To study climate driven effects across an elevational gradient, we selected a continuous and non-fragmented forest area of Mt. Krim (45°58'N, 14°25'E), 10 km south of Ljubljana (central Slovenia), which is part of a continuous montane forest range, extending from Slovenia across the western Balkan Peninsula to Serbia. The area is 140 km², 77% of which is covered with forest and 20% of which is not forested, the remainder being urban areas (i.e. settlements) which are situated only in the lowlands. Mt. Krim is a medium altitudinal mountain in the North Dinaric Alps ranging from 290 to 1108 m a.s.l. The slopes are covered predominantly with mixed temperate forest of Omphalodo-Fagetum s. lat. in which beech (Fagus sylvatica) is the dominant tree species (36%). Other common tree species are silver fir (Abies alba) and Norway spruce (Picea alba). Most of the forest is in an old growth phase, with trees whose trunk diameters are more than 30 cm at breast height. Clearings are small and dispersed, mostly around the settlements (Vrezec 2003; Vrezec and Tome 2004). During the study, Mt Krim was studied on three altitudinal belts (low (300–550 m), middle (550–800 m) and high (800–1100 m). The sampling sites in altitudinal belts were chosen to avoid larger forest clearings and in such a way that they did not differ in type, age, tree species composition, including tree density/abundance, light and other environmental conditions and forest structure.

Fieldwork was conducted in spring, summer and autumn of 2010. During this survey, the altitudinal distribution of three species groups was investigated: carabid beetles, hoverflies and passerine birds. These groups were investigated in the three altitudinal belts.

The carabid beetles were sampled with pitfall traps using vinegar as an attractant (Vrezec and Kapla 2007). A line of five traps at 20 m spacing was set in each altitudinal belt. The traps were set every month for five to seven days in April (19.4–26.4), May (15.5–21.5), June (18.6–23.6), July (21.7–28.7), August/September (28.8–4.9) and October (1.10–8.10). All carabid beetles were collected and identified in the laboratory using determination keys (Mueller 1931; Müller-Motzfeld 2006).

The hoverfly assemblage was assessed using transect counts and malaise traps (Leather 2005). Ten transects of 100 m were established in each altitudinal belt. Transect counts were conducted at the beginning of June, July and August, only on sunny days. Malaise traps were set from 15 April to 31 October in each altitudinal belt. Each trap was filled with alcohol and emptied every 13 to 16 days, 13 times. Although the traps were not replicated per altitudinal belt, it gives an indication of the phenological dynamics. The transects were positioned in areas with mixed deciduous/coniferous forest. All specimens were collected and identified in the laboratory, according to Van Veen (2004).

The passerine birds were counted at 16 points (Bibby et al. 2000). Four points were established in the low, six in the middle and six in the high altitudinal belt. Two counts were conducted: one in March/April and one in April/May. The counts were carried out during the morning hours between 5:00 a.m. and 10:00 a.m. on clear and sunny days. Every count session on the point took 5 minutes. Only forest birds which showed territorial singing behaviour were taken into account (Bibby et al. 2000).

The temperature was measured with a temperature logger (LogTag Trix–8 Temperature Recorder, accuracy ± 0.5 °C). In each altitudinal belt, a logger was placed on the tree. The temperature was measured every six hours during the sample period.

Species assemblage, species richness and abundance per group per altitudinal belt were calculated. Data on carabid beetles and hoverflies were repeated over time and pooled for each altitudinal belt per transect for the hoverflies or trap for the carabid beetles. The relative number of animals/species per day or per 15 days was calculated for the carabid beetles and the hoverflies, respectively. In bird surveys, the maximal abundance from two counts was taken into consideration and expressed as number of territorial birds per point. A permutational MANOVA (PerMANOVA) with the Jaccard dissimilarity index was used to test the differences in species assemblages between the altitudinal belts using only the transect data (Anderson 2001). As a post hoc test, every belt was compared with the other belts using the PerMANOVA and then corrected with the Holm correction (Holm 1979). The results of the PerMANOVA were visualised with NMDS. The species replacement (Simpson similarity index) and the dissimilarity derived from nestedness in the Jaccard dissimilarity index were calculated (Baselga 2012). A rarefaction curve was prepared for every group and altitudinal belt to check whether the sampling effort was sufficient. The species-sample-based R/E curve and sample completeness curve were prepared. The differences in species richness and abundance between altitudinal belts were analysed with a Generalised Linear Model (GLM) using Poisson error distribution (McCullagh and Nelder 1989) using only the data of the transects.

