Management Intensification of Hay Meadows and Fruit Orchards Alters Soil Macro- Invertebrate Communities Differently

Land-use changes and especially management intensification currently pose a major threat to biodiversity both on and beneath the soil surface. With a comparative approach, we investigated how management intensity in orchards and meadows influences soil macro-invertebrate communities in a North-Italian Alpine region. We compared soil fauna assemblies from traditional low-input sites with respective intensively managed ones. As expected, the taxonomical richness and diversity were lower in both intensive management types. Extensive management of both types revealed similar communities, while intensification led to substantial differences between management types. From these results, we conclude that intensification of agricultural practices severely alters the soil fauna community and biodiversity in general, however, the direction of these changes is governed by the management type. In our view, extensive management, traditional for mountain areas, favors soil fauna communities that have adapted over a long time and can thus be viewed as a sustainable reference condition for new production systems that consider the protection of soil diversity in order to conserve essential ecosystem functions.


Introduction
The industrialization of agriculture characterized by high input practices in most cases increases production but severely hampers the overall diversity of life that inhabits managed areas from single fields to whole landscapes [1,2]. This change towards more intensive management is currently causing major losses of biodiversity on a global scale [3,4]. On the other hand, there is a growing acknowledgment that biodiversity sustains the agricultural production with several direct and indirect ecosystem services, from supporting pollination and pest control [5] to the provision of long-term stability and the maintenance of ecosystem functions [6,7]. Consequently, biodiversity loss and the resulting possible absence of ecosystem services are considered to be so profound that they pose a threat to the current and future food provisioning system [8]. Thus, developing sustainable agricultural practices that conserve biodiversity and secure the provision of food in the long run, has become a global goal and challenge [8][9][10][11]. It is of major interest (1) to describe the state of production systems that might sustain a major proportion of biodiversity and are considered sustainable long-term and

