Effects of Flooding and Shade on Survival, Growth, and Leaf Gas Exchange of Bottomland Tree Species across the Great Lakes Region (USA)

By inergency On Mar 13, 2024

[ad_1]

2.1. Study Design

This study was conducted at the USDA Forest Service Northern Research Station Rhinelander Experimental Forest located in Harshaw, WI, USA. Twenty-three species were selected for evaluation based on the input from silviculturists at the Wisconsin Department of Natural Resources (DNR) and other research studies or management treatments (Table 1). These species are currently found in various forest wetland ecosystems within the region or in more southern locations for some species considered for assisted migration [33]. Seedlings were obtained from the Wisconsin State Nursery in Boscobel, WI, USA, stored in a walk-in refrigerator at 4 °C for 35 days, and then potted in early June 2021. Seedlings were potted into plastic containers (10 cm × 10 cm × 30 cm) with a soil mixture of sphagnum peat moss (60%–70%), bark, and perlite (Berger BM7 Bark Mix, Saint-Modeste, QC, Canada).

The experiment was designed as a split–split plot randomized complete block design that included a light reduction treatment as the whole plot factor (light reductions of 0, 40, and 70%), a flood level treatment as the first split–plot factor (water levels at 0, 14, and 27 cm below the soil surface), and tree species as the second split–plot factor. This design allowed for both flooding and shade treatments to be applied to each set of seedlings simultaneously. A single seedling from each species was randomly placed in each of the 72 stock tanks (dimensions of 132 cm × 79 cm × 36 cm) (Figure S1), which were then randomly assigned to a flood and light treatment combination within each block. Within each tank, seedlings were spaced such that the distance between each seedling was maximized. Each treatment combination was replicated eight times.

Flooding treatments were applied by filling the tanks with groundwater until water levels equilibrated at the assigned water table depth. The tanks were checked weekly to ensure that the water depth was maintained over the course of the growing season. Light reduction treatments were applied using shade cloths purchased from a greenhouse supply store. Each shade cloth measured 6 × 7.3 m and reduced ambient light by either 40% or 70%. A photosynthetic active radiation (PAR) sensor (SQ-521, Apogee Instruments, Logan, UT, USA) connected to a datalogger (model Z6, METER Group, Pullman, WA, USA) was used to verify whether the reduced light under the shade cloths matched the expected value. Two cables were hung 2.5 m above the ground and 2 m apart. The shade cloths were draped over the cables and staked out at the corners to form a trapezoidal tent, with the three tanks (each with one of the three flood levels) aligned in the center. This design ensured that flood treatments inside a shade tent were never exposed to direct full sunlight.

Treatments were initiated on June 4th and maintained for 14 weeks. The measured air temperature at the site (model ATMOS 14, METER Group, Pullman, WA, USA) ranged from 4.3 to 30.6 °C outside of the shade tents during the study period. The maximum photosynthetically active radiation in the full sun, 40%, and 70% light reduction treatments was 1999, 923, and 595 µmol m⁻² s⁻¹, respectively. For the purposes of this study, the 0 cm water table depth (fully flooded) was considered the most stressful flooding condition, and 27 cm was considered to be the least, with 14 cm causing intermediate stress.

In the two weeks following treatment initiation, 28% of seedlings were unexpectedly browsed by white-tailed deer (Odocoileus virginianus Zimm.) to varying degrees (Table S1). None of the conifer species were browsed. The initial height and basal diameter measurements of all individuals, both browsed and unbrowsed, were taken after the browse event (using a meter stick and calipers, respectively). To account for any effect of browsing on the response, the leaf gas exchange measurements were balanced between three individuals that were browsed and three that were unbrowsed. In cases where this was not possible, the sample was composed such that the number of browsed and unbrowsed individuals were as balanced as possible.

Leaf gas exchange measurements were taken on a subset of the following 12 species: red maple, silver maple (Acer saccharinum L.), river birch, hackberry (Celtis occidentalis L.), tamarack (Larix laricina (Du Roi) K. Koch), American sycamore, trembling aspen (Populus tremuloides Michx.), swamp white oak, bur oak, bald cypress, northern white cedar (Thuja occidentalis L.), and American elm (Ulmus americana L.). These species were chosen based on the results from previous experiments in a related study [34]. All gas exchange measurements were taken using a portable photosynthesis system (models 6400 and 6800; Licor, Lincoln, NE, USA) on a randomly selected subsample (6 out of 8 total replicates) of each treatment combination. For all measurements, CO₂ concentrations were maintained at 400 µmol CO₂ mol⁻¹ air, the flow rate of air through the leaf chambers was maintained at 600 µmol s⁻¹, and photosynthetic active radiation was maintained at 1500 µmol m⁻² s⁻¹. These measurements were standardized to provide consistent estimates of light-saturated photosynthesis for the valid comparison of effects across treatments. A 2 × 3 cm leaf chamber was used for conifers and a 2 cm² circular chamber for broadleafs. As photosynthesis was recorded on a per-unit-leaf-area basis, the leaf area of conifer needle sprigs was corrected using methods from Bermudez et al. [35]. For broadleaf species, measurements were taken on a leaf in the upper third of the seedling. For conifers, measurements were taken on the previous year’s growth unless none were available, in which case it was taken on the current year’s growth. Gas exchange measurements were taken on a biweekly basis for 10 weeks, totaling 6 sampling periods.

