Philippa Kaur

The ecological baseline in TirezThe geology and the climate of the Tirez region favored the generation and maintenance of a type of hypersaline habitat characterized by extreme seasonality: the sulfate-chlorine waters, with sodium and magnesium cations, showed significant seasonal variations15. The alkaline pH, the low oxidant value for the redox potential of the water column and the highly reduced sediments imposed extreme conditions (see Table 1 and Supplementary Information for details). This extreme seasonality requires to define a valid representative ecological baseline to compare the ecology of the lagoon between 2002 and 2021 and, in this way, set the basis to proposing our model of ecological succession with increasing dryness as a “time-analog” for early Mars. Taxonomic data from 2002 is a snapshot of the community during one season, so we include in our discussion the results presented by Montoya et al. (2013) from a sample campaign carried out in 2005, because they16 analyzed both water and sediment and during both the wet and dry seasons.We consider here only the results obtained by Montoya et al.16 by gene cloning, since those obtained by isolation and sequencing are not comparable. At the level of large groups, no major seasonal differences were observed: Pseudomonadota, followed by Bacteroidetes, were the dominant phyla, in both water and sediments, and both in the dry and the rainy seasons; although Alphaproteobacteria was the dominant class in water, while Gammaproteobacteria was dominant in sediments (in both dry and rainy seasons). With respect to the archaeal domain, all the identified sequences were affiliated to Halobacteriales order, mainly Halorubrum (water) and Halobacterium (sediment), both within Halobacteriaceae family. We can consider these results presented in Montoya et al.16 as the “ecological baseline” for Tirez, however taken with a grain of salt, because only 43 bacterial and 35 archaeal sequences, including rainy and dry seasons and water and sediments, were considered for analysis.Prokaryotic diversity in 2002As can be expected for an extreme environment, the bacterial diversity detected in 2002 was low, although we cannot exclude the possibility that this may reflect the limitation of DNA sequencing techniques at the time. 59% of the obtained clones in the then-wet sediments corresponded to the Malaciobacter genus. Malaciobacter (previously named17 Arcobacter) is an aerotolerant Epsilonproteobacteria. Species within this genus are moderately halophilic, e.g., M. halophilus, capable to grow in up to 4% NaCl. Even though the role that Malaciobacter can play in the environment is not known, it seems to thrive in aquatic systems, like sewage, with a high organic matter content17: e.g., M. canalis, M. cloacae, or M. defluvii.After Malaciobacter-like clones, the next most numerous group belongs to the phylum Bacillota (27% of the sequenced clones; Table 2). Under stressful environmental conditions, members of the genus Virgibacillus produce endospores, a very useful property in an extreme and variable environment (ionic strength, temperature, light intensity), easy to compare with early Mars. Endospores facilitate species survival, allowing them to overcome drastic negative changes, like dry periods, and to germinate when the conditions are favorable again. The closest identified species was the halotolerant V. halodenitrificans, but with low homology, not far from other halotolerant (e.g., V. dokdonensis) or halophilic (e.g., V. marismortui) species within the same genus. The other Gram-positive clones belong to the order Clostridiales. These clones cluster in two taxonomic units related with the strictly anaerobic genus Tissierella.Despite the abundance of Pseudomonadota, their biodiversity was very low, reduced to only two genera within the Epsilon- and Delta-proteobacteria. Six sequences affiliated to Deltaproteobacteria, and clustered in one OTU (salB38, similarity 96.6% with Desulfotignum), were retrieved. Its presence in anaerobic media rich in sulfates (Table 1) seems reasonable. In fact, sulfate-reducing activity was detected using a specific enrichment assay.Finally, one taxon belonging to the phylum Spirochaetota (previously named Spirochaetes) was identified. The presence of Spirochaetota in this system is not strange because members of the genus Spirochaeta are very often found in mud and anaerobic marine environments rich in sulfates18. Moreover, the closest species to SalB63, although with a low similarity of 87%, was Spirochaeta bajacaliforniensis, a spirochete isolated19 from a microbial mat in Laguna Figueroa (Baja California), an extensive hypersaline lagoon with high gypsum content, very similar, although much bigger, than Tirez lagoon.The diversity within the domain Archaea was very low in 2002. The phylogenetic analysis of 96 clones indicate that they correspond to one specie belonging to the obligate halophile genus Methanohalophilus. Their high similarity (99.3%) with several species of Methanohalophilus, such as M. portucalensis (isolated from sediments of a solar saltern in Portugal), M. mahii (isolated from sediments of the Great Salt Lake), or M. halophilus (isolated20 from a cyanobacterial mat at Hamelin Pool, Australia), makes impossible its adscription to any particular species level. Methanohalophilus is strictly methylotrophic, which is consistent with this environment, given that the methylotrophic methanogenesis pathway, non-competitive at low-salt conditions, is predominant at high saline concentrations21. We further confirmed methanogenic activity in Tirez by the measurement of methane by gas chromatography in enrichment cultures.Prokaryotic diversity in context of other studies between 2002 and 2021It was challenging to establish a timeline for the succession of the populations involved, because the scarcity of data harvested and published so far from Tirez. However, combining our results with the few data available in Montoya et al.16 and Preston et al.22 on samplings carried out on 2005 and 2017, respectively, we can see a clear predominance of the phylum Pseudomonadota: Epsilonproteobacteria, i.e. Arcobacter-like, and Deltaproteobacteria, mainly sulfate-reducing bacteria (this work, sampling 2002), and Gammaproteobacteria16 when Tirez maintained a water film, to eventually a final predominance of Gammaproteobacteria, e.g. Chromatiales and Pseudomonadales, in the dry Tirez (this work, 2020 sampling). The Bacillales order has remained widely represented both in the wet and dry Tirez.Regarding the archaeal domain, the few references available (Refs.16,22; this work) confirm that the members of the Halobacteriaceae family are well adapted to both the humid and dry ecosystems of Tirez, being predominant in both conditions. Preston et al.22 found that the second most abundant group of archaea in the dry sediments of Tirez was the Methermicoccaceae family, within the Methanosarcinales order, Methanomicrobia class. Taking into account the results obtained in the dry Tirez (Preston et al.22; and this work, sample 2020), the methanogenic archaea have decreased drastically through time, probably due to salt stress and the competition with sulfate-reducing bacteria.Prokaryotic diversity in 2021From a metabolic point of view, most of the bacteria present today in the sediment are chemoorganotrophs, anaerobes, and halophilic or halotolerant. Scarce information is available about the predominant OTU, Candidate Division OP1. The OP1 division was one of the main bacterial phyla in a sulfur-rich sample in the deepest analyzed samples from the Red Sea sediments under brine pools23. In addition, the phylogenetically related Candidate division KB1 has been observed in deep-sea hypersaline anoxic basins at Orca Basin (Gulf of Mexico), and other hypersaline environments24. Eight of the nine genera identified show coverage greater than 1% of the sequences: i.e., Rubinisphaera, Halothiobacillus, Thiohalophilus, Anaerobacillus/Halolactibacillus, Halomonas, Halothermothrix, and Aliifodinibius are halophilic or halotolerant genera13,25.Regarding archaea, our analyses reveal archaeal groups that seem to thrive in sediments from extreme environments, e.g., marine brine pools/deep water anoxic basins or hypersaline lakes. The most abundant OTU, Thermoplasmata KTK4A, was found prominent and active in the sediment of Lake Strawbridge, a hypersaline lake in Western Australia26, and in soda-saline lakes in China27. The creation of a Candidatus Haloplasmatales, a novel order to include KTK4A-related Thermoplasmata, has been proposed27. On the other hand, both in the aforementioned soda-saline lakes in China27 and in a sulfur-rich section of the sediments from below the Red Sea brine pools23, retrieved sequences were assigned to Marine Benthic Groups B, D, and E. Finally, in the section of nitrogen-rich sediments from the aforementioned Red Sea brine pools, the unclassified lineage ST-12K10A represented the most abundant archaeal group. In the Tirez Lagoon sediment after desiccation, all Methanomicrobia readings belonged to this group.The significance and implications of an ecosystem characterized in 2021 by high diversity, high inequality, and lack of isolated representatives, resides in that Tirez is today an ecosystem in which many (most) of the species/OTUs present are dormant, and they do not play any metabolic role. Hence the high percentage of raretons, greater than 80% for both bacteria and archaea, which are actually present in the lagoon but with only one or two copies each. Only those species adapted to the conditions imposed by the extreme environment are able to actually thrive, and consequently only a few species carry out all the metabolic activity. We conclude that the microbiota in Tirez today represents an ecosystem with a high resilience capacity in the face of environmental changes that may occur.We want to clearly highlight that the technique available in 2002 to study the microbiota of the Tirez lagoon only allowed to obtain a low-resolution image, but that was the state-of-the-art procedure at the time, and the Tirez lagoon cannot be sampled again with the conditions back in 2002, which no longer exist and are not expected to return. Although we have kept in storage several samples of water and sediment from the 2002 Tirez lagoon, it is reasonable to assume that those laboratory microcosms would have chemically and microbiologically changed during the last 20 years, and as such no longer represent reliable replicas of the original lagoon, so we cannot use them for the purposes of this work. Therefore, we are aware that any comparisons of the 2002 laboratory results with the much more robust results obtained by Illumina in 2021 need to be taken with a grain of salt. With all the precautions required, in a high-level, first-order comparison, the most noticeable difference between 2002 and 2021 is a drastic change in the microbial Tirez population. Only some OTUs within Bacillales (Virgibacillus/Anaerobacillus), sulfate-reducing Deltaproteobacteria (Desulfotignum/Desulfobacteraceae-Desulfovibrio), and Spirochaetes are shared among the 2002 and 2021 samples. This comparison is enough for the purposes of this work, as we are interested in the evolution of the lagoon system as a whole to establish a “time-analog” with the wet-to-dry transition on early Mars, and not in the particular outcome of each and every OTU in Tirez. With the results at hand, we conclude that, since 2002, the lacustrine microbiota has shifted to one more adapted to the extreme conditions in the dry sediments, derived from the gradual and persistent desiccation concluding ca. 7 years ago (i.e., completely desiccated in 2015), such as lack of light, absence of oxygen, and lack of water availability. This shift has likely been triggered because organisms that were originally in the lagoon but at low abundance in 2002 became dominant as they were better adapted to desiccation, and because the incoming of new microorganisms transported by birds or wind28.Lipid biomarkers analysis of the desiccated lake sedimentsThe analysis of cell membrane-derived lipid compounds on the dry lake sediments at present allow to provide another perspective of the microbial communities inhabiting the Tirez lagoon, by contributing additional information about the ecosystem and depositional environment. It is important to note that, analyzing only the 2021 lake sediments, we cannot differentiate between lipidic biomarkers of the microorganisms inhabiting Tirez in 2002 and before from those left behind by the microorganisms living in the dried sediments today. Instead, the analyses of lipid biomarkers provide clues about the different microorganisms that have populated Tirez through