Figures
Abstract
Contemporaneous plant communities may retain a mark of past disturbances in their ecological patterns. However, unraveling the history of disturbance on natural systems at a large scale is often unfeasible, due to the complexity of the factors involved and lack of historical data. Here we aim at demonstrating how comparing observed spatial structure of tree assemblages with that expected in a hypothetical, undisturbed scenario can shed light on how natural European forests are. Borrowing an analytical approach developed in the field of network analysis, we assessed how much the observed ecological patterns of nestedness (i.e. positive co-occurrence), segregation (i.e. negative co-occurrence), and modularity in tree assemblages deviate from randomness, and from those projected by Potential Natural Vegetation (PNV) geobotanical expert assessments. We found clear evidence that European forests are far from a natural condition, showing only moderate signals (especially at higher latitudes) of the ecological spatial structure typical of undisturbed vegetation (i.e. nestedness). Our results highlight how taking into account spatial structure along with diversity can be a fundamental tool to address this problem and assess the degree of naturalness in species assemblages.
Citation: Strona G, Mauri A, Veech JA, Seufert G, San-Miguel Ayanz J, Fattorini S (2016) Far from Naturalness: How Much Does Spatial Ecological Structure of European Tree Assemblages Depart from Potential Natural Vegetation? PLoS ONE 11(12): e0165178. https://doi.org/10.1371/journal.pone.0165178
Editor: Filippos A. Aravanopoulos, Aristotle University of Thessaloniki, GREECE
Received: May 12, 2016; Accepted: October 8, 2016; Published: December 22, 2016
Copyright: © 2016 Strona et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and the stable public repository Figshare. Data hosted at Figshare can be found at the following URL: https://figshare.com/collections/A_high-resolution_pan-European_tree_occurrence_dataset/3288407.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Patterns in species spatial arrangements, such as, for example, the tendency for species occurrences to overlap or segregate, can emerge from a variety of ecological processes related to species interactions, habitat filtering, dispersal, colonization and extinction [1–4]. These processes usually work on a long-term temporal scale, which means that structural patterns will emerge slowly and under the condition that the structuring forces remain steady. Conversely, cycles of deforestation, abandonment, and afforestation, as well as natural disasters such as fires and storms, may create, in the most disturbed areas, scenarios where the remnants of the original natural vegetation are hardly detectable.
In a simplified view, this would suggest that undisturbed vegetation should be highly and consistently structured, and that, conversely, different types of disturbance would reduce the strength of structural patterns at landscape level. At local level, a particular disturbance may act as an environmental filter limiting diversity and allowing only few species to persist with very little variation in space (e.g. overgrazing). But, at the landscape level, where typically different types of disturbance operate simultaneously, the overall effect would be a loss of structure.
Due to the variety and complexity of factors responsible for spatial structure in local species assemblages [5–8], identifying likely scenarios expected in absence of disturbance is not straightforward. In this context, expert-based maps of Potential Natural Vegetation (PNV) could offer a possible key of interpretation.
The concept of PNV has been defined in several ways [9]. Here we refer to PNVs as the plant communities that would become established in given areas if all successional sequences were completed under the present local climatic and edaphic conditions without interference by humans [10–12]. Maps of PNV have been produced mostly as a tool to orient conservation strategy [13]. Nevertheless, they could also constitute a valuable source of baseline information to evaluate qualitative and quantitative discrepancies between the observed and the expected structure of plant communities.
Among the various forms of spatial structure, ecologists have been most often interested in investigating species co-occurrence, i.e. in assessing if (and to what extent) the presence of a particular species in a local community is positively or negatively affected by the presence/absence of other species [14–16]. Although, at a local scale, competition can lead to negative co-occurrence between species having similar functional traits/niches, at a larger scale, environmental filtering would promote spatial consistency in associations of species having similar tolerance to abiotic factors and complementary functional traits (e.g. mature vegetation types), and resulting in patterns of positive co-occurrence [17–21]. Of course, resource availability, and hence local diversity, can modulate these patterns, with competitive exclusion being more pronounced in areas with moderate to high diversity [22] than in extreme environments or isolated habitats [23].
