1. Ecology
Download icon

Regularities in species’ niches reveal the world’s climate regions

  1. Joaquín Calatayud  Is a corresponding author
  2. Magnus Neuman
  3. Alexis Rojas
  4. Anton Eriksson
  5. Martin Rosvall
  1. Integrated Science Lab, Department of Physics, Umeå University, Sweden
  2. Departamento de Biología, Geología, Física y Química inorgánica, Universidad Rey Juan Carlos, Spain
Research Article
  • Cited 1
  • Views 838
  • Annotations
Cite this article as: eLife 2021;10:e58397 doi: 10.7554/eLife.58397

Abstract

Climate regions form the basis of many ecological, evolutionary, and conservation studies. However, our understanding of climate regions is limited to how they shape vegetation: they do not account for the distribution of animals. Here, we develop a network-based framework to identify important climates worldwide based on regularities in realized niches of about 26,000 tetrapods. We show that high-energy climates, including deserts, tropical savannas, and steppes, are consistent across animal- and plant-derived classifications, indicating similar underlying climatic determinants. Conversely, temperate climates differ across all groups, suggesting that these climates allow for idiosyncratic adaptations. Finally, we show how the integration of niche classifications with geographical information enables the detection of climatic transition zones and the signal of geographic and historical processes. Our results identify the climates shaping the distribution of tetrapods and call for caution when using general climate classifications to study the ecology, evolution, or conservation of specific taxa.

eLife digest

There are many distinct climates on Earth, from tropical savannas and temperate forests to dry deserts. Historically, each region has been defined by how its annual weather patterns shape the type of vegetation present. For example, hot and humid environments support the growth of evergreen forests that would not survive in drier places.

Identifying the boundaries between climate regions is key to understanding how life is organized on Earth and predicting how climate change will affect different species. Current climate classifications, however, do not account for where animals can be found or how local conditions, such as precipitation and average temperatures, shape the distribution of different animal species. To bridge this gap, Calatayud et al. analyzed the preferred climate of about 26,000 animal species, including amphibians, birds, mammals and reptiles.

For each species, Calatayud et al. calculated the annual rainfall and temperature of its local environment, or ‘niche’, using previously collected data. They then used a computer algorithm to group together climates that had similar species. This identified 16 climate regions which govern the distribution of the animals studied. Calatayud et al. found that these newly defined climatic regions resembled some of the regions classified using plants. This was particularly true for high-energy climates that had lower levels of rainfall and hot temperatures, such as deserts and the tropical savanna.

The animals and plant species living in high-energy regions were found to be fairly consistent across both classification systems. Whereas the species present in milder and colder climates, such as temperate forests or Mediterranean climates, were found to be much more varied. This suggests that temperate climates are harder to classify and may affect the distribution of plants and animals differently. It also implies that less extreme conditions support a larger range of species than harsher climates in which only species with certain adaptations are able to survive.

These findings build the basis for a better understanding of how climates shape ecosystems. More specific climate classifications, based on such analyses, could be used to inform conservation strategies for animal species in the face of climate change.

Introduction

Climate governs the basis of life on Earth. Besides historical contingencies and geographical barriers, abiotic conditions determine species ranges (Woodward, 1987; Hoffmann and Parsons, 1997; Calatayud et al., 2019b) and derived diversity patterns (Hawkins et al., 2003; Kreft and Jetz, 2007; Mendoza and Araújo, 2019). On a global scale, distinctive climate types impose generalized constraints that shape species pools adapted to particular climates (Whittaker, 1962). Identifying the boundaries of these climate types is a fundamental challenge to understanding how life organizes on Earth, with acute implications in diverse disciplines given the current climate change.

Pythagoras proposed a classification of climate regimes of the known world in the 6th century BC (Sanderson, 1999). Still, it was not until the 19th century that geographers laid the foundations for such classifications (Oliver, 1991). By that time, researchers noticed the close relationship between the distribution of various life forms, especially vegetation types, and climate (Oliver, 1991). For instance, Köppen built his long-standing climate classification from pioneer plant classifications, assuming that vegetation types carry information about climatic conditions (Kottek et al., 2006; Thornthwaite, 1943). This assumption has received considerable support (Rohli et al., 2015), and researchers and stakeholders use Köppen’s classification system in a wide range of disciplines, including climatology (Spinoni et al., 2015), geography (Gentine et al., 2012), conservation planning (Tobin et al., 2014), and ecology (Garcia et al., 2014).

But the general climatic conditions derived from plant species may not apply to other organisms. Plants’ and animals’ climatic determinants will likely differ, given their large physiological differences. Although diversity patterns of plants and animals are relatively congruent, the climatic correlates of these patterns vary (Whittaker et al., 2007; Qian and Ricklefs, 2008). Hence, if climatic adaptations vary among taxa, so will the boundaries defining their climate types. In Thornthwaite, 1943 words, the ‘truly active factors’ describing a climate type may vary among organisms, and it remains unknown whether Köppen’s climate classification can indicate the active climatic factors for other organisms than plants. Despite several attempts to refine or propose alternative climate types or regions (Trewartha, 1954; Holdridge, 1947; Thornthwaite, 1948; Netzel and Stepinski, 2016; Gardner et al., 2020), quantitative studies defining climatic regions for other organisms are still lacking.

The current information on species distributions and global climatic variables, together with recent advances in niche modeling and classification techniques, provides an unprecedented opportunity to identify the climatic boundaries shaping the distribution of faunas and floras across the globe. The last decades have witnessed a tremendous collective effort to record occurrences of a large number of species (GBIF, 2019), which has resulted in comprehensive datasets of the distributional ranges of several groups of species (IUCN, 2015; BirdLife, 2015; Roll et al., 2017). Also, data on climatic variables at a global scale have been developed at high spatial resolutions (Fick and Hijmans, 2017; Trabucco and Zomer, 2009). Combining these two sources of information, we can characterize the realized climatic niches of different taxa and find regularities among them. For example, projecting these realized climatic niches into a climatic space (Broennimann et al., 2012) should, if climatic boundaries exist, reveal co-occurring groups of species across particular portions of the climatic space. Thus, identifying these niche domains should uncover the main climatic boundaries shaping the organization of life (Figure 1).

Workflow to identify niche domains and climatic regions.

Using the climatic conditions a given species experiences within its range (A), we project species niches into a climatic space discretized in an optimal number of bins (Appendix 1) (B). We translate the binned data into a weighted bipartite network, where climatic bins and species form the nodes and the probabilities of finding the species in the bins form the weighted links (C). Using a network community detection algorithm, we identify domains of the climatic space with similar species (D, upper). The climatic conditions defining these domains delineate the corresponding climatic regions of the Earth (D, lower). The striped climatic bin links to species classified in both climatic domains, and, therefore, it represents a diffuse transition with low specificity.

In addition to the climate, dispersal barriers and historical contingencies influence the shape of niche domains (Warren et al., 2014; Calatayud et al., 2016; Calatayud et al., 2019b). Therefore, similar climates can have different effects across different geographic regions (Ricklefs, 1987). For instance, while a given climate in some parts of the Earth may lead to specific species pools, the same climate in other parts of the Earth may not. Such a potential lack of specific species can occur, for example, because the required adaptations have not appeared (Flohr et al., 2013), the adapted species have not been able to disperse (Tuomisto et al., 2003), or the area is too small to hold large species pools (Connor and McCoy, 1979). Studying the signature of these historical and geographical processes, known as geographical signals, in niche domains can provide valuable information about the potential mechanisms behind them and their associated climatic regions.

We explore the global climate regions of tetrapods by characterizing the climatic niche domains of amphibians, birds, mammals, and reptiles. Tetrapods are a well-suited group for our purpose. First, comprehensive databases are available, including the distributional ranges of most species in the group (IUCN, 2015; BirdLife, 2015; Roll et al., 2017). Second, the different classes of tetrapods have diverse capabilities to disperse and withstand abiotic conditions, allowing us to investigate whether various capabilities influence climatic classifications. Third, accumulated evidence about the main climatic factors controlling the distribution of these species simplifies the selection of appropriate climatic variables. In particular, the distribution of tetrapods is strongly determined by the water and energy aspects of climate (Hawkins et al., 2003; Currie, 1991; Tingley et al., 2009; Gouveia et al., 2014; Pie et al., 2017; Cooper et al., 2011). Finally, researchers study tetrapods in several disparate fields – from animal husbandry (Abecia et al., 2017) to ecological (Englert Duursma et al., 2019) and evolutionary studies (Rolland et al., 2014) – where a description of their climatic regions can be especially useful.

In our classification approach, we project the realized niche of each tetrapod species onto a binned two-dimensional space representing water and energy, the primary climatic factors impacting their geographic distribution (Figure 1). After translating this climatic niche space into a weighted bipartite network, we apply a network community detection algorithm to identify climatic niche domains with similar species pools. Finally, by mapping the climatic niche domains back onto the Earth’s surface, we provide the climatic regions.

The detected climatic regions support the notion that similar climatic determinants underlie animal and plant distributions in high-energy regions, including deserts, tropical savannas, and steppe regions. However, differences in temperate climates across all groups indicate that specific climatic regions for each group of taxa are required to address ecological, evolutionary, and conservation questions.

Results

Classification of climatic niches

We first identified the species niche domains by calculating the proportion of observations of each species within each bin of a two-dimensional climatic space defined by potential evapotranspiration (PET) and annual precipitation (AP; Figure 1, 'Materials and methods' and Appendix 1). We represented this data as a weighted bipartite network, where climatic bins and species form two disjunct sets of nodes and the probabilities of finding the species in the bins form the link weights. Using a hierarchical network clustering algorithm (Rosvall and Bergstrom, 2008; Rosvall and Bergstrom, 2011), we obtained groups of climatic bins holding similar species, niche domains, and the species associated with them.

