Study site and design
The study sites were in subarctic mountain birch forests near the Abisko Scientific Research Station, northern Sweden (68.35°N 18.82°E). All samples were taken in 2017. The layout was designed to capture as much of the regional climatic variation as possible within a relatively constrained geographic area (<50 km from Abisko) by taking advantage of the considerable climatic gradients created by the mountainous landscape between Abisko and the Kebnekaise complex (Figure 1). We established 9 elevation transects, each with a site near the valley bottom, one near the treeline, and one in the middle, making a total of 27 sites. Further, 8 of the transects were paired as a north- and a south-facing transect at 4 locations to control for the influence of aspect/solar radiation. According to a digital elevation model (DEM) from Lantmäteriet (50 m resolution), regional variation in elevation across the 27 sites is approximately 500 m (minimum 351 and maximum 845 m above sea level), while the minimum and maximum elevational difference along a single transect is 82 and 233 m respectively. To control for aspect, we used the DEM to estimate the total annual solar radiation of 2017 for each site using the “Area Solar Radiation” function in ArcGIS 10.3 (ESRI, Redlands, CA, US). Sites were preferentially established in dry locations with heath-dominated ground vegetation, however, this was not possible in all instances, as some of the wetter transects were dominated by herbaceous ground vegetation (7 out of 27 sites with >50% herbs). The bedrock geology was dominated by granites and other acidic bedrocks, and there was weak-moderate podzolisation at all sites. According to the Swedish Meteorological and Hydrological Institute Luftwebb record (4x4 km resolution) from the last normal period (1961–1990, isolines in Figure 1), the mean annual air temperature across our sites ranged between -2.5±1.03 and -1.4±1.03 °C (mean ± SE), and the mean annual precipitation ranged from 404±75 to 1250±158 mm. For the analyses below, we have used the average of the more recent period of 2000-2014 to derive meteorological variables more comparable to the study period.
The rationale behind the design was to assess the assumption underlying the use of elevation as a climate proxy: If there is a universal relationship between elevation and an ecological variable, the slope of the trendline in an X-Y plot would be similar at regional and local scale. To check the design, we included a plot of a commonly measured ecological variable, foliar N content, against elevation, as we expected this to show a clear and consistent increasing trend with elevation across scales (Körner 1989), although potentially weaker at regional scale (Read et al 2014).