The following traits were investigated: food type (hoverflies: predator, microphagous and phytophagous; birds: seeds and invertebrates), wing length and the body length of the animal. The wing length indicated the dispersal possibility as large animals with large wings having higher dispersal possibility (Delettre 1988). For the same reason, the body length was taken into account, as large animals (also non-flying animals) have a greater dispersal potential than small animals. The data on the different traits per species were extracted from the literature (Snow and Perrins 1998; Müller-Motzfeld 2006; Reemer et al. 2009; Speight 2017) or measured with the animals which were caught. The dissimilarity in traits between the different altitudinal belts was calculated with Fourth Corner Analysis using the Jaccard index (Legendre et al. 1997; Dray and Legendre 2008). For these simulations, model 1 was used, which permutates the values within each column (species) (Dray and Legendre 2008). Each group was investigated separately.

For the seasonal dynamics, the repeated data-sets of the carabid beetles and hoverflies (only malaise trap data) were used. We were only interested in the seasonal dynamics and not differences in abundance between altitudes. Therefore, the species richness and relative abundance data per period were transformed into a percentage of the total number of species/individuals per altitudinal belt.

All analyses were done with the statistics programme R (R Development Core Team 2011). For the PerMANOVA, the “adonis” function in the “vegan” package (Oksanen et al. 2013) was used. For the Fourth Corner Analysis, the “fourthcorner” function of the “ade4” package (Dray and Dufour 2007) and for the rarefaction the package “iNEXT” (Hsieh et al. 2019) were used.

The first question raised was whether the discovered altitudinal patterns are caused by factors other than temperature (McCain 2009). The patterns were not due to the area (Rahbek 1997) and sampling effect (Colwell and Coddington 1994) because the study design was standardised in sampling efforts. Static evolutionary models predict that, at lower altitudes, there is more migration, which also results in higher diversity, while at higher altitudes, there is less dispersion and, thus, a higher extinction rate and lower diversity (Lomolino 2001). Due to the geographical and landscape characteristics of our study area, we were able to rule out several parameters that could contribute to the differentiation of the studied assemblages. First, the area is part of a large continuous non-fragmented forest area in the Northern Dinaric Alps in southern Slovenia, which extends further south to Croatia, so dispersion of forest species is not limited by habitat fragmentation and the extinction risk is therefore low. Secondly, the species found in the assemblages are not isolated relict populations and are distributed widely in the Dinaric Alps and beyond. Thirdly, due to the relatively short distances between sampling sites in the different altitudinal belts (< 3 km), all sites experienced similar weather conditions, especially with respect to precipitation. The patterns are in accordance with the gradient proposed by McCain (2009) and, therefore, it can be assumed that the gradient found here is driven mainly by altitudinal temperature differences.

The contrasting seasonal activity, richness and assemblage patterns observed during this study could be due to the different traits of the investigated species groups. First, the strong difference between the birds and the insect groups could be explained by differences in thermoregulation (Miller and Harley 2009). Insects are ectotherms and are dependent on the environmental temperature for development (Bale et al. 2002). In the case of the carabid beetles, there was a decrease in the number of species and abundance across a temperature gradient. In addition, insects are strongly partitioned in niches on the basis of temperature (Boggs and Inouye 2012). This means that some species are adapted to colder, higher elevation areas and some to warmer, lower elevation areas (Žagar et al. 2018). This is reflected in the strong species turnover observed in both carabid beetles and hoverflies. Birds, on the other hand, did not show any differences in species assemblages between the different altitudinal belts. Endotherms are better adapted to different climates and only exhibit a turnover on a larger scale (Buckley and Jetz 2008). In addition, birds are extremely mobile species and almost half of the recorded species are migratory or at least partly migratory and do not overwinter at their breeding sites. The observed decrease in abundance can only be an indirect effect of temperature. Lepidoptera larvae are an important food source for the observed passerine species (Barbaro and Battisti 2011). As lepidopteran larvae are influenced by temperature, food availability might be lower at higher altitudes, which in turn affects abundances (Randall 1982; Bears et al. 2009), although this is not the case for some other insect groups (de Groot and Kogoj 2015, this study).

