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 of 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. 4).
The climate is seasonal, marked by cold winters (January to March) and warm summers (July to September). The average annual precipitation is 400 mm and yearly precipitation ranges from 150 mm in the south-east part of the park to more than 1000 mm in more central areas. The most of the precipitation occurs from late autumn to early spring. Clearly the winter months with 32.3% of annual rainfall are the moistest months of the year. The area receives 25%, 11.8%, and 30.3% of annual rainfall during spring (April to June), summer, and fall (October to December), respectively. Minimum relative humidity of the region is 60% but increases up to 83% during summer. The average annual temperature varies from +11.8 °C to +18.8 °C. The summers with high temperature in the dry regions can cause really hot and dry conditions in the east, south and northeast and a humid climate in the west part of that region (30).
Fig. 4 Location of Golestan National Park, highlighting the Hyrcanian forests (in green) in the half western part and the surrounding steppes towards the east, north and south. Transitional vegetation zones occur in between and at high altitudes.
GNP lies across the Euro-Siberian and the Irano-Turanian phytogeographical regions (Hyrcanian and Khorassan–Kopet–Dagh provinces, respectively). GNP contains a wide range of flora and fauna, which are unique in many aspects; it is covered by diverse vegetation entities which include the Hyrcanian mesophytic forests, shrublands, scrublands (mixed with C4-composed grasslands sometimes), woodlandsof Juniperus sp., mountain steppes and meadows, Artemisia sp. steppes, and communities composed of halophilous plants (30). Mixed plant communities can be found in between these two phytogeographical regions. We incorporated these plant units into two major habitats types: Hyrcanian closed forests and transitional scrub and Juniper woodlands, where the studied animal vectors are known to be present. We therefore selected study plots within these two major vegetation types; each plot was replicated twice.
At the time of our study, there were about 257 (95% CI: 91-423) red deer (47), 150 roe deer, 6,000 (95% CI: 3,050-9,906) wild boars, and 60 brown bears in the park (Golestan Provincial Department of Environment, 2016).
In our study area, roe deer prefer a closed-forest habitat, which overlaps only slightly with those favoured by the two omnivorous species. Red deer partly share the closed and prairie-forest ecotones with the other three species. Wild boar inhabits a wide range of habitats and brown bears usually prefer mountainous forested sites with high densities of fleshy-fruited shrubs and trees.
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 sample of brown bear and 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 pairs. We restricted dung collection to intact, fresh and wet samples to limit interactions with the environment (11). We prevented contamination by seeds that stuck on the surface of the collected dung samples by removing the lowermost layer of dung samples (Picard et al., 2016). 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 both greenhouse and natural conditions. Because roe deer dung samples were lighter than those of the other three species (average weight of 5.67 ± 2.21 g; Table 1), each individual sample was divided into two equally-sized sub-samples.
Table 1 Summary of the dispersed species assemblages for all vectors combined and by vector. G = greenhouse conditions; N = natural conditions.
Germination experiments
Both the greenhouse and natural experiments had a randomised complete block design with 7 blocks (sampling month) and 4 treatments (animal vectors). Over a 15-month period, we recorded the germinated seedling species and subsequently removed them. To obtain seedling species richness and abundances in each sampling season (spring, summer, autumn), we pooled monthly data of 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 (Picard et al., 2016), and then separately crushed with care to break the pellets. They were mixed with a similar volume of soil and sand, then poured into pots (diameter 20 cm, depth 25 cm) making a layer of approximately 1-2-cm thick. We filled the pots with a 1:2:1 mixture (sand: soil: peat moss), previously sterilised in an autoclave at 120 °C for 45 minutes (10).
The samples were then allowed to grow under natural daylight with daytime temperatures of around 25 °C under greenhouse conditions located in the Isfahan university of technology (32°43′13.6″N latitude and 51°32′52.4″E longitude and at an altitude of 1616 m above sea level). The average minimum temperature fell to 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 months to enable the germination of more deeply buried seeds (48). To control for possible seed bank or seed rain contamination in the greenhouse, 30 control pots (without faecal samples) with a similar substrate were placed among pots with dung samples and kept equally.
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% 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 37°23'53.7"N latitude and 55°47'54.4"E longitude and at an altitude of 450 m above sea level) and the experiment was carried out within the fenced area to prevent disturbance by the grazing animals. We inverted the soil by bringing a deeper layer of soil (from a depth of more than 35 cm) up to the surface of the experimental site to prevent any seeds in the soil seed bank from contaminating the experimental soil (49). Planting pots were filled with the same deep soil and placed on this surface. The faecal samples were not subjected to artificial cold treatment. They were crushed carefully to break the pellets and were placed directly in each planting pot and were allowed to natural cold stratification. To allow the natural soil moisture into the planting pots and to improve rainwater drainage, the bottoms of the pots were removed. Average annual rainfall was about 580 mm during that period. In order to control for air-borne seed input and soil seed bank content, seven soil-only pots with no dung were similarly positioned for each month. Temperature and light were not controlled and no irrigation was applied during the experiment. The samples were completely exposed to natural climatic conditions. Seedlings were identified to the species level whenever possible (11% could only be identified to 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 and were therefore discarded from further analyses.
Data analysis
We built species accumulation curves using Chao 2 estimator to estimate the expected species richness and check if our observed seed-dispersed species richness matched what should be expected in the study area (50).
Greenhouse data analysis
A generalised linear mixed model (GLMM) was used to compare the abundance of seedlings and the number of species between dung samples of the studied vectors, while accounting for potential phenological and habitat variations in the assemblage of dispersed seeds. Negative binomial regression and Poisson models were respectively assigned for seedling abundance and species richness (count response variables) based on over-dispersion in the model. Animal species, sampling season (spring, summer, and autumn) and habitat (forest, prairie-forest ecotone) were fixed factors, and habitat repetition was considered a random effect. The log-transformed weight of each sample was taken as an offset to account for differing sample weights.
The compositions of the germinated species (square root transformed) were compared to the main factors studied (animal vector, sampling season and habitat) through canonical correspondence analysis (CCA). We used Monte-Carlo permutation tests (n = 999 permutations) to test the significance (P < 0.05) of the forward selected variables and the axes of the CCA.
We compared differing plant species composition among animal species, sampling season, and habitat by analysing similarities (ANOSIM procedure), which provided an R statistic ranging from zero (complete species overlap) to one (no species in common). We also used Bray-Curtis similarity indices as they exclude double-zero comparisons and do not weight rare or abundant species (51).
Data analysis for natural versus greenhouse conditions
Poisson regression models were used to test abundance of seedlings and species richness between the planted dung samples of the animal vectors, while accounting for germination conditions (greenhouse vs. natural). Animal species and germination conditions were the fixed effects and dung sub-sample was the random effect. Pairwise comparisons between seedlings abundance of plant species under both conditions were made with the nonparametric Mann-Whitney U test.
We performed all statistical analyses with the R 3.1 software (R Foundation for Statistical Computing, Vienna, AT) using the vegan (52), venndiagram (53), lme4 (54), lsmeans (55) and MuMIn (56) libraries.