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

DataWe tested our models on ten publicly available datasets. In Fig. 4 we show examples of images from each of the datasets. When applicable, the training and test splits were kept the same as in the original dataset. For example, the ZooScan, Kaggle, EILAT, and RSMAS datasets lack a specific training and test set; in these cases, benchmarks come from k-fold cross-validation51,52, and we followed the exact same procedures in order to allow for a fair comparison.Figure 4Examples of images from each of the datasets.(a) RSMAS (b) EILAT (c) ZooLake (d) WHOI (e) Kaggle (f) ZooScan (g) NA-Birds (h) Stanford dogs (i) SriLankan Beetles (j) Florida Wildtrap.Full size imageRSMAS This is a small coral dataset of 766 RGB image patches with a size of (256times 256) pixels each53. The patches were cropped out of bigger images obtained by the University of Miami’s Rosenstiel School of Marine and Atmospheric Sciences. These images were captured using various cameras in various locations. The data is separated into 14 unbalanced groups and whose labels correspond to the names of the coral species in Latin. The current SOTA for the classification of this dataset is by52. They use the ensemble of best performing 11 CNN models. The best models were chosen based on sequential forward feature selection (SFFS) approach. Since an independent test is not available, they make use of 5-fold cross-validation for benchmarking the performances.EILAT This is a coral dataset of 1123 64-pixel RGB image patches53 that were created from larger images that were taken from coral reefs near Eilat in the Red sea. The image dataset is partitioned into eight classes, with an unequal distribution of data. The names of the classes correspond to the shorter version of the scientific names of the coral species. The current SOTA52 for the classification of this dataset uses the ensemble of best performing 11 CNN models similar to RSMAS dataset and 5-fold cross-validation for benchmarking the performances.ZooLake This dataset consists of 17943 images of lake plankton from 35 classes, acquired using a Dual-magnification Scripps Plankton Camera (DSPC) in Lake Greifensee (Switzerland) between 2018 and 2020 14,54. The images are colored, with a black background and an uneven class distribution. The current SOTA22 on this dataset is based on a stacking ensemble of 6 CNN models on an independent test set.WHOI This dataset 55 contains images of marine plankton acquired by Image FlowCytobot56, from Woods Hole Harbor water. The sampling was done between late fall and early spring in 2004 and 2005. It contains 6600 greyscale images of different sizes, from 22 manually categorized plankton classes with an equal number of samples for each class. The majority of the classes belonging to phytoplankton at genus level. This dataset was later extended to include 3.4M images and 103 classes. The WHOI subset that we use was previously used for benchmarking plankton classification models51,52. The current SOTA22 on this dataset is based on average ensemble of 6 CNN models on an independent test set.Kaggle-plankton The original Kaggle-plankton dataset consists of plankton images that were acquired by In-situ Ichthyoplankton Imaging System (ISIIS) technology from May to June 2014 in the Straits of Florida. The dataset was published on Kaggle (https://www.kaggle.com/c/datasciencebowl) with images originating from the Hatfield Marine Science Center at Oregon State University. A subset of the original Kaggle-plankton dataset was published by51 to benchmark the plankton classification tasks. This subset comprises of 14,374 greyscale images from 38 classes, and the distribution among classes is not uniform, but each class has at least 100 samples. The current SOTA22 uses average ensemble of 6 CNN models and benchmarks the performance using 5-fold cross-validation.ZooScan The ZooScan dataset consists of 3771 greyscale plankton images acquired using the Zooscan technology from the Bay of Villefranche-sur-mer57. This dataset was used for benchmarking the classification models in previous plankton recognition papers51,52. The dataset consists of 20 classes with a variable number of samples for each class ranging from 28 to 427. The current SOTA22 uses average ensemble of 6 CNN models and benchmarks the performance using 2-fold cross-validation.NA-Birds NA-Birds58 is a collection of 48,000 captioned pictures of North America’s 400 most often seen bird species. For each species, there are over 100 images accessible, with distinct annotations for males, females, and juveniles, totaling 555 visual categories. The current SOTA59 called TransFG modifies the pure ViT model by adding contrastive feature learning and part selection module that replaces the original input sequence to the transformer layer with tokens corresponding to informative regions such that the distance of representations between confusing subcategories can be enlarged. They make use of an independent test set for benchmarking the model performances.Stanford Dogs The Stanford Dogs dataset comprises 20,580 color images of 120 different dog breeds from all around the globe, separated into 12,000 training images and 8,580 testing images60. The current SOTA59 makes use of modified ViT model called TransFG as explained above in NA-Birds dataset. They make use of an independent test set for benchmarking the model performances.Sri Lankan Beetles The arboreal tiger beetle data61 consists of 380 images that were taken between August 2017 and September 2020 from 22 places in Sri Lanka, including all climatic zones and provinces, as well as 14 districts. Tricondyla (3 species), Derocrania (5 species), and Neocollyris (1 species) were among the nine species discovered, with six of them being endemic . The current SOTA61 makes use of CNN-based SqueezeNet architecture and was trained using pre-trained weights of ImageNet. The benchmarking of the model performances was done on an independent test set.Florida Wild Traps The wildlife camera trap62 classification dataset comprises 104,495 images with visually similar species, varied lighting conditions, skewed class distribution, and samples of endangered species, such as Florida panthers. These were collected from two locations in Southwestern Florida. These images are categorized in to 22 classes. The current SOTA62 makes use of CNN-based ResNet-50 architecture and the performance of the model was benchmarked on an independent test set.ModelsVision transformers (ViTs)31 are an adaptation to computer vision of the Transformers, which were originally developed for natural language processing30. Their distinguishing feature is that, instead of exploiting translational symmetry, as CNNs do, they have an attention mechanism which identifies the most relevant part of an image. ViTs have recently outperformed CNNs in image classification tasks where vast amounts of training data and processing resources are available30,63. However, for the vast majority of use cases and consumers, where data and/or computational resources are limiting, ViTs are essentially untrainable, even when the network architecture is defined and no architectural optimization is required. To settle this issue, Data-efficient Image Transformers (DeiTs) were proposed32. These are transformer models that are designed to be trained with much less data and with far less computing resources32. In DeiTs, the transformer architecture has been modified to allow native distillation64, in which a student neural network learns from the results of a teacher model. Here, a CNN is used as the teacher model, and the pure vision transformer is used as the student network. All the DeiT models we report on here are DeiT-Base models32. The ViTs are ViT-B16, ViT-B32, and ViT-L32 models31.ImplementationTo train our models, we used transfer learning65: we took a model that was already pre-trained on the ImageNet43 dataset, changed the last layers depending on the number of classes, and then fine-tuned the whole network with a very low learning rate. All the models were trained with two Nvidia GTX 2080Ti GPUs.DeiTs We used DeiT-Base32 architecture, using the Python package TIMM66, which includes many of the well-known deep learning architectures, along with their pre-trained weights computed from the ImageNet dataset43. We resized the input images to 224 x 224 pixels and then, to prevent the model from overfitting at the pixel level and help it generalize better, we employed typical image augmentations during training such as horizontal and vertical flips, rotations up to 180 degrees, small zoom up’s to 20%, a small Gaussian blur, and shearing up to 10%. To handle class imbalance, we used class reweighting, which reweights errors on each example by how present that class is in the dataset67. We used sklearn utilities68 to calculate the class weights which we employed during the training phase.The training phase started with a default pytorch69 initial conditions (Kaiming uniform initializer), an AdamW optimizer with cosine annealing70, with a base learning rate of (10^{-4}), and a weight decay value of 0.03, batch size of 32 and was supervised using cross-entropy loss. We trained with early stopping, interrupting training if the validation F1-score did not improve for 5 epochs. The learning rate was then dropped by a factor of 10. We iterated until the learning rate reached its final value of (10^{-6}). This procedure amounted to around 100 epochs in total, independent of the dataset. The training time varied depending on the size of the datasets. It ranged between 20min (SriLankan Beetles) to 9h (Florida Wildtrap). We used the same procedure for all the datasets: no extra time was needed for hyperparameter tuning.ViTs We implemented the ViT-B16, ViT-B32 and ViT-L32 models using the Python package vit-keras (https://github.com/faustomorales/vit-keras), which includes pre-trained weights computed from the ImageNet43 dataset and the Tensorflow library71.First, we resized input images to 128 × 128 and employed typical image augmentations during training such as horizontal and vertical flips, rotations up to 180 degrees, small zooms up to 20%, small Gaussian blur, and shearing up to 10%. To handle class imbalance, we calculated the class weights and use them during the training phase.Using transfer learning, we imported the pre-trained model and froze all of the layers to train the model. We removed the last layer, and in its place we added a dense layer with (n_c) outputs (being (n_c) the number of classes), was preceded and followed by a dropout layer. We used the Keras-tuner72 with Bayesian optimization search73 to determine the best set of hyperparameters, which included the dropout rate, learning-rate, and dense layer parameters (10 trials and 100 epochs). After that, the model with the best hyperparameters was trained with a default tensorflow71 initial condition (Glorot uniform initializer) for 150 epochs using early stopping, which involved halting the training if the validation loss did not decrease after 50 epochs and retaining the model parameters that had the lowest validation loss.CNNs CNNs included DenseNet38, MobileNet39, EfficientNet-B240, EfficientNet-B540, EfficientNet-B640, and EfficientNet-B740 architectures. We followed the training procedure described in Ref.22, and carried out the training in tensorflow.Ensemble learningWe adopted average ensembling, which takes the confidence vectors of different learners, and produces a prediction based on the average among the confidence vectors. With this procedure, all the individual models contribute equally to the final prediction, irrespective of their validation performance. Ensembling usually results in superior overall classification metrics and model robustness74,75.Given a set of n models, with prediction vectors (vec c_i~(i=1,ldots ,n)), these are typically aggregated through an arithmetic average. The components of the ensembled confidence vector (vec c_{AA}), related to each class (alpha ) are then$$begin{aligned} c_{AA,alpha } = frac{1}{n}sum _{i=1}^n c_{i,alpha },. end{aligned}$$
(2)
Another option is to use a geometric average,$$begin{aligned} c_{GA,alpha } = root n of {prod _{i=1}^n c_{i,alpha }},. end{aligned}$$
(3)
We can normalize the vector (vec c_g), but this is not relevant, since we are interested in its largest component, (displaystyle max _alpha (c_{GA,alpha })), and normalization affects all the components in the same way. As a matter of fact, also the nth root does not change the relative magnitude of the components, so instead of (vec c_{GA}) we can use a product rule: (displaystyle max _alpha (c_{GA,alpha })=max _alpha (c_{PROD,alpha })), with (displaystyle c_{PROD,alpha } = prod _{i=1}^n c_{i,alpha }).While these two kinds of averaging are equivalent in the case of two models and two classes, they are generally different in any other case33. For example, it can easily be seen that the geometric average penalizes more strongly the classes for which at least one learner has a very low confidence value, a property that was termed veto mechanism36 (note that, while in Ref.36 the term veto is used when the confidence value is exactly zero, here we use this term in a slightly looser way). More