time, including both older communities inhabiting the former aqueous system and also younger communities better adapted to the present dry conditions. Thus, the lipid biomarkers analysis can be considered as a time-integrative record of the microbial community inhabiting Tirez during the last decades.Based on the molecular distribution of lipid biomarkers, the presence of gram-positive bacteria was inferred from the relative abundance of the monounsaturated alkanoic acid C18:1[ω9], or iso/anteiso pairs of alkanoic acids from 12 to 17 carbons29 with dominance of i/a-C15:0 and i/a-C17:0 (Fig. 3B). In contrast, generally ubiquitous alkanoic acids such as C16:1[ω7], C18:1[ω7], or C18:2[ω6] suggested a provenance rather related to gram-negative bacteria30. The combined detection of the i/a-C15:0 and i/a-C17:0 acids, with dominance of the iso over the anteiso congeners, together with other biomarkers such as the mid-chain branched 10Me16:0, the monounsaturated C17:1, or the cyclopropyl Cy17:0 and Cy19:0 acids, may be associated with a community of SRB31 in today´s dry sediments of Tirez. Specifically, most of those alkanoic acids have been found in a variety of Deltaproteobacteria and/or Bacteroidota (previously named Bacteroidetes). The presence of archaea was deduced from the detection of prominent peaks of archaeol in the polar fraction32 (Fig. 3C), as well as squalene and relatives (dihydrosqualene and tetrahydrosqualene) in the apolar fraction33 (Fig. 3A). Squalene and a variety of unsaturated derivatives are present in the neutral lipid fractions of many archaea with high abundances in saline lakes34. The relative abundance of autotrophs over heterotrophs35 can be estimated by the ratio of the autotrophically-related pristane and phytane over the both autotrophically- and heterotrophically-produced n-C17 and n-C18 alkanes ([Pr + Ph]/[n-C17 + n-C18]). A ratio of 0.56 in the Tirez sediments suggest the presence of a relevant proportion of heterotrophs in the ancient lacustrine system.Furthermore, the lipid biomarkers analysis was able to detect compounds specific of additional microbial sources, such as cyanobacteria36 (n-C17, C17:1, or 7Me-C15 and 7Me-C17), microalgae and/or diatoms (phytosterols37; or C20:5, and C22:6 alkanoic acids30), and other photoautotrophs (phytol and potentially degradative compounds such as pristane and phytane31). A relatively higher preservation of the cell-membrane remnants (i.e., lipids) compared to the DNA-composing nucleic acids may contribute to explain the lack of detection of cyanobacteria, diatoms and microalgae, and other phototrophs by DNA analysis (a deficit in our results shared with Montoya et al.16, and Preston et al.22). Although abundant in higher plants38, sterols such as those detected here (i.e., the sterols campesterol, stigmasterol, and β-sitosterol, as well as ergosterol) are also major sterols in some microalgal classes37 (such as Bacillariophyceae, Chrysophyceae, Euglenophyceae, Eustigmatophyceae, Raphidophyceae, Xanthophyceae, and Chlorophyceae), cyanobacteria (β-sitosterol), and fungi (ergosterol39).The carbon isotopic composition of lipid biomarkers provides a rapid screening of the carbon metabolism in a system, by recognizing the principal carbon fixation pathways used by autotrophs. The range of δ13C values measured in the Tirez sediments (from − 33.9 to − 16.1‰) denotes a mixed use of different carbon assimilation pathways, involving mostly the reductive pentose phosphate (a.k.a. Calvin–Benson–Bassham or just Calvin) cycle (from − 19 to − 30‰), and in lesser extent the reductive acetyl-CoA (a.k.a. Wood–Ljungdahl) pathway (from − 28 to − 44‰), and/or the reverse tricarboxylic acid (rTCA) cycle (from − 12 to − 21‰).The lipids synthesized by microorganisms using the Calvin or reductive acetyl-CoA pathway are typically depleted relative to the bulk biomass, particularly those produced via de latter pathway. In the dry Tirez sediments, the majority of the lipid compounds are more depleted in 13C than the bulk biomass (Fig. 4). In particular, the branched alkane DiMeC18 (Fig. 4A) and the SRB-indicative 10Me16:0 acid (Fig. 4B) showed the most depleted δ13C values and suggested the use of the reductive acetyl-CoA pathway. The rest of lipid compounds showed isotopic signatures (from − 16.1 to − 31.4‰) compatible with the prevalence of the Calvin pathway. These values may directly reflect the autotrophic activity of microorganisms fixing carbon via the Calvin cycle or heterotrophic activity of microorganisms growing on their remnants. Thus, the saturated and linear alkyl chains of lipids (i.e., n-alkanes, n-alkanoic acids, and n-alkanols) showing the most negative δ13C values (e.g., alkanes n-C17 and C17:1; or acid C18:1[ω7]) reflect prokaryotic sources of Calvin-users autotrophs (e.g., cyanobacteria or purple sulfur bacteria), while the rest of compounds with slightly less negative δ13C values instead stem from the autotrophic activity of eukaryotes also users of the Calvin cycle (unsaturated fatty acids and sterols) or from the metabolism of heterotrophs such as SRB (iso/anteiso-, other branched, and cyclopropyl fatty acids) and haloarchaea (isoprenoids, phytanol, and archaeol). All in all, the compound-specific isotope composition of the dry sediments in the today´s Tirez lagoon may indirectly reflect the prevailing autotrophic mechanisms in the present lacustrine system of Tirez, by showing isotopic signatures of secondary lipids similar to their carbon source40.In addition, the use of a number of lipid molecular ratios or proxies allow further characterization of the lacustrine ecosystem and depositional environment. For example, the average chain length of the n-alkanes (24.1) suggests a relevant presence of eukaryotic biomass in the lacustrine sediments, since long-chained alkanes ( > C20) are known to originate from epicuticular leaf waxes in higher plants41. Highlighting the relevance of eukaryotes and their ecological roles is one of the major contributions of this work, because previous studies on the microbial ecology of hypersaline environments have been focused primarily on prokaryotes42.The proportion of odd n-alkanes of high molecular chain (i.e., n-C27, n-C29, and n-C31) over even n-alkanes of low molecular chain (i.e., n-C15, n-C17, and n-C19) provides an estimate of the relative abundance of terrigenous over aqueous biomass43, which in Tirez is TAR = 1.8. The Paq index may also be used to differentiate the proportion of terrigenous versus aquatic (emergent and submerged) plant biomass44. A Paq of 0.3 in the Tirez sediments from 2021 supported the relative abundance of land plants. Finally, the depositional environment in the lacustrine system of Tirez may be also characterized analyzing the ratio of pristane over phytane (Pr/Ph), which is higher than 1 when phytol degrades to pristane under oxic conditions45. Assuming that both isoprenoids in the Tirez sediments derived from phytol31, according to their similarly depleted δ13C (Fig. 4A), we can conclude that the sediments in the Tirez lagoon were deposited under predominantly oxic conditions (i.e., Pr/Ph ratio of 1.1).In summary, the lipid biomarkers study revealed useful information about the depositional environment and lacustrine ecosystem, including the presence of active or past autotrophic metabolisms involving prokaryotes (e.g., cyanobacteria and purple sulfur bacteria) and eukaryotes (plants, diatoms and other microalgae), as well as heterotrophic metabolisms of likely SRB and haloarchaea growing on Calvin-users exudates. These results are quite in agreement with the microbial community previously reported16 in sediments from the wet and dry seasons: abundant Gammaproteobacteria and Alphaproteobacteria, together with Algae and Cyanobacteria, dinoflagellates and filamentous fungi, Bacillota, Actinomycetota (previously named Actinomycetes), and a halophilic sulfate-reducing Deltaproteobacteria.Tirez as the first astrobiological “time-analog” for early MarsEarly Mars most likely had a diversity of environments in terms of pH, redox conditions, geochemistry, temperature, and so on. Field research in terrestrial analog environments contribute to understand the habitability of this diversity of environments on Mars in the past, because terrestrial analogues are places on Earth characterized by environmental, mineralogical, geomorphological, or geochemical conditions similar to those observed on present or past Mars9. Therefore, so far analogs have been referred to terrestrial locations closely similar to any of the geochemical environments that have been inferred on Mars, i.e., they are “site-analogs” that represent snapshots in time: one specific environmental condition at a very specific place and a very specific time. Because of this, each individual field analog site cannot be considered an adequate representation of the changing martian environmental conditions through time. Here we introduce the concept of astrobiological “time-analog”, referred to terrestrial analogs that may help understand environmental transitions and the related possible ecological successions on early Mars. In this sense, they should be “time-resolved analogs”: dynamic analog environments where we can analyze changes over time. To the best of our knowledge, this is the first study that looks at the environmental microbiology of a Mars astrobiological analog site over a significant and long period of change, and try to understand the ecological successions to put them in the context of martian environmental evolution.As Mars lost most of its surface water at the end of the Hesperian5,9,12, this wet-to-dry global transition can be considered the major environmental perturbation in the geological history of Mars, and therefore merits to be the first one to be assigned a “time-analog” for its better understanding and characterization. The drying of Mars was probably a stepwise process, characterized by multiple transitions between drier and wetter environments12,47, and therefore the seasonal fluctuations and eventual full desiccation of Tirez represent a suitable analog to better understand possible ecological transitions during the global desiccation of most of the Mars’s surface before the Amazonian (beginning 3.2 Ga).To introduce Tirez as the first Mars astrobiological “time-analog” of the wet-to-dry transition on early Mars, the objective of this study was threefold: first, we wanted to identify the dominant prokaryotic microorganisms in the active Tirez lagoon 20 years ago, a unique hypersaline ecosystem with an ionic composition different from that of marine environments, and therefore potentially analogous to ancient saline lacustrine environments on Mars during the Noachian and into the Hesperian46,47. Our results provide a preliminary basis to hypothesize how the microbial communities on the Noachian Mars could have developed in salty environments with dramatically fluctuating water availability. The requirement to deal with important variations in ionic strength and water availability, involving at times the complete evaporation of the water, could have represented additional constraints48 for microorganisms on early Mars.The second objective of this investigation was the identification of the microbial community inhabiting the desiccated Tirez sediments today, after all the water was lost, as a potential analog to desiccated basins on Mars at the end of the Hesperian1,3,4,47. Our results suggest that hypothetical early microbial communities on early Mars, living with relative abundance of liquid water during the Noachian, would have been forced to adapt to increasingly desiccating surface environments, characterized by extreme conditions derived from the persistent dryness and lack of water availability. Our investigation in Tirez suggest that hypothetical microorganisms at the end of the Hesperian would have needed to evolve strategies similar to those of microorganisms on Earth adapted to living at very low water activity49, to thrive in the progressively desiccating sediments.And the third objective of this investigation was the identification of the lipidic biomarkers left behind by the microbial communities in Tirez, as a guide to searching and identifying the potential leftovers of a hypothetical ancient biosphere on Mars. Lipids (i.e., fatty acids and other biosynthesized hydrocarbons) are structural components of cell membranes bearing recognized higher resistance to degradation relative to other biomolecules, thus with potential to reconstruct paleobiology in a broader temporal scale than more labile molecules50. Our results reinforce the notion that lipidic biomarkers should be preferred targets in the search for extinct and/or extant life on Mars precisely because they are so recalcitrant. More

