Simplicity or complexity? Important aspects of high nature value grassland management in nature conservation

The study area is located in the Turján region of the Great Hungarian Plain (Central Hungary) (coordinates: 47.04023°N; 19.15289°E). This region belongs to the continental forest-steppe zone, with a slight sub-Mediterranean climatic effect. Annual mean temperature is 10.4 °C (Tölgyesi et al. 2016), while annual average rainfall is 520 mm. At the regional level (Turján region), sandy and loamy soils are dominant (Karátson 2002), while in the study area the sandy soils are dominant (Vadász et al. 2016). Potential vegetation of the study area is the Pannonian sandy forest steppe community, but two thirds of the area is covered by the rich mosaic of primary and secondary, semi-natural grasslands grazed by cattle (Hungarian Grey Cattle and Charolais breeds). Our aim was to study spatially scattered primary, ancient grassland sites (or, on the rare occasion, oldfields, abandoned very long ago). This sandy, xero-mesic meadow-steppe habitat is one the dominant grassland types in the study area, being a transitional, species-rich ecotone between Molinia fen meadows (Molinion caeruleae, Natura 2000 code: 6410) and Pannonic sand steppes (Galio veri-Holoschoenetum vulgaris, Natura 2000 code: 6260). Although extensively managed xero-mesic meadow steppes are one of the most species-rich habitat types in Hungary, management-vegetation interactions in this habitat type are understudied from the perspective of nature conservation (Vadász et al. 2016). Adventive and invasive species are rare in the managed grasslands. Dominant, frequent and characterisitic species of this habitat are Chrysopogon gryllus (Poaceae), Molinia caerulea (Poaceae), Serratula tinctoria (Asteraceae), Sanguisorba officinalis (Rosaceae) and Betonica officinalis (Lamiaceae), while many protected species, such as Ophrys scolopax (Orchidaceae) and Iris spuria (Iridaceae), are also present.

This study considers three management factors: (1) type of management (T), (2) intensity of management (I), and (3) spatio-temporal complexity of management (C). Management types considered are mowing, grazing and the combination of them; furthermore, three categories of herbage removal intensity (low, medium and high intensity) and three spatio-temporal complexity categories (low, medium and high complexity) were investigated. The sample arrangement, in addition to the mentioned categories, was determined by following the method of Vadász et al. (2016) (Table 1). We aimed to sample consantly managed meadow steppe habitats in the study area. Every sampled grassland site exceeded at least 5 ha in size, in order to exclude sources of variance that emerge from diversity of management and neighbouring or edge effects. Investigations were undertaken in the summer of 2015, on 17 grassland sites, which were ordered by the three management factors (T, I, C) (for more details, see Table 1, Appendix Table 12). 10 plots were designated on 15 grassland sites and 11 were designated on two sites, because of the greater extent of the meadow-steppe patches to ensure the proper sampling and interpretation required. Thus, in the end, data gathered in N = 172 plots (1 m × 1 m each) were evaluated (for more details, see Appendix Fig. 3). Because of the indistinct shape and variable extent of the sampling sites, it was not possible to use a previously determined random protocol in the arrangement of the plots. Therefore we applied linear sampling, where we placed each plot per site along transects fitted to special forms of every sites. The location of the first plot of the transect in a given site was determined randomly, while further plots were designated at a minimum of four meters along the transects (for more details, see Appendix Fig. 3). Recorded data at every plot included the name and visually estimated percentage cover of each vascular plant species, and the coordinates of the plots were registered using GPS.

In this paper, we refer to management as all human activities attempting to serve nature conservation or economic purposes in grasslands. Herbage removal intensity and spatio-temporal complexity levels were determined by local farmers and by the local national park ranger with expertise in this area (Csaba Vadász and András Máté—authors of the paper, Kiskunság National Park) (see also Vadász et al. 2016). All the studied management factors (type of management (T), herbage removal intensity (I) and spatio-temporal complexity (C) of management), as explanatory variables, have different technical parameters (e.g., timing, frequency and level of intensity) (Table 1), which are referred to as ‘management details’ and ‘levels’ below in the text.

We categorized all observed plant species into eight predefined PFTs, based on habitat requirement, life-history traits, and growth form (Table 2, Borhidi 1995). We used species number and total cover of each PFT, as well as total species number, Shannon diversity and Simpson diversity, as response variables (Table 2, Borhidi 1995). In our analyses, we used percentage cover of all PFTs as like species numbers of them. In case of life history trait classifications, we also used the list of protected species in Hungary (see http#1). In the life history trait category selection we used only those PFTs, which had hypothetically strongest explanatory and indication power of naturalness and conservation value of the sampled meadow steppe communities (see Borhidi 1996; Hargitai 1940).

