2018).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Grabherr, G., Gottfried, M. & Pauli, H. Climate effects on mountain plants. Nature 369, 448 (1994).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Bennett, K. D. & Willis, K. J. Pollen. Tracking Environmental Change Using Lake Sediments (eds Smol, J. P., Birks, H. J. B., Last, W. M., Bradley, R. S. & Alverson, K.) 5–32 (Kluwer Academic Publishers, 2002).Liu, S. et al. Sedimentary ancient DNA reveals a threat of warming-induced alpine habitat loss to Tibetan Plateau plant diversity. Nat. Commun. 12, 2995 (2021).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Rijal, D. P. et al. Sedimentary ancient DNA shows terrestrial plant richness continuously increased over the Holocene in northern Fennoscandia. Sci. Adv. 7, eabf9557 (2021).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Giguet-Covex, C. et al. Long livestock farming history and human landscape shaping revealed by lake sediment DNA. Nat. Commun. 5, 3211 (2014).Article 
ADS 
PubMed 

Google Scholar 
Väre, H., Lampinen, R., Humphries, C. & Williams, P. Taxonomic diversity of vascular plants in the European alpine areas. in Alpine biodiversity in Europe (eds Nagy, L., Grabherr, G., Körner, C. & Thompson, D. B. A.) 133–148 (Springer Berlin Heidelberg, 2003).Theurillat, J.-P. & Guisan, A. Potential impact of climate change on vegetation in the European alps: A Review. Climatic Change 50, 77–109 (2001).Hewitt, G. The genetic legacy of the Quaternary ice ages. Nature 405, 907–913 (2000).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Tribsch, A. & Schönswetter, P. Patterns of endemism and comparative phylogeography confirm palaeo-environmental evidence for Pleistocene refugia in the Eastern Alps. Taxon 52, 477–497 (2003).Article 

Google Scholar 
Rudmann-Maurer, K., Weyand, A., Fischer, M. & Stöcklin, J. The role of landuse and natural determinants for grassland vegetation composition in the Swiss Alps. Basic Appl. Ecol. 9, 494–503 (2008).Article 

Google Scholar 
Walsh, K. et al. A historical ecology of the Ecrins (Southern French Alps): Archaeology and palaeoecology of the Mesolithic to the Medieval period. Quat. Int. 353, 52–73 (2014).Article 

Google Scholar 
Walsh, K. & Giguet-Covex, C. Encyclopedia of the World’s Biomes 555–573 (Elsevier, 2020).Schwörer, C., Henne, P. D. & Tinner, W. A model-data comparison of Holocene timberline changes in the Swiss Alps reveals past and future drivers of mountain forest dynamics. Glob. Chang. Biol. 20, 1512–1526 (2014).Article 
ADS 
PubMed 

Google Scholar 
Henne, P. D. et al. An empirical perspective for understanding climate change impacts in Switzerland. Reg. Environ. Change 18, 1–17 (2017).
Google Scholar 
Niedrist, G., Tasser, E., Lüth, C., Dalla Via, J. & Tappeiner, U. Plant diversity declines with recent land use changes in European Alps. Plant Ecol. 202, 195–210 (2009).Article 

Google Scholar 
Lasanta-Martínez, T., Vicente-Serrano, S. M. & Cuadrat-Prats, J. M. Mountain Mediterranean landscape evolution caused by the abandonment of traditional primary activities: A study of the Spanish Central Pyrenees. Appl. Geogr. 25, 47–65 (2005).Article 

Google Scholar 
Nautiyal, S. & Kaechele, H. Adverse impacts of pasture abandonment in Himalayan protected areas: Testing the efficiency of a Natural Resource Management Plan (NRMP). Environ. Impact Assess. Rev. 27, 109–125 (2007).Article 

Google Scholar 
Karger, D. N., Nobis, M. P. & Normand, S. CHELSA-TraCE21k v1. 0. Downscaled transient temperature and precipitation data since the last glacial maximum. Climate of the Past (2021).Landolt, E. et al. Flora indicativa: Okologische Zeigerwerte und biologische Kennzeichen zur Flora der Schweiz und der Alpen (Haupt, 2010).Heiri, O., Brooks, S. J., Birks, H. J. B. & Lotter, A. F. A 274-lake calibration data-set and inference model for chironomid-based summer air temperature reconstruction in Europe. Quat. Sci. Rev. 30, 3445–3456 (2011).Article 
ADS 

Google Scholar 
Heiri, O., Ilyashuk, B., Millet, L., Samartin, S. & Lotter, A. F. Stacking of discontinuous regional palaeoclimate records: Chironomid-based summer temperatures from the Alpine region. Holocene 25, 137–149 (2015).Article 
ADS 

Google Scholar 
Ivy-Ochs, S. et al. Latest Pleistocene and Holocene glacier variations in the European Alps. Quat. Sci. Rev. 28, 2137–2149 (2009).Article 
ADS 

Google Scholar 
Finsinger, W. & Tinner, W. Pollen and plant macrofossils at Lac de Fully (2135 m a.s.l.): Holocene forest dynamics on a highland plateau in the Valais, Switzerland. Holocene 17, 1119–1127 (2007).Article 
ADS 

Google Scholar 
Baroni, C. et al. Last Lateglacial glacier advance in the Gran Paradiso Group reveals relatively drier climatic conditions established in the Western Alps since at least the Younger Dryas. Quat. Sci. Rev. 255, 106815 (2021).Article 

Google Scholar 
Schibler, J., Elsner, J. & Schlumbaum, A. Incorporation of aurochs into a cattle herd in Neolithic Europe: Single event or breeding? Sci. Rep. 4, 5798 (2014).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Schimmelpfennig, I. et al. A chronology of Holocene and Little Ice Age glacier culminations of the Steingletscher, Central Alps, Switzerland, based on high-sensitivity beryllium-10 moraine dating. Earth Planet. Sci. Lett. 393, 220–230 (2014).Article 
ADS 
CAS 

Google Scholar 
Ilyashuk, E. A., Heiri, O., Ilyashuk, B. P., Koinig, K. A. & Psenner, R. The Little Ice Age signature in a 700-year high-resolution chironomid record of summer temperatures in the Central Eastern Alps. Clim. Dyn. 52, 1–15 (2018).
Google Scholar 
Willerslev, E. et al. Fifty thousand years of Arctic vegetation and megafaunal diet. Nature 506, 47–51 (2014).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Alsos, I. G. et al. Ancient sedimentary DNA shows rapid post-glacial colonisation of Iceland followed by relatively stable vegetation until the Norse settlement (Landnám) AD 870. Quat. Sci. Rev. 259, 106903 (2021).Article 

Google Scholar 
Pansu, J. et al. Reconstructing long-term human impacts on plant communities: an ecological approach based on lake sediment DNA. Mol. Ecol. 24, 1485–1498 (2015).Article 
PubMed 

Google Scholar 
Varotto, C. et al. A pilot study of eDNA metabarcoding to estimate plant biodiversity by an alpine glacier core (Adamello glacier, North Italy). Sci. Rep. 11, 1208 (2021).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Parducci, L. et al. Proxy comparison in ancient peat sediments: Pollen, macrofossil and plant DNA. Philos. Trans. R. Soc. Lond. B Biol. Sci. 370, 20130382 (2015).Article 
PubMed 
PubMed Central 

