The study was conducted in Scots pine forest located in the Siemianice Experimental Forest near Biadaszki village (51° 14.87′ N, 18° 06.35′ E, elevation 150 m), SW Poland, which belongs to the Poznań University of Life Sciences. The experimental stands were established in 1974, in the podsolic, sandy, and nutrient-poor soil, in vegetation typical of oligotrophic coniferous forests Leucobryo-Pinetum (Ceitel 1982). The mature Scots pine stand was clear-cut, stumps, and coarse roots were dug up and removed and deeply plowed to depths of 60–70 cm. In the spring of 1974, 2-year-old Scots pine seedlings were planted at nine different spacings (3 replicates/spacing; area of each plot was 0.11 ha, 27 m × 41 m; 3.07 ha in total with buffer zone), with initial stand densities from 2500 to 20,833 trees per ha. No cleanings and thinnings were done in the study area from the onset of the experiment. Stand densities changed only as a result of natural mortality (Kamczyc et al. 2019).
The climate of the study site is transitional between maritime and continental. Mean annual precipitation was 591 mm, while the mean annual temperature was 8.2 °C (weather data recorded 300 m from the field site from 1968 to 1997) (Reich et al. 2005; Hobbie et al. 2006). During the study which was conducted in 2009, average monthly temperatures ranged from − 6.6 °C in January 2010 to 19.4 °C in July 2009, while monthly precipitation sums ranged from 9.4 mm in April 2009 to 224.6 mm in July 2009 (Kamczyc et al. 2019). Climatic data analyzed in the present study come from the nearest weather station in Syców Forest District (51° 15′ 48.6000″ N, 17° 40′ 27.8400″ E, 189 m a.s.l.). Temperature and precipitation were recorded with an accuracy of ± 0.01. The mean monthly temperatures recorded in 2009 when the litterbag experiment was conducted were in the same range that was recorded for a longer period from 1999 to 2011. Mean monthly temperature for the longer period ranged from − 1.8 ± 3.0 °C in January to 18.9 ± 1.7 °C in July (Fig. 1A). Also, the mean annual temperature in 2009 was similar to the values recorded for the 1999–2011 period (Fig. 1B). Total monthly precipitation in 2009 was slightly higher in June, July, and October than values recorded for 1999–2011 (Fig. 1C). Total annual precipitation was slightly higher in 2009 than in the longer period (Fig. 1D).
We established the decomposition experiment within the three research stands (plots), covering ca. 35-year-old Scots pine stands, with an initial density of 11,111 trees ha−1. We chose only three plots to exclude the influence of initial stand density on ecosystem functioning, especially light availability and nutrient inputs (Jagodziński and Oleksyn 2009a, b, c). The stands were characterized by eight variables, i.e., mean (± SE) diameter at breast height (9.4 ± 0.28 cm), mean tree height (12.9 ± 0.15 m), stand basal area (37.4 ± 0.90 m2 ha−1), stand density (4908 ± 399 trees ha−1), litter biomass of the organic horizon (30.45 ± 2.10 Mg ha−1), annual litterfall (2.89 ± 0.16 Mg ha−1) and pHH2O of the Ol horizon (4.71 ± 0.09), and pHH2O of the Of horizon (3.91 ± 0.06) (Kamczyc et al. 2019).
Litterbag experiment design
The litter of 11 tree species for the litterbag experiment was collected from plots of the common garden experiment located ca. 500 m from the Scots pine forest. The litter included seven broadleaved and four coniferous species. Litter traits were characterized in detail by Hobbie et al. (2006). The broadleaved species were as follows: Norway maple (Acer platanoides L.), sycamore maple (A. pseudoplatanus L.), European hornbeam (Carpinus betulus L.), European beech (Fagus sylvatica L.), small-leaved lime (Tilia cordata Mill.), English oak (Quercus robur L.), and invasive Northern red oak (Q. rubra L.). The coniferous species were as follows: silver fir (Abies alba Mill.), European larch (Larix decidua Mill.), Norway spruce (Picea abies (L.) H. Karst.), and Scots pine (Pinus sylvestris L.). The litter for the experiment was oven-dried (at 65 °C) to constant mass. This also eliminated all living organisms, which could influence our inference about litter colonization. We placed homogenous litter of each tree species in nylon bags (mesh size of 1 mm) to allow free access of living animals to migrate into the sample with organic matter. The litterbags (size of 18 × 18 cm) were randomly distributed within study plots. The experiment started on 14 October 2008. The litterbags overwintered and were sampled in equal numbers (99 litterbags), seven times at monthly intervals in the vegetation season 2009, on 15.04, 18.05, 18.06, 14.07, 17.08, 16.09, and 19.10. In total, we collected 693 litterbags (11 tree species × 3 plots × 3 replications per plot × 7 sampling periods).
After extraction of mites from litterbags (see Section 2.3), the collected samples were dried to a constant weight at 65 °C, after which any additional material such as other vegetation, insects, sand, etc., was removed from each sample manually using tweezers. Samples were then weighed with an accuracy of 0.001 g to determine leaf mass loss for each litterbag. Litter mass losses (%) used in this paper are the average values of real mass loss of nine samples obtained for each leaf litter type (species) at each collection date.