The general tendency for species co-occurence can be assimilated to the concept of nestedness, which is a pattern where any species assemblage tends to be a subsample of richer assemblages, and which can be promoted by differential extinction, dispersion and colonization [24]. There is no consensus on whether or not a gradient in species richness among sites (as well as a gradient of variation in species abundance among sites) is a necessary condition for nestedness [24–26]. In our reasoning, however, the consistency of species association can be considered a mark of ecosystem maturity (and hence stability) regardless of differences in species diversity. Thus, we will stick to the concept of nestedness as in [25, 26], considering nestedness simply as a measure of overlap in species composition per site and in species site occurrence. Instead, we will refer to the opposite tendency for non-overlap as segregation, again as in [26].
Another interesting pattern that is becoming more and more popular is modularity [27], which occurs when clusters of similar species composition can be identified among species assemblages [28]. Modularity can originate from discontinuous (and synchronous) responses of species to environmental gradients, when differences in resource or habitat specialization among species—and environmental heterogeneity—promote species repartition among different, spatially separated clusters [27]. Although a better understanding of patterns and processes associated with modularity is of great biogeographical interest, the study of modularity has received little attention, probably because of the lack of appropriate methodological tools [29, 30]. Only recently, the increasing incorporation of network theory into ecological studies [31] has opened the way.
Understanding the processes that can lead to nestedness, segregation, and modularity is a complex task. Nevertheless, it is intuitive that any structural pattern emerging from natural ecological processes would become more and more evident as these remain active without interruption. At any time, external disturbances could stop such processes and thus modify the structural patterns built up to that moment. Therefore, it should be possible to quantify the exposure to disturbance of a given species assemblage as a function of its spatial structure or lack thereof.
Sources of disturbance could be either natural phenomena (such as fires, storms, earthquakes, landslides, epidemic diseases [32]), and human activities. However, the first kind of processes is most often limited to a local scale, making anthropic influence the main shaper of large scale patterns. Furthermore, ‘natural’ disasters, despite their unpredictability and stochasticity, are frequently due to human activities [33]. Unwise management often creates risky situations that make the happening of certain dramatic events more likely. For example, growing exotic trees may increase the risk of fires [34], and/or the risk of epidemics due to moving alien pathogens [35]. The risk of epidemics may be severely increased also by reducing local diversity, by selective growing [36, 37], as well as deforestation may increase the risk of landslides and floods [38].
Landscapes in Europe have been modified for millennia through clearance of forests to create croplands and pastures, and by intensive tree collection for fuel wood and construction materials [39]. Although the industrial revolution has certainly played a major role in this process, many European regions underwent profound deforestation thousands of years before that period and several areas have been altered since the mid-Holocene due to the establishment of the first European agricultural societies [40]. As a consequence, there are almost no intact forest landscapes left in Europe, with the exception of small patches in Fennoscandia and North Eastern territories [41]. Thus, we should expect to find much evidence of past disturbances in the current spatial structure of tree assemblages.
Quite surprisingly, only a few studies have focused on this topic. Moreover, these have been conducted only at a regional scale, and gave contrasting explanations to the observed nestedness patterns (see, for example, [42, 43]). Here we try to fill in this gap by (i) developing an innovative approach to investigate ecological spatial structure, and (ii) by applying it to the most comprehensive dataset of tree species available for Europe [44], which includes almost one million occurrences. In doing this, we show that comparing the spatial structural patterns observed for actual vegetation (ACV) with those predicted by PNV assessment can provide valuable insight as to how much contemporaneous tree assemblages depart from naturalness.
Materials and Methods
Use of PNV as a benchmark
The concept of PNV has been the focus of much debate regarding both its formal definition and meaning (see [45] for a review). In particular, it has often been questioned whether it is reasonable to use PNV as a qualitative benchmark to evaluate the status of actual vegetation (ACV) [9]. We agree that multiple factors can potentially affect species composition and diversity of the vegetation observed in a given area in a very site-specific way, so that PNV is not an exact representation of natural mature vegetation. Due to this, some researchers have questioned whether the concept of PNV is useful in any way [9]. Despite this concern, we put aside the conceptual problems of defining exact PNVs, and we built on the idea that the communities depicted by PNVs, in principle, should conform to one among the various ‘reasonable’ ecological outcomes for a given area. Here, the term ‘reasonable’ indicates compatible with the local environmental features. Thus, even without assuming that the prediction of PNV represents the exact mature vegetation (in terms of species composition) that would emerge under the assumption of environmental stability and in the absence of human disturbance [11], it is appropriate to consider a particular PNV as the best available description of the plausible vegetation of an area [12, 45, 46] in terms of ecological structure.
