The effects of increasing land use intensity on soil nematodes: A turn towards specialism

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2019 The Authors. Functional Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society 1Wageningen University and Research, Wageningen, The Netherlands 2Netherlands Institute of Ecology (NIOO‐ KNAW), Wageningen, The Netherlands 3National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands


| INTRODUC TI ON
Humans derive multiple benefits from the soil system (Sarukhan et al., 2005). The delivery of these 'ecosystem services' depends on a number of soil processes driven by different groups of soil dwelling fauna (Ferris & Tuomisto, 2015;Haygarth & Ritz, 2009;Kibblewhite, Ritz, & Swift, 2008). Soil biota acts as decomposers, nutrient transformers, ecosystem engineers and bio-controllers (Kibblewhite et al., 2008). For example, earthworms, enchytraeids and fungi can act as ecosystem engineers (functional group) by restructuring soil material (ecosystem process) which in turn affects erosion, water quality and water supply (ecosystem services) (Pulleman et al., 2012).
Microbes are also important nutrient cyclers and decomposers, and changes in the microbial community can lead to reduced decomposition rates, affecting the provision of food, fibre and water, as well as the capacity of the soil to reduce pollutant concentrations (Bardgett & van der Putten, 2014). The soil food web drives these nutrient transformations, but can also be impacted by (human induced) changes in soil quality and resilience, which in turn can reduce the capacity of soil processes, functions and ecosystem service delivery (Haygarth & Ritz, 2009;Wagg, Bender, Widmer, & Heijden, 2014).
While there have been many attempts to quantify the effect of land use change and land use intensity (LUI) on the diversity of soil biota, none have distinguished these effects on the proportion of generalists and specialists, most probably due to the difficulty in establishing a soil organisms' niche width (Bardgett, 2002).
Traditional methods of calculating a species fundamental niche width (the set of environmental conditions a species can potentially live and reproduce in) require measuring the 'variance in performance measured in common garden or controlled experiments' (Futuyma & Moreno, 1988;Venail et al., 2008). There are two reasons why this approach is not appropriate for soil biota: firstly, due to the high physical and chemical heterogeneity and microclimatic characteristics at small scales in the soil, which result in a myriad of niches (Bardgett, 2002;Ettema & Wardle, 2002); and secondly, more than 5,000 species (belonging to more than 770 genera) of soil and fresh water dwelling nematodes have been described world-wide (Andrassy, 1992). Selecting and manipulating the soil variables that limit species occurrence and setting up individual controlled experiments would become a daunting (time and resource consuming) task.
There are examples, however, of classifications of soil organisms according to traits that are often associated with niche widths. Nematode families, for example, have been classified into the c-p scale, from colonizers to persisters (Bongers, 1990). Bongers (1990) considered nematodes that rapidly increase in number under favourable conditions, with a short life cycle, high colonization ability and a high tolerance to disturbance as colonizers. At the other end of the spectrum are the 'persisters', nematodes with a low reproduction rate, a long life cycle, a low colonization ability and sensitive to disturbance. This classification has served as a starting point to calculate diverse ecological indices to assess, for example the successional stage, disturbance level or nutrient status of the soil (Bongers, systems, however, specialists may also suffer from negative effects of land use intensity. 5. This co-occurrence method to assess niche width opens the door to estimating the soil community's niche breadth, for which resource-based methods are difficult to implement. 1990;Ferris, Bongers, & Goede, 2001). Colonizers fit the typical description of generalists, while persisters are better associated with specialism. There is, however, a lack of consensus regarding the directionality of the relationship between biological traits and specialization (Büchi & Vuilleumier, 2016). This classification into the c-p scale might not be representative of the niche width of the different nematode families, since other factors (such as pH tolerance, host breadth, resting phase) might also limit niche width.
A different method of estimating niche width is calculating the realized niche width (RNW; the set of conditions a species occupies) rather than the fundamental niche width, using diversity metrics or multivariate techniques (Devictor et al., 2010;Futuyma & Moreno, 1988). This approach is not biased by the choice of measured variables or the availability and ease of collection of environmental data (Fridley, Vandermast, Kuppinger, Manthey, & Peet, 2007), problems that are often found when using resource-based methods (see Gaston, Blackburn, & Lawton, 1997 for a review). One such approach uses large-scale co-occurrence matrices under the assumption that extreme specialist species will occur always in the company of the same species, while extreme generalists will occur in very different communities (Fridley et al., 2007;Manthey & Fridley, 2009;Zelený, 2009). Co-occurrence, resource and trait-based methods can lead to similar niche width estimates, but deviations can occur when one species is limited by a resource that is not limiting to others (Carboni, Zelený, & Acosta, 2016;Pannek, Manthey, & Diekmann, 2016).
While co-occurrence methods provide no direct insight into the underlying mechanisms that determine the target species' niche width, they offer the opportunity to study the effects of LUI on communities' overall specialism by calculating an index of community specialization (I CS ) which can be used as an indicator of agricultural intensification (Fried et al., 2010). In this paper, we calculated the RNW of different nematode taxa using data gathered in multiple habitats in the Netherlands. We explored its relationship with the aforementioned c-p scale, as well as other functional traits, in order to understand what determines nematode niche width. We then studied nematode richness, abundance, diversity and the I CS in different habitats in the Netherlands. Finally, we assess the effects of LUI on these four indices and hypothesize that with increasing LUI, there will be a decrease in the I CS , diversity and richness.

