Study sites, climate, and geology
The study was conducted in seven European beech forests of mature age along an elevational gradient between 310 and 800 m a.s.l. on the eastern slopes of the Rothaar Mountains in the state of Hesse, central Germany. The elevation transect had a length of approx. 30 km in east-west direction and ranged from the colline/submontane to the montane belt, covering a gradient in mean annual temperature (MAT) of about 2.4°C. All stands belonged to the Luzulo-Fagetum forest community (beech forests on acidic soils) and stocked on acidic bedrock (Triassic sandstone or Paleozoic clay shale) in level to slightly inclined terrain (Table 1). In the forests, study plots of 30 m x 30 m size were selected in sections with closed canopy. While mean diameter at breast height (DBH) varied only between 32 cm and 45 cm in the seven stands and tree ages ranged mostly between 100 and 180 years according to information of the forest offices, stem density was more variable (150–578 stems ha− 1). Mean tree height as measured in 15 trees per plot with a Vertex III height meter (Haglöf, Längsele, Sweden) with at least three measurements taken per tree from different directions decreased with elevation from 33 m to 20 m. The study region has a cool-temperate humid climate with MAT decreasing from 8.4 to 6.0°C and mean annual precipitation (MAP) increasing from 600 to 1200 mm yr− 1 from the colline to the montane zone (German Weather Service, period 1981–2010; Table 1).
During the study period from spring 2014 to winter 2015, air and soil temperature were continuously measured in 2015 with iButton sensors (Maxim, Dallas, USA) installed at 1.5 m height above the forest floor and in the topsoil (3 cm depth) in the seven stands. The sensors were read every 60 min. Dendrometer tapes (type D1, UMS, Munich, Germany; precision of 0.1 mm) were permanently installed at 1.5 m height on 15 trees per plot to determine DBH and annual stem diameter increment through annual DBH recording.
Soil chemical analyses
To characterize soil chemistry, each five topsoil samples were collected with a 6.6 cm-diameter corer at random position in the study plots in summer 2015. Subsequently, the thickness of the organic layer was measured in the cored hole. All soil samples were separated into organic layer and mineral topsoil material (0–10 cm), transferred to the laboratory in a cooling box, sieved (< 5 mm for organic layer material; < 2mm for mineral soil), and stored in polyethylene bags at 4°C for further processing. Subsamples were analyzed in field-moist condition for pH (measured in H2O: 10 g fresh soil in 25 ml deionized water, or in CaCl2: 10 g soil suspended in 0.01 M CaCl2) after 1 h of equilibration. Additional subsamples were dried (60°C, 48 h), ground, and analyzed for total carbon and nitrogen concentrations through gas chromatography with an elemental analyzer (vario EL III, Elementar Analysensysteme GmbH, Hanau, Germany). Since all soils were highly acidic, total C content was assumed to be organic C (SOC). The total P content was determined by ICP-OES analysis (Perkin Elmer Optima 5300 DV) after acid-pressure digestion (65% HNO3 at 195°C for 6 h) of the ground soil samples. Plant-available phosphorus was estimated with the resin bag method according to Bowman and Cole (1987) using Dowex 1 x 8–50 anion exchange gel (Dow Water& Process Solutions, USA) that was placed for 16 h in a solution of 1 g of soil material suspended in 30 ml water (Sibbesen 1977). Extracted P was re-exchanged with NaCl and NaOH solutions and the P concentration measured in a spectrophotometer (Libra S 21, Biochrom, UK) at 712 nm after adding 5 mM hexaammonium-heptamolybdate solution (Murphy and Riley 1962). The water content of the topsoil was determined gravimetrically in each five soil samples collected at random position in the plots every month from March to December 2015. The sampling in August and September was conducted synchronously with the collection of root exudates.
Fine root biomass and root morphology
In November 2018, each 12 root samples were taken at random locations in the 30 m x 30 m plots using a soil corer (6.6 cm in diameter) and the material separated into organic layer and mineral topsoil (0–10 cm) material. Samples were transported in a cooling box to the laboratory where they were kept at 4°C and processed within four weeks. Only fine roots (diameter < 2 mm) of beech were considered in the analysis. All fine root segments were picked out by hand and sorted into live and dead fine root mass under a stereomicroscope (40x magnification). Root vitality was assessed by means of root color and structure of the root surface, root elasticity and turgescence, branching structure, and the degree of cohesion of cortex, periderm, and stele (for criteria, see Persson, 1978; Meier & Leuschner, 2008). Standing FRB was expressed as profile total (organic layer and uppermost 10 cm of mineral soil; in g m− 2). Specific root length (SRL, m g− 1), specific root surface area (SRA, cm2 g− 1), root tissue density (TD, mg cm− 3), and root tip frequency (RTF, number of root tips per fine root mass; n g− 1) were determined for the root material using a flatbed scanner and the software WinRhizo (Régents Instruments, Quebec, Canada). Fine roots used for exudate collection were clipped off and measured for the total root surface area of the root segment. Root biomass was determined by drying (48h, 78°C) and weighing the sorted root samples.
