Current theory holds that the intensity of biotic interactions decreases with increases in latitude and elevation; however, empirical data demonstrate great variation in the direction, strength, and shape of elevational changes in herbivory. The latitudinal position of mountains may be one important source of this variation, but the acute shortage of data from polar mountains hampers exploration of latitude effects on elevational changes in herbivory. Here, we reduce this knowledge gap by testing the prediction that a decrease in herbivory occurs with increasing elevation from forest to alpine tundra. We examined six elevation gradients located in three Arctic mountain ranges. Across the ten most abundant evergreen and deciduous woody plant species, relative losses of foliage to insect herbivores were 2.2-fold greater at the highest elevations (alpine tundra) than in mid-elevation birch woodlands or low-elevation coniferous forests. Plant quality for herbivores (quantified by specific leaf area) significantly decreased with elevation across all studied species, indicating that bottom-up factors were unlikely to shape the observed pattern in herbivory. An experiment with open-top chambers established at different elevations showed that even a slight increase in ambient temperature enhances herbivory in Arctic mountains. Therefore, we suggest that the discovered increase in herbivory with elevation is explained by higher temperatures at the soil surface in open habitats above the treeline compared with forests at lower elevations. This explanation is supported by the significant difference in elevational changes in herbivory between low and tall plants: herbivory on low shrubs increased 4-fold from forest to alpine sites, while herbivory on trees and tall shrubs did not change with elevation. We suggest that an increase in herbivory with an increase in elevation is typical for high-latitude mountains, where inverse temperature gradients, especially at the soil surface, are common. Verification of this hypothesis requires further studies of elevational patterns in herbivory at high latitudes.

The study was replicated at four hierarchical levels: three mountain ranges, six gradients, three elevations within each gradient corresponding to three vegetation types, and ten plant species. We selected our uppermost sites in alpine tundra (high elevation hereafter), at the upper distribution limit of our study trees (Scots pine, Norway spruce and mountain birch). Our intermediate sites (mid-elevation, hereafter) were located in subalpine birch woodland. The lowest sites (low elevation, hereafter) were chosen in closed canopy coniferous forests at the foot of the respective mountains. The difference in elevation between the tundra and forest sites ranged from 170 to 325 m (Table S1).

We measured herbivory in ten plant species: five trees and tall shrubs (evergreen P. abies, P. sylvestris and Juniperus communis L.; deciduous B. pubescens and Salix phylicifolia L.) and five low shrubs (evergreen V. vitis-idaea and deciduous V. myrtillus, V. uliginosum L., S. glauca L. and B. nana L.). Four of these species (B. pubescens and three Vaccinium species) were present in all sites, whereas other species each occurred on 5‒17 plots (Table S2).

The branches of the study plants were collected at the beginning of autumn (13‒17 August 2012, 8‒12 August 2013 and 8‒15 August 2014), when most insect herbivores had completed their development. We haphazardly (on a first found, first sampled basis) selected five mature individuals (or patches) of each of our study species present at a site while standing at a distance of 5‒10 m away to avoid unconscious selection bias. Each selected plant/patch was located at least 10 m apart from others of its species, and each collected branch (or group of stems) generally contained 100 to 200 leaves. In 2012, at sites in the BG and HI gradients, we also collected samples of S. glauca, B. nana, V. uliginosum, V. myrtillus and V. vitis-idea from the study plots enclosed in open-top chambers.

In the laboratory, each leaf was carefully examined for the presence of damage imposed by chewing insect herbivores (both miners and defoliators). In conifers, we searched for traces of insect feeding in 50 or 100 current-year needles, starting from the tip of the branch, and we counted needles that were missing from the shoot. As in the previous study (Zvereva et al. 2020), we attributed the loss of entire needles on current-year shoots to herbivory, because undamaged needles of conifers in our study region persist for several years (Kozlov et al. 2009). Mechanical damage to leaves and needles rarely occurs in our study region, and it can be easily distinguished from herbivory. Both our observations and published data on the plant-feeding organisms of Northern Europe indicate that all the types of damage recorded in the course of our study were imposed by insects.

Following a widely used methodology (Alliende 1989, Kozlov et al. 2015), each leaf/needle (leaf hereafter) was assigned to one of the damage classes according to the percentage of the leaf area that was consumed or otherwise damaged by insects: 0 (intact leaves), 0.01–1%, 1–5%, 5–25%, 25–50%, 50–75%, 75–99% and 100%. The last class included petioles of fully consumed leaves and missing needles. The foliage lost to insects (i.e. the leaf herbivory level) was calculated as follows: the numbers of leaves in each damage class were multiplied by the respective median values of the damaged leaf area (i.e. 0 for intact leaves, 0.5% for the damage class 0.01–1%, 3% for the damage class 1–5%, etc.); the obtained values were then summed for all damage classes and divided by the total number of leaves (including undamaged ones) in a sample.

In 2012, we also measured the specific leaf area ( SLA ) from all seven species of leaf-bearing plants. For this purpose, we made 2‒5 discs 4.5 or 8 mm in diameter (depending on leaf size) from five haphazardly selected leaves, dried the discs for 24 h at 105ºC and then weighed them to the nearest 0.1 mg. SLA was then calculated as total area of the discs (mm²) divided by dry weight (mg).