| Dataset
Most of the data used for this study were gathered as part of the Netherlands Soil Monitoring Network (NSMN [Rutgers et al., 2009]; Table 1). This monitoring network incorporated abiotic and biotic measurements taken in sites representative of the more common land use/soil texture categories in the Netherlands (Rutgers et al., 2008(Rutgers et al., , 2009. Due to the overwhelming number of samples belonging to dairy farms on sand (115 out of the NSMN 458 sites), and to prevent bias in the niche width calculation deriving from an uneven site selection (Fridley et al., 2007), the dataset was complemented with data (195 sites) from survey studies carried out in the Netherlands in different habitats (Table 1). For sites with several replicates, only one was chosen at random to be part of the dataset. Soil samples were taken from the top 10-20 cm, and nematode extraction was done using an Oostenbrink elutriator. Further information on the sampling procedures can be found in the literature cited in Table 1.
Each sample was categorized according to its land use and soil texture category, resulting in a combination of arable farms, organic and conventional dairy farms, horticulture, city parks, extensively managed grasslands, heathland, dune systems, and coniferous and deciduous forests in a combination of loess, sandy, clayey, loamy and peaty soils (Table 1). We split the dataset (655 sites) into two sets, one was used to calculate the RNW of the target nematode taxa (preliminary set; 229 sites), and the other to test the hypotheses (test set; 426 sites, of which different subsets were selected for further analysis depending on the question at hand). Site selection for the preliminary set is further discussed in the Section 2.4 of the methods.

| Weather data
Soil biota goes through seasonal changes in abundance and composition. In arable and grass fields, microbial and nematode biomasses are highest in the early summer (Buchanan & King, 1992) and lowest in the winter (Bardgett, Leemans, Cook, & Hobbs, 1997;Bardgett, Lovell, Hobbs, & Jarvis, 1999). Water content in the soil can affect nutrient availability, and strong rainfall can lead to nutrient losses through leaching (Bhadoria, Kaselowsky, Claassen, & Jungk, 1991;Kuchenbuch, Claassen, & Jungk, 1986). The effects of temperature and rainfall on the soil's condition may further depend on soil texture (Bhadoria et al., 1991) or the presence of vegetation (Green, Harding, & Oliver, 1984).
To account for differences in sampling season and year, each data point was complemented with information on the long-term (3 months) and short-term (a week) weather prior to sampling.
Average daily temperature (°C), number of freezing days and cumulative precipitation (mm) for the two periods of interest were extracted from the Royal Netherlands Meteorological Institute (KNMI) from the closest available weather station. Due to strong collinearity between short-and long-term weather, only long-term average daily temperature, long-term cumulative precipitation and log-transformed short-term cumulative precipitation were used as explanatory variables in the models (Equations 2 and 3). For five of the sites in the test set, exact coordinates were not available, so rather than local weather data, we used national averages.
Identification was not always possible beyond the family level. If members of a specific family were rarely identified to genus level, further calculations were done at family level (this was the case for Criconematidae, Dolichodoridae, Neodiplogasteridae, Qudsianematidae, Rhabditidae, Thornenematidae and Trichodoridae). If, however, members of a family had been identified to genus level more often than to family level, the nematodes identified into family level were allocated to the genera (within said family) present in the same site. Calculations were done at genus level. This was done to prevent an overestimation of taxon diversity. Prodorylaimus and Mesodorylaimus were grouped prior to analysis. Dauer larvae were analysed as a separate taxon, since they represent a common response to a stressor or environmental cue.