We found similarities among tetrapods classes in the detected niche domains but also observed some differences (Figure 2). For instance, the number of major domains with 50 or more species in the lower hierarchical level is similar (ranging from 13 to 15) across tetrapods classes. However, mammals and birds show a domain of low-energy inputs, whereas reptiles present some domains across arid conditions, that is, with high energy and low water availability (Figure 2). Finally, we classified the climatic space of tetrapods by using all species jointly. The clustering algorithm divided the niche space of tetrapods into 16 main domains that are similar to those of the independent classes, and some of the particularities described above did not appear (Figure 2).

Tetrapods’ niche domains across the climatic space.

The climatic niche domains of each group shown across a space defined by potential evapotranspiration (PET) as a surrogate of energy and annual precipitation (AP) as a surrogate of water inputs. Tetrapods’ domains are labeled so that E and W represent energy and water, respectively, and superscripts H, M, and L mean high, medium, and low, respectively. Numerical subscripts differentiate domains of similar climates. Numbers between 0 and 1 indicate the bootstrap support. The dotted line represents the domains at the highest hierarchical level. Domains of less than 50 species are colored dark gray. Domain colors across groups only indicate similar climatic regions. To characterize the climatic space, we used 17 divisions of both PET and AP for all groups except for amphibians, where we used 18 (see Appendix 1).

Figure 2—source data 1

niches domains and the species associated to them.

https://cdn.elifesciences.org/articles/58397/elife-58397-fig2-data1-v2.zip

Since uncertainties related to the ranges of species exist, we employed a bootstrap and a significance clustering procedure (Rosvall and Bergstrom, 2010; Calatayud et al., 2019a) to assess the domains’ robustness ('Materials and methods'). While several domains were well supported, we found that the niche domains corresponding to intermediate energy (between approximately 1000 and 1500 PET units; EM climates in Figure 2) and low to moderate water (up to approximately 800 ml; WL to WM) were less robust. This robustness analysis shows that these niche domains are more challenging to classify.

Animal vs previous plant-based climate regions

While Köppen’s climate classification is based on expert knowledge on vegetation physiognomy and the distribution of vegetation types, its wide use makes it worthwhile to compare the climate regions derived from this classification system to the ones produced here. Hence, we studied the geographic location of the climatic conditions associated to niche domains, the climatic regions shown in Figures 1 and 3, which allowed for a more precise comparison between groups and Köppen’s plant-based regions. The similarities among the regions of tetrapods classes measured as adjusted mutual information (AMI) ranged from 0.57 to 0.68, with mean AMI = 0.62 (Table 1). Moreover, the regions based on the niche domains of all tetrapods together were to some extent congruent with the regions of its independent classes (mean AMI = 0.71, ranging from 0.66 to 0.77). Köppen’s regions were more dissimilar to the regions of all tetrapods together (AMI = 0.44) and the regions of each class of tetrapods independently (mean AMI = 0.44, ranging from 0.40 to 0.47).

Tetrapod groups and Köppen’s climatic regions.

(A) Geographic location of tetrapods’ niche domains and Köppen’s climatic regions. See also Appendix 1—figure 1. Colors according to Figure 2. (B) Tetrapods’ climatic regions are labeled according to Figure 2.

Figure 3—source data 1

rasters with climate regions for each studied group.

https://cdn.elifesciences.org/articles/58397/elife-58397-fig3-data1-v2.zip
Table 1
Similarity of climatic regions measured by adjusted mutual information.
KöppenTetrapodsAmphibiansReptilesBirds
Tetrapods0.44
Amphibians0.400.66
Reptiles0.450.700.58
Birds0.450.770.610.64
Mammals0.470.720.570.660.68

Focusing on particular regions, we saw that climates of high energy (EH) were consistent among tetrapod groups and Köppen’s classification. Desert climates (high energy and low water, EHWL, Figure 3, and Appendix 1—figure 1) were the most similar across all groups. Tropical savanna and steppe climates (high energy and medium water, EHWM) were also consistently defined, though both of these Köppen regions were classified together for all groups but reptiles (Figures 2 and 3 and Appendix 1—figure 1). Similarly, Köppen’s tropical rainforest and tropical monsoon climates were for the most part well recovered (Appendix 1—figure 1). However, we found three different tropical-humid regions, each one mostly corresponding to one of the three larger masses of tropical rainforests: Amazonian, African, and Southeast Asian rainforests; EHWH1, EHWH2, and EHWH3, respectively (Figures 2 and 3). Regarding regions of low energy, we found a slightly higher level of disagreement between Köppen’s and tetrapods’ regions (Figure 3). Finally, temperate climates (medium energy EM) were the least congruent between tetrapod groups and Köppen’s regions. These regions of medium energy were at the same time the least congruent among groups and the least supported by the bootstrap analyses, suggesting that these climates could impose less restrictive conditions in general and allow the appearance of idiosyncratic and variable adaptations.

Climatic transition zones

A complete understanding of niche domains and their associated climatic regions entails the exploration of whether their boundaries represent abrupt or diffuse transitions. Climatic conditions corresponding to diffuse transitions should present low specificity levels to the domain where they belong (Figure 1). Hence, we can consider specificity as the opposite of transitivity. Our network approach allows us to calculate this specificity by the dual classification of climatic bins and species into the same niche domains (Figure 1). We computed the specificity of each climatic bin as the ratio between the link weights of the species classified in the same domain and the total link weights (Bernardo-Madrid et al., 2019; Calatayud et al., 2019b). Then, we projected these values geographically. As expected, lower specificity values were in general associated with the boundaries of the climatic regions (Figure 4A and Appendix 1—figure 2). Beyond boundaries, our results also revealed that harsh conditions, such as desert and continental-polar climates (EHWL and ELWL), present the highest specificity levels, regardless of the group (Figure 4A and Appendix 1—figure 2), reflecting the difficulty of colonizing these climates. Conversely, temperate regions showed the lowest levels of specificity. These regions were also weakly supported in the bootstrap analyses, and we found that bootstrap p-values and mean specificity were significantly correlated (standard generalized linear mixed model [GLMM] coeff. 6.21; p < 0.001 conditional = 0.29, see 'Materials and methods'). Together with the higher variability of these regions across groups, this result further supports the idea that these climatic conditions could impose less restrictive conditions to tetrapods.

The geographic location of climatic domains and their associated species provides insights into the mechanism underlying the climatic regions.

(A) Geographic projection of the specificity of climatic bins to their niche domain. (B) An example showing a bird’s niche domain with a low geographical signal. The distribution of the climatic conditions (black line) and the species (colored richness values) belonging to the same niche domain was mostly congruent. (C) An example of an amphibian’s niche domain showing a high geographical signal, reflected in a substantial mismatch between the distribution of climatic conditions and species belonging to the same domain. (D) A quantitative approximation of the geographical signal, ranging between 0 and 1, for the different taxonomic groups (see 'Materials and methods').

Geographical signal in climatic regions

Historical and geographical processes may lead to species pools adapted to a given climate in some regions of the Earth but not in others. To search for this geographical signal, we first compared the distribution of the climatic conditions and species grouped within the same niche domain. A geographic mismatch between species and climate distributions would point to portions of the climatic regions that are defined by species occurring in other geographic areas. Exploring these patterns for each niche domain revealed notable geographic agreement between species and climatic conditions of the same domain (Figure 4B and Appendix 1—figures 37). Nevertheless, we found some differences across groups and regions. More extreme climates showed larger mismatches between the distribution of species and climates. For instance, for all groups but reptiles, desert climate (EHWL) was mostly defined by species inhabiting Australia and to a lesser extent by species from the Namibian desert and the Horn of Africa, with few or no species inhabiting the Sahara desert (Figure 4C and Appendix 1—figures 37). Similarly, the northern climatic regions of amphibians and reptiles were defined by species at lower latitudes (Appendix 1—figures 3 and 4). Approaching the geographical signal more quantitatively (see 'Materials and methods'), we found a stronger signal for the worst dispersers, amphibians and reptiles, than for mammals and birds (Figure 4D), suggesting that dispersal capabilities can contribute to the geographical signal in the niche domains. Finally, all tetrapods together showed the lowest geographical signal, which suggests that, in addition to dispersion, increased evolutionary time can reduce the geographical signal.

Discussion

We detected 16 climatic regions governing the distribution of tetrapods. Despite the substantial physiological and functional differences among the groups, some of their niche domains and climatic regions are consistent. These climatic regions also resemble Köppen’s regions and correspond, in general, with extreme climates, being arid climates specially congruent across groups. These climates also presented overall low levels of transitivity, showing that species adapted to other climates have more difficulty withstanding these conditions. These results suggest that extreme climates impose strong adaptive barriers, producing a filtering of species (Butterfield, 2015; Cadotte and Tucker, 2017) even across distinctive evolutionary lineages.

Conversely, milder climatic conditions, especially temperate climates, showed the lowest statistical support and congruence across groups, as well as the highest transitivity. These results indicate that temperate climates are more difficult to classify (low support and congruence) due to the overlap in the climatic space of species pools with different climatic optima (high transitivity). Two complementary reasons can explain this ambiguity: first, while we used two variables widely recognized to shape the distributions of tetrapods, alternative variables, such as seasonal changes in energy and precipitation (Köppen and Geiger, 1930), may also influence species inhabiting temperate regions. Including these variables might help to further separate temperate species pools across the climatic space. Second, the climatic conditions of these domains may not prevent the colonization of species with other realized optima or preferences, which would generate the observed overlap in the climatic space (i.e., high transitivity) across milder conditions. Questions remain about the relative contribution of each process. Similarly, the exact causes and consequences of the climatic transition zones call for future investigation.