The dispersal ability or mobility of a species is another aspect which could result in differences in species assemblages (Hubbell 2001). This could explain the difference in assemblage patterns between the studied insect groups (the brachypterous carabid beetles and macropterous hoverflies). The non-flying carabids had a distinct assemblage for each altitude. Forest carabid beetles in our study were all brachypterous and therefore they disperse on the ground within smaller distances and have consequently smaller ranges (Kotze and O’Hara 2003; Jopp and Reuter 2005; Pizzolotto et al. 2016). Hoverflies, on the other hand, exhibited an overlap in altitudinal ranges, with the assemblages of the low and the high altitudes overlapping slightly and the assemblages of the mid altitudinal belt overlapping to a large extent with those of both the low and the high altitudinal belts. Hoverflies can fly several kilometres and some species are even migratory (Rotheray and Gilbert 2011). Therefore, they can spread over large areas relatively faster than flightless carabids. Given this, we suggest that altitude has a greater effect on the turnover of less mobile than mobile species.

On the other hand, the abundance and species richness of the hoverflies increased with altitude in forested areas. One of the reasons could be that there is competition with hymenopteran species for food resources. It was observed that, towards the north, a higher percentage of plants are pollinated by flies, because bees have their optimum at higher temperatures (Rotheray and Gilbert 2011). The abundance of Diptera species (e.g. Cheilosia fasciata), was found to be even negatively correlated with temperature (de Groot and Kogoj 2015), which might also be the case in other syrphid or other dipteran species (Miličić et al. 2018). Competitive relationship between hymenopteran and dipteran pollinators and even predators should, however, be explored more.

As predicted, the carabid beetles and hoverflies exhibited different patterns of activity over the season. The phenology of hoverflies was strongly correlated with the weather. The earlier flying species showed a delay in flying with increasing altitude, which was also observed with butterflies (Illán et al. 2012). Flying in hoverflies takes a large amount of energy and mainly occurs on sunny days, when they fly in large numbers (Gilbert 1984; Gilbert 1985). The altitudinal difference in temperature was much higher in spring than in summer. This would, therefore, also influence the abundance and the species richness more strongly for hoverflies. The flightless carabid beetles showed a more stable pattern, where there was no altitudinal difference in species richness, while activity was dependent on the temperature. Additionally, carabid beetles showed a higher species turnover towards higher altitudes, which means a higher proportion of high elevation and cold adapted specialists in the carabid assemblage compared to the hoverfly assemblage. This could contribute to greater differences in phenological response across altitude in the more generalist hoverflies than in the more specialised carabids.

Altitudinal patterns can be used to predict future patterns in a continuous habitat under the influence of climate change (Botes et al. 2006; Korner 2007; Illán et al. 2012). We found that endotherms and more mobile species had a smaller turnover than ectotherms and less mobile species, from which we can predict that the former species will undergo a less extreme shift than the latter. This was indeed found in birds, which showed no shift (Chen et al. 2011), but in butterflies, there was a strong shift, which could amount to a maximum of 100 m in 50 years (Konvicka et al. 2003; Wilson et al. 2007) or a median speed of 11 m per year (Chen et al. 2011) and for hoverflies, there was a shift predicted for several species (Miličić et al. 2018). On a latitudinal gradient, it was found that many species’ ranges move northwards (Parmesan et al. 1999); however, birds have a smaller climate debt than butterflies (Devictor et al. 2012). Across a latitudinal range, the connectivity between habitats plays a very important role in species’ response to climate change (Settele et al. 2008). In connected and non-fragmented natural habitats, such as temperate forests with a small altitudinal range, temperature affected insects more than birds. Therefore, it is expected that, over small ranges, birds will be less affected by climate change than insects, such as hoverflies and carabid beetles.

The results of study were constrained in time and space, as the sampling only occurred for one year and only on one mountain. As pointed out, the dynamics of the species groups can be heavily affected by the temperature and this could give different results for the different years. However, because the different belts were relatively close to each other, large annual differences would be equally impacting all the different altitudinal belts. In addition, the different belts of Mt. Krim were sampled with more transects, point counts or traps. However, this case study confirmed expected temperature driven mechanisms in assemblage changes. It is therefore important to note for future studies that additional mountains should be sampled in the same way for more years.

When examining altitudinal shifts in patterns, it is important to consider that climate change will affect different functional groups with different traits in different ways. Species that are expected to be most affected by climate change, such as ectotherms and species with poor dispersal ability, should be prioritised, as they are the best indicators for monitoring and conservation management purposes. Current monitoring and conservation programmes are mainly focused on large and charismatic species (e.g. large mammals and birds), which are usually at the top of the food chain in the ecosystem (Pereira and David Cooper 2006), but are, at least with respect to climate change, less affected and, consequently, less threatened and might not be appropriate indicators for climate-driven ecosystem changes.