Soil type
Land use TA B L E 1 Land use type, soil texture and number of independent sites used to calculate nematodes' realized niche width (preliminary set), to study the variables that affect the nematode's index of community specialization (test set) and related references

Sites in test set Reference
Nematode taxa were assigned a c-p value (Bongers & Bongers, 1998), feeding group (Yeates, Bongers, DeGoede, Freckman, & Georgieva, 1993), functional guild (Ferris et al., 2001), and metabolic footprint (Ferris, 2010), which were extracted from Nemaplex (Ferris, 1999;last accessed November 2018). For average body mass values, we used values reported by Mulder and Vonk (2011), which include the weight of males, females and juveniles extracted from soils belonging to the NSMN. The averages reported by Ferris (2010) are unlikely to be representative of our sample, since they are based on average female weights (which in the case of endoparasitic nematodes cannot be extracted following the procedures in the present work) and have recently been reported to grossly overestimate the average size of nematodes extracted from the soil (Zhao et al., 2019).
Filenchus, Aphelenchoides and Ditylenchus were classified as fungus feeders. Body mass was log-transformed prior to analysis.

| Realized niche width
To quantify the nematodes' RNW, we used the protocol developed by Fridley et al. (2007) with some adjustments. Our data selection for the preliminary set is a fair representation of Dutch habitats. The Netherlands uses up to 60% of its land for agriculture, and only slightly above 12% of the country is considered to be woodland or nature (CBS, 2016). To prevent bias towards one or another habitat (Fridley et al., 2007), the preliminary set was made out of no more than 10 sites per land use/soil texture category (Table 1; resulting in a total of 229 sites), under the assumption that different soil textures and land use types and management styles (organic vs. conventional) provide distinct habitats for soil life (Freckman & Ettema, 1993;de Goede & Bongers, 1994;Jiao et al., 2015;Quist, 2017;Quist et al., 2016).
This protocol is known to be biased when the community is or appears saturated, that is when an increase in landscape (gamma) diversity does not lead to an increase in local (alpha) diversity (Manthey & Fridley, 2009;Zelený, 2009 [Oksanen et al., 2018]). This random selection procedure was repeated 100 times, and we took the average Jaccard's dissimilarity in these 100 repetitions as the target taxon's RNW (theta [θ] in the initial protocol [Fridley et al., 2007]). As a consequence, taxa present in more sites will have a more accurate estimate of RNW.
We analysed differences between RNW and the aforementioned traits using either Spearman's rank order correlation (using the 'cor. test' function in r; [R Core Team, 2017]) for continuous variables or Kruskal-Wallis' rank sum test (Hollander & Wolfe, 1999) (Jenks, 1967).
This division into groups is intended to facilitate the calculation of the I CS . Nematodes classified into specialists simply have a narrower niche width than those classified as generalists. Goodness of variance fit (GVF), a measure based on sum of squares deviation between values and mean, which ranges from 0 (worst fit) to 1 (perfect fit), was used to evaluate the split. Both tests were carried out using the 'classInt' package for r (Bivand, 2017).

| Nematode diversity indices
To monitor the nematode community, we calculated nematode abundance (in number of nematodes per 100 g fresh weight), nematode richness (defined as the number of taxa present in a site), nematode diversity and the I CS .
We calculated taxon diversity using the Shannon-Weaver index and S is the total number of taxa identified per site (the site's richness; Hill (1973)). We used the function 'diversity' from the vegan package (Oksanen et al., 2018).
The I CS was calculated such that: where s i is the abundance of specialist nematodes in site i, and g i is the abundance of generalist nematodes in site i.
We selected (from the test set) land use/soil texture combinations with 10 or more replicates and tested whether different land use/soil texture combinations have a different I CS using the Kruskal-Wallis' rank sum test for categorical variables (Hollander & Wolfe, 1999) (using the aforementioned function in r). We assessed differences between the groups using Dunn's test for multiple comparisons with Bonferroni adjustment for p-values using the 'posthoc.