We also found some domains that were well supported but unique for each group. These differences between groups seem to relate to the particular physiological adaptations of each group. For instance, homeothermic birds and mammals defined a region of low energy, consistent with Köppen’s polar climates, that reptiles and amphibians lacked. Similarly, reptiles, a group that includes several groups adapted to arid environments (Pie et al., 2017), defined some regions of low precipitation and high PET. Hence, our results stress that caution is needed when generalizing the climate classification to other groups of organisms, and question the validity of using plant-based classifications for studies dealing with animals.

Beyond niche domains, our results also show differences in the geographical signal across groups. That amphibians – the group with the lowest dispersal capacity – showed the highest geographical signal suggests that dispersal processes play an essential role: species that are worse dispersers have more difficulties tracking their preferred climates (Araújo and Pearson, 2005), limiting the colonization of disjoint areas with similar climates. Moreover, all tetrapods together showed the lowest geographical signal, suggesting that increased evolutionary time can reduce this geographic pattern. That is, an extended evolutionary time that enables the appearance of convergent adaptations to similar climates in different geographic regions (Mazel et al., 2017) may reduce the geographical signal in niche domains. Alternatively, tetrapods include taxa with varying dispersal strategies – such as the high active dispersal of birds and the elevated survival probabilities of reptiles to passive transoceanic dispersal – which may also influence the geographical signal. In any case, the ultimate causes and consequences of this signal require further attention. Why are some amphibians able to inhabit arid conditions in the Australian desert but not in the Sahara desert (Appendi 1—figure 3)? Why can some reptiles withstand cold climates around the Himalayan mountains but not in the northern hemisphere (Appendi 1—figure 4)? These are some examples of the emerging fundamental questions related to the geographical signal. Answering them would require studying the adaptations that allow the species to tolerate particular climatic conditions and the abiotic, biotic, and historical differences between regions (Moncrieff et al., 2015) that favor these adaptations in some places but not in others. Moreover, the geographical signal idea could also help unravel idiosyncratic processes occurring at smaller scales by comparing local climate regions with global or regional ones. These examples form exciting avenues for future historical biogeographical and evolutionary studies.

Our results and data-driven methodology have the potential to bring us closer to a definition of climatic regions that represent active factors for the spatial organization of life. Nevertheless, it would be interesting to improve some aspects in future studies. First, while we used a notable number of species (about 26,000), they are taxonomically biased and only represent a small fraction of terrestrial organisms. Second, we used two climatic variables widely known to affect the distribution and diversity patterns of animals and plants in general (Currie, 1991; Hawkins et al., 2003), but other climatic variables might refine some of the least supported regions. Finally, our domains represent portions of the realized climatic niche space, which entails two sources of uncertainty. On the one hand, the estimation of realized niches depends on the data quality, scale, and treatment choices. On the other hand, the realized niche space may be influenced by historical, geographical, and biotic factors beyond pure climate (Warren et al., 2014; Calatayud et al., 2019b; Soberón, 2007). Hence, using fundamental rather than realized niches may also improve the accuracy of defining climatic regions. At the current pace of biological data accumulation and computational development, it is reasonable to expect that some of these limitations will soon be overcome. Meanwhile, the considerable congruence of several climatic regions across the studied groups and Köppen’s system provides confidence in their robustness. Hence, it is likely that using better data would not produce regions substantially different from those presented here.

Regardless of how generalizable the results are, the niche domains and their associated species pools and climatic regions can be used as a basis for ecological, evolutionary, and conservation studies concerning tetrapods. Some of the many questions that the results reported here (data available in source data of Figures 2 and 3) can help to answer include: Are the species belonging to different niche domains similarly conserved or protected, or both (Hanson et al., 2020)? To what extent do the differences between the four classes of tetrapods reflect phylogenetic and functional differences, and are such differences to be found in other taxonomic groups? Is the adaptation to niche domains evolutionarily constrained? Do diversification, extinction, or speciation rates differ among the species associated with different domains? Moreover, combining niche domains with bioregions based on pools of species (Bernardo-Madrid et al., 2019) or lineages (Holt et al., 2013) can also help answer several relevant questions: Are introductions more common between bioregions with the same climate types? To what extent does the current climate govern bioregions? Do ecological and evolutionary processes differ between different bioregions with same climate types (Moncrieff et al., 2015)?

In conclusion, our data-driven climate classification reveals major climatic boundaries organizing the distribution of tetrapods on Earth. Questions remain regarding the mechanism underlying the differences between groups in the climatic regions and the geographical signal. Nevertheless, these differences across groups highlight the need for classifications of climates specific to each group to study their ecology, evolution, or conservation.

Materials and methods

Data

We obtained the distribution ranges of mammals and amphibians from the IUCN, 2015, of birds from BirdLife, 2015, and of reptiles from Roll et al., 2017. We included only the native range of terrestrial species in the analyses in all instances. In the case of birds, we only used the breeding ranges. Moreover, since there is a higher uncertainty when determining the realized niches of narrow-ranging species (Lehmann et al., 2002), we arbitrarily removed the species whose ranges were less than five grid cells of 0.5°. After this cleaning of the data, we used 3657 amphibians, 7204 reptiles, 4574 mammals, and 10,684 birds, for a total of 26,119 tetrapod species.

We approximated the species’ Grinnellian niches (Soberón, 2007) with two climatic variables that represent energy and water inputs. While we could have used several other variables, we chose energy and water as they best explain climatic effects on species distributions (Hawkins et al., 2003). As surrogates for energy and water inputs, we used mean annual PET and AP, respectively. Both variables have been shown to be important factors for tetrapod species distributions (Currie, 1991; Tingley et al., 2009; Gouveia et al., 2014). Moreover, they have also been used in previous climate classifications (Thornthwaite, 1948) and are regularly used to derive other drivers of species distributions such as the UNEP aridity index (Unep and Thomas, 1992; Fuller et al., 2016). We obtained PET from Trabucco and Zomer, 2009 and AP from Fick and Hijmans, 2017, both at a 0.08° resolution. Finally, we obtained Köppen’s climatic regions from Kottek et al., 2006 and Rubel et al., 2017.

Niche characterization

Request a detailed protocol

We characterized the realized climatic niche of each species using an approach similar to the one proposed in Broennimann et al., 2012. We divided the climatic space formed by PET and AP into bins and calculated the proportion of occurrences a given species has in each climatic bin. Both the shape of the divisions and the number of divisions of each climatic axis affect the result. For instance, dividing the axis into regular intervals can destroy critical information if the climatic values more important for the species distributions are skewed toward any extreme of the distribution or if the climatic values are represented non-uniformly across the globe (as for AP, Appendix 1—figure 8). Similarly, the grain size to divide the climatic space may affect subsequent results (Levin, 1992; Daru et al., 2020). Dividing the space into too few intervals destroys information, whereas using too many divisions can generate niche domains with only a few species. To overcome the first issue, we divided the axes into quantiles based on the distribution of climatic values across the Earth. By doing so, we obtained an almost uniformly divided PET axis (Appendix 1—figure 8). Contrarily, the number of divisions of the AP axis was skewed toward low values, which resulted in a higher resolution over the presumably more relevant low-precipitation conditions (Appendix 1—figure 8). To solve the second issue, we selected the lowest number of divisions that maximized the gain in information (see Appendix 1). The optimal number of axis divisions was 17 in all cases but amphibians, where it was 18.

Next we accounted for potential commission errors, which may affect the estimated climates a species experiences. Specifically, range maps can overestimate the area occupied by a species, which directly influences the niche characterization (Rondinini et al., 2006). Because range maps typically represent the species’ maximum geographical extent (Rondinini et al., 2006; La Sorte and Hawkins, 2007), extracting the climatic values that a species range covers from a well-fitting 0.08° high-resolution climatic raster can reduce commission errors at the species range’s borders. But with many pixels inside the species range, the noise from extreme and unrepresentative climatic values can increase the error (La Sorte and Hawkins, 2007). Conversely, extracting climatic values from a coarser raster can reduce the commission errors inside a range by averaging out extreme values at the cost of increasing commission errors over the borders. To alleviate the effects of these potential errors, we first extracted the climatic values from the high-resolution rasters (0.08°). Then, we computed the average climatic values among selected raster pixels located within cells of 0.5°. In this way, we reduced the effects of commission errors both at the borders of and inside species ranges. Finally, mean climatic values may not accurately represent the cells’ climates when there is high climate variability or the values are non-normally distributed. Comparing results obtained from different cell sizes is an indirect way to asses the influence of distorted mean values. Instead, we chose to directly explore the effects of high climatic variability and non-normal distributions with non-parametric bootstrap analysis: we resampled climatic values within 0.5° cells with replacement (see below).

Niche domains and climatic regions identification

Request a detailed protocol

We employed a network community detection approach to identify the niche domains and the species mainly associated with them. For each group of species, we first generated a weighted bipartite network where species and climatic bins formed the disjoint sets of nodes, and the proportion of occurrences of species in intervals of the climatic values corresponding to the climatic bins formed the weighted links. To identify the niche domains, we used the hierarchical version of the community detection algorithm known as Infomap (Rosvall and Bergstrom, 2008). Infomap capitalizes on the minimum description length principle of information theory, which equates finding regularities and compression: the model that finds most regularities in a given set of data can compress the data the most (Rissanen, 1978). In our case, modules of highly interconnected climatic bins and species form the regularities, and describing the network with an optimal set of communities corresponds to minimizing the description length (Bernardo-Madrid et al., 2019; Rosvall and Bergstrom, 2008). Among the many community-detection algorithms available, we used Infomap because it can find hierarchically nested communities and is known for its high performance (Lancichinetti and Fortunato, 2009), also for regionalizations (Bernardo-Madrid et al., 2019; Bloomfield et al., 2018; Vilhena and Antonelli, 2015). We ran the algorithm 1000 times, selecting the network partition with the best quality.