Height and basal diameter were measured at the start and end of the growing season. Height was measured from the soil surface to the tip of the uppermost living tissue, and basal diameter was measured 1 cm above the root collar. Negative height values resulted from stem dieback, and negative basal diameter values resulted from stem shrinkage in response to water stress [36]. Survival was also assessed at this time; dead seedlings were identified as those exhibiting no visible green tissue.

2.2. Data Analysis

Each species was analyzed independently for all measured responses to facilitate the interpretation of the results. The effect of water table depth and light on seedling survival was assessed using generalized linear mixed models with a binomial distribution. Water table depth, light, and their interaction were included as fixed effects. For broadleaf species, browse was included as a binary random effect to account for the variation introduced by the early season browse event. For conifers, none of which were browsed, block was included as a random effect to account for the variation introduced by ambient conditions in the field. Block was not included as a random effect in the broadleaf models due to issues with model convergence arising from the inclusion of two random variables. Random effects were removed entirely from some models due to convergence problems. Survival models would not converge for the following species due to low mortality: American elm, northern red oak (Quercus rubra L.), trembling aspen, hackberry, bur oak, black spruce (Picea mariana (Mill.) Britton, Stearns & Poggenb.), northern white cedar, and bald cypress. Survival for the rest of the species was analyzed with mixed model regressions using the glmmTMB function in the ‘glmmTMB’ package [37] in R 4.0.3 software (R Core Development Team, Vienna, Austria, 2018) and fitted by the Type II Wald chi-square test. When significant effects were observed, pairwise multiple comparisons were conducted with the ‘emmeans’ package [38] using Tukey’s HSD adjustment.

The effect of water table depth and light on absolute height and basal diameter growth was analyzed using linear mixed-effects models with a normal distribution. Water table depth, light treatment, and their interaction were included as fixed effects. Block was included as a random effect for both broadleafs and conifers. In addition, browse was included as a binary random effect for broadleafs. Models were run for each species individually. Pre-treatment height or basal diameter was included as a covariate. Mixed models were applied using the lme function in the ‘nlme’ package [39] in R 4.0.3 software. Pairwise multiple comparisons were conducted with the ‘emmeans’ package using Tukey’s HSD adjustment. In all models, an alpha level of 0.1 was chosen to account for high variability in the response variables and reduce the likelihood of a Type 2 error.

Treatment effects on photosynthesis, stomatal conductance, and transpiration were analyzed using linear mixed-effects repeated measures analyses with normal distributions. Water table depth, light, measurement week, and their interactions were included as fixed effects. Browse was included as a binary random effect for broadleafs, and block was included as a random effect for conifers. An AR(1) covariance matrix was used to account for serial correlations among the measurements. The ‘predictmeans’ package [40] in R 4.0.3 software was used to conduct Cook’s distance tests to identify outliers that were then removed from the dataset [41]. Mixed models were applied using the lme function in the ‘nlme’ R package. Pairwise multiple comparisons were conducted with the ‘emmeans’ package using Tukey’s HSD adjustment. The “powerTransform” tool from the ‘car’ R package [42] was used to identify the most appropriate data transformations for data not conforming to a normal distribution and/or with heteroscedastic residuals. Due to the number of plants browsed in the first sampling period and the removal of outliers resulting from measurement issues, the first and last sampling periods had to be removed from the models of several species to solve rank deficiency. For all analyses, model assumptions of normality and homogenous variances were visually assessed and confirmed. In all models, an alpha level of 0.1 was used.

When significant effects were observed, relative responses to convey the trends of responses among species and within treatment factors were used. The reported percentage decrease in response variables was calculated by dividing the smaller value by the larger value, multiplying by 100, and then subtracting from 100. All percentages were rounded to the nearest 5% and listed as an average to account for the variability in the value. To ease the interpretation of response to water table depth in Section 3, species were grouped into four patterns for each response variable separately, excluding survival (Table 2).

[ad_2]