| Effects of land use intensity on the nematode community
While the test set did not permit complete combinations of all land use categories and soil textures, it did allow to test differences in nematode diversity indices due to LUI in (a) sandy soils, where data were available for land use classes with ascending LUI (shrubland-  (Mulder, Zwart, Wijnen, Schouten, & Breure, 2003).
Data were analysed following the protocols proposed by Zuur, Ieno, and Elphick (2010) and Zuur and Ieno (2016). Collinearity between explanatory variables was assessed using correlation plots.
Variance inflation factors (VIF) were calculated for the remaining independent variables using the 'corvif' function for r, which was below 3 for all variables, and none were removed (Zuur, Ieno, Walker, Saveliev, & Smith, 2009). The initial models were such that: where Response is either I CS , abundance, richness or diversity; LUI i is the LUI category in sandy soil; LTAvgT i and LTCPP i are the long-term average temperature and cumulative precipitation; STCPP i is the short-term cumulative precipitation (log-transformed); Ctot i is total carbon (%, determined by thermogravimetric analysis); and PAL i is the extractable phosphorus (determined using an ammonium lactate extraction and expressed in mg P 2 O 5 /100 g dry soil), in site i.
Model selection processes were done following (Zuur et al., 2009), starting with all variables under study and ecologically motivated interactions, terms were dropped using the AIC criterion. Since I CS is restricted from 0 to 1, we used beta regressions (Ferrari & Cribari-Neto, 2004) to test the relationship between I CS and explanatory variables in Equations (1) and (2), using the 'betareg' package in r (Cribari-Neto & Zeileis, 2010). Beta distributions are restricted from 0 to 1, but include neither of these values, and thus, we transformed I CS so that zeroes and ones became numbers close to 0 and 1 respectively, such that: where I CS is the I ′ CS without zeroes or ones, I ′ CS is the index of community specialization calculated using Equation (1), and n is the total number of sites in the analysis (Cribari-Neto & Zeileis, 2010).
To study the relationships between the explanatory variables and (a) taxon richness, (b) nematode abundance and (c) taxon diversity, we used for (a) a Poisson generalized linear model, using the 'glm' function of the 'stats' package (R Core Team, 2017); for (b) a negative binomial generalized linear model, and models were fit using the 'glm.nb' function of the 'mass' package (Venables & Ripley, 2002); and in the case of (c), we fit different regressions to test the models presented in Equations (2) and (3). After checking the residual plots, and performing a log-likelihood ratio test to compare models (L. Ratio, p-value), a model allowing for variable variances per land use type (fit using the 'varIdent' function of the 'nlme' package (Pinheiro, Bates, DebRoy, & Sarkar, 2017) such that 1|LUI i ) was preferable for

Equation (2), while a linear regression was used to test Equation (3).
When LUI was a significant explanatory variable in the resulting models, we carried out a Wald test to compare two linearly restricted models to assess whether LUI classes were different from one another (Fox, 1997) using the 'linear.hypothesis' function from the 'car' package (Fox et al., 2012).

| Nematode realized niche width
There were 45 target taxa (occurring in at least in 23 sites) belonging to 26 families. These target nematode taxa occurred with an average of 19 other taxa per site (alpha diversity) and can co-occur with an average of 69 taxa in 20 sites (gamma diversity). Realized niche width (quantified using Jaccard's dissimilarity) was 0.63, ranging from 0.524 (Psilenchus) to 0.689 (Heterocephalobus) ( Table S1). Realized niche width showed no significant relationships with putative feeding, c-p value, herbivore guild or the average taxon mass (Figure 1).
Mean community specialization (I CS ) was lowest in forests, followed by heathland, semi-natural grasslands, arable fields and highest in dairy farms (Figure 2; Kruskal-Wallis, χ 2 = 284, df = 10, p-value <.01). We observed different mean I CS in different land use/texture classes (Dunn's test for multiple comparisons with Bonferroni adjustment for p-values; Figure 2). Variations in I CS were driven by an increase in nematodes with a narrower niche width ( Figure S1).

| Land use intensity in sandy soils
After initial model validation, a point with a very large Cook's distance and high generalized leverage (an extensive dairy farm on sandy soil with 0 generalists) was taken out and the model was refit. This had no strong effects on the coefficients, but increased the precision parameter. Further model validation showed no underlying problems. In sandy soils, long-term daily average temperature as well as long-term cumulative precipitation had a significant effect on the proportion of specialist nematodes in the community (beta regression, Pseudo R 2 = .66, log-likelihood of 228.2 on 16 degrees of freedom; Table S2), such that drier, warmer conditions relate to the highest I CS (Figure 3a). I CS was significantly lower in shrublandwoodland systems than in the rest of the land use types (Table 2). This land use type had a higher proportional abundance of generalist nematodes ( Figure S2a). Non-target taxa (taxa for which we did not n calculate a RNW, since they were not present in enough sites) made up a higher proportion of the total population in shrubland-woodland systems (46 taxa) than in semi-natural grasslands (33 taxa), dairy farms (34 taxa) and arable farms (7 taxa) ( Figure S2).
Land use intensity explained a significant part of the variation in nematode richness (29% of the variation), which was unaffected by weather. Long-term average daily temperature had different effects on the abundance and diversity of nematodes in soils under different LUIs (Table S2; Figure 3). Increases of temperature led to an increase in nematode abundance in shrubland-woodland systems and a slight decrease in semi-natural grasslands, but had no effects on the nematode abundance in dairy and arable farms. Temperature and LUI explained 48% of the variation in nematode abundance (Table S2).