To consider the uncertainty associated with both the species ranges and the community detection, we conducted a bootstrap analysis. For each species, we resampled with replacement from the distribution of climatic values within species ranges at a resolution of 0.08°. We averaged climatic values occurring within 0.5° cells and calculated the proportion of occurrences in each climatic bin. With resampled data from all species, we generated a bootstrapped network and clustered it with Infomap 1000 times. We repeated this procedure for 100 bootstrap networks and followed the community-stability approach proposed in Calatayud et al., 2019a to calculate the niche domains’ support. For each identified domain, we calculated the proportion of bootstrap networks with a domain more similar to Jaccard index 0.5 (Calatayud et al., 2019a).

With obtained niche domains, we detected the climatic regions by identifying areas across the Earth’s surface that hold the climatic conditions grouped within each niche domain. Finally, to compare climatic regions across tetrapod groups and with Köppen’s classification, we calculated the AMI. AMI measures the mutual information between two partitions, the classifications of raster pixels into climatic regions in this case, correcting for the similarity between partitions that are just due to chance (Vinh et al., 2010). The index is 1 when the partitions are equal and tends to 0 otherwise.

Climatic transition zones

Request a detailed protocol

The joint classification of climatic bins and species into domains D allowed us to calculate the bins’ specificity. Though species belong to single domains, typically together with the bins to which they have many and strong links, they may also have links to bins in other domains. Bins that contain species from different domains have low specificity and form a transition zone between domains (Figure 1Calatayud et al., 2019b; Bernardo-Madrid et al., 2019). We calculated the specificity SiD of a climatic bin i in domain D as the sum of link weights wi,j from bin i to species j assigned to the same domain as the bin, divided by the sum of link weights from bin i to all species j such that

(1) SiD=jDwi,jjwi,j for iD.

This index is 1 when the bin connects only to species in the same domain and tends to 0 otherwise.

We projected the specificity values into the geographic space. To all geographical raster cells q with the climate of climatic bin i, which we call raster cell set Qi, we assigned bin i’s specificity. The projected specificity is SqP.

Finally, we explored the relationship between the average SD and bootstrap support. We fitted a logistic GLMM of bootstrap p-values as a function of the mean SD and the taxonomic group. We used the R (R Development Core Team, 2018) package lme4 (Bates et al., 2015) with the mean SD as a fixed term and the taxonomic group as a random intercept term.

Geographical signal

Request a detailed protocol

To quantify the geographical signal, we compared the geographically projected specificity SP with the actual specificity based on the pool of species that co-occur geographically. Large differences between the species co-occurring in the climatic and geographic spaces indicate a strong geographic signal. For instance, when species assigned to the same domain co-occur only in a portion of the corresponding geographical space, geographical areas with mismatching species contribute to a strong geographical signal. Areas that host most of the species associated with a niche domain have a higher actual specificity than the projected indicates. Areas that are not, or only scarcely, colonized by these species have a lower actual specificity than the projected indicates.

Using Equation 1, we calculated the actual specificity of a geographical raster cell q, whose corresponding climatic bin i is in domain D (qQi and iD), as the ratio between the sum of link weights from bin i to species in raster cell q that belong to its associated domain and the total link weight from bin i to all species in q,

(2) SqA=jD,qwi,jjqwi,j for qQi and iD.

To calculate the geographical signal G, we averaged the absolute difference between the projected and actual specificity of each climatic raster cell q at 0.5° resolution,

(3) G=1Nq=1N|SqA-SqP|,

where N is the total number of raster cells. This index is 0 when there is no geographical signal and tends to 1 for high signals.

Appendix 1

Selecting the number of bins

Dividing the climatic space into an optimal number of bins is critical because it can affect subsequent classifications. On the one hand, an insufficient number of bins may destroy information. For instance, in the extreme case of using only one bin, all the species will belong to the same climatic niche. On the other hand, an excess of bins may lead to niche domains with only a few specialist species, which are not suitable for a global climate classification. The optimal value is in between these extremes. Intuitively, the differences in the species’ climatic niches increase with the number of bins. Yet, when the bins are enough to capture the main differences among species’ niches, increasing the number of bins will not increase these differences. In other words, when the number of bins is sufficient to describe the main information of the system, more bins will not significantly increase the information contained in the species niches. Therefore, when the number of bins already captures the main information, the slope of the relationship between the differences in species’ niches and the number of bins becomes less pronounced.

To measure the average differences between the niches, we used the well-known Jensen–Shannon divergence (JSD). Assuming that each species niche is a probability distribution Pi across the discretized climatic niche space,

(4) JSD=H(i=1nPin)-i=1nH(Pi)n,

 where H(P) is the Shannon entropy for the species niche P and n is the number of species. The JSD is bounded between 0, when the probability distributions are equal, and log2(n), when the distributions are completely dissimilar. To make results comparable, we normalized JSD by log2(n).

We discretized the climatic space by dividing PET and AP in quantiles, ranging from 5 to 50 and increasing in steps of 5. We used the same number n of quantiles in both variables, so the 2D climatic space has n2 bins. For each division, we computed the proportion of observations of each species in each bin and calculated the JSD using all species. Finally, we fitted a piecewise regression of the increments of JSD (ΔJSD) as a function of the number of divisions. We found that the optimal number of divisions was 17 in all cases but amphibians, where it was 18 (Appendi 1—figure 9).

Supporting figures

Appendix 1—figure 1
Tetrapod groups and Köppen’s climatic regions.

(A) Geographic location of tetrapods’ niche domains and Köppen’s climatic regions. (B) Alluvial diagram comparing the climatic regions between classifications. In the stacks of boxes, one for each classification, a box represents a climatic region, and the height of the box is proportional to the number of raster cells of the climatic region. Streamlines connecting boxes in different stacks depict the number of raster cells shared between climatic regions of different classifications. Colors according to Figure 2.

Appendix 1—figure 2
Geographic projection of the specificity of the climatic bins (SP) for the studied groups.
Appendix 1—figure 3
Distribution of the climatic conditions (black line) and the species (colored richness values) belonging to the same niche domains of amphibians.
Appendix 1—figure 4
Distribution of the climatic conditions (black line) and the species (colored richness values) belonging to the same niche domains of reptiles.
Appendix 1—figure 5
Distribution of the climatic conditions (black line) and the species (colored richness values) belonging to the same niche domains of birds.
Appendix 1—figure 6
Distribution of the climatic conditions (black line) and the species (colored richness values) belonging to the same niche domains of mammals.
Appendix 1—figure 7
Distribution of the climatic conditions (black line) and the species (colored richness values) belonging to the same niche domains of tetrapods.
Appendix 1—figure 8
Distribution of (a) annual precipitation and (b) potential evapotranspiration values.

Seventeen divisions of each climatic variable are presented with red dotted lines.

Appendix 1—figure 9
Predictions of piecewise regression of the increment in Jensen–Shannon divergence (ΔJSD) as a function of the number of divisions in the variables defining the climatic space of all studied groups.

The breakpoints and the coefficient of determinations are also provided.

Data availability

Data used is freely available in the sources detailed in the Methods section. Data generated in this study is provided as source data associated to Figures 2 and 3.

The following previously published data sets were used
    1. Roll U
    2. Feldman A
    3. Novosolov M
    4. Allison A
    5. Bauer AM
    6. Bernard R
    7. Böhm M
    8. Castro-Herrera F
    9. Chirio L
    10. Collen B
    (2017) Dryad Digital Repository
    The global distribution of tetrapods reveals a need for targeted reptile conservation.
    https://doi.org/10.5061/dryad.83s7k

References

  1. Book
    1. BirdLife
    (2015)
    Bird Species Distribution Maps of the World
    BirdLife Data Zone.
  2. Conference
    1. GBIF
    (2019)
    What is GBIF?
    The Global Biodiversity Information Facility.
  3. Report
    1. IUCN
    (2015)
    The IUCN Red List of Threatened Species
    IUCN.
  4. Book
    1. Köppen W
    2. Geiger R
    (1930)
    Handbuch Der Klimatologie
    Gebrüder Borntraeger.
  5. Software
    1. R Development Core Team
    (2018) R: A Language and Environment for Statistical Computing
    R Foundation for Statistical Computing, Vienna, Austria.
  6. Report
    1. Thornthwaite CW
    (1948)
    An Approach Toward a Rational Classification of Climate
    LWW.
  7. Conference
    1. Trabucco A
    2. Zomer RJ
    (2009)
    Global aridity index (global-aridity) and global potential evapo-transpiration (global-pet) geospatial database
    CGIAR Consortium for Spatial Information.
  8. Book
    1. Trewartha GT
    (1954)
    An Introduction to Climate
    McGraw-Hill Book Compaky, inc.
  9. Book
    1. Unep NM
    2. Thomas D
    (1992)
    World Atlas of Desertification
    Edward Arnold.

Decision letter

  1. Barnabas Daru
    Reviewing Editor; Texas A&M University-Corpus Christi, United States
  2. Christian Rutz
    Senior Editor; University of St Andrews, United Kingdom
  3. Kevin Cazelles
    Reviewer; University of Guelph
  4. Serban Proches
    Reviewer
  5. Barnabas Daru
    Reviewer; Texas A&M University-Corpus Christi, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The climatic determinants shaping animal distributions are largely unknown. Network modelling of realized niche domains of 26,000 vertebrates, accounting for the climatic conditions a species experiences within its range, reveals new classification of global climate regimes.