Root exudate collection
Root exudates were collected in three sampling campaigns in July of 2014, August 2015, and September 2015 in nine soil pits excavated at each site in at least 3 m distance to the nearest mature beech tree, employing the cuvette-based in situ-collection approach (after Phillips et al. 2008; Freschet et al. 2021). Beech fine root strands were carefully extracted from the topsoil and cleaned with fine forceps and deionized water. The terminal strand sections (average cumulative length of all parts of the strand c. 16.5 cm and mean diameter c. 0.45 mm) were embedded in root cuvettes filled with sterile 2 mm-diameter glass beads to simulate the porosity of the soil and mechanical impedance in a matrix free of carbon. The beads covering the root were moistened with a carbon-free nutrient solution (0.5 mM NH4NO3, 0.1 mM KH2PO4, 0.2 mM K2SO4, 0.15 mM MgSO4, 0.3 mM CaCl2).
The roots in the cuvettes were equilibrated for 48 h before flushing the cuvettes with a diluted nutrient solution under low pressure. New carbon-free nutrient solution was added and after an equilibration of approx. 24 h, these trap solutions were collected using low-pressure vacuum. These solutions were subsequently filtered through sterile glass fiber filters (GE Healthcare Life Sciences Whatman, Glass Microfibre Filters, Grade GF/F) and stored at -20°C until further analysis. Control samples were taken from rootless cuvettes treated similarly. The samples were analyzed for their dissolved organic carbon using a total carbon analyzer (Shimadzu TOC-L CPH/CPN; Shimadzu Scientific Instruments, Duisburg, Germany). Taking fresh root biomass as a calculation basis, net mass-specific exudation rates (µg C g− 1 h− 1) and annual C fluxes with exudation per root mass or ground area (mg g− 1 yr− 1, g C m− 2 g− 1) were calculated, the latter by multiplying daily exudation rates by the length of the growing season (defined here as the number of days with average temperatures over 10°C). To estimate growing season length for the seven sites, we used gridded temperature data provided by the German Weather Service (DWD). Growing season length decreased with decreasing MAT from 170 to 125 d between 310 and 800 m a.s.l. (Table 1).
Statistical analyses
All statistical analyses were conducted with SPSS software. The data was tested for fit to normal distribution using a Shapiro-Wilk test. Normally distributed data were tested for homogeneity of variances with a Levene test. Site differences between means of edaphic (total C and N content, C/N-ratio, pH (H2O), pH (CaCl2), total and plant-available P content) and root morphological variables (TD, SRA, SRL, and RTF), and exudation rates (net mass-specific exudation rate and annual C flux with exudation) were examined with one-way analysis of variance for parametric data and a Kruskal-Wallis test for non-parametric data. ANOVAs were followed by a Scheffé or a Dunnett T3 test, if homogeneity of variances was not given. Kruskal-Wallis tests were followed by pairwise comparisons to locate differences.
Pearson correlations were used for investigating the relation between root exudation rate and elevation, climatic and edaphic variables, and root morphological traits. If data were non-normally distributed, Spearman rank correlation analysis was employed. Correlations were tested for the variables long-term mean temperature and precipitation, average summer temperatures of 2014 and 2015, soil water content at the date of exudate sampling, and air and soil temperatures averaged over the seven days prior to sampling. The p-values were adjusted by the Benjamini-Hochberg procedure for multiple testing.
Multiple regression analyses with backward variable elimination were conducted to test for significant independent predictors of root exudation rates and the estimated annual C flux with exudation per m2 ground area. In the initial model, prevailing temperature, soil moisture, stem density, pH, soil C, N and P content, and the four root morphological traits were included. At each elimination step, the variable showing the smallest contribution to the model was deleted until all variables remaining in the model produced significant F statistics. The p-values were calculated via the bootstrapping method because most of the data showed no fit to normal distribution. Variables were tested for multi-collinearity and were excluded when they were highly correlated and collinearity diagnostics (variance inflation factor and tolerance) were critical.