Buxó, R. & Piqué, R. Arqueobotánica: Los Usos de las Plantas en la Península Ibérica. (Grupo Planeta GBS, 2008).Miller, N. F., Spengler, R. N. & Frachetti, M. Millet cultivation across Eurasia: Origins, spread, and the influence of seasonal climate. Holocene 26, 1566–1575 (2016).ADS 

Google Scholar 
James, T. K., Rahman, A., McGill, C. R. & Trivedi, P. D. Biology and survival of broomcorn millet (Panicum miliaceum) seed. N. Z. Plant Prot. 64, 142–148 (2011).
Google Scholar 
Kirleis, W., Dal Corso, M. & Filipović, D. Millet and What Else?: The Wider Context of the Adoption of Millet Cultivation in Europe. vol. 14 (Sidestone Press, 2022).Sherratt, A. Water, soil and seasonality in early cereal cultivation. World Archaeol. 11, 313–330 (1980).
Google Scholar 
Rachie, K. O. The Millets: Importance, Utilization and Outlook. 74 (International Crops Research Institute for the Semi-Arid Tropics, 1975).Moreno-Larrazabal, A., Teira-Brión, A., Sopelana-Salcedo, I., Arranz-Otaegui, A. & Zapata, L. Ethnobotany of millet cultivation in the north of the Iberian Peninsula. Veg. Hist. Archaeobot. 24, 541–554 (2015).
Google Scholar 
Liu, L. et al. The origins of specialized pottery and diverse alcohol fermentation techniques in Early Neolithic China. Proc. Natl. Acad. Sci. USA 116, 12767–12774 (2019).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Tereso, J. P. et al. Agriculture in NW Iberia during the Bronze Age: A review of archaeobotanical data. J. Archaeol. Sci. Rep. 10, 44–58 (2016).
Google Scholar 
Liu, C., Kong, Z. & Lang, S. D. A discussion on agricultural and botanical remains and the human ecology of Dadiwan site, in Chinese. Zhongyuan Wenwu 4, 25–29 (2004).
Google Scholar 
Zhao, Z. New archaeobotanic data for the study of the origins of agriculture in China. Curr. Anthropol. 52, S295–S306 (2011).
Google Scholar 
Crawford, G. W., Xuexiang, C., Fengshi, L. & Jianhua, W. A Preliminary analysis on plant remains of the Yuezhuang site in Changqing District, Jinan City, Shandong Province. Jianghan Archaeol. 2, 107–113 (2013).
Google Scholar 
Frachetti, M. D. Multiregional emergence of mobile pastoralism and nonuniform institutional complexity across Eurasia. Curr. Anthropol. 53, 2–38. https://doi.org/10.1086/663692 (2012).Article 

Google Scholar 
Ventresca Miller, A. R. & Makarewicz, C. A. Intensification in pastoralist cereal use coincides with the expansion of trans-regional networks in the Eurasian Steppe. Sci. Rep. 9, 8363 (2019).ADS 
PubMed 
PubMed Central 

Google Scholar 
Jones, M. et al. Food globalisation in prehistory: The agrarian foundations of an interconnected continent. J. Br. Acad. 4, 73–87 (2016).
Google Scholar 
Spengler, R. et al. Early agriculture and crop transmission among Bronze Age mobile pastoralists of Central Eurasia. Proc. Biol. Sci. 281, 20133382 (2014).PubMed 
PubMed Central 

Google Scholar 
Hermes, T. R. et al. Early integration of pastoralism and millet cultivation in Bronze Age Eurasia. Proc. Biol. Sci. 286, 20191273 (2019).CAS 
PubMed 
PubMed Central 

Google Scholar 
Frachetti, M. D., Spengler, R. N., Fritz, G. J. & Maryashev, A. N. Earliest direct evidence for broomcorn millet and wheat in the central Eurasian steppe region. Antiquity 84, 993–1010 (2010).
Google Scholar 
Motuzaite-Matuzeviciute, G., Richard, A. S., Hunt, H. V., Liu, X. & Jones, M. K. The early chronology of broomcorn millet (Panicum Miliaceum) in Europe. Antiquity 338, 1073–1085 (2013).
Google Scholar 
Filipović, D. et al. New AMS 14C dates track the arrival and spread of broomcorn millet cultivation and agricultural change in prehistoric Europe. Sci. Rep. 10, 13698 (2020).ADS 
PubMed 
PubMed Central 

Google Scholar 
Hunt, H. V. et al. Millets across Eurasia: Chronology and context of early records of the genera Panicum and Setaria from archaeological sites in the Old World. Veg. Hist. Archaeobot. 17, 5–18 (2008).PubMed 
PubMed Central 

Google Scholar 
Brudenell, M., Fosberry, R., Phillips, T. & Kwiatkowska, M. Early cultivation of broomcorn millet in southern Britain: Evidence from the Late Bronze Age settlement site of Old Catton, Norfolk. Antiquity 2022, 1–6 (2022).
Google Scholar 
Weber, S. A. & Fuller, D. Q. Millets and their role in early agriculture. Pragdhara 18, 69–90 (2007).
Google Scholar 
Shelton, C. P. & White, C. E. The hand-pump flotation system: A new method for archaeobotanical recovery. J. Field Archaeol. 35, 316–326 (2010).
Google Scholar 
Barboff, M. Le millet au Portugal. In Millet– Hirse–Millet. Actes du Congres d’Aizenay (ed. Hörandner, E.) 113–122 (Grazer Beitra¨ge zur 731 europa¨ischen Ethnologie, 1995).Reddy, S. N. If the Threshing Floor Could Talk: Integration of Agriculture and Pastoralism during the Late Harappan in Gujarat, India. J. Anthropol. Archaeol. 16, 162–187 (1997).
Google Scholar 
Dayakar Rao, B. et al. Nutritional and health benefits of millets. In ICAR_Indian Institute of Millets Research (IIMR), Rajendranagar, Hyderabad 112 (2017).Mariotti-Lippi, M., Pisaneschi, L., Sarti, L., Lari, M. & Moggi-Cecchi, J. Insights into the Copper-Bronze Age diet in Central Italy: Plant microremains in dental calculus from Grotta dello Scoglietto (Southern Tuscany, Italy). J. Archaeol. Sci. Rep. 15, 30–39 (2017).
Google Scholar 
Lu, H. et al. Phytoliths analysis for the discrimination of Foxtail millet (Setaria italica) and Common millet (Panicum miliaceum). PLoS ONE 4, e4448 (2009).ADS 
PubMed 
PubMed Central 