Decision letter after peer review:

Thank you for submitting your article "Regularities in species' niches reveal the world's climate regions" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Barnabas Daru as the Guest Reviewing Editor and Reviewer #3, and the evaluation has been overseen by Christian Rutz as the Senior Editor. The following individuals involved in the review of your submission have agreed to reveal their identity: Kevin Cazelles (Reviewer #1); Serban Proches (Reviewer #2).

The reviewers have seen each other's reviews as part of a post-review consultation session. Since there were no disagreements, we have decided to append their separate original reports to this decision letter. Please address the comments as completely as possible in a revised manuscript.

Since revisions are required before the work can be published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Reviewer #1:

My comments are meant to be constructive, and I hope they will be helpful as you revise your manuscript.

Overall opinion

In this manuscript, Calatayud et al. use a network-based approach to identify climatic regions based on tetrapods occurrence. With these in hand, they assess their congruence with the regions of the Köppen-Geiger climate classification (based on climate and plant distributions) and they also assess their specificity. Overall, I think that this study has a lot of merit and I truly value the work done. However, the current version of manuscript requires some work before being publishable. I have several major concerns that I develop below.

Lack of clarity

The lack of clarity is twofold. First, some sentences are just not clear and need to be rephrased. Below are three examples that all concern the description of the method:

– Subsection “Climatic transition zones” I had to read this several times to understand the meaning of it. A formal definition of D is missing (what are the species included in it).

– Subsection “Geographical signal” I think the equations are valid for 1 domain, but this is not made clear (e.g. how what makes index $i$ disappear from (2) to (1))

– The text in Appendix 1 is hard to read.

Because of this lack of clarity, I am not 100% sure about what has actually been done. It might just be a matter of being careful with the annotations and carefully detailing all the steps, so it might not be much work, but this needs to be done.

Second, some details are missing in critical parts. For example:

– "Using a hierarchical network clustering algorithm (Rosvall and Bergstrom, 2008, 2011)". This is a critical step and the procedure is published, but I think we need more details to understand what is done, i.e. the rational of the algorithm and why it is suited for this analysis. Also, I don't understand where the assessment of the domain's robustness is done in the Appendix.

– AMI is not (yet) frequently used in ecology so I would remind the reader what is it otherwise the reader may not understand the meaning of a AMI of 0.7 (e.g.).

My point being that the right level of methodological is not always provided. A few more sentences may do the jobs.

Why this is not all trivial?

Full disclaimer, I don't think the results are trivial. But at first sight, one may think so.

After my first read of the manuscript, I thought that the two following assertions of the Introduction were contradictory

– "abiotic conditions determine species ranges"

– "However, the fact that plant species are good indicators of general climatic conditions does not necessarily imply that such conditions shape the distribution of other organisms in the same manner"

If "abiotic conditions determine species ranges" and if the reference classification is based on plant distributions, given the variety of ecological links between plants and tetrapodes, then the distribution of tetrapodes communities (assuming there are properly defined) and the reference distribution (Köppen) should be very similar. And actually they are similar as it is showed in Figure 3. But after thinking more about this I convinced myself otherwise, partly thanks to the manuscript, but not only. And this is the issue: the authors should better explain why this is not all trivial in the Introduction, they should explain why one could expect significant discrepancies between the two distributions.

Impact statement

Importantly enough, I don't think that the impact statement : "Similarities in climatic niches of terrestrial vertebrates indicate the Earth's climate regions, which substantially differ from previous plant-based climate classifications." is a fair depiction of the results and it is actually contradictory with, inter alia, the results presented in Figure 3 (the caption of which reads "Tetrapoda groups and Köppen's climatic regions are largely congruent"). Unless I have miss something.

Discussion

I must say that the Discussion section includes various bold statements that the authors, for instance:

"Our results bring us closer to a definition of climatic regions that represent active factors for the organization and evolution of life."

Quite frankly, after reading the manuscript, I am not entirely sure why. I would rather discourage the authors from making such statements.

Reviewer #2:

This is a brilliant piece of work, filling in a massive gap in biogeography. I think for a first shot, including just one measure of energy and one for water is perfectly fine. The writing is good and the logic mostly consistent, with very slight lapses in the discussion.

Reviewer #3:

This study presents climate zones based on realized niches of the world's terrestrial vertebrates using network methods. In general, I found the analyses to be sound and conclusions valid, although I do have some few reservations. For example, the authors explored the effect of mean annual potential evapotranspiration and annual precipitation as proxies for energy and water inputs, and therefore determinants of tetrapod diversity. The choice of these metrics comes a bit “off-the-shelf” without unpacking the structural reasons for the potential of other variables such as elevation/altitude or latitude driving tetrapod niches, all of which are key determinants for tetrapod diversity. Other aspects that might be worth incorporating or at least discussing is how other regionalizations such as the recent update of Wallace's zoogeographic regions by Holt et al., 2013, fit into the climate niches identified in this study. How can they inform conservation in the real-world? I also provide some suggestions for improvement throughout the manuscript as detailed below.

Abstract:

Please remove the comparison to plants to avoid any controversy on whether or not plants capture the climate of a region than animal taxa or vice versa. The Abstract will read just fine without the last sentence.

Introduction:

The effect of measurement scale on patterns of biodiversity is increasingly becoming a topic of growing interest in ecology and not acknowledging it can lead to spurious conclusions (see some examples of addressing this topic):

Levin, S. A. The problem of pattern and scale in ecology. Ecology 73, 1943-1967 (1992).

Rahbek, C. & Graves, G. R. Multiscale assessment of patterns of avian species richness. Proc. Natl Acad. Sci. USA 98, 4534-4539 (2001).

Jarzyna, M. A. & Jetz, W. Taxonomic and functional diversity change is scale dependent. Nat. Commun. 9, 2565 (2018).

Daru, B.H., Farooq, H., Antonelli, A. & Faurby, S. (2020) Endemism patterns are scale dependent. Nature Communications 11: 2115

It would be important to discuss how the analysis done here can be sensitive to the measurement scale.

Figure 1: This is a nice figure and would help the reader get the message of the paper even without reading the entire manuscript! I like it. However, it appears that the coloring in panel D is switched. It should be the other way around.

Results

Animal vs plant climatic regions

I found the comparison of the Tetrapoda classes in this study to that of Köppen's plant-based regions strange. Köppen's plant-based regions were based on pioneer plant classifications and not necessarily based on the realized niches of the over 380,000 species of plants across the globe, whereas the current study uses carefully vetted updated range maps of over 26,000 vertebrates from the IUCN. I suggest the authors minimize such plant vs animal comparisons.

Figure 3: "Tetrapoda". Since this is the central figure of the paper, the authors should also indicate the names (or suggested names) of the 16 climatic regions directly on the map, perhaps in a separate blown-out figure.

Discussion:

In addition to the caveats already discussed, another caveat worthy of discussion is how sensitive are the climatic regions to the measurement scale along grain size, spatial extent or even taxonomic treatment? See similar comments above.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Regularities in species' niches reveal the world's climate regions" for further consideration by eLife.

Your article has been re-evaluated, and although we found it much improved, there are some remaining issues that need to be addressed before acceptance, as outlined below:

Reviewer #2:

I think the responses to reviewers have been very thorough.

Reviewer #3:

The work presented by Calatayud and co-authors on a zoogeographic classification of climate regimes is novel. The climatic determinants shaping animal distributions is largely unknown. Network modelling of realized niche domains of 26,000 vertebrates – accounting for the climatic conditions a species experiences within its range – reveals new classification of global climate regimes.

I enjoyed this second draft, just like I did the first time. The authors addressed some of my comments in the previous version of the manuscript. Thank you. However, a few other comments were not well addressed at least not entirely:

1) The issue of scale: The authors acknowledge incorporating how they have dealt with sensitivity of the analyses to the scale of measure in the Materials and methods. However, not only did I feel like my comments were not well understood, these were also not discussed in the Discussion. How do the findings vary across grain sizes or spatial extents?

For instance, the authors stated that "Extracting the climatic values that a species range covers from a high-resolution climatic raster (such as 0.08°) may reduce commission errors at the species range's borders, but otherwise increases this error. Extracting the climatic values that a species range covers from a high-resolution climatic raster (such as 0.08°) may reduce commission errors at the species range's borders, but otherwise increases this error."

While this assumption may be true to some extent, however, the authors did not test this empirically.

Another example of the need to test for scale effects is how the authors stated that: "To alleviate the effects of these potential errors, we first extracted the climatic values from the high-resolution rasters (0.08°). Then, we computed the average climatic values among selected raster pixels located within cells of 0.5 degrees. In this way, we reduced the effects of commission errors both at the borders of and inside species ranges. We also conducted a bootstrap significance test that takes the uncertainty of species ranges into account (see below)."

In terms of grain size for instance, it is well known that significant climate variation can exist even within a 0.5 degree cell (equivalent to ~50 km) that spans mountain peaks and valleys, and where temperatures can vary over 20°C. In a case where the average of 0.08 degree cells are taken within a 0.5 degree cell, is it unlikely that the average represents the climatic conditions of the grid cell. I imagine this will also be quite different at a grid resolution of 1.0 degree, 2.0 degrees, and/or 5 degrees etc. It is not clear how this was addressed in the Materials and methods and discussed in the Discussion.

In terms of the spatial extent, will the findings still hold up if this was analyzed at a regional continental scale or local country level (and varying along the grain sizes mentioned above)? Adding a test and a discussion along these lines would be important.

