Whitehall Forest Soil Warming Facility Leaf Traits and Environmental Data

Abstract We used a long-term, in situ soil warming experiment, nested within a canopy-based habitat treatment, to investigate how leaf traits of temperate forest woody species, namely juvenile trees and lianas (woody vines), may respond to changes in soil temperature, across natural gradients in soil moisture and light availability. The Whitehall Forest Soil Warming Facility is a long-term soil warming experiment, established in 2009, where warming began in 2010. The site consists of 24 experimental plots, each 18.2 m2 in area, with eight environmental treatment levels combining experimentally manipulated soil temperature and canopy cover. Plots were built with supplementary open-top chambers, constructed with sheets of clear greenhouse polyvinyl chloride (PVC) and supplemented with 240 V electrical resistance buried soil warming cables, manually installed 10-cm below the soil surface in 18 of the 24 plots. The temperature treatments include six plots each of targeted ambient, mild, and hot temperatures, plus a non-chambered control (NCC) treatment, which consists of chicken wire rather than greenhouse plastic and does not possess buried soil warming cables. The NCC treatment was used to account for potential “chamber effects” and soil disturbance caused by warming cable installation. Half (three) of the plots of each temperature level (including NCC) were placed in a ~50m-by-50m open canopy, that was manually cleared in 2009 (which we have termed the ‘gap’ treatment). Methods In this study, data was collected from the fifteen most abundant tree species and two liana species. The trees within the plots have been tagged and their annual growth tracked throughout the duration of the experiment (i.e., since 2010). Annual growth refers to measures of lead stem length and diameter at 5-cm above soil surface (D5) or diameter at breast height (DBH, when over 2 m tall). Our two liana species, cross-vine (Bignonia capreolata) and muscadine (Vitis rotundifolia), were tagged and measured for the first time in 2023. Up to five healthy, mature, intact leaves were chosen from each plant, preferentially using full-sized leaves closest to stem tips. Many individuals of each species were small seedlings or ramets with less than five leaves, in which case, all full-sized, intact leaves were measured. For each individual plant, leaf length, width, and thickness were obtained using a flexible measuring tape (following the topology of the leaf; Nuomi, Foshan City, China) and a Rexbeti Digital Micrometer (0.001 mm resolution, ±0.0025 mm accuracy; Rexbeti, Los Angeles, CA). Plant diameter was measured at 5 cm above the soil surface (henceforth ‘D5’) using digital calipers (General Tools USA, Secaucus, NJ). Plant length was measured as the length of the lead stem from the ground (or for vines, from the node connecting the lead stem and the root runners on or just below the soil surface), including bends or kinks in the stem, using jointed metersticks. Measured stem lengths varied from 4.3 to 230.6 cm across our study, where individuals with lengths greater than 240 cm were not measured. Due to the limited number of leaves on many of the juvenile plants within the experimental plots, we calculated allometric estimations of traits from plant measurements, rather than destructively sampling leaves for direct measurements. Leaf areas were estimated based on species-specific correction factors and leaf shape via Schrader et al. (2021), and estimations were verified with 160 independent leaves collected from the six most abundant species at WFWF (collected outside the experimental plots) using a photocopier and ImageJ software. A linear regression was generated for estimated leaf area against scanned leaf area for each species, and for species with trendline slopes of >1.05 or <0.95, new correction factors were calculated following the methods of Schrader et al. (2021) on a per-species and per-site basis. To obtain leaf area, we multiplied leaf length, width, and the corresponding correction factor together. We then similarly calculated our own site-specific correction factors for estimating dry mass by oven-drying and weighing the leaves collected outside the experimental plots and dividing leaf dry mass by leaf area times leaf thickness. Finally, using leaf area and dry mass calculations for the original 692 plants within the experimental plots, we calculated estimates of specific leaf area (SLA) as the ratio of leaf area to leaf biomass. For each plot, mean temperature delta (ΔT) values were calculated as the difference in mean temperature of the chambered plots from that of the nearest control (NCC) plot. Soil volumetric water content (VWC; a unitless ratio of the volume of water to a measured volume of soil) and temperature measurements were recorded hourly using one CS616 Water Content Reflectometer per plot and two to three CSI thermistors per plot (Campbell Scientific, Lincoln, NE, USA), respectively. Photosynthetically active radiation (PAR; in the form of photosynthetic photon flux density) was recorded hourly using one Apogee SQ-110 Quantum Sensor (Apogee Instruments, Logan, UT, USA) in the center of each plot and averaged over the same time periods. SLA relationships were explored using soil ΔT, soil VWC, and PAR (1-year and 5-year) for the most abundant species using linear regression models generated with the stats package in R. Non-parametric difference-of-means tests (Dunn’s Kruskal-Wallis multiple comparisons and Mann-Whitney Wilcoxon) were used to compare plant traits across habitat and temperature treatments. All analyses were performed in R version 4.4.0. To better quantify the combined impacts of soil warming, soil moisture, light level, and other potentially influential variables (e.g., species identity, plot, individual plant, stem length), we also fit the data to mixed-effects multiple-regression models using the glmmTMB package in R with a log-linked gamma distribution function. Model fit was evaluated based on Akaike Information Criteria (AIC) and residual deviation tests (using DHARMa). Individual species whose models did not fit due to small sample sizes were not considered.