Tree species data
Actual tree distribution was obtained from Euforest, a new tree occurrence dataset provided by the European Commission [44], which includes almost 1 million occurrence data for 242 tree species across all Europe. Euforest combines the information from three large datasets provided by the European Forest Data Centre of the European Commission (http://efdac.jrc.ec.europa.eu) and particularly: (1) presence-absence data from 22 National Forest Inventories, (2) plot information from the so-called Level I and Level II schemes set up by countries for the monitoring of atmospheric pollution on forests in the context of Regulation No. 2152/2003 (Forest Focus, [47]), and (3) the Biosoil project [48], in which forest tree biodiversity was sampled in 3,379 plots in Europe. All data were aggregated at a spatial accuracy of 1 square km, by assigning plot presences to the cell centroids of an INSPIRE compliant 1 km × 1 km grid, specifically designed for pan-European mapping (see [44] for details).
Then, we generated a dataset mapping the potential occurrences (PNV) using the map of Bohn et al. [49]. We included in the study only the 190 tree species present in both Bohn et al.’s map and Euforest. We assembled the PNV dataset by projecting the positions of the Euforest plots on the map of Bohn et al. [49], which consists of large numbers of polygons, each associated with a certain vegetation type that, in turn, corresponds to a particular set of plant species. For each plot location, we evaluated the presence/absence of the 190 tree species from the set of species associated with the corresponding polygon. In this way, we obtained two comparable datasets. In order to make clear that we used only tree species in our analyses we will henceforth use the abbreviations ACVT for the actual tree vegetation, and PNVT for the potential tree vegetation.
Because our analysis focuses on tree species interactions, we limited our study to forested areas, in order to exclude localities (such as shrublands) where trees, even if listed in our dataset, are likely to occur at densities too low for species interaction to be a structuring process. For this, we filtered both ACVT and PNVT datasets using the global land cover classification by [50], in order to include only those located in areas categorized as (1) evergreen needleleaf forest; (2) evergreen broadleaf forest; (3) deciduous needleleaf forest; (4) deciduous broadleaf forest; and (5) mixed forest and woodland.
To investigate spatial structure we used a moving window [51] consisting of a regular 10 × 10 grid having size 1° latitude × 1° longitude, with each grid cell being 0.1° × 0.1°. The grid was moved across all of Europe by displacements of 0.1° latitude or longitude at a time (Fig 1). At each repositioning, the grid was superimposed on both ACVT and PNVT distribution maps (i.e. the two tree distribution maps based, respectively, on the three above mentioned datasets, and on the map of PNV [49]) in order to create two species × site matrices including the presence-absence of each tree species of the dataset in each one of the grid cells (n = 100 “sites”) according to, respectively, ACVT and PNVT distribution maps. Only non-empty columns (i.e. grid cells–i.e. “sites”—hosting at least one tree species) and non-empty rows (i.e. tree species found in at least one grid cell) were included in each resulting matrix. To ensure robustness of results, only matrices including at least 5 columns (i.e. “sites”) and 5 rows (i.e. tree species) were retained (leading to a total of 34,353 matrices).
The window consisted of a regular square grid of 100 cells (each having size 0.1° latitude × 0.1° longitude). The window was moved across all of Europe by displacements of 0.1° latitude or longitude at a time. At each repositioning, the window was superimposed on both ACVT and PNVT distribution maps in order to create two species × site matrices.
Each structural measure computed on a given matrix (corresponding to a 1° latitude × 1° longitude area) was then assigned to the centroid of the corresponding window. In this way, we obtained information at a spatial resolution of 0.1 × 0.1 latitudinal per longitudinal degrees.
Quantifying ecological structure
We quantified community spatial structure in both ACVT and PNVT assemblages by using the statistic [26]. The metric aims at assessing whether species in ecological networks and food webs tend to share interacting partners/resources more (or less) than random expectation. The metric is computed as the average, normalized deviation between the observed and the expected number of partners shared by any pair of nodes. Although first conceived for network analysis, this measure can be easily applied to species presence-absence matrices since, in principle, these correspond to bipartite networks, i.e. networks where nodes can be divided into two disjoint sets, one having only in-coming links (e.g. localities inhabited by some species), and the other one having only out-going links (e.g. species inhabiting some localities). The applicability of structural measures conceived for community matrix analysis to bipartite networks (and vice versa) has been widely investigated in studies dealing with nestedness [52], species co-occurrence [53], and modularity [27].
The measure makes it possible to investigate, in a single analytical step, patterns of nestedness, segregation (here meant as the tendency of a species/area matrix towards ‘checkerboardness’, as a potential result of competitive exclusion [14]) and modularity.
