Data for: Local adaptation of switchgrass drives trait relations to yield and differential responses to climate and soil environments
Ricketts, Michael et al. (2023), Data for: Local adaptation of switchgrass drives trait relations to yield and differential responses to climate and soil environments, Dryad, Dataset, https://doi.org/10.5061/dryad.7sqv9s4w9
Switchgrass, a potential biofuel crop, is a genetically diverse species with phenotypic plasticity enabling it to grow in a range of environments. Two primary divergent ecotypes, uplands and lowlands, exhibit trait combinations representative of acquisitive and conservative growth allocation strategies, respectively. Whether these ecotypes respond differently to various types of environmental drivers remains unclear but is crucial to understanding how switchgrass varieties will respond to climate change. We grew two upland, two lowland, and two intermediate/hybrid cultivars of switchgrass at three sites along a latitudinal gradient in the central United States. Over a 4-year period, we measured plant functional traits and biomass yields and evaluated genotype-by-environment (G´E) interaction effects by analyzing switchgrass responses to soil and climate variables. We found substantial evidence of G´E interactions on biomass yield, primarily due to deviations in the response of the southern lowland cultivar Alamo, which produced more biomass in hotter and drier environments relative to other cultivars. While lowland cultivars had the highest potential for yield, their yields were more variable year-to-year compared to other cultivars, suggesting greater sensitivity to environmental perturbations. Models comparing soil and climate principal components as explanatory variables revealed soil properties, especially nutrients, to be most effective at predicting switchgrass biomass yield. Also, conservative plant traits such as high stem mass and tiller height became increasingly positively associated with biomass yield at lower latitudes where the climate is hotter and drier, regardless of ecotype. Lowland cultivars, however, showed a greater predisposition to exhibit these conservative traits. These results suggest switchgrass trait allocation trade-offs that prioritize aboveground biomass production are more tightly associated in hot, dry environments and that lowland cultivars may exhibit a more specialist life strategy relative to other cultivars. Altogether, this research provides essential knowledge for improving the viability of switchgrass as a biofuel crop.
In 2016, switchgrass field sites were established in three locations; Temple, Texas (TX) at the Agricultural Research Service (ARS) Grassland, Water and Soil Research Laboratory, United States Department of Agriculture (USDA; 31.045109°, -97.348135°); Columbia, Missouri (MO) at the Bradford Research Center, University of Missouri (38.899708°, -92.213984°); and Batavia, Illinois (IL) at the Fermi National Accelerator Laboratory (Fermilab) National Environment Research Park (41.836701°, -88.239708°). Before establishment, each site had varying long-term, land-use histories, including periods of cultivation, fallow, or pasture. The soil types and characteristics differ substantially among sites (Figure S1 and Table S1). Briefly, the Houston Black clay at the TX site is a moderately well-drained, very slowly permeable vertisol that was formed in clayey residuum weathered from calcareous mudstone of Upper Cretaceous age and exhibits notable cracking during dry periods; the Mexico silt loam at the MO site is a poorly drained alfisol (with a relatively shallow claypan) that formed in deep loess over pre-Illinoian till; and the Mundelein silt loam at the IL site is a somewhat poorly drained mollisol formed in loess deposited over stratified calcareous loamy Wisconsinan glacial outwash (https://soilseries.sc.egov.usda.gov/).
Similarly, the three sites differed considerably in climatic conditions (Figure S2 and Table S2). We used the National Oceanic and Atmospheric Administration (NOAA) website (https://www.ncei.noaa.gov/products/land-based-station/us-climate-normals) to identify meteorological (MET) stations that were close to each site and had 30-year (1991-2020) mean annual precipitation (MAP) and air temperature (MAT) norms (Table S2). We also collected daily total precipitation (PRCP), average maximum air temperature (Tmax), average minimum air temperature (Tmin), and average daily wind speed (ADWS) from MET stations specific to each field site from 12/1/2016 through 11/30/2020. Lastly, we obtained average daily solar radiation (SOLRAD), and dew point temperatures (Tdew) from the same time period from the National Solar Radiation Database (NSRDB) using the U.S. & Americas dataset with 60-minute temporal resolution and 4 km spatial resolution (https://nsrdb.nrel.gov/data-sets/us-data; Sengupta et al., 2018). The collected data was then used to calculate the average daily potential evapotranspiration (PET) using the Penman-Monteith equation (R package SPEI) using Tmin, Tmax, ADWS, SOLRAD, Tdew, elevation, and latitude for each site, with the crop argument set to “tall” to account for growth taller than 0.5 m (Allen et al., 1994; Beguería & Vicente-Serrano, 2017; Walter et al., 2002).