Google Scholar 
Clarke, C. L. et al. A 24,000-year ancient DNA and pollen record from the Polar Urals reveals temporal dynamics of arctic and boreal plant communities. Quat. Sci. Rev. 247, 106564 (2020).Article 

Google Scholar 
Niemeyer, B., Epp, L. S., Stoof-Leichsenring, K. R., Pestryakova, L. A. & Herzschuh, U. A comparison of sedimentary DNA and pollen from lake sediments in recording vegetation composition at the Siberian treeline. Mol. Ecol. Resour. 17, e46–e62 (2017).Article 
CAS 
PubMed 

Google Scholar 
Wilson, J. B., Peet, R. K., Dengler, J. & Pärtel, M. Plant species richness: the world records. J. Veg. Sci. 23, 796–802 (2012).Article 

Google Scholar 
Wick, L., van Leeuwen, J. F. N., van der Knaap, W. O. & Lotter, A. F. Holocene vegetation development in the catchment of Sägistalsee (1935 m asl), a small lake in the Swiss Alps. J. Paleolimnol. 30, 261–272 (2003).Article 
ADS 

Google Scholar 
Lotter, A. F. et al. Holocene timber-line dynamics at Bachalpsee, a lake at 2265 m a.s.l. in the northern Swiss Alps. Veg. Hist. Archaeobot. 15, 295–307 (2006).Article 

Google Scholar 
Thöle, L. et al. Reconstruction of Holocene vegetation dynamics at Lac de Bretaye, a high-mountain lake in the Swiss Alps. Holocene 26, 380–396 (2016).Article 
ADS 

Google Scholar 
Heiri, O., Lotter, A. F., Hausmann, S. & Kienast, F. A chironomid-based Holocene summer air temperature reconstruction from the Swiss Alps. Holocene 13, 477–484 (2003).Article 
ADS 

Google Scholar 
Garcés-Pastor, S., Cañellas-Boltà, N., Clavaguera, A., Calero, M. A. & Vegas-Vilarrúbia, T. Vegetation shifts, human impact and peat bog development in Bassa Nera pond (Central Pyrenees) during the last millennium. Holocene 27, 553–565 (2017).Article 
ADS 

Google Scholar 
Aeschimann, D., Lauber, K., Moser, D. M. & Theurillat, J. P. Flora Alpina: Atlas des 4500 Plantes Vasculaires des Alpes (Belin, 2004).Sønstebø, J. H. et al. Using next-generation sequencing for molecular reconstruction of past Arctic vegetation and climate. Mol. Ecol. Resour. 10, 1009–1018 (2010).Article 
PubMed 

Google Scholar 
Diekmann, M. Species indicator values as an important tool in applied plant ecology—a review. Basic Appl. Ecol. 4, 493–506 (2003).Article 

Google Scholar 
Giesecke, T. et al. Postglacial change of the floristic diversity gradient in Europe. Nat. Commun. 10, 5422 (2019).Article 
ADS 
PubMed 
PubMed Central 

Google Scholar 
Colombaroli, D. & Tinner, W. Determining the long-term changes in biodiversity and provisioning services along a transect from Central Europe to the Mediterranean. Holocene 23, 1625–1634 (2013).Article 
ADS 

Google Scholar 
Schwörer, C., Colombaroli, D., Kaltenrieder, P., Rey, F. & Tinner, W. Early human impact (5000–3000 BC) affects mountain forest dynamics in the Alps. J. Ecol. 103, 281–295 (2015).Article 

Google Scholar 
Furtwängler, A. et al. Ancient genomes reveal social and genetic structure of Late Neolithic Switzerland. Nat. Commun. 11, 1915 (2020).Article 
ADS 
PubMed 
PubMed Central 

Google Scholar 
Gilck, F. & Poschlod, P. The origin of alpine farming: A review of archaeological, linguistic and archaeobotanical studies in the Alps. Holocene 29, 1503–1511 (2019).Article 
ADS 

Google Scholar 
Tinner, W., Nielsen, E. H. & Lotter, A. F. Mesolithic agriculture in Switzerland? A critical review of the evidence. Quat. Sci. Rev. 26, 1416–1431 (2007).Article 
ADS 

Google Scholar 
Berthel, N., Schwörer, C. & Tinner, W. Impact of Holocene climate changes on alpine and treeline vegetation at Sanetsch Pass, Bernese Alps, Switzerland. Rev. Palaeobot. Palynol. 174, 91–100 (2012).Article 

Google Scholar 
Hafner, A. & Schwörer, C. Vertical mobility around the high-alpine Schnidejoch Pass. Indications of Neolithic and Bronze Age pastoralism in the Swiss Alps from paleoecological and archaeological sources. Quat. Int. https://doi.org/10.1016/j.quaint.2016.12.049 (2017).Oveisi, M. et al. Potential for endozoochorous seed dispersal by sheep and goats: Risk of weed seed transport via animal faeces. Weed Res. 61, 1–12 (2021).Article 

Google Scholar 
Bardgett, R. D. & Wardle, D. A. Herbivore-mediated linkages between aboveground and belowground communities. Ecology 84, 2258–2268 (2003).Article 

Google Scholar 
Scherrer, D. & Körner, C. Topographically controlled thermal-habitat differentiation buffers alpine plant diversity against climate warming. J. Biogeogr. 38, 406–416 (2011).Article 

Google Scholar 
Giguet-Covex, C. et al. New insights on lake sediment DNA from the catchment: Importance of taphonomic and analytical issues on the record quality. Sci. Rep. 9, 14676 (2019).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Andres, B. Alpine settlement remains in the Bernese Alps (Switzerland) in medieval and modern times. Historical Archaeologies of Transhumance across Europe (eds Costello, E. & Svensson, E.) 155–169 (Routledge, 2018).eTopoi. Journal for Ancient Studies. 3, 279–283 (2012).Grime, J. P. Competitive exclusion in herbaceous vegetation. Nature 242, 344–347 (1973).Article 
ADS 

Google Scholar 
Yuan, Z. Y., Jiao, F., Li, Y. H. & Kallenbach, R. L. Anthropogenic disturbances are key to maintaining the biodiversity of grasslands. Sci. Rep. 6, 22132 (2016).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Spiegelberger, T., Matthies, D., Müller-Schärer, H. & Schaffner, U. Scale-dependent effects of land use on plant species richness of mountain grassland in the European Alps. Ecography 29, 541–548 (2006).Article 

Google Scholar 
Maurer, K., Weyand, A., Fischer, M. & Stöcklin, J. Old cultural traditions, in addition to land use and topography, are shaping plant diversity of grasslands in the Alps. Biol. Conserv. 130, 438–446 (2006).Article 

Google Scholar 
Kampmann, D. et al. Mountain grassland biodiversity: Impact of site conditions versus management type. J. Nat. Conserv. 16, 12–25 (2008).Article 

Google Scholar 
Pellegrini, E., Buccheri, M., Martini, F. & Boscutti, F. Agricultural land use curbs exotic invasion but sustains native plant diversity at intermediate levels. Sci. Rep. 11, 8385 (2021).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Bakker, E. S., Ritchie, M. E., Olff, H., Milchunas, D. G. & Knops, J. M. H. Herbivore impact on grassland plant diversity depends on habitat productivity and herbivore size. Ecol. Lett. 9, 780–788 (2006).Article 
PubMed 

