Study area
Golestan National Park (GNP) is situated in the north-eastern part of Iran (37°16'43''N 55°43'25''E-37°31'35''N 56°17'48''E) and is among the oldest and most diverse protected areas in the Middle East. It covers around 920 km2 of eastern Iranian Caspian forests with altitudes ranging from 450 to 2411 m above sea level (Fig. 5).
The average annual temperature varies from +11.8 °C to +18.8 °C. The climate is seasonal, marked by cold winters (January, mean temperature –0.8˚C) and warm summers (July, mean temperature 23.3˚C). Summers with high temperatures in the dry regions can cause extremely hot, dry conditions in the east, south and northeast and a humid climate in the western part of the region (34).Yearly precipitation ranges from 150 mm in the south-eastern part of the park to more than 1000 mm in the more central areas. The area receives 32.3%, 25.6%, 11.8%, and 30.3% of its annual rainfall during winter, spring, summer and fall, respectively.
Fig. 5 Location of Golestan National Park, highlighting the Hyrcanian forests (in green) in the western half and the surrounding steppes towards the east, north and south. Transitional vegetation zones occur in between and at high altitudes.
The GNP lies across the Euro-Siberian and Irano-Turanian phytogeographical regions (Hyrcanian and Khorassan–Kopet–Dagh provinces, respectively). The park contains a wide range of flora and fauna, which are unique in many aspects. It encompasses diverse vegetation entities including Hyrcanian mesophytic forests, shrublands, scrublands (occasionally mixed with C4-composed grasslands), Juniperus sp., woodlands, mountain steppes and meadows, Artemisia sp. steppes, and communities composed of halophilous plants (34). We divided these vegetation entities into two major habitat types where the target animal vectors are known to be present: Hyrcanian closed forests (hereafter, forests) and transitional scrub and Juniper woodlands (hereafter, prairie-forest ecotone). We therefore located our study plots within these two major habitat types, replicated twice; resulting in four sampling sites.
At the time of our study, there were about 257 (95% CI: 91-423) red deer (51), 150 roe deer, 6,000 (95% CI: 3,050-9,906) wild boars (52), and 60 brown bears in the park (Golestan Provincial Department of Environment, 2016).
In our study area, the roe deer typically prefer a closed-forest habitat, which overlaps only slightly with the habitats favoured by the two omnivorous species. Red deer partly share the closed-forest and the prairie-forest ecotones with the other three species. The wild boar inhabit a wide range of habitats and brown bears usually prefer mountainous forested sites with high densities of fleshy-fruited shrubs and trees.
Home range (HR) sizes have not been evaluated in GNP for the four target animal species, however, other studies can provide information on the gradient of HR size among species (e.g. 17 ha, 81 ha, 283 ha and 5000 ha, respectively for roe deer, red deer, wild boar and brown bear (14, 53)).
Dung collection and treatment
Dung samples were collected monthly from mid-May to November 2016 (spanning the seeding period) along random transects in the two habitat types. We could not find any faecal samples for brown bear or roe deer during certain months; therefore, samples were allocated to the following three seasons (spring, summer and autumn) to obtain at least two samples for each season-animal pair. We restricted dung collection to intact, fresh wet samples to limit post-dispersal modifications (7). We prevented contamination from seeds sticking to the surface of the samples by removing the lowermost layer of the collected dung (15). A small number of wild boar dung samples had been hollowed out by coprophagous beetles (5%) and were therefore discarded. The collected samples were air dried in paper bags for 10 days and weighed to the nearest 0.01 g. For red deer, wild boar and brown bear, we extracted two 20-g paired sub-samples from each faecal sample to investigate seedling emergence and plant establishment under greenhouse versus natural conditions. Because samples were lighter for roe deer (average weight of 5.67 ± 2.21 g; Table 4) than for the other three species, each individual roe deer dung sample was divided into two equally-sized sub-samples.
Table 4 Summary of the dispersed species assemblages by animal vector. G = greenhouse conditions; N = natural conditions. Sample size was similar for greenhouse and natural conditions.
Germination experiments
Both the greenhouse and the natural experiments had a randomised block design with seven blocks (corresponding to sampling month) and four treatments (corresponding to animal vector). Over a 15-month period, we recorded the germinated seedling species and then removed them. To obtain seedling species richness and abundance for each sampling season (spring, summer, autumn), we pooled the monthly data from May-June, July-August and October-November for each site and each animal vector.
Greenhouse germination conditions
The samples were stored at 3-5°C until field collections were completed (15), then each sample was carefully crushed to break apart the pellets. Each crushed sample was mixed with a similar volume of soil and sand and poured into pots (diameter 20 cm, depth 25 cm), making a layer approximately 1-2-cm thick. We then filled the pots with a 1:2:1 mixture (sand: soil: peat moss), which had previously been sterilised in an autoclave at 120 °C for 45 minutes (54).