Our approach uses hypothetical 30 cm fixed depth samples taken at three successive time points (t0, t1, and t2) with prescribed changes in SOC (1.4% to 1.6%) and BD (1.5–1.1 g cm−3) over these time points (Table 1). The 30 cm soil depth is the common international standard for sampling and analysis required for SOC stock assessment and adhered to by carbon accounting and market organizations6,18. The changes we adopted (a 27% decrease in BD and a 14% increase in SOC) while relatively large, are consistent with those reported in the literature. For example, Reganold and Palmer reported a 25% decrease in BD (1.2–0.9 g cm−3) in neighboring farms with differing management practices23, and Syswerda et al. observed a 17% increase in SOC concentration (10.4–12.2 g C kg soil−1) when converting from a conventionally to organically managed row crop rotation21.Table 1 Hypothetical changes in bulk density (BD) and soil organic carbon (SOC) concentration in 30 cm fixed depth samples at time points t0, t1 and t2 along with calculated values of SOC stock and total soil mass and mineral soil mass.Full size tableIn Table 1, the total soil mass, mineral soil mass, and the SOC stock of the fixed depth samples were calculated by equations as described in the introduction from our prescribed changes in BD and SOC values.ScenariosWe compared hypothetical ESM correction scenarios with our 30 cm fixed depth sample at each time point (Table 2, Figs. 2, 3).Table 2 Hypothetical ESM scenarios showing variation with depth for bulk density (BD) and soil organic carbon (SOC) at each sampling time point, along with the sample depth intervals investigated.Full size tableFigure 2Flow chart of the definition, sampling, and SOC stock correction for a theoretical data set at time points t0, t1, and t2 for scenarios s1 with linear distributions of BD and SOC and s2 with a linear increase in BD and exponential decrease in SOC with depth. Scenario s2 is sampled at (a) 10 cm, (b) 15 cm, and (c) 30 cm intervals.Full size imageFigure 3Scenarios (S1 and S2), showing (a) bulk density variation (BD, g cm−3), and (b) soil organic carbon (SOC, %) variation by depth (0–30 cm) at each time point (t0, t1, and t2). For scenario 2, the single 30 cm depth interval was used (2c). See Table 1 and 2 for details.Full size imageScenario 1We carried out the ESM correction on a 30 cm sample and assumed that the sample was homogenous throughout the profile, with constant SOC and BD values at each time point.To correct for the error in SOC stock estimation when using fixed depth soil sampling, we used  Eqs.2a, 2b and 2c that consider changes in BD28,35. The adjusted soil depth resulting from the change in BD is calculated as:$${mathrm{M}}_{mathrm{n}}= {mathrm{M}}_{mathrm{i}}$$
(2a)
$${mathrm{D}}_{mathrm{a}}*{mathrm{BD}}_{mathrm{n}}*left(1-mathrm{k}*{mathrm{SOC}}_{mathrm{n}}right)={mathrm{D}}_{mathrm{i}}*{mathrm{BD}}_{mathrm{i}}*left(1-mathrm{k}*{mathrm{SOC}}_{mathrm{i}}right)$$
(2b)
$${mathrm{D}}_{mathrm{a}}={mathrm{D}}_{mathrm{i}}*frac{{mathrm{BD}}_{mathrm{i}}}{{mathrm{BD}}_{mathrm{n}}}*frac{1-mathrm{k}*{mathrm{SOC}}_{mathrm{i}}}{1-mathrm{k}*{mathrm{SOC}}_{mathrm{n}}}$$
(2c)
where Mi = Initial mineral soil mass per area (left[frac{M}{{L}^{2}}right]) , Mn = New mineral soil mass per area (left[frac{M}{{L}^{2}}right]) , Da = Adjusted soil surface depth (left[Lright]) , BDi = Initial bulk density (left[frac{M}{{L}^{3}}right]) , BDn = New bulk density (left[frac{M}{{L}^{3}}right]) , SOCi = Initial SOC as a decimal percent (left[frac{M}{M}right]) , SOCn = New SOC as a decimal percent (left[frac{M}{M}right]) , Di = Initial depth (left[Lright]).To conform with Eq. (2a), an increase in SOC over time results in a displacement of some soil mineral mass from the sample, whereas a decrease in SOC over time requires some soil mineral mass to be replaced34. Multiplying the BD by the mineral fraction of the soil (left(1-mathrm{k}*{mathrm{SOC}}right)) for each time point allowed us to compare equivalent mineral mass28. The effect of a change in SOC on mineral mass is small, with a 1% change in SOC equating to approximately a 2% change in apparent depth. This adjustment relates SOC per unit of mineral mass of the fine fraction ( 2 mm)20. The corrected apparent depth can then be used to calculate the corrected SOC stock of a single layer, fixed depth sample (Eq. 3).$$SO{C}_{stock}={D}_{a}*BD*SOC$$
(3)
Scenario 2In ESM correction scenarios 2a, 2b, and 2c, we imposed variable, dynamic BD and SOC values with depth over time (Table 2, Figs. 2, 3). To investigate these profiles, we determined the SOC and BD values throughout the soil depth by separating the soil into one (1) cm depth increments (i.e., 0–1 cm, 1–2 cm, etc.). We refer to this calculated incremental profile as the scenario 2 baseline. We assumed that our prescribed SOC concentration varied with depth following an exponential decay. To represent this decay, we simulated the global average distribution of SOC concentration with depth on crop land36, following the distribution from Hobley and Wilson37 (Eq. 4),$$SOCleft(dright)=SO{C}_{infty }+left(SO{C}_{o}-SO{C}_{infty }right)times {e}^{-dk}$$
(4)
where SOC (d) is the SOC concentration at depth (d), ({SOC}_{infty }) is the infinity SOC concentration, SOC0 is the SOC concentration at the soil surface, and k is the decay rate. We solved for the decay rate, initial SOC0, and infinity ({SOC}_{infty }) to fit the global average distribution for the 30 cm profile36 and then scaled the SOC concentration to our 30 cm fixed depth sample’s average SOC (1.4%) at t0 (Fig. 3).In scenarios 2a, 2b, and 2c, the BD increased linearly with depth38,39. At the initial time point (t0), we varied the BD values by ± 10% of the BD average over the 30 cm depth, such that for example, BD at t0 (profile average of 1.5 g cm−3) was 1.35 g cm−3 and 1.65 g cm−3 for the upper (0–1 cm) and lower (29–30 cm) depth increment, respectively. For each sequential time point, as the average BD decreased, the soil expanded. To determine the expansion, the depth of the initial sample (e.g., at t0) that filled the 30 cm depth in the subsequent sample (e.g., at t1) was calculated as the initial depth multiplied by the ratio of the average initial BD over the average new BD (e.g., 1.5/1.3 = 1.15 for t0/t1).The linear increase in BD with depth of each following time point maintained the average BD of scenario 1. We then varied the new BD by ± the percent change in the average BD between the time periods (see annotated scripts “main.R” and “functions.R” in Supplementary Material 1 for the development of the theoretical dataset). We then divided each initial BD increment (using soil mass for every 1 cm depth increment) by the new BD in the expanded increment (using soil mass for every  > 1 cm depth increment) to determine the expanded depth of each increment. The SOC value at the initial time represented the same, now expanded, ( > 1 cm) increments, as SOC is a ratio of mass. We used a linear decay rate that was twice that of the percent change in BD between time points to maintain an average BD that was consistent with scenario 1. To model the subsequent fixed depth sample, the BD and SOC concentration values of this expanded soil profile were then interpolated back to the 30 × 1 cm increments of the scenario 2 baseline depth. This calculation preserved the prescribed average BD of the new time point by only expanding the initial SOC concentration.We adjusted the SOC concentration of the next time point to maintain the average SOC concentrations of the 30 cm fixed depth sample, (see annotated scripts “main.R” and “functions.R” in Supplementary Material 1). Because the BD changed between time points and because the SOC stock in the 30 cm fixed depth sample was known, we determined the change in SOC stock between time points by subtracting the average SOC stock in the prior sample from the new sample. We then weighted this change across the 30 cm profile using the distribution of the global soil SOC in the top 30 cm to simulate SOC stratification with reduced tillage or agricultural intensification40. We then multiplied this change by the BD to convert back to SOC concentration and added the delta ((Delta )) SOC value to the prior sample. A worked example is shown in Supplementary Material 2 “Correction Example”.At each time point we split the soil profile at 10 cm and 15 cm depth intervals to create samples for scenarios 2a (3 soil intervals) and 2b (2 soil intervals), respectively. Note that scenario 2c is mathematically equivalent to scenario 1—with only one sample depth interval (30 cm) the sample contains no data on varying SOC or BD. The samples for 2a and 2b were generated by summing the total mass per area and SOC stock values from the scenario 2 baseline to produce single sample values of total soil mass per area and SOC concentration values per depth interval (as would be determined in a laboratory) and calculating BD and mineral mass.In scenario 2, any required additional mineral mass and the associated SOC values were ‘placed’ at the base of the sample to represent a soil profile that had expanded below the fixed 30 cm depth. To account for this, we calculated the increase in adjusted sample depth and accumulated additional soil mineral mass with the lowest sample depth interval of each split sample (Eqs. 5 and 6).$$mathrm{Delta D}={D}_{a}-{D}_{i}$$
(5)
$$SO{C}_{stock}={(D}_{1}*B{D}_{1}*SO{C}_{1}+dots + {(D}_{j}+Delta D)*B{D}_{j}*SO{C}_{j}))*10^2 (mathrm{g}/mathrm{cm}^{2})/(mathrm{Mg}/mathrm{ha})$$
(6)
where (mathrm{ Delta D}) is the apparent change in depth needed to generate the same mineral mass of the initial sample and the subscript j is the number of sample depth intervals from 1 to j.Varying BD linearly with depth introduces additional complexity in the calculation of the apparent depth. Each sample depth interval may expand (or contract in cases not explored here) at differing rates. Here, the over or under sampling of soil mineral mass is no longer constant with depth and the correction for apparent depth (Da) is estimated with linear interpolation using the BD of each sampling depth interval (i.e., 10 cm, 15 cm, or 30 cm). To do so, we calculated the mineral mass in each depth interval, determined their difference between the initial sample time point and new sample time point, and converted the change in mineral mass to a depth, where:$${mathrm{D}}_{mathrm{a}}={mathrm{D}}_{mathrm{i}}+frac{left(mathrm{sum}left({mathrm{D}}_{mathrm{ij}}*{mathrm{BD}}_{mathrm{ij}}*(1-mathrm{k}*{mathrm{soc}}_{mathrm{ij}}right))- mathrm{sum}left({mathrm{D}}_{mathrm{nj}}*{mathrm{BD}}_{mathrm{nj}}*(1-mathrm{k}*{mathrm{soc}}_{nmathrm{j}})right)right)}{{mathrm{BD}}_{{mathrm{nj}}_{mathrm{bottom}}}*1-mathrm{k}*{mathrm{soc}}_{n{mathrm{j}}_{mathrm{bottom}}}}$$
(7)
where jbottom is the lowest sample depth interval, and other terms are as previous. Using Eqs. (5), (6), and (7), with variable BD and SOC values, SOC stock can be corrected using samples split into the 10 cm and 15 cm sampling depth intervals. More