Google Scholar 
Speed, J. D. M., Austrheim, G., Hester, A. J. & Mysterud, A. Elevational advance of alpine plant communities is buffered by herbivory. J. Veg. Sci. 23, 617–625 (2012).Article 

Google Scholar 
Filazzola, A. et al. The effects of livestock grazing on biodiversity are multi-trophic: A meta-analysis. Ecol. Lett. 23, 1298–1309 (2020).Article 
PubMed 

Google Scholar 
Evans, D. M. et al. The cascading impacts of livestock grazing in upland ecosystems: A 10-year experiment. Ecosphere 6, art42 (2015).Article 

Google Scholar 
Alexander, J. M., Diez, J. M. & Levine, J. M. Novel competitors shape species’ responses to climate change. Nature 525, 515–518 (2015).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Mathieu, J. Eine Agrargeschichte der inneren Alpen. Graubünden, Tessin, Wallis 1500–1800 (Chronos, 1992).Aerni, K, Egli, H. R & Fehn, K. Siedlungsprozesse an der Höhengrenze der Ökumene: am Beispiel der Alpen: Referate der 16 Tagung des” Arbeitskreises für genetische Siedlungsforschung in Mitteleuropa” vom 20.−23. (Siedlungsforschung: Spiez, 1991).Brugger, S. O. et al. Alpine glacier reveals ecosystem impacts of Europe’s prosperity and peril over the last millennium. Geophys. Res. Lett. 48, e2021GL095039 (2021).Merkt, J. & Streif, H. Stechrohr-Bohrgeräte für limnische und marine Lockersedimente. Geologisches Jahrbuch 88, 137–148 (1970).Lamb, A. L. Determination of organic and carbonate content in soils and sediments by loss on ignition (LOI), NERC Isotope Geosciences Laboratory Report, 197 (2004).Reimer, P. J. et al. The IntCal20 Northern Hemisphere radiocarbon age calibration curve (0–55 cal kBP). Radiocarbon https://doi.org/10.1017/RDC.2020.41 (2020).Blaauw, M. & Christen, J. A. Flexible paleoclimate age-depth models using an autoregressive gamma process. Bayesian Anal. 6, 457–474 (2011).Article 
MathSciNet 
MATH 

Google Scholar 
Brooks, S. J., Langdon, P. G. & Heiri, O. The identification and use of Palaearctic Chironomidae larvae in palaeoecology. Quat. Res. Assoc. i-vi, 1-276 (2007).Schulze, E. A Key to the Larval Chironomidae and their Instars from Austrian Danube Region Streams and Rivers with Particular Reference to a Numerical Taxonomic Approach. Part I. In: Wasser und Abwasser, Supplementband 3/93. Hrsg.: Bundesamt für Wassergüte, Wien-Kaisermühlen. Schriftenleitung: Werner Kohl. Selbstverlag, 1993, 514 S., öS 562. Acta Hydrochim. Hydrobiol. 22, 191–191 (1994).Article 

Google Scholar 
Juggins, S. C2: Software for ecological and palaeoecological data analysis and visualisation (user guide version 1.5). Newcastle upon Tyne: Newcastle University (2007). https://www.staff.ncl.ac.uk/stephen.juggins/software/code/C2.pdf.Moore, P. D., Webb, J. A. & Collison, M. E. Pollen Analysis, edn 2 (Blackwell, 1991).Stockmarr & Ja Tabletes with spores used in absolute pollen analysis. Pollen Spores 13, 615–621 (1971).
Google Scholar 
Reille, M. Pollen et spores d’Europe et d’Afrique du Nord (Laboratoire de Botanique historique et Palynologie, Marseille, 1992).van Geel, B. et al. Environmental reconstruction of a Roman Period settlement site in Uitgeest (The Netherlands), with special reference to coprophilous fungi. J. Archaeol. Sci. 30, 873–883 (2003).Article 

Google Scholar 
Bennett, K. D. Determination of the number of zones in a biostratigraphical sequence. N. Phytol. 132, 155–170 (1996).Article 
CAS 

Google Scholar 
Tinner, W. et al. Pollen and charcoal in lake sediments compared with historically documented forest fires in southern Switzerland since AD 1920. Holocene 8, 31–42 (1998).Article 
ADS 

Google Scholar 
Adolf, C. et al. The sedimentary and remote-sensing reflection of biomass burning in Europe. Glob. Ecol. Biogeogr. 27, 199–212 (2018).Article 

Google Scholar 
Tinner, W. & Hu, F. S. Size parameters, size-class distribution and area-number relationship of microscopic charcoal: Relevance for fire reconstruction. Holocene 13, 499–505 (2003).Article 
ADS 

Google Scholar 
Parducci, L. et al. Ancient plant DNA in lake sediments. N. Phytol. 214, 924–942 (2017).Article 
CAS 

Google Scholar 
Alsos, I. G. et al. The treasure vault can be opened: Large-scale genome skimming works well using herbarium and silica gel dried material. Plants 9, 432 (2020).Article 
CAS 
PubMed Central 

Google Scholar 
Taberlet, P. et al. Power and limitations of the chloroplast trnL (UAA) intron for plant DNA barcoding. Nucleic Acids Res. 35, e14 (2007).Article 
PubMed 

Google Scholar 
Voldstad, L. H. et al. A complete Holocene lake sediment ancient DNA record reveals long-standing high Arctic plant diversity hotspot in northern Svalbard. Quat. Sci. Rev. 234, 106207 (2020).Article 

Google Scholar 
Boyer, F. et al. obitools: A unix-inspired software package for DNA metabarcoding. Mol. Ecol. Resour. 16, 176–182 (2016).Article 
CAS 
PubMed 

Google Scholar 
Ficetola, G. F. et al. An in silico approach for the evaluation of DNA barcodes. BMC Genomics 11, 434 (2010).Article 
PubMed 
PubMed Central 

Google Scholar 
Soininen, E. M. et al. Highly overlapping winter diet in two sympatric lemming species revealed by DNA metabarcoding. PLoS One 10, e0115335 (2015).Article 
PubMed 
PubMed Central 

Google Scholar 
Boratyn, G. M. et al. BLAST: A more efficient report with usability improvements. Nucleic Acids Res. 41, W29–W33 (2013).Article 
PubMed 
PubMed Central 

Google Scholar 
Leonard, J. A. et al. Animal DNA in PCR reagents plagues ancient DNA research. J. Archaeol. Sci. 34, 1361–1366 (2007).Article 

Google Scholar 
Deiner, K. et al. Environmental DNA metabarcoding: Transforming how we survey animal and plant communities. Mol. Ecol. 26, 5872–5895 (2017).Article 
PubMed 

Google Scholar 
Ter Braak, C. J. F. & Prentice, I. C. A theory of gradient analysis. Adv. Ecol. Res. 18, 271–317 (Elsevier, 1988).Vieira, D. C., Brustolin, M. C., Ferreira, F. C. & Fonseca, G. segRDA: Anr package for performing piecewise redundancy analysis. Methods Ecol. Evol. 10, 2189–2194 (2019).Article 

Google Scholar 
Simpson, G. L. Modelling palaeoecological time series using generalised additive models. Front. Ecol. Evol. 6, 149 (2018).Wood, S. N. Generalized Additive Models: An Introduction with R (Chapman and Hall/CRC, 2017).Lefcheck, J. S. piecewiseSEM: Piecewise structural equation modelling inr for ecology, evolution, and systematics. Methods Ecol. Evol. 7, 573–579 (2016).Article 

