Use climatic space-for-time substitutions with care: not only climate, but also local environment affect performance of the key forest species bilberry along elevation gradient

An urgent aim of ecology is to understand how key species relate to climatic and environmental variation, to better predict their prospects under future climate change. The abundant dwarf shrub bilberry (Vaccinium myrtillus L.) has caught particular interest due to its uphill expansion into alpine areas. Species' performance under changing climate has been widely studied using the climatic space-for-time approach along elevation gradients, but potentially confounding, local environmental variables that vary along elevation gradients have rarely been considered. In this study, performed in ten sites along an elevation gradient (200–875 m) in W Norway, we recorded species composition and bilberry performance, both vegetative (ramet size and cover) and reproductive (berry and seed production) properties, over one to four years. We disentangled effects of local environmental variables and between-year, climatic variation (precipitation and temperature), and identified shared and unique contributions of these variables by variation partitioning. We found bilberry ramet size, cover, and berry production to peak at intermediate elevations, whereas seed production increased upwards. The peaks were less pronounced in extreme (dry or cold) summers than in normal summers. Local environmental variables explained much variation in ramet size and cover, less in berry production, and showed no relation to seed production. Climatic variables explained more of the variation in berry and seed production than in ramet size and cover, with temperature relating to vegetative performance, and precipitation to reproductive performance. Bilberry's clonal growth and effective reproduction probably explain why the species persists in the forest and at the same time invades alpine areas. Our findings raise concerns about the appropriateness of the climatic space-for-time approach. We recommend including both climatic and local environmental variables in studies of variation along elevation gradients, and conclude that variation partitioning can be a useful supplement to other methods for analysing variation in plant performance.All data are stored in an Excel Workbook, in separate sheets.We studied bilberry performance in 2017–2020 along an elevation gradient in a boreal, bilberry-dominated Scots pine (Pinus sylvestris) forest in Sogndal, Vestland county, W Norway (61°13'N, 7°9'E). The study site extended along a 2.7 km SW-facing ridge, covering the altitudinal interval 200–875 m a.s.l., from dense lowland forest up to the alpine tree line. In 2017, we subjectively placed ten blocks of 5 × 10 m along the elevation gradient, all facing S–SW with an average slope of 10–30 °. Each block spanned the variation from open to shaded sites. In each block, we randomly placed and permanently marked five plots of 0.5 × 0.5 m with a minimum between-plot distance of 1 m. We collected six data sets in the 50 plots; (1) species composition in 2017, (2) bilberry vegetative performance (ramet size) in 2017-2018, (3) bilberry vegetative performance (bilberry cover) in 2017, (4) bilberry reproductive performance (berry production) in 2017-2020, (5) bilberry reproductive performance (seed production) in 2017-2018, and (6) environmental data in 2017. 1. Species composition. We divided each plot into 16 subplots, each 0.0156 m2. In June–July 2017, we recorded the presence or absence of vascular plants, bryophytes, and lichens in each subplot and calculated subplot frequency as a measure of species' abundance. 2. Bilberry vegetative performance - ramet size. We tagged selected bilberry ramets in each plot by applying a split, coloured Hama maxi bead to the main twig of each ramet (Hegland et al., 2010). Ramet selection, performed in 2017, started in one of the four central subplots and continued in an anticlockwise manner by including additional subplots until eight ramets were included. The total number of ramet tagged in 2017 was 272, of which 15 could not be re-located at the 2018 census. That year we included all fertile ramets in each plot, thereby increasing the total number of tagged ramets by 108. A total of 46 and 40 plots contained berry-producing ramets in 2017 and 2018, respectively. In 2017 and 2018, we measured plant height (H), stem diameter (DS), and number of annual shoots (AS) for all tagged ramets and calculated dry mass (DM) of each ramet as a non-destructive estimate of plant size, using the formula described by Hegland et al. (2010): log2(DM) = 1.41700 × log2 (DS) + 0.97104 × log2 (H) + 0.44153 × log2 (AS + 1) –7.52070. The formula is based upon a model developed from a representative, destructively sampled data set (R2 = 0.944, n = 150). 3. Bilberry vegetative performance - bilberry cover. We visually estimated percent cover of bilberry in each plot. 4. Bilberry reproductive performance - berry production. Each year from 2017 to 2020, we counted the total number of berries produced in each 0.25 m2 plot and multiplied this number by four to express berry production as berries per square meter (m-2). 5. Bilberry reproductive performance - seed production. In 2017 and 2018, we also counted seeds. We collected all berries produced by the tagged ramets in each plot, dried them in a drying cabinet at 70 °C for 36 hours, and weighed them individually. For each plot the berry closest to plotwise mean mass was selected, dissected under a stereo microscope, and the number of mature (> 1 mm long, filled) and non-mature seeds (aborted seeds or unfertilised ovules; < 1 mm, flat or negligibly swollen) were counted. The counts were used to calculate two measures of seed production per plot: the number of mature seeds, and the fraction of mature to total (mature and immature) seeds, per berry (Pato and Obeso, 2012a). 6. Environmental data. A set of 16 environmental variables that could potentially influence species composition and bilberry performance, was recorded for each of the 50 plots based on field measurements or soil samples taken in 2017. We recorded soil chemical and physical variables for the 50 plots from soil samples obtained by mixing four soil sub-samples taken from the upper 10 cm of the humus layer in each plot. Soil samples were dried at 35 °C for ten days in drying cabinets before sifting (2 mm mesh width). pH and soil organic matter (SOM) were measured at the soil laboratory at Western Norway University of Applied Sciences, Sogndal, by procedures described by Krogstad (1992). All other soil analyses were carried out at the soil laboratory of the Norwegian University of Life Sciences, Ås. Total N was determined by the Dumas method. Other soil elements were extracted in a NH4NO3 solution (Stuanes et al., 1984) for determination of exchangeable element concentrations by ICP. Among these, we used elements regarded as macronutrients, micronutrients, or toxic to plants (P, K, Al, Ca, Fe, K, Mg, Mn, S and Zn) for further analyses. We recalculated all elements as ppm (parts per million) by multiplication with 1/SOM (Økland, 1988). We recorded soil moisture in August 2017 after four days without rainfall, using an AT Delta-T moisture meter, type HH2 SM300 v 4.0. The mean moisture, used in analyses, was calculated from four measurements per plot. Soil depth was measured as the distance a steel rod (diameter 1 cm) could be driven into the soil at eight fixed positions 10 cm outside the plot border. The median of these eight values was used in the analyses. The Heat index (HI) was calculated according to Heikkinen (1991) by measuring aspect and slope (0–360° scale) using a clinometer compass held at a position considered representative for each plot. We measured canopy openness by use of a convex, spherical densiometer with 24 squares (Lemmon, 1956). Canopy openness (Light) was estimated as the number of squares not containing canopy