Castillo, C. P. et al. Agricultural Land Abandonment in the EU within 2015-2030. (Joint Research Centre (Seville site), 2018).van der Zanden, E. H., Verburg, P. H., Schulp, C. J. E. & Verkerk, P. J. Trade-offs of European agricultural abandonment. Land Use Policy 62, 290–301 (2017).Article 

Google Scholar 
Sun, Z. et al. Dietary change in high-income nations alone can lead to substantial double climate dividend. Nat. Food 3, 29–37 (2022).Article 
CAS 

Google Scholar 
Rasmussen, L. V. et al. Social-ecological outcomes of agricultural intensification. Nat. Sustainability 1, 275–282 (2018).Article 

Google Scholar 
Arneth, A. et al. Post-2020 biodiversity targets need to embrace climate change. Proc. Natl. Acad. Sci. 117, 30882–30891 (2020).Article 
CAS 

Google Scholar 
Seddon, N., Turner, B., Berry, P., Chausson, A. & Girardin, C. A. J. Grounding nature-based climate solutions in sound biodiversity science. Nat. Clim. Change 9, 84–87 (2019).Article 

Google Scholar 
Rey Benayas, J. M. & Bullock, J. M. Restoration of biodiversity and ecosystem services on agricultural land. Ecosystems 15, 883–899 (2012).Article 

Google Scholar 
Malhi, Y. et al. The role of large wild animals in climate change mitigation and adaptation. Current Biology 32, R181–R196 (2022).Article 
CAS 

Google Scholar 
Perino, A. et al. Rewilding complex ecosystems. Science 364, eaav5570 (2019).Article 
CAS 

Google Scholar 
Zhou, Y. et al. Limited increases in savanna carbon stocks over decades of fire suppression. Nature 603, 445–449 (2022).Article 
CAS 

Google Scholar 
Anderegg, W. R. L. et al. Climate-driven risks to the climate mitigation potential of forests. Science 368, eaaz7005 (2020).Article 
CAS 

Google Scholar 
Svenning, J.-C. Rewilding should be central to global restoration efforts. One Earth 3, 657–660 (2020).Article 

Google Scholar 
Dandy, N. & Wynne-Jones, S. Rewilding forestry. Forest Policy Econ. 109, 101996 (2019).Article 

Google Scholar 
Reino, L. et al. Does afforestation increase bird nest predation risk in surrounding farmland. Forest Ecol. Manag. 260, 1359–1366 (2010).Article 

Google Scholar 
Pausas, J. G. & Bond, W. J. Alternative biome states in terrestrial ecosystems. Trends Plant Sci. 25, 250–263 (2020).Article 
CAS 

Google Scholar 
Bakker, E. S. et al. Combining paleo-data and modern exclosure experiments to assess the impact of megafauna extinctions on woody vegetation. Proc. Natl. Acad. Sci. 113, 847–855 (2016).Article 
CAS 

Google Scholar 
Johnston, C. M. T. & Radeloff, V. C. Global mitigation potential of carbon stored in harvested wood products. Proceedings of the National Academy of Sciences 116, 14526–14531 (2019).Article 
CAS 

Google Scholar 
Shukla, P. R. et al. Climate Change and Land: an IPCC special report on climate change, desertification, landdegradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems. Intergovernmental Panel on Climate Change (2019).Hong, S. et al. Divergent responses of soil organic carbon to afforestation. Nat. Sustainability 3, 694–700 (2020).Article 

Google Scholar 
Bala, G. et al. Combined climate and carbon-cycle effects of large-scale deforestation. Proc. Natl. Acad. Sci. 104, 6550–6555 (2007).Article 
CAS 

Google Scholar 
Rohatyn, S., Yakir, D., Rotenberg, E. & Carmel, Y. Limited climate change mitigation potential through forestation of the vast dryland regions. Science 377, 1436–1439 (2022).Article 
CAS 

Google Scholar 
Beer, C., Zimov, N., Olofsson, J., Porada, P. & Zimov, S. Protection of permafrost soils from thawing by increasing herbivore density. Sci Rep-Uk 10, 4170 (2020).Article 
CAS 

Google Scholar 
Johnson, C. N. et al. Can trophic rewilding reduce the impact of fire in a more flammable world. Philos. Trans. R. Soc. B: Biological Sci. 373, 20170443 (2018).Article 

Google Scholar 
Kristensen, J. A., Svenning, J.-C., Georgiou, K. & Malhi, Y. Can large herbivores enhance ecosystem carbon persistence? Trends Ecol. Evol. https://doi.org/10.1016/j.tree.2021.09.006 (2021).Granado-Díaz, R., Villanueva, A. J. & Gómez-Limón, J. A. Willingness to accept for rewilding farmland in environmentally sensitive areas. Land Use Policy 116, 106052 (2022).Article 

Google Scholar 
Broughton, R. K. et al. Long-term woodland restoration on lowland farmland through passive rewilding. PloS one 16, e0252466 (2021).Article 
CAS 

Google Scholar 
Carver, S. et al. Guiding principles for rewilding. Conserv. Biology 35, 1882–1893 (2021).Article 

Google Scholar  More

Biodiversity on Earth is threatened by land-use changes, overexploitation of resources, pollution, biological invasions, and current and projected climate change. Understanding how species will respond to these stressors is difficult, in part because stressors don’t occur in isolation, and because responses can trickle through ecological networks due to interactions among species. More