The general formula to compute the measure for a pair of nodes Vi and Vj (i.e. species or localities) is: (1) Where Sij is the actual number of neighbors shared by Vi and Vj (for example, the number of localities where both species i and j occur, when computing species overlap; or the number of species occurring at both locality i and j, when computing site overlap). Pij is the expected number of shared neighbors, di and dj are the respective node degrees (i.e. the number of localities where species i and j occur respectively, or the number of species respectively found at locality i and j). Ωij is a standardization parameter corresponding to the maximum theoretical value of Ɲij (see [26] for details). is then computed as the average of all Ɲij pairs, while the standard deviation of Ɲij values provides a measure of modularity.
The key point of the method is the application of a probabilistic approach to compute the expected number of shared nodes: (2)
The metric can vary between -1 (indicating complete spatial segregation) and 1 (indicating perfect nestedness).
For each matrix, we computed both and modularity. In the context of our analysis, Ɲij values represent the degree of overlap in distribution between two tree species, or the degree of overlap in species composition between two localities (i.e. grid cells). We assessed significance of values using a Z test [26].
Comparing the structure of ACV to PNV
The difference in ecological structure between ACVT and PNVT provided by , was mapped at a resolution of 0.1 latitudinal × longitudinal degrees. In addition, to investigate changes for regions with more homogeneous environmental conditions, species communities and ecological processes, we averaged the statistics per eco-region [54] and constructed ecoregion-based maps of the differences between of ACVT and of PNVT. The use of ecoregions has also been suggested as an informative approach for biodiversity conservation assessment [55].
Intact forest landscapes in Europe
Europe is extremely poor in terms of intact forest landscapes (IFL) [41]. Moreover, some of the most important old-growth forests (such as the Białowieża Forest in Poland [56]) are not covered by the Euforest dataset [41, 44]. Nevertheless, there is some IFLs for which the Euforest dataset provides tree occurrence data. We used those areas (that we identified by filtering the Euforest dataset using the global IFL map for the year 2000, available at http://www.intactforests.org/index.html) to challenge our starting hypothesis, i.e. to verify if their tree assemblages showed, as expected, a high degree of structure resulting from the absence of disturbance.
Sensitivity analysis
Although the data we have used to define actual tree vegetation (the Euforest dataset, [44]) represent the best information available on European tree distribution, as for any empirical dataset, we cannot consider this information as complete. Moreover, the fact that data were collected at national level, could possibly lead to a heterogeneous distribution of biases across Europe. We investigated whether these potential issues affected our results and conclusions by performing a robust sensitivity analysis. Since the data about PNV can be considered virtually complete, we focused on ACV, replicating the structural analyses (i.e. the computation of nestedness, segregation and modularity) by removing at random 50% of occurrences from the Euforest dataset, and then comparing these results with those obtained using the complete dataset.
Results
Mean values of for PNVT were much higher than those recorded for the ACVT, indicating a greater tendency towards nestedness in the potential tree vegetation than in the actual one (Fig 2). Moreover, although we found cases of = 1 in both ACVT and PNVT, only ACVT showed negative values (Table 1). This is reflected by Z values, which were much higher in PNVT than in ACVT, with negative values only in the ACVT (Table 1). However, in both PNVT and ACVT significant spatial segregation (i.e. values close to -1, Z <-2) was extremely rare, while we found significant nestedness (i.e. values close to 1, Z>2) in many areas (see Figs 3C and 3D and 4C and 4D). In general, PNVT was much more nested than ACVT, and particularly at Northern latitudes (Figs 3A and 3B and 4A and 4B).
Relationships between , Z-score and modularity values computed for ACVT and the corresponding values computed for PNVT in all 1° latitude × 1° longitude matrices. Diagonal lines of equality provide a visual guide for seeing whether ACVT or PNVT had the greater value.
, Z and modularity values computed in all 1° latitude × 1° longitude matrices for both ACVT and PNVT. Red pixels in mid panels correspond to matrices having Z>2, which indicates significance of at p<0.05.
, Z and modularity values computed in all 1° latitude × 1° longitude matrices for both ACVT and PNVT, averaged per eco-region.
At the scale of eco-regions, values in PNVT were fairly homogeneous across Europe and, in general, higher than 0.4, i.e. indicating a tendency towards nestedness (Fig 4B). Conversely, the scenario depicted by values in ACVT is a bit more complex, with higher values in northeastern regions (and in some areas of the Iberian Peninsula) than in the rest of the continent. It is interesting that, as regards PNVT, there are several localities having significant Z values (i.e. >2) in the United Kingdom, but not in mid-southern Europe (Fig 4C and 4D).