Google Scholar 
Chen, W. & Ficetola, G. F. Numerical methods for sedimentary‐ancient‐DNA‐based study on past biodiversity and ecosystem functioning. Environ. DNA 2, 115–129 (2020).Article 

Google Scholar 
Juggins, S. Rioja: Analysis of Quaternary Science Data. R package version 0.9-26. https://cran.r-project.org/web/packages/rioja/index.html (2020).Oksanen, J. et al. vegan: Community Ecology Package. Software http://CRAN.R-project.org/package=vegan (2012).Wickham, H. ggplot2-Elegant Graphics for Data Analysis (Springer, 2016).Fox, J. & Weisberg, S. An R Companion to Applied Regression (Sage, 2019).Tinner, W. & Ammann, B. Long-term responses of mountain ecosystems to environmental changes: Resilience, adjustment, and vulnerability. In Global change and mountain regions. 133–143 (Springer, Dordrecht; 2005). More

The framework of Swin MSIWe established the first Swin Transformer-based mosquito species identification (Swin MSI) model, with the help of self-constructed image dataset and multi-adjustment-test. Gradient-weighted class activation mapping was used to visualize the identification process (Fig. 1a). The key Swin Transformer block was described on Fig. 1b. Based on practical needs, Swin MSI was additional designed to identify Culex pipiens Complex on the subspecies level (Fig. 1c) and novel mosquito (which was defined as ones beyond 17 species in our dataset) classification attribution (Fig. 1d). Detailed results are shown in the following sections.Figure 1The Framework of Swin MSI. (a)The basic architecture for mosquito features extraction and identification. Attention visualization generated by filters at each layer are shown. (b) Details for Swin Transformer block. (c) For mosquito within our dataset 17 species, output is the top 5 confidence species. (d) For mosquito beyond 17 species (defined as novel species), whether the output is a species or a genus is decided after comparing with confidence threshold.Full size imageMosquito datasetsWe established the highest-definition and most-balanced mosquito image dataset to date. The mosquito image dataset covers 7 genera and 17 species (including 3 morphologically similar subspecies in the Cx. pipiens Complex), which covers the most common and important disease-transmitting mosquitoes at the global scale, with a total of 9,900 mosquito images. The image resolution was 4464 × 2976 pixels. The specific taxonomic status and corresponding images are shown in Fig. 2. Due to the limitation of field collection, Ae. vexans, Coquillettidia ochracea, Mansonia uniformis, An. vagus and Toxorhynchites splendens only have females or only have males. In addition, each mosquito species included 300 images of both sexes, which was large enough and same number for each species, in order to balance the capacity and variety of training sets.Figure 2Taxonomic status and index of mosquito species included in this study Both male and female mosquitoes were photographed from different angles such as dorsal, left side, right side, ventral side, etc. Except for 5 species, each mosquito includes 300 images of both sexes, and the resolution of mosquito photos were 4464 × 2976. Cx. pipiens quinquefasciatus, Cx. pipiens pallens, and Cx. pipiens molestus (subspecies level, in dark gray background) were 3 subspecies in Cx. pipiens Complex (species level).Full size imageWorkflow for mosquito species identificationA three-stage flowchart of building best deep learning model for identification of mosquito species model was adopted (Fig. 3). The first learning stage was conducted by three CNNs (the Mask R-CNN, DenseNet, and YOLOv5) and three transformer models (the Detection Transformer, Vision Transformer, and Swin Transformer). Based on the performance of the first-stage model and the real mosquito labels, the second learning stage involved adjusting the model parameters of the three Swin Transformer variants (T, B, and L) to compare their performances. The third learning stage involved testing the effects of inputting differently sized images (384 × 384 and 224 × 224) to the Swin Transformer-L model; finally, we proposed a deep learning model for mosquito species identification (Swin MSI) to test the recognition effects of different mosquito species. The model was validated on different mosquito species, with a focus on the identification accuracy of three subspecies within the Cx. pipiens Complex and the detection effect of novel mosquito species.Figure 3Flowchart of testing deep learning model for intelligent identification of mosquito species.Full size imageComparison between the CNN model and Transformer model results (1st round of learning)Figure 4a shows the accuracies obtained for the six different computer vision network models tested on the mosquito picture test set. The test results show that the transformer network model had a higher mosquito species discrimination ability than the CNN.Figure 4Comparison of mosquito recognition effects of computer vision network models and variants. (a) Comparison of mosquito identification accuracy between 3 CNNs and 3 Transformer; (b) The best effect CNN (YOLOv5) training set loss curve(blue), validation set loss curve(green) and validation set accuracy curve(orange); (c) The best effect Transformer (Swin Transformer) training set loss curve, validation set loss curve and validation set accuracy curve. (d) Swin-MSI-T test result confusion matrix; (e) Swin-MSI -B test result confusion matrix; (f) Swin-MSI -L test result confusion matrix. Confusion matrix of mosquito labels in which odd numbers represent females and even numbers represent males. The small squares in the confusion matrix represent the recognition readiness rate, from red to green, the recognition readiness rate is getting higher and higher An. sinensis: 1, 2; Cx. pipiens quinquefasciatus: 3, 4; Cx. pipiens pallens: 5, 6; Cx. pipiens molestus: 7,8 Cx. modestus: 9,10; Ae. albopictus: 11, 12 Ae. aegypti: 13, 14; Cx. pallidothorax: 15, 16 Ae. galloisi: 17,18 Ae. vexans: 19, 20; Ae. koreicus: 21, 22 Armigeres subalbatus: 23, 24; Coquillettidia ochracea: 25, 26 Cx. gelidus: 27, 28 Cx. triraeniorhynchus: 29, 30 Mansonia uniformis: 31, 32 An. vagus: 33, 34 Ae. elsaie: 35,36 Toxorhynchites splendens: 37, 38.Full size imageIn the CNN training process (applied to YOLOv5), the validation accuracy requires more than 110 epochs to grow to 0.9, and the validation loss requires 110 epochs to drop to a flat interval; in contrast, during the training step, these losses represent a continuously decreasing process. These results indicate that the deep learning model derived based on the Swin Transformer algorithm was able to achieve a higher recognition accuracy in less time than the rapid convergence ability of the CNN during the iterative process (Fig. 4b).The Swin Transformer model exhibited the highest test accuracy of 96.3%. During the training process, the loss of this model could stabilize after 30 epochs, and its validation accuracy could grow to 0.9 after 20 epochs; during the validation step, the loss can drop to 0.36 after 20 epochs, after which the loss curve fluctuated but did not produce adverse effects (Fig. 4c). Based on the excellent performance of the Swin Transformer model, this model was used as the baseline to carry out the subsequent analyses.Swin Transformer model variant adjustment (2nd round of learning)Following testing performed to clarify the superior performance of the Swin Transformer algorithm, we chose different Drop_path_rate, Embed_dim and Depths parameter settings and labeled the parameter sets as the Swin Transformer-T, Swin Transformer-B, and Swin Transformer-L variants. Drop_path is an efficient regularization method, and an asymmetric Drop_path_rate is beneficial for supervised representation learning when using image classification tasks and Transformer architectures. The Embed_dim parameter represents the image dimensions obtained after the input red–green–blue (RGB) image is calculated by the Swin Transformer block in stage 1. The Depths parameter is the number of Swin Transformer blocks used in the four stages. The parameter information and test results are shown in Table 1. Due to the increase in the Swin Transformer block and Embed_dim parameters in stage 3, the recognition accuracies of the three variants were found to be 95.8%, 96.3%, and 98.2%, Correspondingly, the f1 score were 96.2%, 96.7% and 98.3%; thus, these variants could effectively improve the mosquito species identification ability in a manner similar to the CNN by increasing the number of convolutional channels to extract more features and improve the overall classification ability. In this study, the Swin Transformer-L variant, which exhibited the highest accuracy, was selected as the baseline for the next work.Table 1 Parameters and test accuracy of three variants of Swin Transformer.Full size tableBy plotting a confusion matrix of the test set results derived using the three Swin Transformer variants, we clearly obtained the proportion of correct and incorrect identifications in each category to visually reflect the mosquito species discrimination ability (Fig. 4d–f). In the matrix, the darker diagonal colors indicate higher identification rate accuracies of the corresponding mosquito categories. Among