Data reportingThe data underpinning this study is a compilation of existing datasets and therefore, no statistical methods were used to predetermine sample size, the experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. The measurements were taken from distinct samples, repeated measurements from the same sites were averaged in the main analysis.Inclusion & ethicsData were primarily collected from individual archives of contributing co-authors. The data collection initiative was openly announced via the mailing list of the 10th International Seminar on Apterygota and via social media (Twitter, Researchgate). In addition, colleagues from less explored regions (Africa, South America) were contacted via personal networks of the initial authors group and literature search. All direct data providers who collected and standardised the data were invited as co-authors with defined minimum role (data provision and cleaning, manuscript editing and approval). For unpublished data, people who were directly involved in sorting and identification of springtails, including all local researchers, were invited as co-authors. Principal investigators were normally not included as co-authors, unless they contributed to conceptualisation and writing of the manuscript. All co-authors were informed and invited to contribute throughout the research process—from the study design and analysis to writing and editing. The study provided an inclusive platform for researchers around the globe to network, share and test their research ideas.Data acquisitionBoth published and unpublished data were collected, using raw data whenever possible entered into a common template. In addition, data available from Edaphobase47 was included. The following minimum set of variables was collected: collectors, collection method (including sampling area and depth), extraction method, identification precision and resources, collection date, latitude and longitude, vegetation type (generalized as grassland, scrub, woodland, agriculture and other for the analysis), and abundances of springtail taxa found in each soil sample (or sampling site). Underrepresented geographical areas (Africa, South America, Australia and Southeast Asia) were specifically targeted by a literature search in the Web of Science database using the keywords ‘springtail’ or ‘Collembola’, ‘density’ or ‘abundance’ or ‘diversity’, and the region of interest; data were acquired from all found papers if the minimum information listed above was provided. All collected datasets were cleaned using OpenRefine v3.3 (https://openrefine.org) to remove inconsistencies and typos. Geographical coordinates were checked by comparing the dataset descriptions with the geographical coordinates. In total, 363 datasets comprising 2783 sites were collected and collated into a single dataset (Supplementary Fig. 1).Calculation of community parametersCommunity parameters were calculated at the site level. Here, we defined a site as a locality that hosts a defined springtail community, is covered by a certain vegetation type, with a certain management, and is usually represented by a sampling area of up to a hundred metres in diameter, making species co-occurrence and interactions plausible. To calculate density, numerical abundance in all samples was averaged and recalculated per square metre using the sampling area. Springtail communities were assessed predominantly during active vegetation periods (i.e., spring, summer and autumn in temperate and boreal biomes, and summer in polar biomes). Our estimations of community parameters therefore refer to the most favourable conditions (peak yearly densities). This seasonal sampling bias is likely to have little effect on our conclusions, since most springtails survive during cold periods38,48. Finally, we used mean annual soil temperatures49 to estimate the seasonal mean community metabolism (described below) and tested for the seasonal bias in additional analysis (see Linear mixed-effects models).All data analyses were conducted in R v. 4.0.250 with RStudio interface v. 1.4.1103 (RStudio, PBC). Data was transformed and visualised with tidyverse packages51,52, unless otherwise mentioned. Background for the global maps was acquired via the maps package53,54. To calculate local species richness, we used data identified to species or morphospecies level (validated by the expert team). Since the sampling effort varied among studies, we extrapolated species richness using rarefaction curves based on individual samples with the Chao estimator51,52 in the vegan package53. For some sites, sample-level data were not available in the original publications, but site-level averages were provided, and an extensive sampling effort was made. In such cases, we predicted extrapolated species richness based on the completeness (ratio of observed to extrapolated richness) recorded at sites where sample-level data were available (only sites with 5 or more samples were used for the prediction). We built a binomial model to predict completeness in sites where no sample-level data were available using latitude and the number of samples taken at a site as predictors: glm(Completeness~N_samples*Latitude). We found a positive effect of the number of samples (Chisq = 1.97, p = 0.0492) and latitude (Chisq = 2.07, p = 0.0391) on the completeness (Supplementary Figs. 17–19). We further used this model to predict extrapolated species richness on the sites with pooled data (435 sites in Europe, 15 in Australia, 6 in South America, 4 in Asia, and 3 in Africa).To calculate biomass, we first cross-checked all taxonomic names with the collembola.org checklist55 using fuzzy matching algorithms (fuzzyjoin R package56) to align taxonomic names and correct typos. Then we merged taxonomic names with a dataset on body lengths compiled from the BETSI database57, a personal database of Matty P. Berg, and additional expert contributions. We used average body lengths for the genus level (body size data on 432 genera) since data at the species level were not available for many morphospecies (especially in tropical regions), and species within most springtail genera had similar body size ranges. Data with no genus-level identifications were excluded from the analysis. Dry and fresh body masses were calculated from body length using a set of group-specific length-mass regressions (Supplementary Table 1)58,59 and the results of different regressions applied to the same morphogroup were averaged. Dry mass was recalculated to fresh mass using corresponding group-specific coefficients58. We used fresh mass to calculate individual metabolic rates60 and account for the mean annual topsoil (0–5 cm) temperature at a given site61. Group-specific metabolic coefficients for insects (including springtails) were used for the calculation: normalization factor (i0) ln(21.972) [J h−1], allometric exponent (a) 0.759, and activation energy (E) 0.657 [eV]60. Community-weighted (specimen-based) mean individual dry masses and metabolic rates were calculated for each sample and then averaged by site after excluding 10% of maximum and 10% of minimum values to reduce impact of outliers. To calculate site-level biomass and community metabolism, we summed masses or metabolic rates of individuals, averaged them across samples, and recalculated them per unit area (m2).Parameter uncertaintiesOur biomass and community metabolism approximations contain several assumptions. To account for the uncertainty in the length-mass and mass-metabolism regression coefficients, in addition to the average coefficients, we also used maximum (average + standard error) and minimum coefficients (average—standard error; Supplementary Table 1) in all equations to calculate maximum and minimum estimations of biomass and community metabolism reported in the main text. Further, we ignored latitudinal variation in body sizes within taxonomic groups62. Nevertheless, latitudinal differences in springtail density (30-fold), environmental temperature (from −16.0 to +27.6 °C in the air and from −10.2 to +30.4 °C in the soil), and genus-level community compositions (there are only few common genera among polar regions and the tropics)55 are higher than the uncertainties introduced by indirect parameter estimations, which allowed us to detect global trends. Although most springtails are concentrated in the litter and uppermost soil layers20, their vertical distribution depends on the particular ecosystem63. Since sampling methods are usually ecosystem-specific (i.e. sampling is done deeper in soils with developed organic layers), we treated the methods used by the original data collectors as representative of a given ecosystem. Under this assumption, we might have underestimated the number of springtails in soils with deep organic horizons, so our global estimates are conservative and we would expect true global density and biomass to be slightly higher. To minimize these effects, we excluded sites where the estimations were likely to be unreliable (see data selection below).Data selectionOnly data collection methods allowing for area-based recalculation (e.g. Tullgren or Berlese funnels) were used for analysis. Data from artificial habitats, coastal ecosystems, caves, canopies, snow surfaces, and strong experimental manipulations beyond the bounds of naturally occurring conditions were excluded (Supplementary Fig. 1). To ensure data quality, we performed a two-step quality check: technical selection and expert evaluation. Collected data varied according to collection protocols, such as sampling depth and the microhabitats (layers) considered. To technically exclude unreliable density estimations, we explored data with a number of diagnostic graphs (Supplementary Table 2; Supplementary Figs. 12–20) and filtered it, excluding the following: (1) All woodlands where only soil or only litter was considered; (2) All scrub ecosystems where only ground cover (litter or mosses) was considered; (3) Agricultural sites in temperate zones where only soil with sampling depth 90% of cases were masked on the main maps; for the map with density-species richness visualisation, two corresponding masks were applied (Fig. 2).To estimate spatial variability of our predictions while accounting for the spatial sampling bias in our data (Fig. 1a) we performed a spatially stratified bootstrapping procedure. We used the relative area of each IPBES79 region (i.e., Europe and Central Asia, Asia and the Pacific, Africa, and the Americas) to resample the original dataset, creating 100 bootstrap resamples. Each of these resamples was used to create a global map, which was then reduced to create mean, standard deviation, 95% confidence interval, and coefficient of variation maps (Supplementary Figs. 4–7).Global biomass, abundance, and community metabolism of springtails were estimated by summing predicted values for each 30 arcsec pixel10. Global community metabolism was recalculated from joule to mass carbon by assuming 1 kg fresh mass = 7 × 106 J80, an average water proportion in springtails of 70%58, and an average carbon concentration of 45% (calculated from 225 measurements across temperate forest ecosystems)81. We repeated the procedure of global extrapolation and prediction for biomass and community metabolism using minimum and maximum estimates of these parameters from regression coefficient uncertainties (see Parameter uncertainties).Path analysisTo reveal the predictors of springtail communities at the global scale, we performed a path analysis. After filtering the selected environmental variables (see above) according to their global availability and collinearity, 13 variables were used (Supplementary Fig. 9b): mean annual air temperature, mean annual precipitation (CHELSA database67), aridity (CGIAR database68), soil pH, sand and clay contents combined (sand and clay contents were co-linear in our dataset), soil organic carbon content (SoilGrids database73), NDVI (MODIS database72), human population density (GPWv4 database74), latitude, elevation69, and vegetation cover reported by the data providers following the habitat classification of European Environment Agency (woodland, scrub, agriculture, and grasslands; the latter were coded as the combination of woodland, scrub, and agriculture absent). Before running the analysis, we performed the Rosner’s generalized extreme Studentized deviate test in the EnvStats package82 to exclude extreme outliers and we z-standardized all variables (Supplementary R Code).Separate structural equation models were run to predict density, dry biomass, community metabolism, and local species richness in the lavaan package83. To account for the spatial clustering of our data in Europe, instead of running a model for the entire dataset, we divided the data by the IPBES79 geographical regions and selected a random subset of sites for Eurasia, such that only twice the number of sites were included in the model as the second-most represented region. We ran the path analysis 99 times for each community parameter with different Eurasian subsets (density had n = 723 per iteration, local species richness had n = 352, dry biomass had n = 568, and community metabolism had n = 533). We decided to keep the share of the Eurasian dataset larger than other regions to increase the number of sites per iteration and validity of the models. The Eurasian dataset also had the best data quality among all regions and a substantial reduction in datasets from Eurasia would result in a low weight for high-quality data. We additionally ran a set of models in which the Eurasian dataset was represented by the same number of sites as the second-most represented region, which yielded similar effect directions for all factors, but slightly higher variations and fewer consistently significant effects. In the paper, only the first version of analysis is presented. To illustrate the results, we averaged effect sizes for the paths across all iterations and presented the distribution of these effect sizes using mirrored Kernel density estimation (violin) plots. We marked and discussed effects that were significant at p  More