Linear mixed models, namely LMER and GLMER were used to evaluate the explanatory power and informativity of the three, studied management factors and their combination on response variables. First, to investigate main effects of management factors (type, intensity, and spatio-temporal complexity) vegetation, we ran three separate models for each response variable: as a function of management type (1: T), as a function of herbage removal intensity (2: I), and as a function of spatio-temporal complexity (3: C). Site was included as a random factor to account for non-independence of plots. Second, to investigate combined effect of management factors, we ran four additional models for each response variable: 4: T + I; 5: T + C; 6: I + C; and 7: T + I + C. Normality of residuals were checked with the Shapiro–Wilk normality test. In case of non-normally distributed residuals, GLMER models were applied with Gamma family. In some cases, where convergence problem occurred, “log” link was applied in models. After simple model evaluations (T, I and C), Tukey post hoc tests were applied with “glht” function to evaluate significant differences between levels of different management factors. Bonferroni adjustment method was applied in the post hoc tests to counteract the problem of multiple comparisons. In the models, AICc (corrected Akaike information criterion) values were used to determine model parsimony. Application of AICc values was important and adequate for estimating the relatively few real combinations of the explanatory variables (Burnham and Anderson 2002). Candidate models with strong explanatory power (ΔAICc ≤ 4) were used for model averaging and model selection. Unadjusted beta R² values (with ‘r2beta’ function) of models were used to evaluate the amount of explained variance and fit, following the recommendations of Posada and Buckley (2004). Analyses were run with R 3.5.1 (R Core Team 2018) software environment and by ‘MuMIn’, ‘lme4’, ‘goft’, ‘r2glmm’, ‘multcomp’and the ‘vegan’ packages.

Management type (T) as fixed factor, didn’t affected strongly plant diversity and most of the PFT covers and numbers. Only generalist species number and disturbance tolerant species cover had relatively strong relationship with T (Tables 4, 5), but significant differences between different management types wasn’t experienced (Appendix Table 6). Herbage removal intensity (I) had much more stronger effects on diversity, mainly on Shannon diversity and species number and had significant effect on grass, generalist and forbs species coverages (Tables 3, 4). Parallelly, spatio-temporal complexity of management (C) had the strongest relationships with diversity indices and species number. C had powerful impact on generalist species cover and number as like forb, specialist and disturbance tolerant species numbers (Tables 4, 5). In several cases, simple, univariate models were most parsimonious and had best fits. In case of Shannon and Simpson diversity, simple C model was the most parsimonious. Forb, generalist and disturbance tolerant species coverage had most parsimonious models in relation with single management factors. Similarly, simple models were most parsimonious in case of grass, disturbance tolerant, natural competitor, ruderal competitor, specialist and protected species numbers. On the other hand, in case of species number, grass, natural and ruderal competitor and protected species cover, complex models were more parsimonious (Tables 3, 4, 5). Regarding the aforementioned dependent variables, R² values of complex models with the best fit increased by ΔR² = 0.006–0.131 when compared to the best fitting single-variable models (Tables 3, 4).

For five of the 19 dependent variables, the most complex model (T + I + C) was the best: species number, and the cover of natural competitors, ruderal competitors, and protected species as like generalist species number (Tables 3, 4, 5). For the dependent variables forb species number and grass species cover, explanatory power increased by applying a combination of T, I and/or C (Tables 4, 5).

In our study, none of the studied response variables differed significantly by effect of mowing, grazing or combined management based on LMER and GLMER post hoc tests. On the other hand, different levels of I had a strongly negative differential effect on the total number of species, as like on forb species cover. Parallelly, factor I had negative effect on forb, generalist disturbance tolerant species numbers. In contrast, high level of I had strong, positive effect on grass and ruderal competitor species coverages (Appendix Tables 7, 10).

High or medium level of C had strong, positive effect on grassland diversity (Fig. 1, Appendix Tables 8, 11) and on the generalist PFT group coverage (Appendix Table 8). Parallelly, higher levels of factor C had strong, positive effects on forb, disturbance tolerant, generalist and specialist species numbers. In contrast, C had strong, negative effect on grass species cover (Fig. 1, Appendix Table 11).

Increasing or decreasing level of I and C were the two, most powerfully forming management factors with mostly opposite effects on plant diversity and PFTs. This phenomenon in details is pictured in Fig. 2 below.