| Land use intensity in dairy farms
Within dairy farms, I CS increased slightly with increasing PAL and precipitation, but there was a significant negative interaction between long-term cumulative precipitation and PAL (beta regression, pseudo R 2 = .39; log-likelihood of 105.7 on 10 degrees of freedom; Figure 4a). I CS was significantly higher on peaty soils, which tend to have a lower proportion of generalists than clayey and sandy soils (Table S3; Figure S3).
None of the studied independent variables could explain the vari-

| D ISCUSS I ON
In this paper, we estimated the niche width of soil nematodes using co-occurrence data, studied the effects of land use on the nematode community, and the vulnerability of specialist nematodes to LUI. Habitat generalists were similarly abundant in all studied systems, while nematodes with a narrower niche width (opposite to our expectations) dominated agricultural landscapes.
The I CS was lowest in forests, and higher in the other land use types. The communities' specialization in dairy farms increased with increasing PAL, but the overall role of nutrient availability in determining I CS appears to be dependent on external factors such as weather conditions.

| Realized niche width
Although the protocol to calculate RNW using co-occurrence matrices was initially developed and applied to tree communities, it has since been used to calculate the RNW of, for example, vertebrates (Ducatez et al., 2014). The suitability of this method to calculate the RNWs of such different organisms resides in the simplicity of the idea behind it: a habitat specialist will occur in the company of species that can inhabit the same habitat. A generalist might appear in this and other habitats in the company of diverse sets of species. It is widely accepted that nematode communities differ from one another under different environmental conditions, even in environments that are already extreme (Kerfahi et al., 2017), implying at least community level habitat specialization. As such, the protocol could also be used to assess the RNW of other soil fauna for which similar trends F I G U R E 3 Expected index of community specialization (ICS; a), abundance (c) and diversity (d); and observed richness (b), under different land use intensities. Lines correspond to the predicted values in different models, with significant explanatory variables in the x axis: average temperature (a, c, d) and cumulative precipitation (CPP) (a) in the 3 months prior to sampling. Shaded areas represent the 95% confidence interval for each model. Letters denote significant differences between groups through pairwise comparisons of linearly restricted models have been observed (such as earthworms [Decaëns, Margerie, Aubert, Hedde, & Bureau, 2008], enchytraeids or collembolans [van Dijk et al., 2009]). It is, however, not suitable for communities that are or appear saturated (which might be the case for soil bacteria [Raynaud & Nunan, 2014]) although Zelený (2009) proposed a Beals transformation of the zeroes in the co-occurrence matrix prior to the calculation the niche width of saturated communities.
There were no significant relationships between RNW and the studied traits, (i.e. nematode life-history and feeding groups).
While there is a possibility that there is no relationship between these factors, the sample size of the groups in our study may be too small to pick up any sort of significant pattern. The target taxa contained, for example, only three predatory nematodes and five omnivores. A decrease in dispersal ability has been associated with specialism in the past, particularly in the case of seed dispersal (Fridley et al., 2007). Similarly to seeds, smaller nematodes are more likely to be wind-blown than larger nematodes (Ptatscheck, Gansfort, & Traunspurger, 2018), but what this might mean to dispersal rates is not clear. There is no knowledge on how wind dispersal compares to crawling, or how either of these compare to other forms of passive dispersal. Furthermore, while dispersal might play a role in RNW, it might be small compared to that of food availability, pH tolerance or host breadth in the case of plant parasites.
Niche breadth is a result of differences in environmental stability: a stable environment leads to specialization, and a heterogeneous environment will lead to different generalist strategies depending on the time-scale at which organisms experience disturbance (Clavel et al., 2011;Levins, 1968). This means that the same disturbance is experienced differently between below-and above-ground organisms, but also by soil free-living nematode species with different life spans. For example, a species with a 7-day life span (Rhabditis terracoli) or one with a 138-day life span (in Cephalobus dubius) (Gems, 2000) will experience a week of flooding as either the nature of the environment (in the first case) or a temporary disturbance (in the latter), and this event will have a very different impact in their evolutionary history.
TA B L E 2 Differences in the nematode index of community specialization (I CS ), taxon richness, abundance (nematodes per 100 g of fresh weight) and Shannon diversity between different land use categories in sandy soils Contrary to previous studies (Carboni et al., 2016;Ducatez et al., 2014;Fridley et al., 2007), community level specialization increased with increasing LUI. Lowest levels of community specialization were observed on shrubland-woodland systems. These are characterized by having a very low human impact, due in part to the poor nutrient availability and acidity of the soils. Nematode habitat generalism might be a reflection of any of these characteristics, a tolerance to a wide pH range or the possibility to survive under different nutrient regimes.