Google Scholar 
Lucarini, G., Radini, A., Barton, H. & Barker, G. The exploitation of wild plants in Neolithic North Africa. Use-wear and residue analysis on ground stone tools from the Farafra Oasis, Egypt. Quat. Int. 410, 77–92 (2016).
Google Scholar 
Madella, M., Lancelotti, C. & García-Granero, J. J. Millet microremains—an alternative approach to understand cultivation and use of critical crops in Prehistory. Archaeol. Anthropol. Sci. 8, 17–28 (2016).
Google Scholar 
Yang, X. et al. From the modern to the archaeological: Starch grains from millets and their wild relatives in China. J. Archaeol. Sci. 39, 247–254 (2012).
Google Scholar 
Lightfoot, E., Liu, X. & Jones, M. K. Why move starchy cereals? A review of the isotopic evidence for prehistoric millet consumption across Eurasia. World Archaeol. 45, 574–623 (2013).
Google Scholar 
Armendariz, A. In Las cuevas sepulcrales del País Vasco. (Tesis Doctoral Inédita, Universidad del País Vasco-Euskal Herriko Unibertsitatea, 1992).Vazquez-Varela, J. M. El cultivo del mijo, (Panicum miliaceum, L.), en la cultura castreña del noroeste de la peninsula iberica. Cuad. Estud. Gallegos 1, 65–73 (1994).
Google Scholar 
Patterson, N. et al. Large-scale migration into Britain during the Middle to Late Bronze Age. Nature 601, 588–594 (2021).ADS 
PubMed 
PubMed Central 

Google Scholar 
Olalde, I. et al. The genomic history of the Iberian Peninsula over the past 8000 years. Science 363, 1230–1234 (2019).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Arias, P. & Armendariz, A. Aproximación a la Edad del Bronce en la región cantábrica. A Idade do Bronce en Galicia: novas perspectivas. Cadernos Semin. Sargadelos 77, 47–80 (1998).
Google Scholar 
DeNiro, M. J. Postmortem preservation and alteration of in vivo bone collagen isotope ratios in relation to palaeodietary reconstruction. Nature 317, 806–809 (1985).ADS 
CAS 

Google Scholar 
Ambrose, S. H. Preparation and characterization of bone and tooth collagen for isotopic analysis. J. Archaeol. Sci. 17, 431–451 (1990).
Google Scholar 
van Klinken, G. J. Bone collagen quality indicators for palaeodietary and radiocarbon measurements. J. Archaeol. Sci. 26, 687–695 (1999).
Google Scholar 
Nehlich, O. & Richards, M. P. Establishing collagen quality criteria for sulphur isotope analysis of archaeological bone collagen. Archaeol. Anthropol. Sci. 1, 59–75 (2009).
Google Scholar 
Cristiani, E., Radini, A., Edinborough, M. & Borić, D. Dental calculus reveals Mesolithic foragers in the Balkans consumed domesticated plant foods. Proc. Natl. Acad. Sci. USA 113, 10298–10303 (2016).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Henry, A. G. & Piperno, D. R. Using plant microfossils from dental calculus to recover human diet: A case study from Tell al-Raqā’I, Syria. J. Archaeol. Sci. 35, 1943–1950 (2008).
Google Scholar 
Minagawa, M. & Wada, E. Stepwise enrichment of 15N along food chains: Further evidence and the relation between δ15N and animal age. Geochim. Cosmochim. Acta 48, 1135–1140 (1984).ADS 
CAS 

Google Scholar 
Hedges, R. E. M. & Reynard, L. M. Nitrogen isotopes and the trophic level of humans in archaeology. J. Archaeol. Sci. 34, 1240–1251 (2007).
Google Scholar 
López-Costas, O., Müldner, G. & Martínez-Cortizas, A. Diet and lifestyle in Bronze Age Northwest Spain: The collective burial of Cova do Santo. J. Archaeol. Sci. 55, 209–218 (2015).
Google Scholar 
Jones, J. R. et al. Investigating prehistoric diet and lifeways of early farmers in central northern Spain (3000–1500 CAL BC) using stable isotope techniques. Archaeol. Anthropol. Sci. 11, 3979–3994 (2019).
Google Scholar 
DeNiro, M. J. & Epstein, S. Influence of diet on the distribution of carbon isotopes in animals. Geochim. Cosmochim. Acta 42, 495–506 (1978).ADS 
CAS 

Google Scholar 
O’Leary, M. H. Carbon isotope fractionation in plants. Phytochemistry 20, 553–567 (1981).
Google Scholar 
Chisholm, B. S., Nelson, D. E. & Schwarcz, H. P. Stable-carbon isotope ratios as a measure of marine versus terrestrial protein in ancient diets. Science 216, 1131–1132 (1982).ADS 
CAS 
PubMed 

Google Scholar 
de Blas Cortina, M. Á. De la caverna al lugar fortificado: Una mirada a la edad del bronce en el territorio Astur-Cántabro. Quad. Prehist. Arqueol. Castelló 29, 105–134 (2011).
Google Scholar 
Nehlich, O. The application of sulphur isotope analyses in archaeological research: A review. Earth-Sci. Rev. 142, 1–17 (2015).ADS 
CAS 

Google Scholar 
Richards, M. P., Fuller, B. T. & Hedges, R. E. M. Sulphur isotopic variation in ancient bone collagen from Europe: Implications for human palaeodiet, residence mobility, and modern pollutant studies. Earth Planet. Sci. Lett. 191, 185–190 (2001).ADS 
CAS 

Google Scholar 
González-Rabanal, B. et al. Diet, mobility and death of Late Neolithic and Chalcolithic groups of the Cantabrian Region (northern Spain). A multidisciplinary approach towards studying the Los Avellanos I and II burial caves. J. Archaeol. Sci. Rep. 34, 1–13 (2020).
Google Scholar 
McGovern, P. E. et al. Fermented beverages of pre- and proto-historic China. Proc. Natl. Acad. Sci. USA 101, 17593–17598 (2004).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Fernández-Crespo, T., Ordoño, J., Bogaard, A., Llanos, A. & Schulting, R. A snapshot of subsistence in Iron Age Iberia: The case of La Hoya village. J. Archaeol. Sci. Rep. 28, 1–10 (2019).
Google Scholar 
Hedges, R. E. M. On bone collagen?apatite-carbonate isotopic relationships. Int. J. Osteoarchaeol. 13, 66–79 (2003).
Google Scholar 
Arias, P. Determinaciones de isótopos estables en restos humanos de la región Cantábrica. Aportación al estudio de la dieta de las poblaciones del Mesolítico y el Neolítico. Munibe Antropol.-Arkeol. 57, 359–374 (2005).
Google Scholar 
Palencia-Madrid, L. et al. Ancient mitochondrial lineages support the prehistoric maternal root of Basques in Northern Iberian Peninsula. Eur. J. Hum. Genet. 25, 631–636 (2017).CAS 
PubMed 
PubMed Central 