Materials and Methods
The study region (South Tyrol, Italy) is an Alpine region where agriculture is characterized by small farms, with orchards dominating the agricultural practice in the valley floors and hay meadows on intermediate elevations [29]. Hay meadows we defined as managed grasslands mowed for hay production at least once a year, with no or limited pasturing occurring in late summer and autumn after the hay harvest.
The extensive and traditional management form of both practices was locally performed for several hundred and up to thousands of years and is expected to harbor a specialized and diverse community [30,31]. Both production types underwent land-use changes (intensification or abandonment) over the course of the last decades and intensive management has broadly replaced the extensive form, especially on lower elevations [32].
A full factorial design was chosen with a set of four fields in six different locations ( Figure 1). Each location comprised two meadows and two orchards, each with intensive and extensive management; overall, 24 fields were investigated. The selection of the locations was limited by the requirement that all four management types had to be close together (based on geographical information systems). The single fields were selected on-site based both on the farmer's statement regarding the management intensity and a brief inspection of the vegetation cover. Data concerning the mowing frequencies, manuring type and frequency, grazing type, and intensity were compiled as precisely as possible (Table S1). With these parameters, a simplified version of the land-use index (LUI) following the procedure proposed by Blüthgen [33] and Fischer [34] was performed. The only difference from this approach concerned the use of manuring frequency rather than the amount of nitrogen used to fertilize since precise data could not be obtained for each field. Further, site parameters were recorded locally (coordinates, elevation, slope, exposition), while chemical parameters of soil samples (such as SOM, pH, C/N, N, and P content) were analyzed by a specialized laboratory at the Laimburg Research Center (South Tyrol). Physical parameters (soil texture) were analyzed using the automated PARIO system (meter group [35]) (for soil parameters and site characteristics see Table S1; for correlation among these factors see Figure S1). The soil samples for the chemical and physical analyses were collected in each field by taking six subsamples in 5-10 cm depth (the medium depth of the macrofauna samples) at least 10 m apart from each other (three in autumn 2018 and three in spring 2019, at the same time of sampling of the soil fauna community). These six subsamples were mixed to receive one homogenized sample per field for further analyses.
Agronomy 2020, 10, x FOR PEER REVIEW 3 of 9 laboratory at the Laimburg Research Center (South Tyrol). Physical parameters (soil texture) were analyzed using the automated PARIO system (meter group [35]) (for soil parameters and site characteristics see Table S1; for correlation among these factors see Figure S1). The soil samples for the chemical and physical analyses were collected in each field by taking six subsamples in 5-10 cm depth (the medium depth of the macrofauna samples) at least 10 m apart from each other (three in autumn 2018 and three in spring 2019, at the same time of sampling of the soil fauna community). These six subsamples were mixed to receive one homogenized sample per field for further analyses. In total, 213 single soil core samples (20 × 20 × 15 cm square frame) were collected at three different time points (autumn 2018, spring 2019, and autumn 2019) with three sub-replicates per field and date. The three sub-replicate samples were collected within a minimum of 10 m distance from each other and in orchards within a distance of one to two meters from the tree base (e.g., for intensive apple orchards in the middle of the driveway). We added an additional sampling in autumn 2019 (except for one single field where the access was not granted by the farmer), since the very dry summer 2018 may have influenced the community composition in autumn 2018. The soil macroinvertebrates were extracted using a heat extractor (modified after [36]), quantified and identified to family level (except for Symphyla) using suitable identification keys [37][38][39][40][41] and following the taxonomic information of Fauna Europea Database [42]. For the ordination analysis, we treated adult and larval stages of beetles as different taxa since they differ functionally.
Sampling coverage was investigated for each single-season and field type using species accumulation curves based on the single soil core sample as sampling unites (INEXT [43]; Figure S2). The three sub-replicates were then averaged to one single sample per sampling time and used as such for the following analysis (see the full community in Table S2). Taxon diversity is represented by the Shannon-Wiener diversity index (computed with function diversity in VEGAN; [44]). For the statistical analysis of taxon richness and diversity, a linear mixed model with season and regions (nested in season) modeled as random factors was constructed. Scores were extracted with the function predict and used to generate boxplots. Further, a mixed-model constrained ordinations (function capscale in VEGAN; [44] using Bray-Curtis distances) were used to investigate the community composition changes and a PERMANOVA (function adonis in VEGAN; [44]) was used to verify if the field type centroids differ significantly. For these computations, the community matrix was reduced by excluding ants, for which the sampling methodology is not appropriate, they differ inlife history and occurrence. All variables were standardized. The season was modeled as a random factor to account for temporal autocorrelation and the region was modeled as a random factor nested within the season In total, 213 single soil core samples (20 × 20 × 15 cm square frame) were collected at three different time points (autumn 2018, spring 2019, and autumn 2019) with three sub-replicates per field and date. The three sub-replicate samples were collected within a minimum of 10 m distance from each other and in orchards within a distance of one to two meters from the tree base (e.g., for intensive apple orchards in the middle of the driveway). We added an additional sampling in autumn 2019 (except for one single field where the access was not granted by the farmer), since the very dry summer 2018 may have influenced the community composition in autumn 2018. The soil macro-invertebrates were extracted using a heat extractor (modified after [36]), quantified and identified to family level (except for Symphyla) using suitable identification keys [37][38][39][40][41] and following the taxonomic information of Fauna Europea Database [42]. For the ordination analysis, we treated adult and larval stages of beetles as different taxa since they differ functionally.
Sampling coverage was investigated for each single-season and field type using species accumulation curves based on the single soil core sample as sampling unites (iNext [43]; Figure S2). The three sub-replicates were then averaged to one single sample per sampling time and used as such for the following analysis (see the full community in Table S2). Taxon diversity is represented by the Shannon-Wiener diversity index (computed with function diversity in vegan; [44]). For the statistical analysis of taxon richness and diversity, a linear mixed model with season and regions (nested in season) modeled as random factors was constructed. Scores were extracted with the function predict and used to generate boxplots. Further, a mixed-model constrained ordinations (function capscale in vegan; [44] using Bray-Curtis distances) were used to investigate the community composition changes and a PERMANOVA (function adonis in vegan; [44]) was used to verify if the field type centroids differ significantly. For these computations, the community matrix was reduced by excluding ants, for which the sampling methodology is not appropriate, they differ inlife history and occurrence.
All variables were standardized. The season was modeled as a random factor to account for temporal autocorrelation and the region was modeled as a random factor nested within the season to account for spatial autocorrelation between the fields. We chose single factors to be tested in the constrained ordination, which (1) represented the site characteristics, the management intensity, the chemical and the physical soil properties, (2) did not correlate significantly with each other, and (3) had a priori expected effect on the soil community. We tested these factors with a permutation test upon the full model and only significant parameters were implemented in the final ordination. All computations and graphics were performed and generated using the statistical programming software R [45]. Graphics were produced using the package ggplot2 [46], the correlation plot was generated using the package corrplot [47].

