Phenotypic Measurements and Statistics
We measured 12 morphological traits of different A. viridiflorapopulations in the experimental field during the full flowering stage.
To ensure the accuracy of the measurements, we randomly selected 5
plants from each population to measure the number of inflorescences and
plant height; among these 5 plants, 6 flowers and 6 leaves were randomly
selected from each plant, and the corolla diameter, pistil length,
stamen length, leaf area, leaf perimeter and chlorophyll content were
measured and recorded. Among these 6 flowers, we randomly selected 3
petals and calyxes to measure the petal length (not including spurs),
calyx length, spur length, angle between the petals and spurs (referred
to simply as angle hereafter) and calyx length.
To reduce the error caused by different measurement batches, we used a
mixed linear model to evaluate traits in the lme4 (Bates, Mächler,
Bolker, & Walker, 2014) package in R according to the following
regression model equation:
\begin{equation}
Y_{i}=\mu+\beta_{1}+\beta_{2}P+\varepsilon_{i}\nonumber \\
\end{equation}where \(Y_{i}\) represents the traits of different populations, \(\mu\)is the actual measurement, \(\beta_{1}\) and \(\beta_{2}\) are
regression coefficients, A represents the measurement batches, P is the
person conducting the measurement, and \(\varepsilon_{i}\) is the
residual variance. The evaluation results were used for ANOVA and
K-means cluster analysis in R.