We used litter decomposition changes over the experiment as the background for the study of mite assemblages. At the beginning of the litterbag experiment (April), the lowest litter mass loss was found for A. alba (11.2%), F. sylvatica (14.0%), and P. sylvestris (15.2%), whereas the highest for A. platanoides (22.4%), Q. rubra (21.5%), and A. pseudoplatanus (21.3%). At the end of the experiment (October), the lowest litter loss was in F. sylvatica (19.3%), then in A. alba (19.4%), P. abies (20.2%), T. cordata (23.3%), Q. robur (23.5%), P. sylvestris (25.1%), C. betulus (25.3%), L. decidua (25.8%), A. pseudoplatanus (30.6%), Q. rubra (32.6%), and the highest in A. platanoides (35.8%). This means that the litter decomposition rate of A. platanoides is two times higher than of F. sylvatica (Fig. 2).
Mite extraction and identification
Litterbags were carefully placed in a portable cooler and transported to the laboratory. Mites were extracted from samples in Tullgren funnels, according to the recommendations for studies concerning organic substrata such as those in the Pinus sylvestris forest floors in this study (Crossley and Blair 1991; Edwards 1991). We set samples within Tullgren funnels (with 40 W bulbs) as quickly as possible (5 h after sampling) and the extraction procedure lasted 7 days until the samples were dry. Then, we selected mesostigmatid mites from the samples, and we identified them to the species level and developmental stages using taxonomical keys of Karg (1993), Ghilarov and Bregetova (1977), and Micherdziński (1969). Mite species nomenclature follows Błoszyk (2008) and Skorupski (2008).
All statistical analyses were conducted using R software (R Core Team 2019). To avoid pseudoreplications, we pooled all mite records coming from the same plots, sampling date, and litter types to allow conclusions about diversity within sample plots. This produced three replications (pooled values) of each study date and litter type which gave 231 (3 replications × 3 plots × 7 sampling periods) data points for the analysis. We evaluated species richness as number of taxa recorded within the study plot and litter type, we accounted for species alpha diversity using the Shannon index, and we calculated abundance per sample. Data were presented as mean values followed by the standard error (SE).
To assess the impact of weather conditions (mean temperature and precipitation sums of sampling month and the month before) and litter quality (expressed by its identity, which can be linked with measured litter traits and decomposition constants), we used generalized linear mixed models (GLMM). We assumed Poisson distributions for mite abundance and species richness and a normal distribution for Shannon’s index. Abundance was not recalculated per sample mass, as applied Poisson distributions assume integer values. In the models, we accounted for random effects connected with sample dependencies (study plot and collection date), to exclude plot-specific and date-specific factors, which could bias the inference. Models were developed using the lme4 package (Bates et al. 2015) while the statistical significance of variables was calculated using z-values implemented in the lmerTest package (Kuznetsova et al. 2017). For all GLMMs, we evaluated the parsimony of models using Akaike’s Information Criterion (AIC). We also provided AIC0 – AIC of models with intercept and random effects only. To evaluate differences between litter origin and collection dates in the models we used Tukey posteriori tests. We also calculated marginal (R2m) and conditional (R2c) coefficients of determination, expressing amount of variance explained by fixed effects only and by both fixed and random effects jointly, respectively (Nakagawa and Schielzeth 2013). These coefficients were calculated using the MuMIn package (Bartoń 2017). Due to high collinearity, we did not include decomposition constant and species identity together in analyses, but we tested variants with each of them separately, to avoid variance inflation, reported by high values of variance inflation factors.
To assess the importance of temperature and precipitation in shaping mite species communities, we used Canonical Correspondence Analysis (CCA), implemented in the vegan package (Oksanen et al. 2018). CCA is the method of constrained ordination of the multivariate data (mite species abundances). In contrast to unconstrained ordination, CCA also allows to evaluate the importance of environmental variables in ordered sample coordinates within reduced analytical space. We tested the importance of temperature and precipitation using permutation analysis of variance (PERMANOVA), also implemented in the vegan package (Oksanen et al. 2018). Before analyses, we transformed species abundances using Hellinger’s square root transformation (Legendre and Gallagher 2001), and we downweighted rare species (i.e., those with total abundance < 5). The selection of variables used to constrain the ordination (environmental variables) was based on forward selection and variable elimination to decrease AIC.
We described relationships between litter origin and mite species using bipartite network metrics (bipartite package in R), assuming litter species as the lower-level group and mite species as the higher-level group in the data processing. We also calculated network metrics—connectiveness and coefficient of network specialization H2’. Connectiveness is a proportion of links between species related to all possible links in the network. Network specialization H2’ is an index describing the level of so-called complementarity specialization of the whole network. The H2’ index describes how much observed interactions deviate from those that would be expected given the species marginal totals. Therefore, higher values of H2’ indicate higher selectiveness and specialization of species (Blüthgen et al. 2006; Dormann et al. 2009). At the species level, we determined the number and proportion of litter (among 11 litter types) where a particular mite species was recorded. We also calculated species diversity in the litter, i.e., diversity of litter species where a particular species was recorded, using the Shannon index, where species abundance on a particular litter type was used as a weight (Dormann 2011). To assess the level of specialization, we determined specialization index d’, derived from Kulback-Leibler distance, which indicates the strength of a species deviation from a random sampling of all available taxa from the lower-level group (in our case—from litter origin species). Specialization index d’ ranges from 0 to 1, and higher values indicate a higher level of specialization (Blüthgen et al. 2006; Dormann 2011). All network analyses were conducted using a bipartite package (Dormann et al. 2008): for abundances, species richness and Shannon diversity, litter decomposition and climatic data for the analyses—check complete dataset (Kamczyc et al. 2022).