Modularity was more evident for ACVT than for PNVT (Table 1), with most of central and southern Europe having values close to 1 (Fig 3E and 3F). The high modularity values are indicative of the following: (1) species and localities occurred in distinct sets; (2) species of a given set tended to have an overlapping geographic distribution with species of the same set; (3) species of a given set tended to have a disjoint distribution with species of any other set; (4) localities of the same set tended to have similar species composition; and (5) localities belonging to different sets tended to have different species composition. In particular, high modularity of PNVT occurred in central Europe, whereas several localities in Fennoscandia were characterized by very low modularity (Fig 3E and 3F).
Geographical patterns of modularity are even clearer when observed at the level of eco-region. As regards PNVT, we found high modularity in mountain regions, and particularly in the Iberian peninsula, and along the Alps, the Apennines and the Carpathians (Fig 4E and 4F). Conversely, we found, on average, low values in Fennoscandia. A strong difference in modularity can be observed between regions above and below the southern coastlines of the North Sea and of the Baltic Sea, with the latter having higher values (especially in the Mediterranean areas) (Fig 4E and 4F).
Actual tree assemblages in localities falling within the few European intact forest landscapes were characterized by higher nestedness and lower modularity in comparison with non-intact forest landscapes (Student’s t-tests with Welch’s correction: t = -9.277, df = 169.939, P < 2.2×10−16 for nestedness; t = 6.198, df = 169.759, P < 4.2×10−9 for modularity, respectively).
The measure resulted extremely robust to data availability. Both the values of nestedness and modularity obtained using the whole Euforest dataset were virtually identical to those obtained using a random sample of half the tree occurrence records, with R2 values equal, respectively, to 0.91 and 0.78, and both regression lines having a slope close to one and an intercept close to zero (y = 0.87x + 0.15 for nestedness; y = x − 0.08 for modularity).
Discussion
The idea that ecological communities are inherently structured by various processes extends far back in the history of ecological research, at least to Clements' early studies of succession and plant communities as well-organized “super-organisms” [57]. This perspective gained momentum and eventually lead to deterministic views on community structure [14]. Although the idea that communities are structured by species interactions and assembly rules has been challenged by the neutral theory of biodiversity [58, 59], it is reasonable to assume that, given enough time, a certain degree of spatial structure at some scale is likely to emerge in absence of unnatural external disturbance by humans [60, 61]. More cautiously, we could at least assume that natural communities that are frequently disrupted by anthropogenic influences are not as likely to retain structure as are communities that are not disrupted (see, for example, [62]).
Thus, our starting hypothesis was that the continuous and prolonged management of European forests [40] could have led to strong alterations in the natural processes of species interactions and other structuring forces. Although this idea is intuitive, testing it was challenging due to the almost complete lack of information about the level of structure expected in the absence of human disturbance.
Taking advantage of the PNV concept, we were able to fill this knowledge gap, showing that management has set apart the ecological and spatial patterns of contemporary tree communities from those expected in a hypothetical undisturbed scenario. In general, the spatial arrangement of European ACVT resulted highly modular and weakly nested, in clear contrast with that of PNVT, that showed high nestedess and low modularity.
The patterns observed in the PNVT are well consistent with the relatively high values of nestedness observed in the ACVT of the few intact forest landscapes (and, in general, at higher latitudes) when compared to those of other European areas. This provides strong support to our idea that mature vegetation should show spatial consistency in species associations, resulting in an overall tendency for positive co-occurrence. Conversely, the scale of our analysis was large enough to rule out the effect of local competition, as confirmed by the fact that we did not detect any patterns of segregation in either ACVT or PNVT.
Besides the straightforward idea that nestedness reflects consistency in species associations in mature tree assemblages, our results can also reflect complex biogeographical processes. In a perfectly undisturbed scenario, such as that depicted by PNVT, one could expect species assemblages to be driven by colonization and extinction dynamics. These are responsible for species dispersal and replacement along environmental gradients and hence contribute to generating nested patterns. Conversely, nested patterns are likely to be disrupted by human induced biodiversity loss and habitat fragmentation, with the consequent reduction of diversity gradients and, possibly, an increase in modularity. Regardless of their interpretation, however, our results strongly suggest that current European forests are either too ‘young’ to have evolved into a structured system, or, more likely, too much managed to retain the structure predicted by the PNVT.