2) Climatic transition zones

The authors analyzed an important issue about climate transition zones which can represent places where different climate regimes admix and potentially provide source for novel species interactions, and potentially high diversity. However, the authors miss a great opportunity to return to this aspect later in the Discussion. I was hoping for some real discussion about this where you discussed regions of milder climates. Do the transition zones correspond to these regions of milder climates, and/or what do they mean?

https://doi.org/10.7554/eLife.58397.sa1

Author response

Reviewer #1:

My comments are meant to be constructive, and I hope they will be helpful as you revise your manuscript.

Overall opinion

In this manuscript, Calatayud et al. use a network-based approach to identify climatic regions based on tetrapods occurrence. With these in hand, they assess their congruence with the regions of the Köppen-Geiger climate classification (based on climate and plant distributions) and they also assess their specificity. Overall, I think that this study has a lot of merit and I truly value the work done. However, the current version of manuscript requires some work before being publishable. I have several major concerns that I develop below.

Lack of clarity

The lack of clarity is twofold. First, some sentences are just not clear and need to be rephrased. Below are three examples that all concern the description of the method:

Thanks for raising this concern. We have rephrased several parts of the methodology section for clarity and precision.

– Subsection “Climatic transition zones” I had to read this several times to understand the meaning of it. A formal definition of D is missing (what are the species included in it).

We have rewritten the section Climatic transition zones for clarity, and included a definition of D. The beginning of the section now reads:

“The joint classification of climatic bins and species into domains D allowed us to calculate the bins’ specificity. Though species belong to single domains, typically together with the bins to which they have many and strong links, they may also have links to bins in other domains. Bins that contain species from different domains have low specificity and form a transition zone between domains (Calatayud et al., 2019; Bernardo-Madrid et al., 2019)…”

– Subsection “Geographical signal” I think the equations are valid for 1 domain, but this is not made clear (e.g. how what makes index $i$ disappear from (2) to (1))

We have also rewritten the section Geographic signal and revised the equation. We define the raster cell set Qi as the set of all geographical raster cells q with the climate of climatic bin i in the end of section Climatic transition zones. The beginning of section Geographical signal now reads:

“To quantify the geographical signal, we compared the geographically projected specificity SP with the actual specificity based on the pool of species that cooccur geographically. Large differences between the species co-occurring in the climatic and geographic spaces indicate a strong geographic signal. For instance, when species assigned to the same domain co-occur only in a portion of the corresponding geographical space, geographical areas with mismatching species contribute to a strong geographical signal. Areas that host most of the species associated with a niche domain have a higher actual specificity than the projected indicates…”

– The text in Appendix 1 is hard to read.

We improved Appendix 1 following your suggestions. We believe the appendix is clearer now.

Because of this lack of clarity, I am not 100% sure about what has actually been done. It might just be a matter of being careful with the annotations and carefully detailing all the steps, so it might not be much work, but this needs to be done.

Your useful suggestions helped us clarify the methodology.

Second, some details are missing in critical parts. For example:

– "Using a hierarchical network clustering algorithm (Rosvall and Bergstrom, 2008, 2011)". This is a critical step and the procedure is published, but I think we need more details to understand what is done, i.e. the rational of the algorithm and why it is suited for this analysis.

Thanks for raising this concern. We added an explanation of the basis of the algorithm and our reasons for choosing it in the Materials and methods section.

“To identify the niche domains, we used the hierarchical version of the community detection algorithm known as Infomap (Rosvall and Bergstrom, 2008, 2011). Infomap capitalizes on the minimum description length principle of information theory, which equates finding regularities and compression: The model that finds most regularities in a given set of data can compress the data the most (Rissanen, 1978). In our case, modules of highly interconnected climatic bins and species form the regularities, and describing the network with an optimal set of communities corresponds to minimizing the description length Bernardo-Madrid et al. (2019); Rosvall and Bergstrom (2008). Among the many community detection algorithms available, we used Infomap because it can find hierarchically nested communities and is known for its high performance (Lancichinetti and Fortunato, 2009), also for regionalizations (Bernardo-Madrid et al., 2019; Bloomfield et al., 2018; Vilhena and Antonelli, 2015).”

Also, I don't understand where the assessment of the domain's robustness is done in the Appendix.

Thanks for pointing out this problem. It was a mistake since we explain the robustness analysis in the Materials and methods section. We changed accordingly.

– AMI is not (yet) frequently used in ecology so I would remind the reader what is it otherwise the reader may not understand the meaning of a AMI of 0.7 (e.g.).

Agree. We included an explanation about AMI:

“AMI measures the mutual information between two partitions, the classifications of raster pixels into climatic regions in this case, correcting for the similarity between partitions that are just due to chance (Vinh et al., 2010). The index is 1 when the partitions are equal and tends to 0 otherwise.”

My point being that the right level of methodological is not always provided. A few more sentences may do the jobs.

Why this is not all trivial?

Full disclaimer, I don't think the results are trivial. But at first sight, one may think so.

After my first read of the manuscript, I thought that the two following assertions of the Introduction were contradictory

– "abiotic conditions determine species ranges"

– "However, the fact that plant species are good indicators of general climatic conditions does not necessarily imply that such conditions shape the distribution of other organisms in the same manner"

If "abiotic conditions determine species ranges" and if the reference classification is based on plant distributions, given the variety of ecological links between plants and tetrapodes, then the distribution of tetrapodes communities (assuming there are properly defined) and the reference distribution (Köppen) should be very similar. And actually they are similar as it is showed in Figure 3. But after thinking more about this I convinced myself otherwise, partly thanks to the manuscript, but not only. And this is the issue: the authors should better explain why this is not all trivial in the Introduction, they should explain why one could expect significant discrepancies between the two distributions.

In the Introduction, we clarified that while abiotic conditions can determine species ranges, it is expectable that climatic determinants of plant and animal ranges differ due to the large physiological differences between these groups. Moreover, we supported this statement with previous evidence on different climatic correlates of diversity patterns of plants and animals. We hope the reviewer find this justification more compelling.

“…the fact that plant species are good indicators of general climatic conditions does not necessarily imply that such conditions shape the distribution of other organisms in the same manner. For instance, given the large physiological differences between plants and animals, it is expectable that the climatic determinants of their ranges will also be different. Indeed, while diversity patterns of plant and animals are to some extent congruent, the climatic correlates of these patterns vary (Whittaker et al., 2007; Qian and Ricklefs, 2008). Hence, if climatic adaptations vary among taxa, the boundaries defining their climate types will also vary.”

Impact statement

Importantly enough, I don't think that the impact statement : "Similarities in climatic niches of terrestrial vertebrates indicate the Earth's climate regions, which substantially differ from previous plant-based climate classifications." is a fair depiction of the results and it is actually contradictory with, inter alia, the results presented in Figure 3 (the caption of which reads "Tetrapoda groups and Köppen's climatic regions are largely congruent"). Unless I have miss something.

We are thankful for this comment. The reviewer is right; these two statements are contradictory. We removed the adjective phrases “substantially” and “largely congruent” so that the readers can form their own opinions.

Discussion

I must say that the Discussion section includes various bold statements that the authors, for instance:

"Our results bring us closer to a definition of climatic regions that represent active factors for the organization and evolution of life."

Quite frankly, after reading the manuscript, I am not entirely sure why. I would rather discourage the authors from making such statements.

We agree that the quoted sentence was too bold. We hedged the statement:

“Our results and data-driven methodology have the potential to bring us closer to a definition of climatic regions that represent active factors for the spatial organization of life.”

We carefully searched for other sentences that may also be controversial and removed or rephrased them.

Reviewer #2:

This is a brilliant piece of work, filling in a massive gap in biogeography. I think for a first shot, including just one measure of energy and one for water is perfectly fine. The writing is good and the logic mostly consistent, with very slight lapses in the discussion.

Reviewer #3:

This study presents climate zones based on realized niches of the world's terrestrial vertebrates using network methods. In general, I found the analyses to be sound and conclusions valid, although I do have some few reservations. For example, the authors explored the effect of mean annual potential evapotranspiration and annual precipitation as proxies for energy and water inputs, and therefore determinants of tetrapod diversity. The choice of these metrics comes a bit “off-the-shelf” without unpacking the structural reasons for the potential of other variables such as elevation/altitude or latitude driving tetrapod niches, all of which are key determinants for tetrapod diversity.

It is certainly true that other variables may turn important to define tetrapods’ niches. Indeed, we explicitly discussed this possibility at various points in the Discussion section:

“…alternative variables, such as seasonal changes in energy and precipitation (Köppen and Geiger, 1930) may also influence species inhabiting temperate regions. Including these variables might help to further separate temperate species pools across the climatic space.”

“…but other climatic variables might refine some of the least supported regions.”

Nevertheless, and in line with reviewer 2’s comments, annual precipitation and mean annual potential evapotranspiration are sufficient variables to capture water and energy aspects of species niches. These aspects are arguably the most important to understand the distribution of species in general (Hawkins et al., 2003; Currie, 1991; Tingley et al., 2009; Gouveia et al., 2014; Pie et al., 2017; Cooper et al., 2011) and, therefore, precipitation and PET are among the best variables to identify global climate regions (Thornthwaite, 1948), which is our final goal.

Finally, we believe there are reasons for not considering elevation and latitude as plausible variables in our analysis. Elevation and latitude are indeed important correlates of tetrapods diversity. Yet, it is widely accepted that they can only be correlates, but not drivers, of such diversity because these variables per se do not have direct biological meaning in most instances. Latitude and elevation are biologically indirect topo/geographical variables that can inform about variations in biologically direct environmental ones such as PET, precipitation, temperature, atmospheric pressure, and so on. The latter directly shape species niches and derived diversity patterns. Thus, while latitude and elevation may work as surrogates of climatic variables, they are not climatic variables (Körner, 2007).