The samples were then allowed to grow under natural daylight with daytime temperatures of around 25 °C in a greenhouse located at the Isfahan University of Technology (1616 m above sea level). The average minimum temperature was 18 °C. The samples were monitored every two days to maintain humidity. To prevent competition, we identified, counted, and removed the emerging seedlings as soon as possible. When no new seedlings emerged, the soil in each pot was thoroughly mixed and the experiment was continued for two more months to enable more deeply buried seeds to germinate (55). To control for possible seed bank or seed rain contamination in the greenhouse, 30 control pots (without faecal samples) containing a similar substrate were placed among the pots with dung samples and were maintained under the same conditions.
Seedlings were identified at species level whenever possible. Overall, 5.3% of the species could only be identified to the family level (seven Poaceae taxa) and 10% only to the genus level (13 taxa). Two seedlings died before they had grown sufficiently to enable identification. We did not observe any contaminating seedlings in the control pots.
Natural germination conditions
To examine germination success under natural conditions, a 10×20 m exclosure was established (located in the Tangrah region: 37°23'53.7"N latitude, 55°47'54.4"E longitude, 450 m above sea level) and the experiment was carried out within the fenced area to prevent disturbance from grazing animals. To prevent any seeds in the soil seed bank from contaminating the experimental soil, we inverted the soil by bringing a layer of soil from a depth of more than 35 cm up to the surface of the experimental site (21). Planting pots were filled with this deep soil and placed on the surface. The faecal samples were carefully crushed to break apart the pellets and were placed directly into each planting pot. To allow natural soil moisture into the planting pots and to improve rainwater drainage, the bottoms of the pots were removed. The faecal samples were not subjected to artificial cold treatment but were exposed to natural temperatures. Average annual rainfall was about 580 mm during the germination period. In order to control for air-borne seed input and soil seed bank content, seven control pots with soil only and no dung were positioned among the pots with dung samples, organized by month. Temperature and light were not controlled and no irrigation was applied during the experiment. The samples were completely exposed to natural climatic conditions. Emerged seedlings were identified to the species level whenever possible (11% could only be identified to the genus level). In the control pots, five species (Hesperis hyrcana, Lamium album, Torilis japonica, Nonea lutea, and Veronica persica) were recorded. These five species occurred more often in the control pots than in the non-control pots and were therefore eliminated from further analyses.
Data analysis
We built species accumulation curves with a Chao 2 estimator to assess how well we sampled the expected species richness (56). These species accumulation curves helped us compare the different animal vectors under both greenhouse and natural conditions.
Greenhouse data analysis
We used generalized linear mixed models (GLMM) to compare seedling abundance and the number of species among dispersal vectors, season and habitat type (dung sample as the statistical unit). Negative binomial and Poisson regression models were respectively assigned for seedling abundance and species richness (count response variables) including additional over-dispersion in the model. Animal species (4 species), sampling season (spring, summer and autumn) and habitat (forest and prairie-forest ecotones) were fixed factors, and site within habitat was a random effect. The log-transformed weight of each dung sample was taken as an offset to account for differing sample weights.
We used the lsmeans package to obtain the predicted values for each combination of factors. We then performed a Tukey post-hoc test for pairwise comparisons.
First, we fitted the full model to include all the main factors and Animal × Site and Animal × Season interactions (Additional file 5 & 6). The final model was obtained by backward stepwise selection. Best model selection was based on the lowest Akaike Information Criterion value (AIC).
Data analysis for natural versus greenhouse conditions
We used Poisson regression models to compare seedling abundance and species richness among animal species and between germination conditions (greenhouse vs. natural). First, we fitted the full model to include all the main factors and Animal species × Germination conditions interaction, with dung sample as the statistical unit. Best model selection was based on the lowest Akaike Information Criterion value (AIC). The lsmeans package and Tukey post-hoc test were used to obtain the predicted values for each combination of factors and for pairwise comparisons, respectively.
Pairwise comparisons between greenhouse and natural conditions for seedling abundance of common plant species were made with the nonparametric Mann-Whitney U test.
We used canonical correspondence analysis (CCA) to compare the composition of germinating plants (square root of seedling abundance for each plant species) among the animal vectors and between germination conditions. Due to the high number of plant species, plotting priority was given to most abundant plant species in the dung samples, following Hill’s N2 diversity index. We used Monte-Carlo permutation tests (n = 999 permutations) to test the significance (P < 0.05) of the variables and the axes of the CCA. We compared differing plant species composition among animal vectors, and between germination conditions by an analysis of similarities (ANOSIM), with a Bray-Curtis similarity index ranging from zero (complete species overlap) to one (no species in common). This index excludes double-zero comparisons and does not weight rare or abundant species (57).
We performed all statistical analyses with the R 3.6.2. software (R Foundation for Statistical Computing, Vienna, AT) in the vegan (58), venndiagram (59), lme4 (60), lsmeans (61) and MuMIn (62) libraries.