Making comparisons between PNVs and ACVs is often complicated by the fact that the information provided by the first is, in principle, complete, while the latter is not. This can lead to obvious problems in analyzing and interpreting patterns of richness and diversity. However, an important feature of the methodological approach we have used in this study, is its independence from matrix properties [26], which reduces the potential biases due to differences in data availability. As demonstrated by the sensitivity analysis we conducted by using only half the tree occurrence dataset, the presence of a given structural pattern should emerge even if the information is not complete, which ensures that our conclusions are very robust against potential biases due to unequal sampling or data deficiency. This point is key to understanding why using PNVs as null models for ecological structure is fundamentally different (and less problematic) than using them as null models for vegetation types and/or diversity.
The use of PNV as a benchmark was a crucial aspect of our study. The Map of the Natural Vegetation of Europe that we used in this study [49] is the result of more than two decades of joint work of an international team composed by more than 100 vegetation scientists from 31 European countries [63]. Although remarkable, the effort put in compiling the map is by no means a solution to the many problems highlighted by PNV critics [9], and should not be brought into the debate about the general validity of the concept of PNV itself. Our cautious use of PNV illustrates that there is much value in the concept and data generated from PNV databases. We hope that our approach sheds light on what PNVs can tell us, instead of what they cannot.
Author Contributions
- Conceptualization: G Strona SF AM JSMA JV G Seufert.
- Data curation: G Strona AM G Seufert JSMA.
- Formal analysis: G Strona SF.
- Methodology: G Strona SF JV.
- Resources: JSMA.
- Software: G Strona.
- Supervision: SF JSMA JV.
- Writing – original draft: G Strona SF.
- Writing – review & editing: G Strona SF AM JSMA JV G Seufert.
References
- 1.
Brown JH. Macroecology. Chicago: University of Chicago Press; 1995.
- 2.
Gaston KJ, Blackburn TM. Pattern and Process in Macroecology. Oxford: Blackwell Scientific Publications; 2000.
- 3.
Lomolino MV, Riddle BR, Whittaker RJ, Brown JH. Biogeography, 4th edition. Sunderland (MA): Sinauer Associates, Inc.; 2010.
- 4.
Verhoef HA, Morin PJ. Community ecology: processes, models, and applications. Oxford: Oxford University Press; 2010.
- 5. Diez JM, Pulliam HR. Hierarchical analysis of species distributions and abundance across environmental gradients. Ecology. 2007;88: 3144–3152. pmid:18229848
- 6. Azovsky AI. Structural complexity of species assemblages and spatial scale of community organization: A case study of marine benthos. Ecological Complexity. 2009;3: 308–315.
- 7. Jiménez JJ, Decaëns T, Lavelle P, Rossi JP. Dissecting the multi-scale spatial relationship of earthworm assemblages with soil environmental variability. BMC Ecology. 2014;14: 26. pmid:25476419
- 8. Punchi-Manage R, Wiegand T, Wiegand K, Getzin S, Gunatilleke CS, Gunatilleke IN. Effect of spatial processes and topography on structuring species assemblages in a Sri Lankan dipterocarp forest. Ecology. 2014;95: 376–386. pmid:24669731
- 9. Chiarucci A, Araújo MB, Decocq G, Beierkuhnlein C, Fernández‐Palacios JM. The concept of potential natural vegetation: an epitaph? Journal of Vegetation Science. 2010;21: 1172–1178.
- 10.
Ellenberg H. Vegetation Mitteleuropas mit den Alpen, 4th edn. Stuttgart: Eugen Ulmer; 1963.
- 11.
Westhoff V, van der Maarel E. The Braun-Blanquet approach, Ordination and classification of communities. In: Whittaker RH, editor. Handbook of Vegetation Science Vol. 5, The Hague (NL): Junk; 1973. pp. 617–726.
- 12. Ricotta C, Carranza ML, Avena G, Blasi C. Are potential natural vegetation maps a meaningful alternative to neutral landscape models? Applied Vegetation Science. 2002;5: 271–275.
- 13. Zerbe S. Potential natural vegetation: validity and applicability in landscape planning and nature conservation. Applied Vegetation Science. 1998;1: 165–172.
- 14.
Diamond JM. Assembly of species communities. In: Cody ML, Diamond JM, editors. Ecology and evolution of communities. Cambridge (MA): Harvard University Press; 1975. pp. 342–444.