them, five mosquito species were missing because the Ae. vexans, Coquillettidia ochracea, Mansonia uniformis, An. vagus and Toxorhynchites splendens species were represented in the dataset by only females or only males. The confusion matrix shown in Panel C lists the lowest number of mosquito species identification error points and the lowest accuracy level obtained in each category, suggesting that the Swin Transformer-L model has a better classification performance than the Swin Transformer-T and Swin Transformer-B models.Effect of the input image size on the discrimination ability (3rd round of learning)To investigate the relationship between the input image size and mosquito species identification performance, in this study, we conducted a comparison test between input images with sizes of 224 × 224 and 384 × 384, based on the Swin Transformer-L model, and identified 8 categories of mosquito identification accuracy differences. These test results are shown in Table 2. When using an image size of 224 × 224 pixels, the batch_size parameter was set to 16, and when using an image size of 384 × 384 pixels, the batch_size parameter was set to 4; under these conditions, the proportion of utilized video memory accounted for 67%, as shown in Eq. 4, and this was consistent with the description of the relationship between the size of self-attentive operations during the operation of the Swin Transformer model when 384 × 384 pixels images were used. The time required for the Transformer-L model to complete all the training sessions was excessive, reaching 126 h and even exceeding the 124 h required by the YOLOv5 model, which was found to require the highest computation time during the training process in this work. Long-term training process could more fully reflect the performance differences between models. Fortunately and actually, the response speed of the model will not be affected by the training time. Compared to the accuracy of 98.2% obtained for 224 × 224 inputs, the 384 × 384 input image size derived based on the Swin Transformer-L model provided a higher mosquito species identification accuracy of 99.04%, representing an improvement of 0.84%.$$Omega ({text{W}} – {text{MSA}}) = 4{text{HWC}}^{2} + 2{text{M}}^{2} {text{HWC}}$$
(1)
Table 2 Comparison of recognition accuracy for different input image sizes.Full size tableVisualizing and understanding the Swin MSI modelsTo investigate the differences in the attentional features utilized by the Swin MSI and taxonomists for mosquito species identification, we applied the Grad-CAM method to visualize the Swin MSI attentional areas on mosquitoes at different stages. Because the Swin Transformer has different attentional ranges among its multi-head self-attention steps in different stages, different attentional weights can be found on different mosquito positions. In stage 1, the feature dimension of each patch was 4 × 4 × C, thus enabling the Swin Transformer’s multi-head self-attention mechanism to give more attention to the detailed parts of the mosquitoes, such as their legs, wings, antennae, and pronota. In stage 2, the feature dimension of each patch was 8 × 8 × 2C, enabling the Swin Transformer’s multi-head self-attention mechanism to focus on the bodies of the mosquitoes, such as their heads, thoraces, and abdomens. In stage 3, when the feature dimension of each patch was 16 × 16 × 4C, the Swin Transformer’s multi-head self-attention mechanism could focus on most regions of the mosquito, thus forming a global overall attention mechanism for each mosquito (Fig. 5). This attentional focus process is essentially the same as the process used by taxonomists when classifying mosquito morphology, changing from details to localities to the whole mosquito.Figure 5Attention visualization of representative mosquitoes of the genera Ae., Cx., An., Armigeres, Coquillettidia and Mansonia. This is a visualization for identifying the regions in the image that can explain the classification progress. Images of Toxorhynchites contain only males, with obvious differences in morphological characteristics, are not shown.Full size imageAe. aegypti is widely distributed in tropical and subtropical regions around the world and transmits Zika, dengue and yellow fever. A pair of long-stalked sickle-shaped white spots on both shoulder sides of the mesoscutum, with a pair of longitudinal stripes running through the whole mesotergum, is the most important morphological identification feature of this species. This feature was the deepest section in the attention visualization, indicating that the Swin MSI model also recognized it as the principal distinguishing feature. In addition, the abdominal tergum of A. aegypti is black and segments II-VII have lateral silvery white spots and basal white bands; the model also focused on these areas.Cx. triraeniorhynchus is the main vector of Japanese encephalitis; this mosquito has a small body size, a distinctive white ring on the proboscis (its most distinctive morphological feature), and a peppery color on its whole body. Similarly, the model constructed herein focused on both the head and abdominal regions of this species.An. sinensis is the top vector of malaria in China and has no more than three white spots on its anterior wing margin and a distinct white spot on its marginal V5.2 fringe; this feature was observed in Stage 2, at which time the modelstrongly focused on the corresponding area.The most obvious feature of Armigeres subalbatus is the lateral flattening and slightly downward curving of its proboscis; the observation of the attention visualization revealed that the constructed model focused on these regions from Stage 1 to Stage 3. The mesoscutum and abdominal tergum were not critical and were less important for identification than the proboscis, and the attention visualization results correspondingly show that the neural network focused less on these features.Coquillettidia ochracea belongs to the Coquillettidia genus and is golden yellow all over its body, with the most pronounced abdomen among the analyzed species. The model showed a consistent morphological taxonomic focus on the abdomen of this species.Mansonia uniformis is a vector of Malayan filariasis. The abdominal tergum of this species is dark brown, and its abdominal segments II-VII have yellow terminal bands and lateral white spots, which are more obvious than the dark brown feature on proboscis. Through the attention visualization, we determined that the Swin MSI model was more concerned with the abdominal region features than with the proboscis features.Subspecies-level identification tests of mosquitos in the Culex pipiens ComplexFine-grained image classification has been the focus of extensive research in the field of computer vision25,26. Based on the test set (containing 270 images) constructed herein for three subspecies of the Cx. pipiens Complex, the subspecies and sex identification accuracies were 100% when the Swin MSI model was used.The morphological characteristics of Cx. pipiens quinquefasciatus, Cx. pipiens pallens, and Cx. pipiens molestus within the Cx. pipiens Complex are almost indistinguishable, but their host preferences, self-fertility properties, breeding environments, and stagnation overwintering strategies are very different27. Among the existing features available for morphological classification, the stripes on the abdominal tergum of Cx. pipiens quinquefasciatus are usually inverted triangles and are not connected with the pleurosternums, while those of Cx. pipiens pallens are rectangular and are connected with the pleurosternums. Cx. pipiens molestus is morphologically more similar to Cx. pipiens pallens as an ecological subspecies of the Cx. pipiens Complex. However, taxonomists do not recommend using the unstable feature mentioned above as the main taxonomic feature for differentiation. By analyzing the attention visualization results of these three subspecies (last three rows on Fig. 5), we found that the neural networks of Cx. pipiens quinquefasciatus, Cx. pipiens pallens, and Cx. pipiens molestus still focused on the abdominal regions, as shown in dark red. The area of focus of these neural networks differ from that of the human eye, and the results of this study suggest that the Swin MSI model can detect finely granular features among these three mosquito subspecies that are indistinguishable to the naked human eye.Novel mosquito classification attributionAfter we performed a confidence check on the successfully identified mosquito images in the dataset, the lowest confidence value was found to be 85%. A higher confidence threshold mean stricter evaluation criteria, which can better reflect the powerful performance of the model. Therefore, 0.85 was set as the confidence threshold when judging novel mosquitoes. When identifying 10 unknown mosquito species, the highest derived species confidence level was below 85%; when the results were output to the genus level (Fig. 1d), the average probability of obtaining a correct judgment was 96.26%accuracy and 98.09% F1-score (Table 3). The images tested as novel Ae., Cx. and An. mosquito were from Minakshi and Couret et al.28,29.Table 3 Probability of correct attribution of novel species.Full size table More