Study areaIn Colombia, the Amazon region represents 42.3% of the territory with an estimated area of 483,164 km2. In this area, 14% is dominated by agricultural lands, secondary vegetation and fragmented forests. Currently, 86% of the area corresponds to natural areas in a good state of conservation, where forests are the dominant coverage6. In the northwest area, the region borders the Andean Cordillera and Orinoquía to the north. The political-administrative division includes the departments Amazonas, Caquetá, Guainía, Guaviare, Putumayo and Vaupés, and part of the departments Cauca, Meta, Nariño and Vichada. The human population is estimated at ~ 1.4 million, with a density of 2.5 inhab/km2. Internal conflict and poverty make this region one of the most important population dynamics in the country in terms of displacement36. The geographical location of the study area and the spatial pattern of the loss of forests that occurred between 2002 and 2016 are shown in Fig. 1.Figure 1Study area. Colombian Amazon and location of Amazonian tropical forests that were lost between 2002 and 2016. (Maps were generated using software ArcGis 10.7.1 https://www.esri.com).Full size imageLand cover maps and variables for change analysisThematic land cover maps used in this research were produced by the Colombian Amazon Land Cover Monitoring System (SIMCOBA) of the Amazon Institute for Scientific Research SINCHI (https://siatac.co/simcoba/). SIMCOBA has prepared land cover maps for the periods 2002, 2007, 2012, 2014, 2016 and 2018. Three of the land cover maps prepared were used in this study: 2002, 2016 and 2018 a scale of 1:100,00033. The maps were generated from the visual interpretation of a mosaic of Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) images, using the PIAO technique (Photo Interprétation Assistée par Ordinateur). The classification categories of the land cover maps were based on the Corine land cover methodology adapted for Colombia37.The SIMCOBA system calculates the annual rates of Amazon forest loss (forest loss/ha/annual) by comparing the cover maps of the last two periods and subtracting from the previous map those forests that are no longer present in the most current map (Fig. 3). This process only considers the forests loss and the permanent forests. New forests due to natural regeneration or restoration are omitted in the calculations6.To facilitate the interpretation of changes and cover transitions, the classification categories of the maps were re-categorized into 7 types: “Amazon forests”, “floodplain forests”, “fragmented forests and secondary vegetation”, “grasslands and shrublands”, “water bodies and wetlands”, “pastures and crops” and “urban and artificialized cover”. The land cover maps were resampled at a resolution of 60 m × 60 m to facilitate the computational analysis of the explanatory model, the simulations of the scenarios, and to keep the detailed spatial resolution of the coverage and explanatory variables16.A geospatial database was created with a set of variables for the cover changes to create an explanatory model for each transition. Driving factors of change are grouped into the following variables: (1) accessibility, (2) climate, (3) landscape features, (4) production practices and environmental degradation, (5) landscape management, (6) socioeconomy, and (7) soil characteristics. We considered 41 explanatory variables (see supplementary information Table S1).Accessibility variables such as roads and navigable rivers were obtained from the geodatabase at a scale of 1:100,000 of the Agustín Codazzi Geographical Institute of Colombia (IGAC). Bioclimatic temperature data were obtained from Worldclim v1.438. Cover variables (e.g., patch sizes Amazon forests and distance to pastures and crops) were created using the software ArcGis (v.10.7.1)39 from the 2002 land cover map to understand which drivers were more influential in the dynamics of land-use changes since 2002 that resulted in the distribution of land cover in 2016.Degradation variables, such as advances of the agricultural frontier, were obtained from the Territorial Environmental Information System of the Colombian Amazon (SIAT-AC)40; livestock density data came from the Colombian Agricultural Institute (ICA); the fire density were processed from MODIS and VIIRS images (https://siatac.co/puntos-de-calor/); and the location of mining titles was obtained from the National Mining Agency.The information on the landscape features and socioeconomic variables was obtained from different sources: (1) the limit of the protected natural areas was provided by the National System of Protected Areas (SINAP)41, (2) the Amazon Forest Reserve areas (Second Law of 1959) were obtained from the Ministry of Environment and Sustainable Development (MADS), (3) the location of the indigenous reservations was provided by the Ministry of the Interior, and (4) the limits of the areas of Indigenous Reservations and the legal status of the Amazonian region were obtained from the SINCHI cartographic database40.Socioeconomic information was spatialized from data from the National Administrative Department of Statistics (DANE). Soil-type data were obtained from IGAG, and topographic and altitudinal variables were derived from a DEM at 100 m resolution from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER V003) sensor42. All explanatory variables were resampled at a resolution of 60 m.Patterns of land cover changes and transitionsThe transformation patterns of territory are mainly defined by human intentions and the activities that these groups plan to develop after making the land cover changes, as well as the dynamics of vegetation regeneration43. In this study, these changes in the study area were obtained and analyzed employing the Land Change Modeller (LCM) module of TerrSet34 and using the land cover maps for 2002 and 2016 as input information (Fig. 2).Figure 2(Source: Open Data—SINCHI Institute https://datos.siatac.co/pages/coberturas) (Maps were generated using software ArcGis 10.7.1 ).Land cover maps 2002, 2016 and 2018, produced by the Colombian Amazon Land Cover Monitoring System (SIMCOBA) of the Amazonian Research Institute SINCHIFull size imageTo represent dynamics and changes in the vegetation during the study period, a total of 14 transitions of greater importance in terms of area were considered (transitions with an area  More

Global Forest Resources Assessment 2020 (FAO, 2020).Taubert, F. et al. Global patterns of tropical forest fragmentation. Nature 554, 519–522 (2018).ADS 
CAS 

Google Scholar 
Vancutsem, C. et al. Long-term (1990–2019) monitoring of forest cover changes in the humid tropics. Sci. Adv. 7, eabe1603 (2021).ADS 

Google Scholar 
Foley, J. A. et al. Amazonia revealed: Forest degradation and loss of ecosystem goods and services in the Amazon Basin. Front. Ecol. Environ. 5, 25–32 (2007).
Google Scholar 
Barlow, J. et al. The future of hyperdiverse tropical ecosystems. Nature 559, 517–526 (2018).ADS 
CAS 

Google Scholar 
Brandon, K. Ecosystem services from tropical forests: Review of current science. SSRN J. https://doi.org/10.2139/ssrn.2622749 (2014). Article 

Google Scholar 
Indarto, J. & Mutaqin, D. J. An overview of theoretical and empirical studies on deforestation. MPRA. Paper No. 70178 (2016).Geist, H. J. & Lambin, E. F. Proximate causes and underlying driving forces of tropical deforestation: Tropical forests are disappearing as the result of many pressures, both local and regional, acting in various combinations in different geographical locations. Bioscience 52, 143–150 (2002).
Google Scholar 
Angelsen, A. & Kaimowitz, D. Rethinking the causes of deforestation: Lessons from economic models. World Bank Res. Obs. 14, 73–98 (1999).CAS 

Google Scholar 
Contreras-Hermosilla, A. The Underlying Causes of Forest Decline (Center for International Forestry Research, 2000).
Google Scholar 
Turner, B. L. et al. Two types of global environmental change: Definitional and spatial-scale issues in their human dimensions. Glob. Environ. Change 1, 14–22 (1990).
Google Scholar 
Meyer, W. B. & Turner, B. L. Human population growth and global land-use/cover change. Ann. Rev. Ecol. Syst. 2, 39–61 (1992).
Google Scholar 
Miyamoto, M., Mohd Parid, M., Noor Aini, Z. & Michinaka, T. Proximate and underlying causes of forest cover change in Peninsular Malaysia. For. Policy Econ. 44, 18–25 (2014).
Google Scholar 
Lim, C. L., Prescott, G. W., De Alban, J. D. T., Ziegler, A. D. & Webb, E. L. Untangling the proximate causes and underlying drivers of deforestation and forest degradation in Myanmar. Conserv. Biol. 31, 1362–1372 (2017).
Google Scholar 
Carodenuto, S. et al. A methodological framework for assessing agents, proximate drivers and underlying causes of deforestation: Field test results from southern cameroon. Forests 6, 203–224 (2015).
Google Scholar 
Curtis, P. G., Slay, C. M., Harris, N. L., Tyukavina, A. & Hansen, M. C. Classifying drivers of global forest loss. Science 361, 1108–1111 (2018).ADS 
CAS 

Google Scholar 
Hosonuma, N. et al. An assessment of deforestation and forest degradation drivers in developing countries. Environ. Res. Lett. 7, 044009 (2012).ADS 

Google Scholar 
Köthke, M., Leischner, B. & Elsasser, P. Uniform global deforestation patterns—An empirical analysis. For. Policy Econ. 28, 23–37 (2013).
Google Scholar 
Busch, J. & Ferretti-Gallon, K. What drives deforestation and what stops it? A meta-analysis. Rev. Environ. Econ. Policy 11, 3–23 (2017).
Google Scholar 
Ferrer Velasco, R. F., Köthke, M., Lippe, M. & Günter, S. Scale and context dependency of deforestation drivers: Insights from spatial econometrics in the tropics. PLoS One 15, e0226830 (2020).CAS 