- 15. Diamond JM, Gilpin ME. Examination of the ‘‘null” model of Connor and Simberloff for species co-occurrences on islands. Oecologia. 1982;52: 64–67.
- 16. Gotelli NJ, McCabe D. Species co-occurrence: a meta-analysis of J. M. Diamond's assembly rules model. Ecology. 2002;83: 2091–2096.
- 17. Chase JM. Community assembly: when should history matter? Oecologia. 2003;136: 489–498. pmid:12836009
- 18. Fukami T, Martijn Bezemer T, Mortimer SR, Putten WH. Species divergence and trait convergence in experimental plant community assembly. Ecology letters. 2005;8: 1283–1290.
- 19. Hutchinson GE. Homage to Santa Rosalia or why are there so many kinds of animals? American naturalist. 1959;870: 145–159.
- 20. Leibold MA. Similarity and local co-existence of species in regional biotas. Evolutionary Ecology. 1998;12: 95–110.
- 21. Šímová I, Violle C, Kraft NJB, Storch D, Svenning J-C, Boyle B, et al. Shifts in trait means and variances in North American tree assemblages: species richness patterns are loosely related to the functional space. Ecography. 2015;38: 649–658.
- 22. Dickie IA, Schnitzer SA, Reich PB, Hobbie SE. Spatially disjunct effects of co-occurring competition and facilitation. Ecology Letters. 2005;8: 1191–1200. pmid:21352443
- 23. Karadimou E, Tsiripidis I, Kallimanis AS, Raus T, Dimopoulos P. Functional diversity reveals complex assembly processes on sea-born volcanic islands. Journal of Vegetation Science. 2015;26: 501–512.
- 24. Ulrich W, Almeida-Neto M. On the meanings of nestedness: back to the basics. Ecography. 2012;35: 865–871.
- 25. Staniczenko PPA, Kopp JC, Allesina S. The ghost of nestedness in ecological networks. Nature Communications. 2013;4: 1391. pmid:23340431
- 26. Strona G, Veech JA. A new measure of ecological network structure based on node overlap and segregation. Methods in Ecology and Evolution. 2015;6: 907–915.
- 27. Thébault E. Identifying compartments in presence–absence matrices and bipartite networks: insights into modularity measures. Journal of Biogeography. 2013;40: 759–768.
- 28. Olesen JM, Bascompte J, Dupont YL, Jordano P. The modularity of pollination networks. Proceedings of the National Academy of Sciences. 2007;104: 19891–19896.
- 29. Montoya JM, Pimm SL, Solé RV. Ecological networks and their fragility. Nature. 2006;442: 259–264. pmid:16855581
- 30. Hausdorf B, Hennig C. Null model tests of clustering of species, negative co-occurrence patterns and nestedness in meta-communities. Oikos. 2007;116: 818–828.
- 31. Proulx SR, Promislow DE, Phillips PC. Network thinking in ecology and evolution. Trends in Ecology & Evolution. 2005;20: 345–353.
- 32. Attiwill PM. The disturbance of forest ecosystems: the ecological basis for conservative management. Forest Ecology and management. 1994;63: 247–300.
- 33. Liu J, Dietz T, Carpenter SR, Folke C, Alberti M, Redman CL, et al. Coupled human and natural systems. AMBIO: a journal of the human environment. 2007;36: 639–649.
- 34. Brooks ML, D'antonio CM, Richardson DM, Grace JB, Keeley JE, DiTomaso JM, et al. Effects of invasive alien plants on fire regimes. BioScience. 2004;54: 677–688.
- 35. Perrings C, Dehnen-Schmutz K, Touza J, Williamson M. How to manage biological invasions under globalization. Trends in ecology & evolution. 2005;20: 212–215.
- 36. Haas SE, Hooten MB, Rizzo DM, Meentemeyer RK. Forest species diversity reduces disease risk in a generalist plant pathogen invasion. Ecology letters. 2011;14: 1108–1116. pmid:21884563
- 37. Skelsey P, Rossing WAH, Kessel GJT, Powell J, Van der Werf W. Influence of host diversity on development of epidemics: an evaluation and elaboration of mixture theory. Phytopathology. 2005;95: 328–338. pmid:18943033
- 38. Bradshaw CJ, Sodhi NS, Peh KSH, Brook BW. Global evidence that deforestation amplifies flood risk and severity in the developing world. Global Change Biology. 2007;13: 2379–2395.