Results and Discussion
Taxa richness (on average 18.5 ± 4.9 SD in extensive meadows; 17.1 ± 3.9 SD in intensive meadows; 19.1 ± 5.0 SD in extensive orchards; 14.1 ± 3.7 SD in intensive orchards) as well as diversity (on average 2.33 ± 0.27 SD in extensive meadows; 2.18 ± 0.17 SD in intensive meadows; 2.19 ± 0.27 SD in extensive orchards; 2.04 ± 0.25 SD in intensive orchards) was found to be significantly reduced under more intensive management in both meadows and orchards ( Figure 2; Table 1). This result was expected and has already been shown for both soil [16,17,20] and above ground organisms (e.g., [4,30,48]). Nevertheless, this confirms the severe impact management intensification has on overall soil biodiversity also for the systems investigated in this study. Specifically, for meadows, there is already broad evidence that management intensification alters the soil community and its related functional aspect (e.g., [22,48]). For orchards, on the other hand, the literature is more sparse and inconsistent regarding the supported biodiversity and intensification effects, especially for soil fauna communities (but see [49,50]). Species accumulation curves depict an overall good coverage for each season and further confirm the low scoring of intensive orchards in comparison to extensive managed fields. Intensive meadows score on an intermediate position, due to a higher taxa richness in spring ( Figure S2). to account for spatial autocorrelation between the fields. We chose single factors to be tested in the constrained ordination, which (1) represented the site characteristics, the management intensity, the chemical and the physical soil properties, (2) did not correlate significantly with each other, and (3) had a priori expected effect on the soil community. We tested these factors with a permutation test upon the full model and only significant parameters were implemented in the final ordination. All computations and graphics were performed and generated using the statistical programming software R [45]. Graphics were produced using the package GGPLOT2 [46], the correlation plot was generated using the package CORRPLOT [47].

Results and Discussion
Taxa richness (on average 18.5 ± 4.9 SD in extensive meadows; 17.1 ± 3.9 SD in intensive meadows; 19.1 ± 5.0 SD in extensive orchards; 14.1 ± 3.7 SD in intensive orchards) as well as diversity (on average 2.33 ± 0.27 SD in extensive meadows; 2.18 ± 0.17 SD in intensive meadows; 2.19 ± 0.27 SD in extensive orchards; 2.04 ± 0.25 SD in intensive orchards) was found to be significantly reduced under more intensive management in both meadows and orchards ( Figure 2; Table 1). This result was expected and has already been shown for both soil [16,17,20] and above ground organisms (e.g., [4,30,48]). Nevertheless, this confirms the severe impact management intensification has on overall soil biodiversity also for the systems investigated in this study. Specifically, for meadows, there is already broad evidence that management intensification alters the soil community and its related functional aspect (e.g., [22,48]). For orchards, on the other hand, the literature is more sparse and inconsistent regarding the supported biodiversity and intensification effects, especially for soil fauna communities (but see [49,50]). Species accumulation curves depict an overall good coverage for each season and further confirm the low scoring of intensive orchards in comparison to extensive managed fields. Intensive meadows score on an intermediate position, due to a higher taxa richness in spring ( Figure S2).  Boxplot of the partial effect of habitat (meadow and orchard) and management intensity (intensive and extensive) on taxa richness and diversity (Shannon diversity) (n = 18). The data was extracted from the linear mixed model accounting for the effect of region and season treated as random factors. Boxplot depicting the interquartile range (IQR between the 25% and the 75%) and the median, whiskers extending 1.5 * the IQR. Table 1. Likelihood ratio test of linear mixed-model for taxa richness and diversity (Shannon diversity). Habitat (meadow vs. orchard) and management intensity (extensive vs. intensive) were tested separately. The region was implemented as a random factor nested in the season to account for both spatial and temporal autocorrelation. The community composition was overall significantly different between the four field types (PERMANOVA; 999 permutations: F 3,70 = 2.169, p < 0.001). Further, a surprisingly similar community was found in the extensively managed fields (Figure 3). This was in part unexpected because of the significant role trees are known to exert on the soil community [28]. This similarity is probably due to both the similar land-use intensity (characterized by the factor land use index (LUI) after [33]) and the generally richer and more diverse soil community adapted to the traditional extensive practice. The rather clear separate plotting of the intensive fields both from the extensive ones and from each other further underlines the important effect of management intensification and type.