Google Scholar 
Lambin, E. F. et al. Effectiveness and synergies of policy instruments for land use governance in tropical regions. Glob. Environ. Change 28, 129–140 (2014).
Google Scholar 
Börner, J., Schulz, D., Wunder, S. & Pfaff, A. The effectiveness of forest conservation policies and programs. Ann. Rev. Resour. Econ. 12, 45–64 (2020).
Google Scholar 
Bemelmans-Videc, M.-L., Rist, R. C. & Vedung, E. Carrots, Sticks & Sermons: Policy Instruments and their Evaluation (Transaction Publishers, 1998).
Google Scholar 
Seymour, F. & Harris, N. L. Reducing tropical deforestation. Science 365, 756–757 (2019).ADS 
CAS 

Google Scholar 
Lambin, E. F. et al. The role of supply-chain initiatives in reducing deforestation. Nat. Clim. Change 8, 109–116 (2018).ADS 

Google Scholar 
Wolff, S. & Schweinle, J. Effectiveness and economic viability of forest certification: A systematic review. Forests 13, 798 (2022).
Google Scholar 
Müller, R., Pistorius, T., Rohde, S., Gerold, G. & Pacheco, P. Policy options to reduce deforestation based on a systematic analysis of drivers and agents in lowland Bolivia. Land Use Policy 30, 895–907 (2013).
Google Scholar 
Tegegne, Y. T., Lindner, M., Fobissie, K. & Kanninen, M. Evolution of drivers of deforestation and forest degradation in the Congo Basin forests: Exploring possible policy options to address forest loss. Land Use Policy 51, 312–324 (2016).
Google Scholar 
Hoffmann, C., García Márquez, J. R. & Krueger, T. A local perspective on drivers and measures to slow deforestation in the Andean-Amazonian foothills of Colombia. Land Use Policy 77, 379–391 (2018).
Google Scholar 
Henders, S., Ostwald, M., Verendel, V. & Ibisch, P. Do national strategies under the UN biodiversity and climate conventions address agricultural commodity consumption as deforestation driver?. Land Use Policy 70, 580–590 (2018).
Google Scholar 
Salvini, G. et al. How countries link REDD+ interventions to drivers in their readiness plans: implications for monitoring systems. Environ. Res. Lett. 9, 074004 (2014).ADS 

Google Scholar 
Bos, A. B. et al. Integrated assessment of deforestation drivers and their alignment with subnational climate change mitigation efforts. Environ. Sci. Policy 114, 352–365 (2020).CAS 

Google Scholar 
Fritz, S. et al. A continental assessment of the drivers of tropical deforestation with a focus on protected areas. Front. Conserv. Sci. https://doi.org/10.3389/fcosc.2022.830248 (2022).Article 

Google Scholar 
Lawrence, D. & Vandecar, K. Effects of tropical deforestation on climate and agriculture. Nat. Clim. Change 5, 27–36 (2015).ADS 

Google Scholar 
Fedele, G., Locatelli, B., Djoudi, H. & Colloff, M. J. Reducing risks by transforming landscapes: Cross-scale effects of land-use changes on ecosystem services. PLoS One 13, e0195895 (2018).
Google Scholar 
Yackulic, C. B. et al. Biophysical and socioeconomic factors associated with forest transitions at multiple spatial and temporal scales. Ecol. Soc. https://doi.org/10.5751/ES-04275-160315 (2011).Article 

Google Scholar 
Loran, C., Ginzler, C. & Bürgi, M. Evaluating forest transition based on a multi-scale approach: Forest area dynamics in Switzerland 1850–2000. Reg. Environ. Change 16, 1807–1818 (2016).
Google Scholar 
Moonen, P. C. et al. Actor-based identification of deforestation drivers paves the road to effective REDD+in DR Congo. Land Use Policy 58, 123–132 (2016).
Google Scholar 
Strassburg, B. The tragedy of the tropics: A dynamic, cross-scale analysis of deforestation incentives. Working Paper—Centre for Social and Economic Research on the Global Environment No. 07-02 (2007).López-Carr, D. et al. Space versus place in complex human–natural systems: Spatial and multi-level models of tropical land use and cover change (LUCC) in Guatemala. Ecol. Model. 229, 64–75 (2012).
Google Scholar 
Hoang, N. T. & Kanemoto, K. Mapping the deforestation footprint of nations reveals growing threat to tropical forests. Nat. Ecol. Evol. 5, 845–853 (2021).
Google Scholar 
Pendrill, F. et al. Agricultural and forestry trade drives large share of tropical deforestation emissions. Glob. Environ. Change 56, 1–10 (2019).
Google Scholar 
Ferrer Velasco, R. et al. Towards accurate mapping of forest in tropical landscapes: A comparison of datasets on how forest transition matters. Remote Sens. Environ. 274, 112997 (2022).ADS 

Google Scholar 
Jayathilake, H. M., Prescott, G. W., Carrasco, L. R., Rao, M. & Symes, W. S. Drivers of deforestation and degradation for 28 tropical conservation landscapes. Ambio 50, 215–228 (2021).
Google Scholar 
Minang, P. A. et al. REDD+Readiness progress across countries: Time for reconsideration. Clim. Policy 14, 685–708 (2014).
Google Scholar 
Current pledges | Bonn challenge. https://www.bonnchallenge.org/pledges. Accessed: 15th August 2022.Nansikombi, H. et al. Can de facto governance influence deforestation drivers in the Zambian Miombo?. For. Policy Econ. 120, 102309 (2020).
Google Scholar 
Sullivan, A., York, A., White, D., Hall, S. & Yabiku, S. D. Jure versus de facto institutions: Trust, information, and collective efforts to manage the invasive mile-a-minute weed (Mikania micrantha). Int. J. Commons 11, 171–199 (2017).
Google Scholar 
Busch, J. & Amarjargal, O. Authority of second-tier governments to reduce deforestation in 30 tropical countries. Front. For. Glob. Change https://doi.org/10.3389/ffgc.2020.00001 (2020).Article 

Google Scholar 
Sandström, C., Eckerberg, K. & Raitio, K. Studying conflicts, proposing solutions—Towards multi-level approaches to the analyses of forest conflicts. For. Policy Econ. 33, 123–127 (2013).
Google Scholar 
Hoogstra-Klein, M. A., Permadi, D. B. & Yasmi, Y. The value of cultural theory for participatory processes in natural resource management. For. Policy Econ. 20, 99–106 (2012).
Google Scholar 
de Jong, W., Ruiz, S. & Becker, M. Conflicts and communal forest management in northern Bolivia. For. Policy Econ. 8, 447–457 (2006).
Google Scholar 
Eckerberg, K. & Sandström, C. Forest conflicts: A growing research field. For. Policy Econ. 33, 3–7 (2013).
Google Scholar 
Sierra, R., Calva, O. & Guevara, A. La Deforestación en el Ecuador, 1990–2018. Factores promotores y tendencias recientes, 216 (2021).Wasserstrom, R. & Southgate, D. Deforestation, agrarian reform and oil development in Ecuador, 1964–1994. Nat. Resour. 04, 31 (2013).
Google Scholar 
Wiebe, P. C., Zhunusova, E., Lippe, M., Ferrer Velasco, R. & Günter, S. What is the contribution of forest-related income to rural livelihood strategies in the Philippines’ remaining forested landscapes?. For. Policy Econ. 135, 102658 (2022).
Google Scholar 
Le, H. D., Smith, C. & Herbohn, J. What drives the success of reforestation projects in tropical developing countries? The case of the Philippines. Glob. Environ. Change 24, 334–348 (2014).
Google Scholar 
Carandang, A. P. et al. Analysis of key drivers of deforestation and forest degradation in the Philippines. Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) (2013).Phiri, D., Morgenroth, J. & Xu, C. Four decades of land cover and forest connectivity study in Zambia—An object-based image analysis approach. Int. J. Appl. Earth Obs. Geoinf. 79, 97–109 (2019).ADS 

Google Scholar 
Nansikombi, H., Fischer, R., Kabwe, G. & Günter, S. Exploring patterns of forest governance quality: Insights from forest frontier communities in Zambia’s Miombo ecoregion. Land Use Policy 99, 104866 (2020).
Google Scholar 
Zhang, H., Wang, P. & Wood, J. Does institutional quality matter for the nexus between environmental quality and economic growth?: A tropics perspective. In Business, Industry, and Trade in the Tropics (eds Wood, J. et al.) (Routledge, 2022).
Google Scholar 
Reed, J., Van Vianen, J., Deakin, E. L., Barlow, J. & Sunderland, T. Integrated landscape approaches to managing social and environmental issues in the tropics: Learning from the past to guide the future. Glob. Change Biol. 22, 2540–2554 (2016).ADS 

Google Scholar 
Fischer, R. et al. Interplay of governance elements and their effects on deforestation in tropical landscapes: Quantitative insights from Ecuador. World Dev. 148, 105665 (2021).
Google Scholar 
Torres, B., Vasco, C., Günter, S. & Knoke, T. Determinants of agricultural diversification in a hotspot area: Evidence from colonist and indigenous communities in the Sumaco biosphere reserve Ecuadorian Amazon. Sustainability 10, 1432 (2018).
Google Scholar 
Ojeda Luna, T., Zhunusova, E., Günter, S. & Dieter, M. Measuring forest and agricultural income in the Ecuadorian lowland rainforest frontiers: Do deforestation and conservation strategies matter?. For. Policy Econ. 111, 102034 (2020).
Google Scholar 
Kazungu, M. et al. Effects of household-level attributes and agricultural land-use on deforestation patterns along a forest transition gradient in the Miombo landscapes Zambia. Ecol. Econ. 186, 107070 (2021).
Google Scholar 
Kleemann, J. et al. Deforestation in continental ecuador with a focus on protected areas. Land 11, 268 (2022).
Google Scholar 
Mulenga, M. M. & Roos, A. Assessing the awareness and adoptability of pellet cookstoves for low-income households in Lusaka, Zambia. J. Energy South. Afr. 32, 52–61 (2021).
Google Scholar 
Eguiguren, P., Ojeda Luna, T., Torres, B., Lippe, M. & Günter, S. Ecosystem service multifunctionality: Decline and recovery pathways in the amazon and chocó lowland rainforests. Sustainability 12, 7786 (2020).CAS 