Among the considered management factors, the low and/or medium herbage removal intensity, in line with high spatio-temporal complexity of management resulted in a significantly higher plant species number on the studied grasslands (Appendix Tables 7, 8).

Different management intensity levels affect plant diversity patterns and PFTs’ relations (de Bello et al. 2006; Diaz and Cabido 1997; Imrichova and Vrahnakis 2005). Low or medium herbage removal intensity gives room for (1) greater plant heterogeneity (e.g., different reproduction phases or variable height on a management-unit scale) (Austrheim and Eriksson 2001), and (2) increased diversity of plant species. In contrast, high herbage removal intensity can lead to functional and compositional homogenisation, and reduces diversity of grasslands especially in drier grasslands with relatively lower productivity (as our study area as well) (Fedrigo et al. 2018). In our study, high herbage removal intensity alone negatively affected the species number and cover of forbs, while positively affected grass species’ cover (Fig. 1, Tables 3, 4 and Appendix Table 7). For example, clonal grasses with great competitive power (e.g. Calamagrostis epigeios and Botriochloa ischaemum) are favoured by high herbage removal intensity (see Öckinger et al. 2006; Malatinszky 2016; Ruprecht et al. 2010), are able to completely dominate grassland stands (Szentes et al. 2012) and parallelly, reducing the species diversity and leading to a degraded state of grassland composition (Házi et al. 2011; Szentes et al. 2012). Lower level of herbage removal intensity has a positive effect on the proportion between grass and forb species coverage supporting the preservation of the optimal plant composition and diversity (Appendix Tables 7, 10, Fig. 1).

Spatio-temporal complexity of management, namely the application of diverse combinations of mowing and grazing with changing herbage removal intensities between years and within a year can influence the physiognomy, phenology and generative relationships of the grassland vegetation (Kelemen et al. 2017; Szépligeti et al. 2018), and positively affect the diversity and PFTs’ dominance relations to a considerable extent (e.g. forb, generalist and specialist species numbers, see Table 5 and Appendix Table 11). This variance in management affects the inter-specific competition by creating temporal reproductive niches (niche partitioning and segregation) (Catorci et al. 2014; Marini et al. 2007). A few years with more intensive herbage removal intensity in a spatio-temporally complex grassland management can create a reduced height of the vegetation, resulting in a sunnier and drier microclimate (benefits light-preferring and drought tolerant species) (Wan et al. 2002). Periods with less intensive management (e.g., one mowing per year) favours species that prefer a milder microclimate, are taller and grow faster (Steen 1980). Thus, higher spatio-temporal complexity of grassland management (e.g., leaving unmown refuge strips with location varying from year to year; or grazing with changing regimes and starting time year by year) coupled with lower herbage removal intensity provides a great opportunity for more plant species with various competitive strategies to coexist (see Tables 3 and 4, Appendix Tables 7, 8, Fig. 1) (Austrheim and Eriksson 2001; Duru et al. 2005). Based on other studies, these abovementioned, positive effects also can be helped by the plant structure providing effect of extensive grazing (Kelemen et al. 2017). In summary, herbage removal intensity and spatio-temporal complexity had stronger effects on grassland diversity and on PFT groups than management type (Figs. 1, 2), furthermore their importance is greater in effective conservation practice and planning than management types only.

Studied management factors and their parameters (different levels of intensity and spatio-temporal complexity) can be combined in several ways (theoretically 3³ = 27 ways). At the studied sites 7 combinations were realized (for more detailed information, see Table 1, Appendix Fig. 3, Appendix Table 12). The combination of low herbage removal intensity and high spatio-temporal complexity resulted in the highest number of species by far, positively affecting the cover of forbs and number of specialist species. This combination has the greatest relevance in conservation management practice. Furthermore, based on model selection, a definite hierarchy can be established among the three management factors based on their overall conservation relevance: (1) spatio-temporal complexity of management had the strongest explanatory power on most of the response variables (especially on plant diversity, generalist species cover and species number of forbs and specialists); (2) herbage removal intensity had less, but similarly great importance in conservation, and (3) management type was the third factor in the hierarchical order of the management factors’ importance (Table 3). Based on this relevance, grassland conservation planning should take different levels of spatio-temporal complexity of grassland management into consideration to develop more comprehensive management systems. We recommend to investigate the effects of spatio-temporal complexity, herbage removal intensity of management and different management types in different grassland habitats to develop more effective nature conservation grassland management. Considering the significance of the relevance of management factors, a more comprehensive and detailed conservation planning and management would be achieved and oversimplifying models and approaches would be avoided in conservation studies and practice.