Google Scholar 
Fernández-Crespo, T., Mujika, J. A. & Ordoño, J. Aproximación al patrón alimentario de los inhumados en la cista de la Edad del Bronce de Ondarre (Aralar, Guipúzcoa) a través del análisis de isótopos estables de carbono y nitrógeno sobre colágeno óseo. Trab. Prehist. 73, 325–334 (2016).
Google Scholar 
Higuero Pliego, A. In Análisis Isotópico de Carbono y Nitrógeno en Secuencias de Dentina y de Estroncio en Esmalte Procedente de Restos Humanos Prehistóricos de la Cueva de Los Canes (Cabrales, Asturias). (Tesis Doctoral Inédita, Universidad de Cantabria, 2020).Teira-Brión, A. Traditional millet cultivation in the Iberian Peninsula: Ethnoarchaeological reflections through the lens of social relations and economic concerns. In (eds. Kirleis, W. et al.) Millet and What Else?: The Wider Context of the Adoption of Millet Cultivation in Europe vol. 14 263–278 (Sidestone Press, 2022).Pechenkina, E. A., Ambrose, S. H., Xiaolin, M. & Benfer, R. A. Reconstructing northern Chinese Neolithic subsistence practices by isotopic analysis. J. Archaeol. Sci. 32, 1176–1189 (2005).
Google Scholar 
Hu, Y. et al. Palaeodietary study of Sanxingcun Site, Jintan, Jiangsu. Chin. Sci. Bull. 52, 660–664 (2007).
Google Scholar 
Motuzaite-Matuzeviciute, G., Ananyevskaya, E., Sakalauskaite, J., Soltobaev, O. & Tabaldiev, K. The integration of millet into the diet of Central Asian populations in the third millennium BC. Antiquity 96, 560–574 (2022).
Google Scholar 
Herrscher, E. et al. The origins of millet cultivation in the Caucasus: Archaeological and archaeometric approaches. Préhistoires Méditerr. 2018, 6 (2018).
Google Scholar 
Tafuri, M. A., Craig, O. E. & Canci, A. Stable isotope evidence for the consumption of millet and other plants in Bronze Age Italy. Am. J. Phys. Anthropol. 139, 146–153 (2009).PubMed 

Google Scholar 
Varalli, A., Moggi-Cecchi, J., Moroni, A. & Goude, G. Dietary variability during bronze age in central Italy: First results. Int. J. Osteoarchaeol. 26, 431–446 (2016).
Google Scholar 
Goude, G., Rey, L., Toulemonde, F., Cervel, M. & Rottier, S. Dietary changes and millet consumption in northern France at the end of Prehistory: Evidence from archaeobotanical and stable isotope data. Environ. Archaeol. 22, 268–282 (2017).
Google Scholar 
Fernández-Crespo, T., Ordoño, J., Llanos, A. & Schulting, R. J. Make a desert and call it peace: Massacre at the Iberian Iron Age village of La Hoya. Antiquity 94, 1245–1262 (2020).
Google Scholar 
Lu, H. et al. Earliest domestication of common millet (Panicum miliaceum) in East Asia extended to 10,000 years ago. Proc. Natl. Acad. Sci. USA 106, 7367–7372 (2009).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Li, M. et al. Starch grains from dental calculus reveal ancient plant foodstuffs at Chenqimogou site, Gansu Province. Sci. China Earth Sci. 53, 694–699 (2010).ADS 
CAS 

Google Scholar 
Zhang, J., Lu, H., Wu, N., Yang, X. & Diao, X. Phytolith analysis for differentiating between foxtail millet (Setaria italica) and Green Foxtail (Setaria viridis). PLoS ONE 6, e19726 (2011).ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Ge, Y. et al. Phytolith analysis for the identification of barnyard millet (Echinochloa sp.) and its implications. Archaeol. Anthropol. Sci. 10, 61–73 (2018).
Google Scholar 
Tao, D., Zhang, G., Zhou, Y. & Zhao, H. Investigating wheat consumption based on multiple evidences: Stable isotope analysis on human bone and starch grain analysis on dental calculus of humans from the Laodaojing cemetery, Central Plains, China. Int. J. Osteoarchaeol. 30, 594–606 (2020).
Google Scholar 
Bucchi, A., Burguet-Coca, A., Expósito, I., Aceituno-Bocanegra, F. J. & Lozano, M. Comparisons between methods for analyzing dental calculus samples from El Mirador cave (Sierra de Atapuerca, Spain). Archaeol. Anthropol. Sci. 11, 6305–6314 (2019).
Google Scholar 
Cristiani, E. et al. Wild cereal grain consumption among Early Holocene foragers of the Balkans predates the arrival of agriculture. Elife 10, 1–37 (2021).
Google Scholar 
Bocanegra, F. J. A. & Sáez, J. A. L. Caracterización morfológica de almidones de los géneros Triticum y Hordeum en la Península Ibérica. Trabprehist 69, 332–348 (2012).
Google Scholar 
Hardy, K., Buckley, S. & Copeland, L. Pleistocene dental calculus: Recovering information on Paleolithic food items, medicines, paleoenvironment and microbes. Evol. Anthropol. 27, 234–246 (2018).PubMed 

Google Scholar 
López-Dóriga, I. In The use of plants during the Mesolithic and the Neolithic in the Atlantic coast of the Iberian peninsula. (Tesis Doctoral Inédita, Universidad de Cantabria, 2016).Nava, A. et al. Multipronged dental analyses reveal dietary differences in last foragers and first farmers at Grotta Continenza, central Italy (15,500–7000 BP). Sci. Rep. 11, 1–14 (2021).
Google Scholar 
Pyankov, V. I., Ziegler, H., Akhani, H., Deigele, C. & Lüttge, U. European plants with C4 photosynthesis: Geographical and taxonomic distribution and relations to climate parameters. Bot. J. Linn. Soc. 163, 283–304 (2010).
Google Scholar 
Zapata, L. In La explotación de los recursos vegetales y el origen de la agricultura en el País Vasco. (Tesis Doctora Inédita, Universidad del País Vasco, 2002).Figueiral, I., de-Jesus-Sanches, M. & Cardoso, J. L. Crasto de Palheiros (Murça, NE Portugal, 3rd – 1st millennium BC): From archaeological remains to ordinary life. Estudos Quat. 17, 13–28 (2017).
Google Scholar 
Bettencourt, A. M. S. O povoado da Idade do Bronze da Sola, Braga, norte de Portugal. Cadernos Arqueol. 9, 29–44 (2000).
Google Scholar 
Jesus, A., Tereso, J. P. & Gaspar, R. Interpretative trajectories towards the understanding of negative features using Terraço das Laranjeiras Bronze Age site as a case study. J. Archaeol. Sci. Rep. 30, 1–14 (2020).
Google Scholar 
Alonso-Martínez, N. Registro arqueobotánico de Cataluña occidental durante el II y I milenio a.n.e.. Complutum 11, 221–238 (2000).
Google Scholar 
Tarongi-Chavarri, M. Análisis comparativo de los estudios carpológicos de la Depresión del Ebro durante el III y I milenio a. C. Un estado de la cuestión. Rev. d’arqueologia Ponent 27, 41–59 (2017).
Google Scholar 
Stika, H.-P. & Heiss, A. G. Plant cultivation in the Bronze Age. In The Oxford Handbook of the European Bronze Age (eds. Fokkens, H. & Harding, A.) 348–369 (2013).González-y-Fernández-Valles, J. M. Temas de toponimia asturiana. Archivum 21, 121–140 (1971).
Google Scholar 
de Carvallo, L. A. In Antiguedades y Cosas Memorables del Principado de Asturias. (Julian de Paredes, 1695).MacKinnon, A. T., Passalacqua, N. V. & Bartelink, E. J. Exploring diet and status in the Medieval and Modern periods of Asturias, Spain, using stable isotopes from bone collagen. Archaeol. Anthropol. Sci. 11, 3837–3855. https://doi.org/10.1007/s12520-019-00819-2 (2019).Article 

Google Scholar 
Renfrew, J. M. Palaeoethnobotany: The Prehistoric Food Plants of the Near East and Europe (Columbia University Press, 1973).
Google Scholar 
Bronk Ramsey, C. Bayesian Analysis of radiocarbon dates. Radiocarbon 51, 337–360 (2009).
Google Scholar 
Reimer, P. J. et al. The IntCal20 Northern hemisphere radiocarbon age calibration curve (0–55 cal kBP). Radiocarbon 62, 725–757 (2020).CAS 

Google Scholar 
Richards, M. P. & Hedges, R. E. M. Stable isotope evidence for similarities in the types of marine foods used by late mesolithic humans at sites along the Atlantic Coast of Europe. J. Archaeol. Sci. 26, 717–722 (1999).
Google Scholar 
Sabin, S. & James, A. In Dental Calculus Field-Sampling Protocol (Sabin version) v2 (protocols.io.bqecmtaw). (2020). https://doi.org/10.17504/protocols.io.bqecmtaw.Cristiani, E. et al. Dental calculus and isotopes provide direct evidence of fish and plant consumption in Mesolithic Mediterranean. Sci. Rep. 8, 8147 (2018).ADS 
PubMed 
PubMed Central 

Google Scholar 
Fiorin, E. et al. Combining dental calculus with isotope analysis in the Alps: New evidence from the Roman and medieval cemeteries of Lamon. Italy. Quat. Int. https://doi.org/10.1016/j.quaint.2021.11.022 (2021).Article 