Google Scholar 
Vasco, C., Torres, B., Pacheco, P. & Griess, V. The socioeconomic determinants of legal and illegal smallholder logging: Evidence from the Ecuadorian Amazon. For. Policy Econ. 78, 133–140 (2017).
Google Scholar 
van der Ploeg, J., van Weerd, M., Masipiqueña, A. B. & Persoon, G. A. Illegal logging in the Northern Sierra Madre Natural Park, the Philippines. Conserv. Soc. 9, 202–215 (2011).
Google Scholar 
Liu, D. S., Iverson, L. R. & Brown, S. Rates and patterns of deforestation in the Philippines: Application of geographic information system analysis. For. Ecol. Manag. 57, 1–16 (1993).
Google Scholar 
Boquet, Y. Environmental challenges in the Philippines. In The Philippine Archipelago (ed. Boquet, Y.) 779–829 (Springer International Publishing, 2017). https://doi.org/10.1007/978-3-319-51926-5_22.Chapter 

Google Scholar 
MAGAP. ATPA: Reconversión Agro productiva Sostenible en la Amazonía Ecuatoriana (2014).Jones, K. W. et al. Forest conservation incentives and deforestation in the Ecuadorian Amazon. Environ. Conserv. 44, 56–65 (2017).
Google Scholar 
Lindsey, P. A. et al. Underperformance of African protected area networks and the case for new conservation models: Insights from Zambia. PLoS One 9, e94109 (2014).ADS 

Google Scholar 
Fischer, R. et al. Effectiveness of policy instrument mixes for forest conservation in the tropics – a stakeholder perspective from Ecuador, the Philippines and Zambia. Land Use Policy https://doi.org/10.1016/j.landusepol.2023.106546 (2022).Article 

Google Scholar 
Gurney, G. G. et al. Biodiversity needs every tool in the box: Use OECMs. Nature 595, 646–649 (2021).ADS 
CAS 

Google Scholar 
Maxwell, S. L. et al. Area-based conservation in the twenty-first century. Nature 586, 217–227 (2020).ADS 
CAS 

Google Scholar 
Priebe, J. et al. Transformative change in context—Stakeholders’ understandings of leverage at the forest–climate nexus. Sustain. Sci. 17, 1921–1938 (2022).
Google Scholar 
Höhl, M. et al. Forest landscape restoration—What generates failure and success?. Forests 11, 938 (2020).
Google Scholar 
Köthke, M., Ahimbisibwe, V. & Lippe, M. The evidence base on the environmental, economic and social outcomes of agroforestry is patchy—An evidence review map. Front. Environ. Sci. https://doi.org/10.3389/fenvs.2022.925477 (2022).Article 

Google Scholar 
Fischer, R., Giessen, L. & Günter, S. Governance effects on deforestation in the tropics: A review of the evidence. Environ. Sci. Policy 105, 84–101 (2020).
Google Scholar 
Bare, M., Kauffman, C. & Miller, D. C. Assessing the impact of international conservation aid on deforestation in sub-Saharan Africa. Environ. Res. Lett. 10, 125010 (2015).ADS 

Google Scholar 
Vuohelainen, A. J., Coad, L., Marthews, T. R., Malhi, Y. & Killeen, T. J. The effectiveness of contrasting protected areas in preventing deforestation in Madre de Dios. Peru. Environ. Manag. 50, 645–663 (2012).ADS 

Google Scholar 
Hull, V. & Liu, J. Telecoupling: A new frontier for global sustainability. Ecol. Soc. 23, 41 (2018).
Google Scholar 
Aitchison, J. The statistical analysis of compositional data. J. Roy. Stat. Soc. 44, 139–160 (1982).MathSciNet 
MATH 

Google Scholar 
Norman, G. Likert scales, levels of measurement and the “laws” of statistics. Adv. Health Sci. Educ. 15, 625–632 (2010).
Google Scholar 
Day, M., Gumbo, D., Moombe, K. B., Wijaya, A. & Sunderland, T. Zambia Country Profile: Monitoring, Reporting and Verification for REDD+ Vol. 113 (CIFOR, 2014).
Google Scholar 
Piotrowski, M. Nearing the tipping point. Drivers of Deforestation in the Amazon Region (2019).Sarker, P. K., Fischer, R., Tamayo, F., Navarrete, B. T. & Günter, S. Analyzing forest policy mixes based on the coherence of policies and the consistency of legislative policy instruments: A case study from Ecuador. For. Policy Econ. 144, 102838 (2022).
Google Scholar 
Likert, R. A technique for the measurement of attitudes. Arch. Psychol. 22(140), 55–55 (1932).
Google Scholar 
Altinsoy, M. et al. Ambulatory ECG monitoring for syncope and collapse in United States, Europe, and Japan: The patients’ viewpoint. J. Arrhythm. 37, 1023–1030 (2021).
Google Scholar 
R Core Team. R: A language and environment for statistical computing. https://www.R-project.org/. (R Foundation for Statistical Computing, Vienna, Austria, 2022).Kassambara, A. rstatix: Pipe-friendly framework for basic statistical tests. R package version 0.7.0 (2021).Kassambara, A. & Mundt, F. factoextra: Extract and visualize the results of multivariate data analyses. R package version 1.0.7 (2020).Komsta, L. & Novometsky, F. moments: Moments, cumulants, skewness, kurtosis and related tests. R package version 0.14.1 (2022).Zhang, Y., Zhou, M. & Shao, Y. mvnormalTest: Powerful tests for multivariate normality. R package version 1.0.0 (2020).Kassambara, A. ggpubr: ‘ggplot2’ Based Publication Ready Plots. R package version 0.4.0 (2020).Wickham, H. et al. Welcome to the Tidyverse. JOSS 4, 1686 (2019).ADS 

Google Scholar 
Bache, S. M. & Wickham, H. magrittr: A Forward-Pipe Operator for R. R package version 2.0.3 (2022).Ushey, K., Allaire, J., Wickham, H. & Ritchie, G. rstudioapi: Safely Access the RStudio API. R package version 0.13 (2020).Wickham, H. ggplot2: Elegant Graphics for Data Analysis (Springer International Publishing, 2016). https://doi.org/10.1007/978-3-319-24277-4.Book 
MATH 

Google Scholar 
Wilkins, D. treemapify: Draw Treemaps in ‘ggplot2’. R package version 2.5.5 (2021).Shapiro, S. S. & Wilk, M. B. An analysis of variance test for normality (complete samples). Biometrika 52, 591–611 (1965).MathSciNet 
MATH 

Google Scholar 
Mardia, K. V. Measures of multivariate skewness and kurtosis with applications. Biometrika 57, 519–530 (1970).MathSciNet 
MATH 

Google Scholar 
Kruskal, W. H. & Wallis, W. A. Use of ranks in one-criterion variance analysis. J. Am. Stat. Assoc. 47, 583–621 (1952).MATH 

Google Scholar 
Dunn, O. J. Multiple comparisons using rank sums. Technometrics 6, 241–252 (1964).
Google Scholar 
Conover, W. J. & Iman, R. L. Rank transformations as a bridge between parametric and nonparametric statistics. Am. Stat. 35, 124–129 (1981).MATH 

Google Scholar 
Student,. The probable error of a mean. Biometrika 6, 1–25 (1908).MATH 

Google Scholar 
Tukey, J. W. Comparing individual means in the analysis of variance. Biometrics 5, 99–114 (1949).MathSciNet 
CAS 

Google Scholar 
Jolliffe, I. T. Principal Component Analysis (Springer, 2002).MATH 

Google Scholar  More

The growth behaviour of Linnemannia is strain-specificMost strains showed comparable morphological characteristics on both media as well as in pure and co-culture. However, Linnemannia solitaria and Entomortierella galaxiae produced more aerial mycelium on PDA compared to LcA. There was more/less aerial mycelium in co-cultures with P. helmanticensis compared to pure cultures depending on the strain (Fig. 1, SI Fig. S3).The comparison of Linnemannia and E. galaxiae daily radial growth rates did not support a difference between these genera (p ≥ 0.3). The overall linear model indicated that the fungal daily growth rates mainly differed among species (Table 1). In addition, the effect of strains highlighted the heterogeneity among strains within species (Fig. 2, SI Figs. S4, S5). Although there was no relevant main effect of medium on the daily radial growth rate of the fungi, the medium did affect the fungi in a strain-specific manner (Table 1, Fig. 2, SI Figs. S4, S5). On nutrient poor LcA, the fungal daily radial growth rates were reduced for all species, except for L. solitaria, which grew better on LcA (SI Figs. S3, S4).Table 1 The effect of experimental factors on the fungal daily radial growth rate.Full size tableFigure 2Daily radial growth rate of pure Linnemannia and Entomortierella cultures as well as co-cultures with P. helmanticensis on nutrient rich PDA medium. (a) L. exigua, (b) L. gamsii, (c) L. hyalina, (d) L. sclerotiella, (e) L. solitaria, (f) E. galaxiae.Full size imageThe main effect of co-plating P. helmanticensis on radial growth rate was small, yet significant (0.7%, p  More