- 39. Kaplan JO, Krumhardt KM, Zimmermann N. The prehistoric and preindustrial deforestation of Europe. Quaternary Science Reviews. 2009;28: 3016–3034.
- 40.
Gaillard MJ. Detecting human impact in the pollen record. In Elias SA, editor. Encyclopedia of Quaternary Science. London: Elsevier; 2007. pp. 2570–2595.
- 41. Potapov P, Yaroshenko A, Turubanova S, Dubinin M, Laestadius L, Thies C, et al. Mapping the World's Intact Forest Landscapes by Remote Sensing. Ecology and Society. 2008;13: 51.
- 42. Butaye J, Jacquemyn H, Hermy M. Differential colonization causing non-random forest plant community structure in a fragmented agricultural landscape. Ecography. 2001;24: 369–380.
- 43. Honnay O, Hermy M, Coppin P. Nested plant communities in deciduous forest fragments: species relaxation or nested habitats? Oikos. 1999; 84: 119–129.
- 44. Mauri A, Strona G, San Miguel-Ayanz J. A high-resolution tree occurrence dataset for Europe. Scientific Data. 2016;3:160123.
- 45. Loidi J, Fernández-González F. Potential natural vegetation: reburying or reboring? Journal of Vegetation Science. 2012;23: 596–604.
- 46. Somodi I, Molnár Z, Ewald J. Towards a more transparent use of the potential natural vegetation concept–an answer to Chiarucci et al. Journal of Vegetation Science. 2012;23: 590–595.
- 47.
European Commmission. Regulation (EC) No 2152/2003 of the European Parliament and of the Council of 17 November 2003 concerning monitoring of forests and environmental interactions in the Community (Forest Focus) Official Journal of the European Union. 2003;L 324/1.
- 48.
Durrant T, San-Miguel-Ayanz J, Schulte E, Suarez Meyer A. Evaluation of BioSoil Demonstration Project: Forest Biodiversity. EUR 24777 EN, Publication Office of the European Union, Luxemburg. 2011; 36.
- 49.
Bohn U, Neuhäusle R, Gollub G, Hettwer C, Neuhäuslová Z, Raus T, et al. Map of the natural vegetation of Europe. Explanatory text with CD-ROM. 2003; Bonn: German Federal Agency for Nature Conservation.
- 50. Hansen MC, DeFries RS, Townshend JR, Sohlberg R. Global land cover classification at 1 km spatial resolution using a classification tree approach. International Journal of Remote Sensing. 2000;21: 1331–1364.
- 51.
Lloyd CD. Local models for spatial analysis. 2010; CRC Press.
- 52. Bascompte J, Jordano P, Melián CJ, Olesen JM. The nested assembly of plant–animal mutualistic networks. Proceedings of the National Academy of Sciences. 2003;100: 9383–9387.
- 53. Strona G. The underrated importance of predation in transmission ecology of direct lifecycle parasites. Oikos. 2015;124: 685–690.
- 54. Olson DM, Dinerstein E, Wikramanayake ED, Burgess ND, Powell GV, Underwood ECet al. Terrestrial Ecoregions of the World: A New Map of Life on Earth A new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience. 2001;51: 933–938.
- 55. Kier G, Mutke J, Dinerstein E, Ricketts TH, Küper W, Kreft H, et al. Global patterns of plant diversity and floristic knowledge. Journal of Biogeography. 2005;32: 1107–1116.
- 56. Bobiec A. Living stands and dead wood in the Białowieża forest: suggestions for restoration management. Forest Ecology and Management. 2002;165: 125–140.
- 57.
Clements FE. Plant succession: an analysis of the development of vegetation. 1916; Washington: Carnegie Institution.
- 58.
Hubbell SP. The Unified Neutral Theory of Biodiversity and Biogeography. 2001; Princeton: Princeton University Press.
- 59. McGill BJ. A test of the unified neutral theory of biodiversity. Nature. 2003;422: 881–885. pmid:12692564
- 60. Tilman D. Competition and Biodiversity in Spatially Structured Habitats. Ecology. 1994;75: 2–16.
- 61. Stokstad E. On the Origin of Ecological Structure. Science. 2009;326: 33–35. pmid:19797635
- 62. Steadman DW. Biogeography of Tongan birds before and after human impact. Proceedings of the National Academy of Sciences. 1993;90: 818–822.
- 63. Bohn U, Gollub G. The use and application of the map of the natural vegetation of Europe with particular reference to Germany. Biology and Environment: Proceedings of the Royal Irish Academy. 2006;106B: 199–213.