Google Scholar  More

Gibson, L. et al. Primary forests are irreplaceable for sustaining tropical biodiversity. Nature 478, 378–381. https://doi.org/10.1038/nature10425 (2011).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Pimm, S. L. & Raven, P. Extinction by numbers. Nature 403, 843–845. https://doi.org/10.1038/35002708 (2000).Article 
ADS 
CAS 
PubMed 

Google Scholar 
​FAO. Global Forest Resources Assessment 2020: Main report. 184p (Rome, Italy, 2020).Harvey, C. A. et al. Integrating agricultural landscapes with biodiversity conservation in the Mesoamerican hotspot. Conserv Biol 22, 8–15 (2008).Article 
PubMed 

Google Scholar 
Brouwer, F. & McCarl, B. Agriculture and climate beyond 2015: A New Perspective on Future Land Use Patterns. (2006).Redo, D. J., Grau, H. R., Aide, T. M. & Clark, M. L. Asymmetric forest transition driven by the interaction of socioeconomic development and environmental heterogeneity in Central America. Proc. Natl. Acad. Sci. 109, 8839–8844. https://doi.org/10.1073/pnas.1201664109 (2012).Article 
ADS 
PubMed 
PubMed Central 

Google Scholar 
Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B. & Kent, J. Biodiversity hotspots for conservation priorities. Nature 403, 853–858. https://doi.org/10.1038/35002501 (2000).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Declerck, F. et al. Biodiversity conservation in human-modified landscapes of Mesoamerica: Past, present and future. Biol. Conserv. 143, 2301–2313. https://doi.org/10.1016/j.biocon.2010.03.026 (2010).Article 

Google Scholar 
Miller, K., Chang, E. & Johnson, N. Defining Common Ground for the Mesoamerican Biological Corridor (World Resources Institute, Washington, 2001).
Google Scholar 
Fischer, J. et al. Conservation: Limits of land sparing. Science 334, 593–593. https://doi.org/10.1126/science.334.6056.593-a (2011).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Morecroft, M. D. et al. Agricultural lands key to mitigation and adaptation—Response. Science 367, 518–519. https://doi.org/10.1126/science.aba7577 (2020).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Vidal, A., Kumar, C., Zinngrebe, Y., Dobie, P. & Gassner, A. Trees on farms as a nature-based solution for
biodiversity conservation in agricultural landscapes. Report number: ICRAF Policy brief No 47. 12p. World
Agroforestry Centre. https://doi.org/10.13140/RG.2.2.14852.07045 (2020).César, R. et al. Forest and landscape restoration: A review emphasizing principles, concepts, and practices. Land 10, 28. https://doi.org/10.3390/land10010028 (2020).Article 

Google Scholar 
Stanturf, J. A. et al. Implementing forest landscape restoration under the Bonn Challenge: A systematic approach. Ann. For. Sci. https://doi.org/10.1007/s13595-019-0833-z (2019).Article 

Google Scholar 
VilchezMendoza, S. et al. Consistency in bird use of tree cover across tropical agricultural landscapes. Ecol. Appl. Publ. Ecol. Soc. Am. 24, 158–168. https://doi.org/10.1890/13-0585.1 (2014).Article 

Google Scholar 
Kremen, C. & Merenlender, A. M. Landscapes that work for biodiversity and people. Science 362, eaau6020. https://doi.org/10.1126/science.aau6020 (2018).Article 
CAS 
PubMed 

Google Scholar 
Shaver, I. et al. Coupled social and ecological outcomes of agricultural intensification in Costa Rica and the future of biodiversity conservation in tropical agricultural regions. Glob. Environ. Change 32, 74–86. https://doi.org/10.1016/j.gloenvcha.2015.02.006 (2015).Article 

Google Scholar 
Zermeño-Hernández, I., Pingarroni, A. & Martínez-Ramos, M. Agricultural land-use diversity and forest regeneration potential in human- modified tropical landscapes. Agric. Ecosyst. Environ. 230, 210–220. https://doi.org/10.1016/j.agee.2016.06.007 (2016).Article 

Google Scholar 
Garibaldi, L. A. et al. Working landscapes need at least 20% native habitat. Conserv. Lett. 14, e12773. https://doi.org/10.1111/conl.12773 (2021).Article 

Google Scholar 
Estrada-Carmona, N., Martínez-Salinas, A., DeClerck, F. A. J., Vílchez-Mendoza, S. & Garbach, K. Managing the farmscape for connectivity increases conservation value for tropical bird species with different forest-dependencies. J. Environ. Manag. 250, 109504. https://doi.org/10.1016/j.jenvman.2019.109504 (2019).Article 
CAS 

Google Scholar 
Vandermeer, J. & Perfecto, I. The agroecosystem: A need for the conservation biologist’s lens. Conserv. Biol. 11, 591–592 (1997).Article 

Google Scholar 
Pardon, P. et al. Trees increase soil organic carbon and nutrient availability in temperate agroforestry systems. Agr. Ecosyst. Environ. 247, 98–111. https://doi.org/10.1016/j.agee.2017.06.018 (2017).Article 
CAS 

Google Scholar 
Nair, P. R. The coming of age of agroforestry. J. Sci. Food Agric. 87, 1613–1619. https://doi.org/10.1002/jsfa.2897 (2007).Article 
CAS 

Google Scholar 
Chatterjee, N., Nair, P. K. R., Chakraborty, S. & Nair, V. D. Changes in soil carbon stocks across the Forest-Agroforest-Agriculture/Pasture continuum in various agroecological regions: A meta-analysis. Agric. Ecosyst. Environ. 266, 55–67. https://doi.org/10.1016/j.agee.2018.07.014 (2018).Article 

Google Scholar 
Toledo-Hernández, M., Wanger, T. C. & Tscharntke, T. Neglected pollinators: Can enhanced pollination services improve cocoa yields? A review. Agr. Ecosyst. Environ. 247, 137–148. https://doi.org/10.1016/j.agee.2017.05.021 (2017).Article 

Google Scholar 
Pumariño, L. et al. Effects of agroforestry on pest, disease and weed control: A meta-analysis. Basic Appl. Ecol. 16, 573–582. https://doi.org/10.1016/j.baae.2015.08.006 (2015).Article 

Google Scholar 
Tscharntke, T. et al. Multifunctional shade-tree management in tropical agroforestry landscapes—A review. J. Appl. Ecol. 48, 619–629. https://doi.org/10.1111/j.1365-2664.2010.01939.x (2011).Article 

Google Scholar 
Martínez-Fonseca, J. G., Chávez-Velásquez, M., Williams-Guillen, K. & Chambers, C. L. Bats use live fences to move between tropical dry forest remnants. Biotropica 52, 5–10. https://doi.org/10.1111/btp.12751 (2020).Article 

Google Scholar 
Prevedello, J. A., Almeida-Gomes, M. & Lindenmayer, D. B. The importance of scattered trees for biodiversity conservation: A global meta-analysis. J. Appl. Ecol. 55, 205–214. https://doi.org/10.1111/1365-2664.12943 (2018).Article 

Google Scholar 
INE. Ministerio de Agricultura, Pesca y Alimentación (MAPA)- Gobierno de España-. 2021. Ficha de sectores. Sectores Agricultura y Pesquero. Honduras (2022).MinAmbiente-ICF. Tipologías de Bosques de Honduras. Programa ONU-REDD. Forest Carbon Partnership Facility. Tegucigalpa, Honduras. Secretaria de Energía, Recursos Naturales, Ambiente y Minas (Min Ambiente)/Instituto Nacional de Conservación y Desarrollo Forestal, Areas Protegidas y Vida Silvestre (ICF). (2017).Godinot, F., Somarriba, E., Finegan, B. & Delgado-Rodríguez, D. Secondary tropical dry forests are important to cattle ranchers in Northwestern Costa Rica. Trop. J. Environ. Sci. 54, 20–50 (2020).
Google Scholar 
Zahawi, R. A. Establishment and growth of living fence species: An overlooked tool for the restoration of degraded Areas in the Tropics. Restor. Ecol. 13, 92–102. https://doi.org/10.1111/j.1526-100X.2005.00011.x (2005).Article 

Google Scholar 
Harvey, C. A. et al. Patterns of animal diversity in different forms of tree cover in agricultural landscapes. Ecol. Appl. Publ. Ecol. Soc. Am. 16, 1986–1999. https://doi.org/10.1890/1051-0761(2006)016[1986:poadid]2.0.co;2 (2006).Article 