Seed sampling and germinationSeed from ten trees from each of 21 native Scottish Scots pine populations (Table 1) were collected in March 2007 and germinated at the James Hutton Institute, Aberdeen (latitude 57.133214, longitude −2.158764) in June 2007. Populations were chosen to represent the species’ native range in Scotland and to include three populations from each of the seven seed zones (Fig. 2). There was no selection of seed-trees on the basis of any traits except for the possession of cones on the date of sampling. Ten seed trees were sampled from each population according to a spatial protocol designed to cover a circle of approximately 1 km in diameter located around a central tree. The sampling strategy identified nine points each in a pre-determined random direction from the central point, whilst stratifying the number sampled with increasing distance from the central point in the ratio 1: 3: 5. This strategy avoids over-sampling the areas close to the centre point. For smaller fragments of woodland, or where only a few trees with cones were present, then the directions of the sampled trees from the central tree were maintained to give a wide coverage of the woodland area, but the distances between trees varied but were never closer than 50 m. To break dormancy, seeds were soaked for 24 hours on the benchtop at room temperature, after which they were stored in wet paper towels and refrigerated in darkness at 3–5 °C for approximately 4 weeks. Seeds were kept moist and transferred to room temperature until germination began (approx. 5–7 days), then transplanted to 8 cm × 8 cm × 9 cm, 0.4 L pots filled with Levington’s C2a compost and 1.5 g of Osmocote Exact 16–18 months slow release fertiliser and kept in an unheated glasshouse. The compost was covered with a layer of grit to reduce moss and liverwort growth. Seedlings from the same mother tree are described as a family and are assumed to be half-siblings.Table 1 Locations and basic environmental data for the populations sampled for seed to establish the trial. See the maternal traits dataset15 for precise data for each mother tree sampled.Full size tableExperimental design: nurseriesThe full collection consisted of 210 families (10 families from each of 21 populations) each consisting of 24 half sibling progeny (total 5,040 individuals); needle tissue was sampled from each seedling and preserved for long term storage, one needle on silica gel, 2–5 needles at −20 °C. After transfer into pots, 8 seedlings per family were moved to one of three nurseries (total 1,680 seedlings per nursery): outdoors at Inverewe Gardens in western Scotland (nursery in the west of Scotland: coded NW, latitude 57.775714, longitude −5.597181, Fig. 2); outdoors in a fruit cage (to minimise browsing) at the James Hutton Institute in Aberdeen (nursery in the east of Scotland: NE); in an unheated glasshouse at the James Hutton Institute in Aberdeen (nursery in a glasshouse: NG). Trees were arranged in 40 randomised trays (blocks) in each nursery. Each block contained two trees per population (total 42 trees). Watering was automatic in NG, and manually as required for NE and NW. No artificial light was used in any of the nurseries. In May 2010 the seedlings from NG were moved outdoors to Glensaugh in Aberdeenshire (latitude 56.893567, longitude −2.535736). In 2010 all plants were repotted into 19 cm (3 L) pots containing Levingtons CNSE Ericaceous compost with added Osmocote STD 16–18 month slow release fertilizer.Experimental design: field sitesIn 2012 the trees were transplanted to one of three field sites: Yair in the Scottish Borders (field site in the south of Scotland: FS, latitude 55.603625, longitude −2.893025); Glensaugh (field site in the east of Scotland: FE); and Inverewe (field site in the west of Scotland: FW). All trees transplanted to FS were raised in the NG and all but four of the trees transplanted to FE were raised locally in the NE (the remainder were grown in NG). In contrast, following mortality and ‘beating up’ (filling gaps where saplings had died), the FW experiment ultimately contained cohorts of trees raised in each of the three nurseries as follows: 290 grown locally in the NW; 132 were grown in the NG; and 82 were grown in the NE.Site historiesThe Yair site (FS) had previously been used for growing Noble fir (Abies procera) for Christmas trees and Lodgepole pine (Pinus contorta), a section of the former were felled and chipped to create a clear area prior to planting. The planting site is also adjacent to a large block of commercial Sitka spruce (Picea sitchensis) forestry, and the Glenkinnon Burn Site of Special Scientific Interest (SSSI NatureScot site code 736; EU site code 135445), an area of mixed broadleaf woodland. Prior to planting, major areas of tall weeds were strimmed. The site was protected by a deer fence. The experiment was planted 8–11 October 2012. The Glensaugh site (FE) is in Forestry Compartment 3 of the Glensaugh Research Station, adjacent to Cleek Loch. It is thought to have been cleared of Scots pine and Larch (Larix decidua) around 1917, after which it reverted to rough grazing. An attempt to reseed part of the site in the 1980s was unsuccessful and it quickly reverted to rough grazing for a second time. The whole site (within which the experimental area is embedded) was deer fenced and re-planted under the Scottish Rural Development Programme (SRDP) in 2012. The experimental plot was planted up 7–9 March 2012. The Inverewe site (FW) had previously been a Sitka spruce and Lodgepole pine plantation (50:50 mix) that had been clear-felled in 2010 following substantial windthrow. The experimental site was deer fenced in early 2012, and the experiment was planted 12–16 March 2012, followed by beating up on 27–28 March 2013 and 22–24 October 2013. There had been minimal preparation of the site in line with current practice for restocking sites. The experimental site is included in the Inverewe Forest Plan, which included deer fencing of a larger area (2014) around the experimental site. Planting of this area was completed in early 2015, funded by NTS (National Trust for Scotland), although natural regeneration is also taking place.At each site, trees were planted in randomised blocks at 3 m × 3 m spacing. There are four randomised blocks in both FS and FE and three in FW. A guard row of Scots pine trees was planted around the periphery of the blocks and between blocks B and C at FS. Each block comprised one individual from each of eight (of the 10 sampled) families per 21 populations (168 trees). Although most families (N = 159) were represented at each of the three sites, families with insufficient trees (N = 9) were replaced in one site (FS) with a different family from the same population. Each experimental site was designed with redundancy such that, if thinning becomes necessary as the trees mature, then the systematic removal of trees (i.e. trees 1,3,5,7, etc of row 1, and 2,4,6,8, etc of row 2, 1,3,5,7,etc of row 3) will maintain a balanced design of the experiment, with sufficient family and population representation to provide an ongoing experiment with full geographic coverage.The field sites generally experience different climates, with FW typically warmer and wetter and with more growing degree days per year and a much longer growing season than both FE and FS (Table 2). The coldest site with the shortest growing season is generally FE.Table 2 Average climatic variables at field sites Glensaugh (FE), Inverewe (FW) and Yair (FS) from planting in 2012 until 2020. Climatic variables are derived from data provided by the Met Office (daily mean, minimum and maximum temperatures and monthly rainfall).Full size tablePhenotype assessmentsMaternal traitsFollowing seed collection, a range of traits were measured in the mother trees in order to control for maternal effects in subsequent measurements of their progeny (Table 3). For each mother tree, measurements of height and diameter at breast height (DBH) were taken, and ten cones were collected and assessed in detail. Cone width and length were measured prior to drying the cones (when they were still closed). Cone weight was measured post-drying. Seed removed from each cone was assessed for total weight (after wings had been removed) and for the count and percentage of seeds which were classed as viable (viable seed were those which had both a wing and an obvious seed). No further seed sorting was applied.Table 3 Traits assessed in mother trees, cones, seeds (dataset: Maternal), nursery seedlings (dataset: Nursery) and field trials (dataset: Field). Within the datasets, traits are recorded in a single column for each year using the format Code-Year (e.g. absolute height in 2008 = HA08) except for the maternal traits datasets which were all measured in 2007.Full size tableNursery traitsSeedling phenotype assessments were performed annually from 2007–2010 for three different trait types: phenology (budburst and growth cessation); form (total number of buds, needle length); cumulative growth (stem diameter and height, canopy width). Measurements of tree form and cumulative growth traits were taken after the end of each growing season. Phenology was assessed weekly during the spring and autumn of 2008 for budburst and growth cessation, respectively. Budburst was defined as the number of days from 31 March 2008 to the time when newly emerged green needles were observed (budburst stage 6: Fig. 3). In some seedlings in 2008, a secondary flush of growth occurred from terminal buds that had formed during the summer of that year. Therefore, growth cessation was defined retrospectively as the number of days from 10 September 2008 to the date when a terminal bud had formed on the leading shoot of the seedling, providing no further growth was observed either on the stem below that bud, nor from the bud itself. Canopy width (widest point) was measured at two perpendicular points in the horizontal plane. Needle length was measured for three needles per tree. Mortality was recorded each year from 2007 to 2010.Fig. 3Phenological stages of bud burst in Pinus sylvestris assessed in field trials. Inset numbers correspond to budburst stage. Budburst stage 1: bud dormant; 2: bud swelling and showing signs of linear expansion; 3: scales open at base revealing green tissue. Remaining bud remains encased by smooth bud scales; 4: scales open along length of shoot revealing green tissue and partially visible needles; 5: white tipped needles visible along length of the shoot; 6: green needles emerging away from the shoot (bottle brush appearance) along its entire length; 7: Needles have separated and next year’s terminal bud is usually formed and clearly visible.Full size imageField traitsTree height was measured in the field in the winter after each growing season from 2013 at FE and FW, and from 2014 to 2020 at all sites. Height was taken as the vertical measurement in cm from top bud straight to the ground. Basal stem diameter was measured at the end of the growing season for trees growing at FE and FW from 2014 to 2020 and for FS in 2020.Phenology assessments were performed in spring at each site from 2015 to 2019. Seven distinct stages of budburst (assessed on the terminal bud) were defined (Fig. 3) although only stages 4 to 6 are included in the dataset and considered for analysis due to high proportions of missing data for the early and late stages. Each tree was assessed for budburst stage at weekly intervals from early spring until budburst was complete. In order to allow comparisons within and among sites and years, the date at which each stage of budburst occurred was considered relative to 31 March of that year. For example, 25 May 2019 is recorded as 55 days since 31 March 2019. The duration of budburst (time taken to reach stage 6 from stage 4) was also estimated.When trees progressed through budburst stages rapidly, skipping a stage between assessments, a mean value was taken from the two assessment dates. For example, if a tree was at stage 4 on day 55 and was recorded as stage 6 at the next assessment on day 62, it is assumed to have reached stage 5 at day 58.5. More

