Inferences from studies along natural elevational gradients
Despite the contrasting elevational trends along individual transects in our study (increasing) and the general global trend (decreasing) found by Galmán et al. (2018), our results confirm that the variation in insect herbivory with elevation along individual transects cannot be explained entirely by traditional climate variables (Galmán et al. 2018). Inconsistent relationships between elevation and ecological characteristics, when comparing multiple individual gradients and regional scale variation, have been shown for a multitude of above- and belowground variables (Sundqvist et al. 2013, Read et al. 2014). While adding ecological context, e.g. field-layer vegetation (Zhao et al. 2018) or grazer presence (Bernes et al. 2015, Vowles et al. 2017), may sometimes be sufficient to explain unexpected trends, the complex geometry of mountainous landscapes poses some challenges to the simplistic, implicit assumption of universality in the relationship between elevation and abiotic variables, such as temperature (Körner 2007). Further, we often implicitly assume that the driver and response variable change at similar rates, i.e. are in a steady state, but this is for instance rarely the case in studies of plant community compositional responses to climate change along natural gradients (Damgaard 2019, Hagedorn et al. 2019). Some of the discrepancies in observed elevation patterns likely reflect difference in variables, methods and variation in the extent to which confounding factors obscure shifts solely related to temperature change with elevation. Our study provides a useful test of the power of elevation gradients by recording the same suite of variables using the same methods across multiple gradients within the same ecosystem in a rather constrained geographical area. While the relationship with elevation was rather consistent across scales for some variables (leaf N content), we found considerable variation in elevation trends in others (leaf herbivory level). Yet, identifying useful moderator variables (e.g. solar insulation) to account for these scale-dependent differences in trends, might be a way to allow for quick integration of space-for-time substitution data from different geographical contexts into ecosystem models. This could improve understanding of important long-term responses to climate change (Dunne et al. 2004; Elmendorf et al. 2015), until sufficient understanding of the underlying mechanisms emerge. Thus, to understand the effect of elevation at a broader scale, we need both local elevational gradient data and regional moderator variables. For example, the contrasting relationship between elevation and BIH in the Subarctic, shown in our study, and the trend at lower latitudes, may be incorporated into models by using latitude as a moderator variable. This approach needs further confirmation, but we show that setting up multiple elevation gradients in the same biome within a rather constrained area yield useful data for such assessments.