Google Scholar 
Miceli-Mèndez, C. L., Ferguson, B. G. & Ramìrez-Marcial, N. in Post-Agricultural Succession in the Neotropics (ed Randall W. Myster) 165–191 (Springer New York, 2008).Gaoue, O. G. & Ticktin, T. Patterns of harvesting foliage and bark from the multipurpose tree Khaya senegalensis in Benin: Variation across ecological regions and its impacts on population structure. Biol. Conserv. 137, 424–436. https://doi.org/10.1016/j.biocon.2007.02.020 (2007).Article 

Google Scholar 
Daily, G., Ceballos, G., Pacheco, J., Suzan, G. & Anchez-Azofeifa, A. Countryside biogeography of neotropical mammals: Conservation opportunities in agricultural landscapes of Costa Rica. Conserv. Biol. https://doi.org/10.1111/j.1523-1739.2003.00298.x (2003).Article 

Google Scholar 
Mayfield, M. M. & Daily, G. C. Countryside biogeography of neotropical herbaceous and shrubby plants. Ecol. Appl. 15, 423–439. https://doi.org/10.1890/03-5369 (2005).Article 

Google Scholar 
Sánchez-Merlos, D. et al. Diversidad, composición y estructura de la vegetación en un agropaisaje ganadero en Matiguás, Nicaragua. Rev. Biol. Trop. https://doi.org/10.15517/rbt.v53i3-4.14601 (2005).Article 

Google Scholar 
Sekercioglu, C. H., Loarie, S. R., Oviedo Brenes, F., Ehrlich, P. R. & Daily, G. C. Persistence of forest birds in the Costa Rican agricultural countryside. Conserv. Biol. 21, 482–494. https://doi.org/10.1111/j.1523-1739.2007.00655.x (2007).Article 
PubMed 

Google Scholar 
Wallace, G., Barborak, J. & MacFarland, C. Land use planning and regulation in and around protected areas: A study of best practices and capacity building needs in Mexico and Central America. Nat Conserv 3 (2005).
Rozendaal Danaë, M. A. et al. Biodiversity recovery of Neotropical secondary forests. Sci. Adv. 5, eaau3114. https://doi.org/10.1126/sciadv.aau3114 (2019).Article 
ADS 
PubMed 
PubMed Central 

Google Scholar 
Souza Oliveira, M. et al. Biomass of timber species in Central American secondary forests:
Towards climate change mitigation through sustainable timber harvesting. Forest Ecology and Management 496,
119439. https://doi.org/10.1016/j.foreco.2021.119439 (2021).Article 

Google Scholar 
Gillespie, T. W., Grijalva, A. & Farris, C. N. Diversity, composition, and structure of tropical dry forests in Central America. Plant Ecol. 147, 37–47. https://doi.org/10.1023/A:1009848525399 (2000).Article 

Google Scholar 
Ngo Bieng, M. A. et al. Relevance of secondary tropical forest for landscape restoration. For. Ecol. Manag. 493, 119265. https://doi.org/10.1016/j.foreco.2021.119265 (2021).Article 

Google Scholar 
Souza Oliveira, M. et al. Biomass of timber species in Central American secondary forests: Towards climate change mitigation through sustainable timber harvesting. For. Ecol. Manag. 496, 119439. https://doi.org/10.1016/j.foreco.2021.119439 (2021).Article 

Google Scholar 
Chacón, L. M. & Harvey, C. A. Live fences and landscape connectivity in a neotropical agricultural landscape. Agrofor. Syst. 68, 15. https://doi.org/10.1007/s10457-005-5831-5 (2006).Article 

Google Scholar 
Harvey, C. A. et al. Conservation value of dispersed tree cover threatened by pasture management. For. Ecol. Manag. 261, 1664–1674. https://doi.org/10.1016/j.foreco.2010.11.004 (2011).Article 

Google Scholar 
Suding, K. N. Toward an Era of restoration in ecology: Successes, failures, and opportunities ahead. Annu. Rev. Ecol. Evol. Syst. 42, 465–487. https://doi.org/10.1146/annurev-ecolsys-102710-145115 (2011).Article 

Google Scholar 
Moguel, P. & Toledo, V. M. Biodiversity conservation in traditional coffee systems of Mexico. Conserv. Biol. 13, 11–21. https://doi.org/10.1046/j.1523-1739.1999.97153.x (1999).Article 

Google Scholar 
Harrison, R. D., Harrison, S., Laumonier, Y., Somarriba, E. & Suber, M. Biodiversity monitoring for agricultural landscapes. A protocol using biodiversity metrics to monitor agricultural sustainability under Aichi Target 7. (2019).Heck, K. L. Jr., van Belle, G. & Simberloff, D. Explicit calculation of the rarefaction diversity measurement and the determination of sufficient sample size. Ecology 56, 1459–1461. https://doi.org/10.2307/1934716 (1975).Article 

Google Scholar 
Magurran, A. E. Measuring Biological Diversity (Wiley-Blackwell, New Jersey, 2004).
Google Scholar 
Chao, A. et al. Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies. Ecol. Monogr. 84, 45–67. https://doi.org/10.1890/13-0133.1 (2014).Article 

Google Scholar 
Jost, L. Partitioning diversity into independent alpha and beta components. Ecology 88, 2427–2439. https://doi.org/10.1890/06-1736.1 (2007).Article 
PubMed 

Google Scholar 
Gotelli, N. J. & Colwell, R. K. Quantifying biodiversity: Procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett. 4, 379–391. https://doi.org/10.1046/j.1461-0248.2001.00230.x (2001).Article 

Google Scholar 
Zuur, A. F., Ieno, E. N., Walker, N., Saveliev, A. A. & Smith, G. M. Mixed Effects Models and Extensions in Ecology with R. XXII, 574 (Springer New York, NY, 2009).R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. (2021).Oksanen, J. et al. Vegan: Community Ecology Package. R Package Version 2.2-1 2, 1–2 (2015).Hsieh, T. C., Ma, K. & Chao, A. iNEXT: An R package for rarefaction and extrapolation of species diversity (Hill numbers). Methods Ecol. Evol. https://doi.org/10.1111/2041-210X.12613 (2016).Article 

Google Scholar 
Venables, W. N & Ripley, B. D Modern Applied Statistics with S. Fourth Edition. Springer, New York. ISBN 0-387-
95457-0 (2002)Wickham, H. ggplot2: Elegant graphics for data analysis (Springer, 2009).Book 
MATH 

Google Scholar 
gridExtra: Miscellaneous Functions for “Grid” Graphics. R package version 2.3. (2017). More

Angerbjörn, A. & Flux, J. E. C. Lepus timidus. Mamm. Species 1–11, https://doi.org/10.2307/3504302 (1995).Bergengren, A. On genetics, evolution and history of distribution of the heath-hare, a distinct population of the Arctic hare, Lepus timidus Lin. Swed. Wildl. (Viltrevy) 6, 381–460 (1969).
Google Scholar 
Thulin, C.-G. The distribution of mountain hares Lepus timidus in Europe: a challenge from brown hares L. europaeus? Mamm. Rev. 33, 29–42 (2003).Article 

Google Scholar 
Mills, L. S. et al. Camouflage mismatch in seasonal coat color due to decreased snow duration. Proc. Nat.Acad. Sci. 110, 7360–7365 (2013).Article 
ADS 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Zimova, M. et al. Lack of phenological shift leads to increased camouflage mismatch in mountain hares. Proc.Royal Soc. B: Biol. Sci. 287, 20201786 (2020).Article 

Google Scholar 
Levänen, R., Kunnasranta, M. & Pohjoismäki, J. Mitochondrial DNA introgression at the northern edge of the brown hare (Lepus europaeus) range. Ann Zool Fennici 55, 15–24 (2018).Article 

Google Scholar 
Thulin, C.-G., Winiger, A., Tallian, A. G. & Kindberg, J. Hunting harvest data in Sweden indicate precipitous decline in the native mountain hare subspecies Lepus timidus sylvaticus (heath hare). J. Nat. Conserv. 64, 126069 (2021).Article 

Google Scholar 
Thulin, C.-G., Jaarola, M. & Tegelström, H. The occurrence of mountain hare mitochondrial DNA in wild brown hares. Mol. Ecol. 6, 463–467 (1997).Article 
CAS 
PubMed 