Since we aim to describe climate regions, we think direct climatic variables are superior to latitude and elevation. Indeed, to our knowledge, there is no any previous study where these variables have been used or considered when delimiting climate regions and, thus, we are not totally convinced on the need of explaining their potential for our purposes.

We hope you find this arguments convincing. Otherwise we are open to further suggestions to improve our manuscript.

Other aspects that might be worth incorporating or at least discussing is how other regionalizations such as the recent update of Wallace's zoogeographic regions by Holt et al., 2013, fit into the climate niches identified in this study.

Thanks for this comment, which helped us to improve the discussion. Combining information from previous regionalizations based on species or lineages pools with the climate regions we proposed can definitively be a fruitful exercise. We added some sentences on the potential of such a combination in the Discussion section:

“Moreover, combining niche domains with bioregions based on pools of species (Bernardo-Madrid et al., 2019) or lineages (Holt et al., 2013) can also help answer several relevant questions: Are introductions more common between bioregions with the same climate types? To what extent does the current climate govern bioregions? Do ecological and evolutionary processes differ between different bioregions with same climate types (Moncrieff et al., 2015)?”

We did not directly compare these regionalizations with the climate regions, as this would require a battery of analyses and interpretations far beyond this study’s scope.

How can they inform conservation in the real-world?

This is a good question. To be more conservative regarding using the climatic regions for conservation planning in the real world, we rephrased the paragraph related to it. We changed “conservation plannings” to “conservation studies” since there are direct applications of the climatic regions for such studies. We also reformulated the question related to conservation that using the climatic regions can be answered and cited a relevant study on the topic.

“…the niche domains and their associated species pools and climatic regions can be used as a basis for ecological, evolutionary and conservation studies concerning Tetrapoda. …Are the species belonging to different niche domains similarly conserved or protected or both (Hanson et al., 2020)? …”

I also provide some suggestions for improvement throughout the manuscript as detailed below.

Abstract:

Please remove the comparison to plants to avoid any controversy on whether or not plants capture the climate of a region than animal taxa or vice versa. The Abstract will read just fine without the last sentence.

We rephrased this sentence to avoid the apparent controversy:

“…call for caution when using general climate classifications to study the ecology, evolution or conservation of specific taxa.”

Introduction:

The effect of measurement scale on patterns of biodiversity is increasingly becoming a topic of growing interest in ecology and not acknowledging it can lead to spurious conclusions (see some examples of addressing this topic):

Levin, S. A. The problem of pattern and scale in ecology. Ecology 73, 1943-1967 (1992).

Rahbek, C. & Graves, G. R. Multiscale assessment of patterns of avian species richness. Proc. Natl Acad. Sci. USA 98, 4534-4539 (2001).

Jarzyna, M. A. & Jetz, W. Taxonomic and functional diversity change is scale dependent. Nat. Commun. 9, 2565 (2018).

Daru, B.H., Farooq, H., Antonelli, A. & Faurby, S. (2020) Endemism patterns are scale dependent. Nature Communications 11: 2115

It would be important to discuss how the analysis done here can be sensitive to the measurement scale.

This a critical comment. Indeed, we were very cautious when dealing with potential scales issues at all relevant points of our analyses.

First, we explained the potential commission issues related to the grain size used to characterize the species climatic niches:

“… Extracting the climatic values that a species range covers from a high resolution climatic raster (such as 0.08°) may reduce commission errors at the species range’s borders but otherwise increases this error. Conversely, extracting climatic values from a coarser raster can reduce the influence of commission errors over the areas inside a range but increase them over the borders.”

Moreover, we applied an elaborated procedure (including a bootstrap sensibility analysis) to deal with this issue:

“…we first extracted the climatic values from the high-resolution rasters (0.08°). Then, we computed the average climatic values among selected raster pixels located within cells of 0.5 degrees. In this way, we reduced the effects of commission errors both at the borders of and inside species ranges. We also conducted a bootstrap significance test that takes the uncertainty of species ranges into account.”

Second, we explained the potential issues related to the grain size used to divide the climatic space:

“…dividing the space into too few intervals destroys information, whereas using too many divisions can generate niche domains with only a few species.”

To solve this problem, we developed the optimization procedure explained in Appendix 1.

Third, differences due to taxonomic scale are at the core of the study because we compared the climatic regions of Tetrapoda class with the ones of its subclasses.

Finally, regarding the spatial and climatic extent, we are constrained from above by the study’s global perspective. On the other hand, the potential effects of reducing the extent, which would imply focusing on particular geographic regions, are also at the core of the study because we considered the possible discrepancies between regions when analyzing the geographical signal.

For the reasons above, we think the scale issues related to our analyses have been treated and discussed adequately and extensively.

To further acknowledge the critical role scale plays for analyzing ecological patterns, and why we care so much about it in the analysis, we have added the suggested references when we begin dealing with scale in the section Niche characterization in Materials and methods.

Figure 1: This is a nice figure and would help the reader get the message of the paper even without reading the entire manuscript! I like it. However, it appears that the coloring in panel D is switched. It should be the other way around.

This is right. Thank you very much for noticing this error and the positive comments. We fixed the figure.

Results

Animal vs plant climatic regions

I found the comparison of the Tetrapoda classes in this study to that of Köppen's plant-based regions strange. Köppen's plant-based regions were based on pioneer plant classifications and not necessarily based on the realized niches of the over 380,000 species of plants across the globe, whereas the current study uses carefully vetted updated range maps of over 26,000 vertebrates from the IUCN. I suggest the authors minimize such plant vs animal comparisons.

Indeed, Köppen based his climatic regions on the distribution of vegetation types rather than on species. Still, we think this comparison is valuable because Köppen’s system is nowadays the reference climate classification. It is also the most widely used, even in studies dealing with animals. We think it would be more strange to present a new classification without comparing the previously established one. To include this information and motivate this comparison, we wrote at the beginning of the section at question:

“While Köppen’s climate classification is based on expert-knowledge on vegetation physiognomy and the distribution of vegetation types, its wide use makes it worthwhile to compare the climate regions derived from this classification system to the ones produced here.”

Moreover, to minimize the animal vs. plant comparison, we renamed the section to “Animal versus previous plant-based climatic regions”

Figure 3: "Tetrapoda". Since this is the central figure of the paper, the authors should also indicate the names (or suggested names) of the 16 climatic regions directly on the map, perhaps in a separate blown-out figure.

Agree. Thanks for this comment. We added the names of the regions in Figure 3B.

Discussion:

In addition to the caveats already discussed, another caveat worthy of discussion is how sensitive are the climatic regions to the measurement scale along grain size, spatial extent or even taxonomic treatment? See similar comments above.

As previously mentioned, we discussed and treated the sensitivity to different scale choices in other relevant places in the manuscript. We think we have adequately and extensively considered scale issues as detailed above.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #3:

The work presented by Calatayud and co-authors on a zoogeographic classification of climate regimes is novel. The climatic determinants shaping animal distributions is largely unknown. Network modelling of realized niche domains of 26,000 vertebrates – accounting for the climatic conditions a species experiences within its range – reveals new classification of global climate regimes.

I enjoyed this second draft, just like I did the first time. The authors addressed some of my comments in the previous version of the manuscript. Thank you. However, a few other comments were not well addressed at least not entirely:

1) The issue of scale: The authors acknowledge incorporating how they have dealt with sensitivity of the analyses to the scale of measure in the Materials and methods. However, not only did I feel like my comments were not well understood, these were also not discussed in the Discussion. How do the findings vary across grain sizes or spatial extents?

Thanks for raising this issue again and for the detailed explanation. We must admit that we were surprised by your concerns about scale issues during the first revision round because thorough sensitivity analysis is at the core of our work. For example, our statistical assessments go beyond those in recent studies on related topics and data (e.g. Daru et al., 2017; Holt et al., 2013; Ficetola et al., 2017; Gardner et al., 2020). With your expanded explanation and detailed examples, we realized that we had been victims of the curse of knowledge with unclear descriptions on how we dealt with scale issues. As explained below, we clarified these points throughout the text and included several lines about scale issues in the Discussion.

For instance, the authors stated that "Extracting the climatic values that a species range covers from a high-resolution climatic raster (such as 0.08°) may reduce commission errors at the species range's borders, but otherwise increases this error. Extracting the climatic values that a species range covers from a high-resolution climatic raster (such as 0.08°) may reduce commission errors at the species range's borders, but otherwise increases this error."

While this assumption may be true to some extent, however, the authors did not test this empirically.

Correct, we have not explicitly and empirically tested this argument because it aimed to alleviate uncertainties related to range maps based on most species’ theoretically expected situation. We have rewritten the paragraph to clarify this point, adding some citations to studies with similar reasoning:

“Because range maps typically represent the species’ maximum geographical extent (Rondinini et al., 2006; La Sorte and Hawkins, 2007), extracting the climatic values that a species range covers from a well-fitting 0.08° high-resolution climatic raster can reduce commission errors at the species range’s borders. But with many pixels inside the species range, the noise from extreme and unrepresentative climatic values can increase the error (La Sorte and Hawkins, 2007). Conversely, extracting climatic values from a coarser raster can reduce the commission errors inside a range by averaging out extreme values at the cost of increasing commission errors over the borders.”

Finally, we now mention this uncertainty issue in the Discussion:

“…our domains represent portions of the realized climatic niche space, which entails two sources of uncertainty. On the one hand, the estimation of realized niches depends on the data quality, scale, and treatment choices. On the other hand, the realized niche space may be influenced by historical, geographical, and biotic factors beyond pure climate (Warren et al., 2014; Calatayud et al., 2019b; Soberón, 2007). Hence, using fundamental rather than realized niches may also improve the accuracy of defining climatic regions.”

