Weather variables and time frames prior to the start of vegetative and reproductive growth used to construct the hypothesis-based covariance matrices. Covariance matrices were constructed for, and tested on, three subpopulations and two phenological traits. The correlations between values of these weather variables for genetically identical plants grown in different gardens were used to fill off-diagonal cells of the covariance matrices. Narrow-sense heritabilities for these values at each garden were used for the diagonal cells.

Characterization of the timing of the onset of vegetative (green-up) and reproductive (flowering) growth in the switchgrass diversity panel. (a) Map and trait histograms of green-up and flowering dates across two genetically distinct switchgrass subpopulations and eight common gardens. Purple represents individuals from the Midwest genetic subpopulation, and pink individuals from the Gulf subpopulation; map positions represent the original collection locations for the genotypes, and shapes represent the ecotype of the genotype. Histogram vertical dashed lines indicate the summer solstice. Common gardens are arranged in latitudinal order. (b) Phenotypic correlations between clonal replicates planted at eight common gardens, within and between two genetic subpopulations.

Example Canonical and GxWeather covariance matrices specified in mash and the posterior weights placed on all covariance matrices. (a) Common gardens are arranged in latitudinal order within the matrices. Top row: Four example canonical covariance matrices. Canonical matrices (purple) have simple interpretations, such as equal effects across all common gardens, or effects specific to a single common garden. Bottom row: Five example GxWeather covariance matrices specified for the green-up date or flowering date phenotype; these matrices were created from environment-specific correlations across eight common gardens, and are described in Table 1. (b,d) Total posterior weight placed on each covariance matrix type specified for (b) green-up date and (d) flowering date mash models, within and between two genetic subpopulations. Covariance matrices included in mash that had zero posterior weight in all three mash runs on the genetic subpopulations, such as the identity matrix, are not shown. (c,e) Total posterior weight placed on covariance matrices that were GxWeather or Canonical, for the (c) green-up date phenotype and (e) flowering date phenotype.

Types of GxE present between of pairs of jointly re-estimated SNP effects in eight common gardens, for effects with lfsr < 0.05 at both gardens for each pair of gardens contrasted. a) Examples of effect patterns at three pairs of sites with three types of GxE. All effects are for an alternate allele, with the reference allele effect defined to be zero at both gardens and represented by the dashed vertical line. Sign: Effects that differ in sign at these pairs of gardens (p < 0.05, lfsr). Magnitude: Effects identical in sign (p < 0.05, lfsr) that differ in magnitude by a factor of >0.4. Not Distinguishable: Effects not distinguishable by magnitude nor sign of the effect, with no measurable GxE. Confidence intervals are illustrative that the effect estimate does not overlap zero. b) The fraction of effects with each GxE type for the onset of vegetative growth (green-up date) and reproductive growth (flowering date), within and between two genetic subpopulations. Common gardens are grouped by the larger region they came from: North gardens are within the natural range of the Midwest subpopulation, and include MO, NE, MI, and SD, while Texas gardens are within the natural range of the Gulf subpopulation, and include TX1, TX2, and TX3.

Overlaps of QTL from an outbred pseudo-F2 cross and with jointly re-estimated SNP effects in the 1% tail of significance from a diversity panel. Dotted lines indicate permutation-based significance thresholds for each weather-related function. Stars indicate QTL with significant enrichment for SNPs in the 1% mash tail; G, M, and B indicate which subpopulation had enrichment: G - Gulf subpopulation, M-Midwest subpopulation, B - both subpopulations. Rug plots show genomic locations of SNPs in the 1% mash tail for flowering date for each subpopulation: Both subpopulations are above the plot panel, the Gulf subpopulation is above the x-axis, and the Midwest subpopulation is below the x-axis. a) QTL mapping for the onset of vegetative growth (green-up date), and three weather-related functions of green-up date. b) QTL mapping for the onset of reproductive growth (flowering date), and two weather-related functions of flowering date.