Dimension
Agronomy 2020, 10, x FOR PEER REVIEW 5 of 9 Table 1. Likelihood ratio test of linear mixed-model for taxa richness and diversity (Shannon diversity). Habitat (meadow vs. orchard) and management intensity (extensive vs. intensive) were tested separately. The region was implemented as a random factor nested in the season to account for both spatial and temporal autocorrelation. The community composition was overall significantly different between the four field types (PERMANOVA; 999 permutations: F3,70 = 2.169, p < 0.001). Further, a surprisingly similar community was found in the extensively managed fields (Figure 3). This was in part unexpected because of the significant role trees are known to exert on the soil community [28]. This similarity is probably due to both the similar land-use intensity (characterized by the factor land use index (LUI) after [33]) and the generally richer and more diverse soil community adapted to the traditional extensive practice. The rather clear separate plotting of the intensive fields both from the extensive ones and from each other further underlines the important effect of management intensification and type. Regarding site and soil parameters, we found expectable trends and correlations between the single factors ( Figure S1). Only non-correlating factors representing site characteristics, physical and Regarding site and soil parameters, we found expectable trends and correlations between the single factors ( Figure S1). Only non-correlating factors representing site characteristics, physical and chemical parameters as well as management details were selected to be tested in a constrained ordination. The resulting significant parameters in a permutation test where land-use index (LUI; Permutation test: F 1,57 = 2.93; p = 0.001) and soil nitrogen content (N; Permutation test: F 1,57 = 2.89; p = 0.001) and were implemented in the final ordination (Biplot scores for CAP1: LUI = 0.90; N = 0.31; and for CAP2 LUI = 0.42; N = −0.78; Figure 3). The first constrained axis depicts well the management intensity, separating both the intensely managed fields from the extensive ones. The second axis clearly separates the two intensive management types from each other, contrary to our expectations (Figure 3). We expected intensification to lead to a more simple and similar community as found in several other studies [17,20,49], but this was not the case between orchards and meadows. The differences are apparently directly connected to different management practices. While the recorded management parameters and the sum factor LUI (Table S1) was not able to fully explain these differences, we suspect that (1) the more frequent presence of heavy machinery in intensive orchards (for frequent mowing, working in the fields, picking the fruits) as well as (2) the application of additional chemicals (herbicides and insecticides routinely used in the region: [51]) might further significantly impact the soil community of intensive orchards. The more frequent presence of heavy machines is probably causing soil compaction, limiting the presence of soil macro-invertebrates, and the use of additional chemicals is also known to hamper the soil biodiversity in general [15]. The quantification of both these influences has not been considered in detail in this study, but potentially explains the severe differences between intensive meadows and orchards. Further, N is highly correlating with the SOM content (r = 0.94) and appears to be one factor that differentiates the two intensive forms of management from each other. A decrease in SOM is generally known to be a consequence of soil degradation and a potential threat for soil biodiversity [52], accordingly, richness and diversity of soil invertebrates were lower in intensive orchards where SOM also resulted in lower values. This appears to be true for the intensive orchards but not for the intensive meadows characterized by higher nitrogen and SOM contents even exceeding those of the extensive managed fields, probably due to the higher fertilization with cattle manure.

Dimension
Concluding, we found a more rich and diverse soil macro-invertebrate community in the extensive forms of the two investigated management types (viz. meadows and orchards). Further, we can state (1) that both extensive management types harbor a strikingly similar soil fauna community, (2) that intensification of land-use leads to a substantial change in these communities, and (3) that this change was profoundly influenced by the type of management.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4395/10/6/767/s1, Figure S1: Correlation plot between all parameters. Significant correlations (p < 0.05) where marked with a red frame, Figure S2: Species accumulation plots where computed for each season and management type to visualize the sample coverage and species diversity more in detail, Table S1: Table with the site characteristics, Table S2:  Table with taxon and site mean abundance and standard deviation for each single field. Data was standardized to 1 m 2 of sample and averaged over all nine single subplots sampled per field.

Conflicts of Interest:
The authors declare no conflicts of interest.