Please notice that most of the apparent issues of scale effects we are discussing are related to the estimation of realized niches. Hence, using the fundamental rather than the realized niche would be the best way to overcome them.

Another example of the need to test for scale effects is how the authors stated that: "To alleviate the effects of these potential errors, we first extracted the climatic values from the high-resolution rasters (0.08°). Then, we computed the average climatic values among selected raster pixels located within cells of 0.5 degrees. In this way, we reduced the effects of commission errors both at the borders of and inside species ranges. We also conducted a bootstrap significance test that takes the uncertainty of species ranges into account (see below)."

In terms of grain size for instance, it is well known that significant climate variation can exist even within a 0.5 degree cell (equivalent to ~50 km) that spans mountain peaks and valleys, and where temperatures can vary over 20°C. In a case where the average of 0.08 degree cells are taken within a 0.5 degree cell, is it unlikely that the average represents the climatic conditions of the grid cell. I imagine this will also be quite different at a grid resolution of 1.0 degree, 2.0 degrees, and/or 5 degrees etc. It is not clear how this was addressed in the Materials and methods and discussed in the Discussion.

We agree that this is a potential issue in the many biogeographical and macroecological studies using grid cells with the mentioned grain sizes. While averaged climatic values might not represent well the climates of such large cells, it is generally accepted that the effects of this misrepresentation mostly vanish at large scales. For example, cells in the tropics are on average warmer than cells close to the poles such that the climatic variability between cells at large scales is high enough to override the effects of the within-cell variability. But even when taking this difference into account, we agree that it is crucial to test the impact of distorted climatic means in coarse-grained grids. Using different grid resolutions and comparing the results is one way to test the robustness. To measure the effects of not only within-cell climatic variability but also non-normal distributions, instead, we used non-parametric bootstrap resampling: We resampled climatic values within cells with replacement, reclassified the data, and estimated the probability of obtaining similar climatic regions. Overall, we found high bootstrap probabilities showing that the effects of distorted climatic means from coarse-grained grids are small in our case.

We now explain this sensitivity analysis better in the Materials and methods section:

“… mean climatic values may not accurately represent the cells’ climates when there is high climate variability, or the values are non-normally distributed. Comparing results obtained from different cell sizes is an indirect way to assess the influence of distorted mean values. Instead, we chose to directly explore the effects of high climatic variability and non-normal distributions with nonparametric bootstrap analysis: We resampled climatic values within 0.5 cells with replacement (see below).”

“… we conducted a bootstrap analysis. For each species, we resampled with replacement from the distribution of climatic values within species ranges at a resolution of 0.08°. We averaged climatic values occurring within 0.5 cells and calculated the proportion of occurrences in each climatic bin. With resampled data from all species, we generated a bootstrapped network and clustered it with Infomap 1000 times. We repeated this procedure for 100 bootstrap networks and followed the community-stability approach proposed in Calatayud et al., 2019a, to calculate the niche domains’ support.”

Finally, since this issue is also related to the estimation of realized niches, we now mention them together in the Discussion.

In terms of the spatial extent, will the findings still hold up if this was analyzed at a regional continental scale or local country level (and varying along the grain sizes mentioned above)? Adding a test and a discussion along these lines would be important.

We agree that delineating climate regions at smaller extents can be of interest. However, such an analysis is beyond the scope of our manuscript’s global perspective. As explained above, the bootstrap analysis more directly assesses the issues that would motivate the analyses at different grain sizes. Taken together, and following the first revision round’s suggestions, we have mentioned this idea in the Discussion rather than adding more tests.

“…the geographical signal idea could also help unravel idiosyncratic processes occurring at smaller scales by comparing local climate regions with global or regional ones.”

2) Climatic transition zones

The authors analyzed an important issue about climate transition zones which can represent places where different climate regimes admix and potentially provide source for novel species interactions, and potentially high diversity. However, the authors miss a great opportunity to return to this aspect later in the Discussion. I was hoping for some real discussion about this where you discussed regions of milder climates. Do the transition zones correspond to these regions of milder climates, and/or what do they mean?

We appreciate this new comment as it allows us to clarify the meaning of the transition zones. We have reworded the text to explain the transition zones’ importance in our study. Specifically, we make the connection between “specificity” and “transitivity” more transparent. We believe the Results and Discussion now provide clear answers to your questions:

“Climatic conditions corresponding to diffuse transitions should present low specificity levels to the domain where they belong (Figure 1). Hence, we can consider specificity as the opposite of transitivity… As expected, lower specificity values were in general associated with the boundaries of the climatic regions (Figure 4A and Appendix 1—figure 2). Beyond boundaries, our results also revealed that harsh conditions, such as desert and continental-polar climates (EHWL and ELWL), present the highest specificity levels, regardless of the group (Figure 4A and Appendix 1—figure 2), reflecting the difficulty of colonizing these climates. Conversely, temperate regions showed the lowest levels of specificity.”

In the Discussion, we now write:

“Conversely, milder climatic conditions, especially temperate climates, showed the lowest statistical support and congruence across groups, as well as the highest transitivity. These results indicate that temperate climates are more difficult to classify (low support and congruence) due to the overlap in the climatic space of species pools with different climatic optima (high transitivity).”

Finally, we mention that:

“…the exact causes and consequences of the climatic transition zones call for

future investigation.”

We prefer to be conservative and avoid a detailed but speculative discussion about the causes and consequences of transition zones. Such claims would require an in-depth empirical evaluation of different causal factors beyond the scope of this study.

https://doi.org/10.7554/eLife.58397.sa2

Article and author information

Author details

  1. Joaquín Calatayud

    1. Integrated Science Lab, Department of Physics, Umeå University, Umeå, Sweden
    2. Departamento de Biología, Geología, Física y Química inorgánica, Universidad Rey Juan Carlos, Madrid, Spain
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    j.calatayud.ortega@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9056-4149
  2. Magnus Neuman

    Integrated Science Lab, Department of Physics, Umeå University, Umeå, Sweden
    Contribution
    Conceptualization, Resources, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Alexis Rojas

    Integrated Science Lab, Department of Physics, Umeå University, Umeå, Sweden
    Contribution
    Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1063-9102
  4. Anton Eriksson

    Integrated Science Lab, Department of Physics, Umeå University, Umeå, Sweden
    Contribution
    Formal analysis, Validation, Visualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Martin Rosvall

    Integrated Science Lab, Department of Physics, Umeå University, Umeå, Sweden
    Contribution
    Resources, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared

Funding

Carl Tryggers Stiftelse för Vetenskaplig Forskning

  • Joaquín Calatayud

Stiftelsen Olle Engkvist Byggmästare

  • Martin Rosvall

Svenska Forskningsrådet Formas

  • Magnus Neuman
  • Martin Rosvall

Swedish Research Council (2016–00796)

  • Martin Rosvall

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are thankful to Andrea Briega and Miguel Á. Rodríguez for discussion on early ideas. We are very grateful to Fernanda Alves-Martins, Rafaél Molina-Venegas, Cristina Roquillo, Rubén Bernardo-Madrid, Ana G Garcia, Daniel Fuentes, and Marina Coca for critical reviews. JC was supported by the Carl Tryggers Foundation for Scientific Research (CTS 16:384), AR by the Olle Engkvist Byggmästare Foundation, and MR by the Swedish Research Council, grant 2016–00796.

Senior Editor

  1. Christian Rutz, University of St Andrews, United Kingdom

Reviewing Editor

  1. Barnabas Daru, Texas A&M University-Corpus Christi, United States

Reviewers

  1. Kevin Cazelles, University of Guelph
  2. Serban Proches
  3. Barnabas Daru, Texas A&M University-Corpus Christi, United States

Publication history

  1. Received: April 29, 2020
  2. Accepted: February 7, 2021
  3. Accepted Manuscript published: February 8, 2021 (version 1)
  4. Version of Record published: March 16, 2021 (version 2)

Copyright

© 2021, Calatayud et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 838
    Page views
  • 147
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Ecology
    2. Evolutionary Biology
    Heather E Machado et al.
    Research Article

    To advance our understanding of adaptation to temporally varying selection pressures, we identified signatures of seasonal adaptation occurring in parallel among Drosophila melanogaster populations. Specifically, we estimated allele frequencies genome-wide from flies sampled early and late in the growing season from 20 widely dispersed populations. We identified parallel seasonal allele frequency shifts across North America and Europe, demonstrating that seasonal adaptation is a general phenomenon of temperate fly populations. Seasonally fluctuating polymorphisms are enriched in large chromosomal inversions and we find a broad concordance between seasonal and spatial allele frequency change. The direction of allele frequency change at seasonally variable polymorphisms can be predicted by weather conditions in the weeks prior to sampling, linking the environment and the genomic response to selection. Our results suggest that fluctuating selection is an important evolutionary force affecting patterns of genetic variation in Drosophila.

    1. Ecology
    2. Evolutionary Biology
    David V McLeod, Sylvain Gandon
    Research Article Updated

    The evolution of multidrug resistance (MDR) is a pressing public health concern. Yet many aspects, such as the role played by population structure, remain poorly understood. Here, we argue that studying MDR evolution by focusing upon the dynamical equations for linkage disequilibrium (LD) can greatly simplify the calculations, generate more insight, and provide a unified framework for understanding the role of population structure. We demonstrate how a general epidemiological model of MDR evolution can be recast in terms of the LD equations. These equations reveal how the different forces generating and propagating LD operate in a dynamical setting at both the population and metapopulation levels. We then apply these insights to show how the LD perspective: (i) explains equilibrium patterns of MDR, (ii) provides a simple interpretative framework for transient evolutionary dynamics, and (iii) can be used to assess the consequences of different drug prescription strategies for MDR evolution.