Google Scholar 
Pohjoismäki, J. L. O., Michell, C., Levänen, R. & Smith, S. Hybridization with mountain hares increases the functional allelic repertoire in brown hares. Sci. Rep. 11, 15771 (2021).Article 
ADS 
PubMed 
PubMed Central 

Google Scholar 
Hoekstra, H. E. Genetics, development and evolution of adaptive pigmentation in vertebrates. Heredity (Edinb) 97, 222–234 (2006).Article 
CAS 

Google Scholar 
Hamill, R. M., Doyle, D. & Duke, E. J. Spatial patterns of genetic diversity across European subspecies of the mountain hare, Lepus timidus L. Heredity (Edinb) 97, 355–365 (2006).Article 
CAS 

Google Scholar 
Leach, K., Montgomery, W. I. & Reid, N. Biogeography, macroecology and species’ traits mediate competitive interactions in the order Lagomorpha. Mamm. Rev. 45, 88–102 (2015).Article 

Google Scholar 
Marques, J. P. et al. Data Descriptor: Mountain hare transcriptome and diagnostic markers as resources to monitor hybridization with European hares. Sci. Data 4, 1–11 (2017).Article 

Google Scholar 
NCBI Sequence Read Archive https://identifiers.org/insdc.sra:SRP358660 (2022).Andrews, S. FastQC: a quality control tool for high throughput sequence data. Babraham Bioinformatics. Preprint at http://www.bioinformatics.babraham.ac.uk/projects/fastqc (2010).Martin, M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17, 10–12 (2011).Article 

Google Scholar 
Marques, J. P. et al. An annotated draft genome of the mountain hare (Lepus timidus). Genome Biol. Evol. 12, 3656–3662 (2020).Article 
CAS 
PubMed 

Google Scholar 
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Broad Institute. Picard toolkit. Broad Institute, GitHub repository. Preprint at https://broadinstitute.github.io/picard/ (2019).Garrison, E. & Marth, G. Haplotype-based variant detection from short-read sequencing. arXiv 1207.3907 (2012).Danecek, P. et al. The variant call format and VCFtools. Bioinformatics 27, 2156–2158 (2011).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Browning, S. R. & Browning, B. L. Rapid and accurate haplotype phasing and missing-data inference for whole-genome association studies by use of localized haplotype clustering. Am. J. Hum. Genet. 81, 1084–1097 (2007).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Michell, C. T., Pohjoismäki, J. L. O., Spong, G. & Thulin, C.-G. Mountain- and brown hare genetic polymorphisms to survey local adaptations and conservation status of the heath hare (Lepus timidus sylvaticus, Nilsson 1831), Dryad, https://doi.org/10.5061/dryad.3bk3j9kmp (2022).Khan, A. & Mathelier, A. Intervene: a tool for intersection and visualization of multiple gene or genomic region sets. BMC Bioinformatics 18, 287 (2017).Article 
PubMed 
PubMed Central 

Google Scholar 
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/.Jombart, T. & Ahmed, I. adegenet 1.3-1: new tools for the analysis of genome-wide SNP data. Bioinformatics 27, 3070–3071 (2011).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Jombart, T. adegenet: a R package for the multivariate analysis of genetic markers. Bioinformatics 24, 1403–1405 (2008).Article 
CAS 
PubMed 

Google Scholar 
Dierckxsens, N., Mardulyn, P. & Smits, G. NOVOPlasty: De novo assembly of organelle genomes from whole genome data. Nucleic Acids Res. 45 (2017).Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: Improvements in performance and usability. Mol. Biol. Evol. 30 (2013).Trifinopoulos, J., Nguyen, L. T., von Haeseler, A. & Minh, B. Q. W-IQ-TREE: a fast online phylogenetic tool for maximum likelihood analysis. Nucleic Acids Res. 44 (2016).Kalyaanamoorthy, S., Minh, B. Q., Wong, T. K. F., von Haeseler, A. & Jermiin, L. S. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat. Methods 14, 587–589 (2017).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Stamatakis, A. RaxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312–1313 (2014).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Paradis, E. & Schliep, K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics 35, 526–528 (2019).Article 
CAS 
PubMed 

Google Scholar 
Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32, 3047–3048 (2016).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Kamvar, Z. N., Tabima, J. F. & Grünwald, N. J. Poppr: an R package for genetic analysis of populations with clonal, partially clonal, and/or sexual reproduction. PeerJ 2, e281 (2014).Article 
PubMed 
PubMed Central 

Google Scholar 
Li, H. Minimap2: Pairwise alignment for nucleotide sequences. Bioinformatics 34, 3094–3100 (2018).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Levänen, R., Thulin, C.-G., Spong, G. & Pohjoismäki, J. L. O. Widespread introgression of mountain hare genes into Fennoscandian brown hare populations. PloS One 13, e0191790 (2018).Article 
PubMed 
PubMed Central 

Google Scholar 
Giska, I. et al. The evolutionary pathways for local adaptation in mountain hares. Mol. Ecol. 31, 1487–1503 (2022).Article 
PubMed 
PubMed Central 

Google Scholar 
Thulin, C.-G., Isaksson, M. & Tegelström, H. The origin of Scandinavian mountain hares (Lepus timidus). Gibier Faune Savage/Game and Wildlife 14, 463–475 (1997).
Google Scholar 
Ferreira, M. S. et al. The legacy of recurrent introgression during the radiation of hares. Syst. Biol. 70, 593–607 (2021).Article 
PubMed 

Google Scholar  More

Sacristán, D., Peñarroya, B., Recatalá, L. Increasing the Knowledge on the Management of Cu-Contaminated Agricultural Soils by Cropping Tomato (Solanum Lycopersicum L.). Land Degrad. Dev. 26, 587–595 (2015).FAO. Land Degradation Assessment in Drylands. Manual for Local Level Assessment of Land Degradation and Sustainable Land Management. Part 1: Planning and Methodological Approach, Analysis and Reporting. https://www.fao.org/3/i6362e/i6362e.pdf (Food and Agriculture Organization of the United Nations, 2011).Vlachodimos, K., Papatheodorou, E. M., Diamantopoulos, J. & Monokrousos, N. Assessment of Robinia pseudoacacia cultivations as a restoration strategy for reclaimed mine spoil heaps. Environ Monit. Assess. 185, 6921–6932 (2013).Article 
CAS 
PubMed 

Google Scholar 
Misano, G. & Di Pietro, R. Habitat 9250 “Quercus trojana woods” in Italy. Fitosociologia 44, 235–238 (2007).
Google Scholar 
Biondi, E. et al. A contribution towards the knowledge of semideciduous and evergreen woods of Apulia (south-eastern Italy). Fitosociologia 41(1), 3–28 (2004).MathSciNet 

Google Scholar 
Brunetti, G. et al. Remediation of a heavy metals contaminated soil using mycorrhized and non-mycorrhized Helichrysum italicum (Roth) Don. Land Degrad. Dev. 29, 91–104 (2017).Article 

Google Scholar 
Poblador, S. et al. The influence of the invasive alien nitrogen-fixing Robinia pseudoacacia L. on soil nitrogen availability in a mixed Mediterranean riparian forest. Eur. J. For. Res. 138, 1083–1093 (2019).Article 
CAS 

Google Scholar 
Vítková, M., Müllerová, J., Sádlo, J., Pergl, J. & Pyšek, P. Black locust (Robinia pseudoacacia) beloved and despised: A story of an invasive tree in Central Europe. For. Ecol. Manag. 384, 287–302 (2017).Article 

Google Scholar 
Doran, J.W., Parkin, T.B. Quantitative indicators of soil quality: a minimum data set. in Methods for Assessing Soil Quality (eds. Doran, J.W., Jones, A.J.). 25–37 (Soil Science Society of America, 1996).Gil-Sotres, F., Trasar-Cepeda, C., Leirós, M. C. & Seoane, S. Different approaches to evaluating soil quality using biochemical properties. Soil Biol. Biochem. 37, 877–887 (2005).Article 
CAS 

Google Scholar 
Andriani, G. F. & Walsh, N. An example of the effects of anthropogenic changes on natural environment in the Apulian karst (southern Italy). Environ. Geol. 58, 313–325 (2009).Article 
ADS 

Google Scholar 
Bisantino, T., Pizzo, V., Polemio, M. & Gentile, F. Analysis of the flooding event of October 22–23, 2005 in a small basin in the province of Bari (Southern Italy). J. Agric. Eng. 531, 197–204 (2016).Article 