Pre-planting soil measurements
Prior to planting in 2016, soil samples were collected from six plots that were pre-selected to capture and characterize the range of variability in soil physical and chemical properties at each site (Figure S1 and Table S1). Soil cores (3.8-cm diameter) were taken with a hydraulic coring machine to a depth of 1.5 to 2 m, depending on soil conditions. The cores were split into 25-cm depth increments, and the surface 25-cm increment was divided further into 0–5 cm, 5–15 cm, and 15–25 cm increments. After removing roots and large pieces of organic debris, samples were oven dried at 65°C and passed through a 2-mm sieve. Gravimetric soil moisture (% dry mass basis) and bulk density (BD; g cm-3) were calculated for each depth increment. Subsamples from three of the six cores from each site were analyzed for pH (1:1 soil:water), total phosphorus (TP; salicylic-sulfuric acid digestion; g kg-1), Mehlich-3 extractable phosphorus (M3-P; g kg-1), ammonium and nitrate nitrogen (NH4+-N and NO3- -N; 2 M potassium chloride extraction; g kg-1), electrical conductivity (EC; saturated paste; ds m-1), cation exchange capacity (CEC; summation of neutral 1 M ammonium acetate exchangeable cations plus neutralizable acidity; meq 100 g-1), and texture (sand, silt, and clay; hydrometer method; %) at the Kansas State University Soil Testing Laboratory. Another subsample was finely ground in a ball mill (SPEX SamplePrep Mixer/Mill, SPEX SamplePrep LLC, Metuchen, NJ, USA), and total organic carbon and total nitrogen concentrations (TOC and TN, respectively; g kg-1) were determined at Argonne National Laboratory by dry combustion (vario MAX cube, Elementar Americas, Inc., Ronkonkoma, NY, USA) at 900°C for TN and 650°C for TOC (Provin, 2014).
Experimental design and plant establishment
At each site we established identical common gardens of switchgrass cultivars, consisting of 30 6´6 m sward plots separated by 3 m mowed alleys. Each common garden was arranged in a randomized complete block design, where plots within each block were randomly assigned to one of six cultivars (5 blocks ´ 6 genotypes). The cultivars spanned the ecotypic diversity of switchgrass with the lowland ecotype represented by Alamo and Kanlow cultivars, the upland ecotype represented by Blackwell and Cave-in Rock (CIR) cultivars, and intermediate ecotypes represented by Carthage (a coastal ecotype) and Liberty (a hybrid of upland and lowland ecotypes).
To control for potential site biases commonly associated with early plant establishment, development, and vigor in switchgrass experiments, seedlings of each cultivar were germinated in pots (3.8 cm ´ 21.0 cm SC10 Conetainers™, Stuewe & Sons, Tangent, OR) from commercial seed sources, except for Liberty, which was provided by the UDSA-ARS, Lincoln, NE. For germination, pots were filled with a soil mixture (Turface, Profile Products LLC, Buffalo Grove, IL and Pro-Mix BX, Premier Tech, Rivière-du-Loup, Québec). Seed was spread on the surface of dampened soil, covered with vermiculite, stratified in the dark at 4 ºC for 7 days and subsequently placed in a greenhouse at the University of Texas at Austin (30°C/22°C, 14-hour photoperiod). During early growth, pots were thinned to one seedling, then watered as necessary and allowed to grow for ~7 weeks. Each field plot was tilled to ~15 cm depth and planted with 169 individual plants spaced 0.5 m apart in a 13 ´ 13 grid. Seedlings were irrigated by broadcast sprinklers at each site as necessary in coordination with precipitation over two months to ensure establishment. Additionally, pre-emergent and broadleaf herbicides (e.g., Prowl H2O, Simazine, 2,4-D) were applied at the beginning of each growing season to control weed growth according to the specific needs of each site. During the growing season, plots were maintained by supplemental hand weeding within, and regular mowing between, plots, as necessary.
Cultivar trait measurements
We performed repeated measurements of leaf, stem, whole tiller, and whole plant traits throughout the 2017–2020 growing seasons. Every 2–6 weeks following cultivar green-up, we recorded plant height and number of tillers from five preselected plants in each plot. We also harvested five tillers from each of three additional plants per plot. We separated leaves from stems and measured their areas using leaf area meters (LI-3100C Area Meter in TX and MO, LI-3000C Portable Area Meter with LI-3050C Transparent Belt Conveyer in IL, LI-COR Biosciences, Lincoln, NE), then dried the samples at 65°C for at least 72 hours before determining dry weights. Leaves and stems were combined and ground in a Wiley mill, and then subsamples of the homogenized tissue were finely ground with a cyclone mill for analysis of tiller C and N concentrations by dry combustion at 900°C in the elemental analyzer as described above.