Each of the 23 key variables can be used for analysis. To validate the dataset, we used five plant-related variables (diversity of order, family, genus, species and species endemic to China) to demonstrate the process of using the dataset for analysis as follows:(1) For the four variables of plant taxa “order”, “family”, “genus” and “species”, the similarity and difference in spatial distribution pattern of diversity of different taxa in the Qinling-Daba Mountains climate transition zone were analyzed. The spatial distribution pattern of the diversity of the four taxa is shown in Fig. 3, which is increasingly lower from south (low latitude) to north (high latitude). This result is consistent with the classical latitudinal gradient model of plant diversity. The boundary between higher diversity in the south and lower diversity in the north is roughly located in the area of Funiu Mountains in the eastern Qinling-Daba Mountains, Taibai Mountains in the central Qinling-Daba Mountains and Baishui River in the western Qinling-Daba Mountains. However, with the reduction in taxon scale, the spatial distribution pattern of diversity tends to be complex. Orders (Fig. 3a) and families (Fig. 3b) can be divided by lines, while genera (Fig. 3c) need thicker lines, and species (Fig. 3d) can only be divided by polygons. Figure 3 shows that the taxonomic groups of families are more clearly divided, while species can only be divided by staggered bands. Therefore, when dividing the north–south boundary, the family taxon scale is appropriate, whereas the species scale is more appropriate when studying the north–south transition zone.Fig. 3Spatial distribution of diversity of orders, families, genera and species. (a) The blue dotted line is basically the dividing line of the order diversity of 50 species. The order diversity to the north of the blue dotted line is lower than 50 species, and the order diversity to the south of the blue dotted line is higher than 50 species. (b) The blue dotted line is basically the dividing line of the family diversity of 150 species. The family diversity to the north of the blue dotted line is lower than 150 species, and the family diversity to the south of the blue dotted line is higher than 150 species. (c) The thicker blue dotted line is basically the dividing line of genus diversity of 578–681 species. The genus diversity to the north of the blue dotted line is lower than 578 species, and the genus diversity to the south of the blue dotted line is higher than 681 species. (d) The blue area is basically the dividing line of species diversity of 1385–1618 species. The species diversity to the north of the blue dotted line is lower than 1385 species, and the species diversity to the south of the blue dotted line is higher than 1618 species.Full size imageThe dataset can also count the orders, families and genera that appear in 58 nature reserves, indicating that these orders, families and genera are widely distributed in this area, while the orders, families and genera that only appear in a single nature reserve indicate that these taxa are unique to this nature reserve in this area, reflecting their locality and uniqueness, which is helpful to understanding the specific distribution of plants in detail. The relevant statistics are as follows:
There are 28 orders present in every nature reserve:
Liliales, Dipsacales, Lamiales, Fabales, Ericales, Poales, Saxifragales, Malpighiales, Malvales, Asterales, Fagales, Gentianales, Geraniales, Ranunculales, Rosales, Solanales, Apiales, Cornales, Brassicales, Caryophyllales, Dioscoreales, Santalales, Myrtales, Asparagales, Celastrales, Sapindales, Alismatales, and Boraginales.The order that only appears in one nature reserve is Petrosaviales, which appears in the Dabashan Nature Reserve in Chongqing.
There are 51 families present in every nature reserve:
Liliaceae, Primulaceae, Plantaginaceae, Lamiaceae, Euphorbiaceae, Cannabaceae, Juncaceae, Fabaceae, Poaceae, Elaeagnaceae, Betulaceae, Apocynaceae, Violaceae, Malvaceae, Crassulaceae, Campanulaceae, Asteraceae, Orchidaceae, Polygonaceae, Orobanchaceae, Onagraceae, Gentianaceae, Geraniaceae, Ranunculaceae, Rubiaceae, Rosaceae, Caprifoliaceae, Thymelaeaceae, Apiaceae, Cyperaceae, Cornaceae, Paeoniaceae, Brassicaceae, Amaryllidaceae, Caryophyllaceae, Rhamnaceae, Santalaceae, Asparagaceae, Celastraceae, Sapindaceae, Adoxaceae, Araliaceae, Berberidaceae, Hydrangeaceae, Scrophulariaceae, Convolvulaceae, Urticaceae, Salicaceae, Papaveraceae, Iridaceae, and Boraginaceae.There are 15 families that only appear in one nature reserve, as shown in Table 2.Table 2 Endemic families of the nature reserves in the Qinling-Daba Mountains and surrounding areas.Full size table
There are 54 genera present in every nature reserve:
Patrinia, Polygonum, Sanicula, Plantago, Allium, Delphinium, Euphorbia, Juncus, Cynanchum, Trigonotis, Artemisia, Sorbus, Polygonatum, Scutellaria, Cirsium, Viburnum, Ajuga, Viola, Galium, Geranium, Salix, Epilobium, Gentiana, Ranunculus, Malus, Acer, Rubia, Rosa, Torilis, Lonicera, Adenophora, Philadelphus, Cornus, Paeonia, Rhamnus, Rumex, Carex, Thalictrum, Asparagus, Carpesium, Clematis, Potentilla, Euonymus, Eleutherococcus, Berberis, Spiraea, Rubus, Populus, Vicia, Silene, Iris, Poa, Aster, and Buddleja.There were 225 genera that only appeared in one nature reserve, as shown in Figshare file 269.(2) For the “species endemic to China” variable of plants, we can see from the diversity distribution pattern of species endemic to China in this region (Fig. 4) that the number of endemic species in the Qinling-Daba Mountains is higher than that of species outside of the region, which reflects the strong transition zone in the Qinling-Daba Mountains. The variables of species endemic to China obtained from the Qinling-Daba Mountains and their surroundings were clustered by the Bray–Curtis dissimilarity measure70 and Ward’s minimum variance (the clustering method recommended for plant cluster analysis). The clustering results are shown in Fig. 5a. At the same time, the clustering results are displayed in space. Figure 5b shows that category 3 extends from the east outside the Qinling-Daba Mountains to the Baishuijiang Nature Reserve inside the western Qinling-Daba Mountains, which is consistent with the fact that the Qinling-Daba Mountains are an important ecogeographical “corridor” connecting the east and the west.Fig. 4Spatial distribution of diversity of species endemic to China in the Qinling-Daba Mountains and adjacent areas.Full size imageFig. 5(a) Clustering results of Ward’s connection aggregation of species endemic to China in 58 nature reserves. (b) Spatial distribution of clustering results of species endemic to China; the larger the dot and the darker the color, the earlier it is merged into this category, and the smaller the dot and the lighter the color, the later it is merged into this category.Full size image More