Google Scholar 
Soil Survey Staff. Keys to Soil Taxonomy 12th edn. (USDA-Natural Resources Conservation Service, 2014).
Google Scholar 
Tartarino, P. Inventario dei Boschi Spontanei e dei Rimboschimenti delle Provincie BAT e Bari e Stima del Loro Volume Legnoso e della sua Frazione Prelevabile nel Prossimo Ventennio. (Rapporto Tecnico Scientifico, 2011).Ismail, A. et al. Chemical composition and biological activities of Tunisian Cupressus arizonica Greene essential oils. Chem. Biodivers. 11, 150–160 (2014).Article 
CAS 
PubMed 

Google Scholar 
Navarro, A. et al. Feasibility of SRC Species for growing in Mediterranean conditions. Bioenerg. Res. 9, 208–223 (2015).Article 

Google Scholar 
Perrino, E. V., Brunetti, G. & Farrag, K. Plant communities in multi-metal contaminated soils: A case study in the National Park of Alta Murgia (Apulia Region-Southern Italy). Int. J. Phytoremediat. 16, 871–888 (2014).Article 
CAS 

Google Scholar 
VV AA Perizia Studi per il Riequilibrio Socio-Economico dell’area Interessata dall’invaso sul Torrente Locone. Consorzio Di Bonifica Apulo Lucano (1986).Lavarra, P. et al. Il Sistema Carta della Natura della Regione Puglia. (ISPRA, Serie Rapporti 204, 2014).Sparks, D. L. et al. Method of Soil Analysis: Part 3 (American Society of Agronomy Inc, 1996).Book 

Google Scholar 
Brink, R. H. Jr., Dubach, P. & Lynch, D. L. Measurement of carbohydrates in soil hydrolyzates with anthrone. Soil Sci. 89, 157–166 (1960).Article 
ADS 
CAS 

Google Scholar 
Lowry, O. H., Rosebrough, N. J., Farr, A. L. & Randall, R. J. Protein measurement with the folin phenol reagent. J. Biol. Chem. 193, 265–275 (1951).Article 
CAS 
PubMed 

Google Scholar 
García, C., Hernandez, T. & Costa, F. Potential use of dehydrogenase activity as an index of microbial activity in degraded soils. Commun. Soil Sci. Plant Anal. 28, 123–134 (1997).Article 

Google Scholar 
Vance, E. D., Brookes, P. C. & Jenkinson, D. S. An extraction method for measuring soil microbial biomass C. Soil Biol. Biochem. 19, 703–707 (1987).Article 
CAS 

Google Scholar 
Gregorich, E. G., Wen, G., Voroney, R. P. & Kachanoski, R. G. Calibration of a rapid direct chloroform extraction method for measuring soil microbial biomass C. Soil Biol. Biochem. 22, 1009–1011 (1990).Article 
CAS 

Google Scholar 
Nannipieri, P., Ceccanti, B., Cervelli, S. & Matarese, E. Extraction of phosphatase, urease, protease, organic carbon and nitrogen from soil. Soil Sci. Soc. Am. J. 44, 1011–1016 (1980).Article 
ADS 
CAS 

Google Scholar 
Tabatabai, M.A. (1994) Soil enzymes. in Methods of Soil Analysis. Part 2. Microbiological and Biochemical Properties (eds. Weaver, R.W. et al.). 775–833 (Soil Science Society of America, Inc., 1996)Traversa, A., Said-Pullicino, D., D’Orazio, V., Gigliotti, G., & Senesi, N. Properties of humic acids in Mediterranean forest soils (Southern Italy): Influence of different plant covering. Eur. J. For. Res. 130, 1045–1054 (2011)De Marco, A. et al. Decomposition of black locust and black pine leaf litter in two coeval forest stands on Mount Vesuvius and dynamics of organic components assessed through proximate analysis and NMR spectroscopy. Soil Biol. Biochem. 51, 1–15 (2012).Article 
CAS 

Google Scholar 
Wei, G. et al. Invasive Robinia pseudoacacia in China is nodulated by Mesorhizobium and Sinorhizobium species that share similar nodulation genes with native American symbionts. FEMS Microbiol. Ecol. 68, 320–328 (2009).Article 
CAS 
PubMed 

Google Scholar 
Schulze, E. D., Gebauer, G., Ziegler, H. & Lange, O. L. Estimates of nitrogen fixation by trees on an aridity gradient in Namibia. Oecologia 88, 451–455 (1991).Article 
ADS 
PubMed 

Google Scholar 
Zahran, H. H. Rhizobium-legume symbiosis and nitrogen fixation under severe conditions and in an arid climate. Microbiol. Mol. Biol. Rev. 63, 968–989 (1999).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Veste, M. & Kriebitzsch, W. U. Influence of drought stress on photosynthesis, transpiration, and growth of juvenile black locust (Robinia pseudoacacia L.). Forstarchiv 84, 35–42 (2013).
Google Scholar 
Nicolescu, V. N. et al. Ecology, growth and management of black locust (Robinia pseudoacacia L.), a non-native species integrated into European forests. J. For. Res. 31, 1081–1101 (2020).Article 
CAS 

Google Scholar 
Sposito, G. The Chemistry of Soil (Oxford University Press, 2008).
Google Scholar 
Margalef, O. et al. Global patterns of phosphatase activity in natural soils. Sci. Rep. 7, 1337. https://doi.org/10.1038/s41598-017-01418-8 (2017).Prescott, C. E. & Grayston, S. J. Tree species influence on microbial communities in litter and soil: Current knowledge and research needs. For. Ecol. Manag. 309, 19–27 (2013).Article 

Google Scholar 
Frankenberger, W. T. & Dick, W. A. Relationships between enzyme, activities and microbial growth and activity indices in soil. Soil Sci. Soc. Am. J. 47, 945–951 (1983).Article 
ADS 
CAS 

Google Scholar 
Frankenberger, W.T., Tabatabai, M.A. Amidase activity in soils III. Stability and distribution. Soil Sci. Soc. Am. J. 45, 333–338 (1981).Nannipieri, P., Trasar-Cepeda, C. & Dick, R. P. Soil enzyme activity: A brief history and biochemistry as a basis for appropriate interpretations and meta-analysis. Biol. Fertil. Soils 54, 11–19 (2018).Article 
CAS 

Google Scholar 
Pascual, J. A., Garcia, C., Hernandez, T., Moreno, J. L. & Ros, M. Soil microbial activity as a biomarker of degradation and remediation processes. Soil Biol. Biochem. 32, 1877–1883 (2000).Article 
CAS 

Google Scholar 
García-Gil, J. C., Plaza, C., Solker-Rovira, P. & Polo, A. Long-term effects of municipal solid waste compost application on soil enzyme activities and microbial biomass. Soil Biol. Biochem. 32, 1907–1913 (2000).Article 

Google Scholar 
Insam, H. & Domsch, K. H. Relationship between soil organic carbon and microbial biomass on chronosequences of reclamation sites. Microb. Ecol. 15, 177–188 (1988).Article 
CAS 
PubMed 

Google Scholar 
Acosta-Martinez, V. & Tabatabai, M. Enzyme activities in a limed agricultural soil. Biol. Fertil. Soils 31, 85–91 (2000).Article 
CAS 

Google Scholar 
Uselman, S. M., Qualls, R. G. & Thomas, R. B. A test of a potential short cut in the nitrogen cycle: the role of exudation of symbiotically fixed nitrogen from the roots of a N-fixing tree and the effects of increased atmospheric CO2 and temperature. Plant Soil 210, 21–32 (1999).Article 
CAS 

Google Scholar 
De Marco, A., Esposito, F., Berg, B., Zarrelli, A. & Virzo De Santo, A. Litter inhibitory effects on soil microbial biomass activity, and catabolic diversity in two paired stands of Robinia pseudoacacia L. and Pinus nigra Arn. Forest 9, 766. https://doi.org/10.3390/f9120766 (2018).Article 

Google Scholar 
Haghverdi, K. & Kooch, Y. Effects of diversity of tree species on nutrient cycling and soil-related processes. CATENA 178, 335–344 (2019).Article 
CAS 

Google Scholar 
Anderson, H. T. Microbial eco-physiological indicators to assess soil quality. Agric. Ecosyst. Environ. 98, 285–293 (2003).Article 

Google Scholar 
Jenkinson, D.S., Ladd, J.N. Microbial biomass in soil: Measurement and turnover. in Soil Biochemistry (eds. Paul, E.A., Ladd, J.N.). 415–471 (Marcel Dekker Inc., 1981) More