From these repeated measurements, we derived several tiller, whole plant, and plot-level traits. Stand aboveground biomass production (Mg ha-1) was estimated by multiplying the average tiller (leaf + stem) dry mass by the average number of tillers per plant and the number of plants (n=169) per plot. Specific leaf area (SLA; cm2 g-1) was calculated as the ratio of leaf area to dry leaf biomass. Finally, for each plot and within each year, we considered the time point with the highest estimated biomass to be “peak biomass”. Trait values measured during this “peak biomass” sampling period were used for all subsequent analyses.
Cultivar aboveground biomass yield (i.e., total annual yield; Mg ha-1) was measured by collecting and weighing aboveground biomass at the end of each growing season, typically mid-to-late November at all sites. Biomass was collected from two 2×2 m quadrats (16 plants each) within each plot, which were left undisturbed throughout each year of the study. Plants were cut at roughly 10–15 cm above the soil surface, loaded onto a tarp, and weighed immediately using a hanging scale (OP-926, Optima Scale Inc., Rancho Cucamonga, CA). Harvested wet weights were converted to dry weights by weighing subsamples before and after drying at 65°C and computing a moisture correction factor. The remaining aboveground biomass in each plot was mowed and removed to simulate agricultural harvest practices. To provide further context to switchgrass biomass yields, all individual plants at each site were examined during the early growing season of 2020 to determine survival rates after four years of growth.
Data processing and statistical analysis
All data were consolidated into single, averaged values per plot for statistical analysis (R version 4.1.0; R Core Team, 2021) R studio 2022.07.1+554 and all data files and code are openly available in this repository. Because our data were not normally distributed, we used non-parametric tests (i.e., Kruskal Wallis and Wilcox posthoc tests with Benjamini & Hochberg p-value corrections) to compare biomass yields among cultivars and ecotypes at each site regardless of year, among sites within each cultivar, and year-to-year changes in yield. We tested variation in yield across sites and years using a modified signed-likelihood ratio test (SLRT) for equality of coefficients of variation (cv; R package cvequality Version 0.2.0; Krishnamoorthy & Lee, 2014; Marwick & Krishnamoorthy, 2019).
To characterize the soils and climates of these sites, we performed principal components analyses (PCA) separately on soil and climate data. Our soil PCA included the following variables measured in the topsoil (0–25 cm depth) before planting in 2016: BD, TOC, TN, TOC:TN, NH4+-N, NO3--N, TP, M3-P, pH, EC, CEC, % Sand, % Silt, and % Clay. For each site, topsoil values were determined by calculating the weighted average of measured values for the 0–5 cm, 5–15 cm, and 15–25 cm depth increments from each of the three soil cores with sufficient soil to analyze all variables (Table S1). Our climate PCA included the four following site-level variables: Tmax, Tmin, PRCP, and PET. Averages (and sums in the case of PRCP) were calculated for the 270 days preceding the collection of switchgrass biomass yield to represent the yearly climate data for each growing season.
To determine the effects of G´E interactions on switchgrass biomass yield, we developed mixed effects models (R package lme4; Bates, Mächler, Bolker, & Walker, 2015). The models included fixed effects of cultivar and either soil, climate, or soil and climate PC scores, as well as their interactions with cultivar. Block and year of harvest were fitted as random effects. All models were evaluated for best fit, predictive power, and overall quality using marginal R2 (mR2; Nakagawa & Schielzeth, 2013) and Akaike information criterion (AIC). To determine the relative importance of site soils and climate on switchgrass biomass yield, and whether G´E interactions occurred, we used a two-step approach. First, we fit a full model containing main effects and all interactions of cultivar with the first two soil and climate PC axes. Interactions involving PC axes from the same PCA (e.g., soil PC1 and soil PC2), which are, by definition, orthogonal, were excluded from the model. The various time scales of averaged climate data were evaluated to find the time scale which achieved the best fit. We then fit four reduced models: (1) soil PCs and their interactions with cultivar, (2) climate PCs and their interactions with cultivar, (3) soil PCs and cultivar without interactions, and (4) climate PCs and cultivar without interactions. These reduced models were compared to the full model using mR2 and delta-AIC (△AIC = AIC Reduced model - AIC Full model). All models are presented in the supplementary materials (Equations S1–S5).
To evaluate the contributions of plant leaf and tiller traits to switchgrass biomass yield, we calculated simple linear regressions between biomass yield and each plant trait. Data were analyzed by combining site and ecotype regardless of year. Plant traits included stem mass (g), leaf mass (g), specific leaf area (cm2 g-1), whole tiller C:N, tiller height (cm), and tiller number per plant.
R 4.1.0+ (open source)
Microsoft Excel (open source = LibreOffice)
U.S. Department of Energy, Award: DE-SC0014156
U.S. Department of Energy, Award: DE-SC0021126
U.S. Department of Energy, Award: DEAC02-06CH11357