| Land use intensity on sandy soils
Once we accounted for weather variations, we discovered that the I CS was still lowest in shrubland-woodland systems, but did not differ significantly between semi-natural grasslands, dairy farms and arable farms in sandy soils. The management of productive soils in the Netherlands strives to provide stable conditions for the plant's growing season, conditions that will also favour nematode growth by maintaining high nutrient inputs and minimizing impacts such as drought or flooding. In this context, a growing season in a managed system might represent an unstable environment for organisms with life spans longer than a growing season, but a stable one for nematodes, which might explain a more specialized community.
Take for example, the case of Rhabditidae, a taxon classified as a generalist. A flush of nutrients often leads to a rapid increase in the number of Rhabditidae (Ettema & Bongers, 1993). When the nutrients are scarce, the new generation might go into a resting stage (dauer larvae) to wait for better conditions. The ability to go into a temporary developmental stage is one of the expected outcomes of evolution in a system with coarse environmental variability (a disturbance regime that affects only some members of the population at a time) (Levins, 1968). Dauer larvae, which we treated as its own target taxon, had a relatively narrow niche width. This might be a result of farmland management practices, which stimulate the growth of Rhabditidae, and the subsequent appearance of dauer larvae, while conditions in natural systems rarely allow for a flush of Rhabditidae.
To confirm this, we calculated the I CS without dauer larvae did not observe any change in the observed trends ( Figure S4).

| Land use intensity in dairy farming
Dairy farms on peat soils (with a higher carbon content) had a more specialized community than those on sand or clay.
Increased soil carbon content was also related to an increase in nematode abundance, in line with previous results (Briar, Grewal, Somasekhar, Stinner, & Miller, 2007;Ferris, Venette, & Lau, 1996 result of a higher cattle density grazing and/or a higher frequency of mechanical manure applications, both of which could lead to compaction of the soil, but also an increase in nitrogen in the soil (Bilotta, Brazier, Haygarth, & Sparks, 2007;Matches, 1992;Mulholland & Fullen, 1991;Scholefield & Hall, 1986).We found a slight positive relationship between PAL and I CS but also an interaction between PAL and long-term cumulative precipitation that led to a decrease in I CS . Manure applications have been shown to lead to an increase in the total number of nematodes (Forge, Bittman, & Kowalenko, 2005), which can lead to unevenness in the nematode community (which will result in a decrease in diversity) if some nematodes are benefited more than others (in line with our results). While previous studies have observed an interaction between the effect of nitrogen additions and rainfall to the nematode communities, such that an increase of these two factors led to a decrease in nematode abundance and a change in composition (Sun et al., 2013), we could find no studies linking PAL and precipitation with any effects on the nematode community, nor can we provide a suitable explanation without further study.

| CON CLUS IONS
Co-occurrence-based methods of niche width estimation offer a great opportunity for soil ecology, as well as a potential tool for biological soil quality assessment. Soil biota is often difficult to culture and manipulate, and much of its ecology is still to be discovered. Contrary to our expectations, the highest levels of community specialization were found on farmland systems. The average taxon composition in farmland highlights the environmental homogeneity of such environments (particularly during the growing season), a fact also supported by the decrease in species rarity in these systems.
We provide the first look into the RNW of soil nematodes, a soil biota group with a relatively well known ecology, but we suggest that future work should look into the niche width of other soil biota groups. From our work, there are strong indications that belowground community level specialization is a result of human activity, but that different activities might have different effects on the overall specialization of the community.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support this work were mainly collected by the