SamplesThe intestine and muscle samples from 22 wild boars were collected between September 4 and March 2, 2020, in Namie town in Fukushima prefecture. Furthermore, control intestine samples were collected from three wild boars in Hyogo prefecture. Each location is depicted in Fig. 1. In each case, after the licensed hunters slaughtered the wild boar to be exterminated, only the tissue was transferred to the study.Measurement of radioactivityRadioactivity in the muscle samples was determined by gamma-ray spectrometry using high-purity germanium (HPGe) detectors (Ortec Co., Oak Ridge, TN, USA), as described in our previous report3. Gamma rays from 137Cs were observed.Exposure dose estimationIn order to estimate internal and external dose rates of the wild boars according to the ICRP publication 10826, we supposed the shapes of wild boars as prolate spheroids whose long axis was to be their body lengths. The short axis was given from their weight assuming their specific gravities were the same
as water. The dose rates were calculated from the contribution of 137Cs, not including
natural radionuclides. The energy deposition to the spheroids by beta and gamma rays from radionuclides were calculated by the numerical simulation with the use of the Particle and Heavy Ion Transport code System (PHITS)27. For the sake of simplicity, we supposed the spheroids consisted of only muscle, which would give overestimated values because muscle contains more radio cesium than other organs. The external exposure dose was calculated from the air dose rates which were observed from the monitoring post near the boars captured place. The average values of the air dose rates were obtained from fitting observed data of two years with decay curve. The background due to the natural radionuclides was estimated to be 0.05 µGy/h which was observed before the Fukushima Daiichi accident, and was removed before the fittings. The half-lives of the air dose rates were 2000–3000 days depending on the environment. Assuming the external exposure dose was ascribed to the 137Cs included in the surface of the ground. The amount of the 137Cs was calculated so as to reproduce the observed air does rates. Since the maximum range of the beta ray from 137Cs is a few millimeters, almost all of the beta ray from inside the body should be absorbed in the boar’s body, but the beta ray from outside the body would stop in its fur. The beta rays contribute 100% to internal exposure dose but 0% to external one. Since the linear attenuation coefficient for gamma rays from 137Cs is 0.084 cm−1 = (12 cm)−1, some of the gamma rays cannot stop in the body depending on the size of the body. The numerical simulation suggested that 65–90 percent of the gamma rays from 137Cs inside the body would go out, and 40–65 percent of the gamma rays from 137Cs outside would go through the body.Pathological analysisA piece of the small intestine was fixed in 10% neutral formalin at 4 °C for 24–48 h. Then, paraffin blocks were prepared for pathomorphological examination using hematoxylin and eosin (HE) staining.Gene expression analysisTotal RNA was extracted from the whole tissue of the intestine using TRIzol Reagent (Life Technologies, Inc., Frederic, MD, USA) according to the manufacturer’s instructions. RNA concentration was measured using a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE, USA), and cDNA was synthesized using random primers and SuperScript II (Life Technologies, Inc.). Real-time PCR for IFN-γ, TLR3, and CyclinG1 was performed using Brilliant SYBR Green QPCR Master Mix III (Stratagene, La Jolla, CA, USA) with an AriaMx system (Agilent Technologies, Santa Clara, CA, USA). Primer sequences were designed using Primer-BLAST with sequences obtained from GenBank as described in the previous report4. Amplification conditions were 95 °C for 3 min, 40 cycles at 95 °C for 5 s, and 60 °C for 20 s. Fluorescence signals measured during the amplification were analyzed. Ribosomal RNA primers were used as an internal control, and all data were normalized to constitutive rRNA values. Quantitative differences between the groups were analyzed using the AriaMx software (Agilent Technologies).Statistical analysisAll data are presented as mean ± standard error (SE) for each treatment group. Differences in mRNA expression among the groups were determined using the unpaired t-test with Welch’s correction. (Prism: GraphPad Software Inc., La Jolla, CA, USA). Differences were considered to be statistically significant at a P value of  More