flwr_traits_2019.csv

Dataset created to compare floral morphometric traits and flower nectar quantity and quality of three Cucurbita sister species. Staminate and pistiallate flowers were collected in the states of Jalisco, Guerrero and Michoacan in Mexico. 

To determine whether floral attributes differ between cucurbit species, we compared the size of each floral trait using a generalized linear mixed model (lme4 function ‘glmer’, Bates 2010, and stats function ‘glm’, R Core Team 2020) followed by a post hoc Tukey test (emmeans function ‘emmeans’, Lenth 2021). Floral trait measurements were log10 transformed to obtain normality. Pollinator attraction traits corolla (CD, TL, CL, TD1, TD2, TD3; Fig. 1A and B) were evaluated individually as response variables and the interaction between cucurbit species and floral sex as fixed factor. Plant reproductive traits, androecium (AD, StL, AL; Fig. 1C), gynoecium (SD, SL, PL, OL, OD; Fig. 1D) and nectary diameter (NDm and NDf; Fig. 1C and D), were evaluated individually as the response variables and the cucurbit species as the predictor variable. Field sites were set as a random variable to account for site level variation. Because multiple flowers were measured from the same plant, plant ID was nested within the site random variable to account for individual level variation within each plant at each site. Significance level was found using type III Wald chi-square tests (car function ‘Anova’, Fox and Weisberg 2019). Additionally, to analyze the individual variation of each floral trait, we calculated the coefficient of variation (CV = s / x̅) for one randomly selected flower from each plant. All floral attribute analysis were completed in the R program (version 4.0.1) (R Core Team, 2020).

 To analyze whether volume and sugar concentration of nectar differs between cucurbits species, floral sex, and time, we performed generalized linear mixed models (nectar volume: lme4 function ‘glmer’, Bates 2010, nectar sugar concentration: glmmTMG function ‘glmmTMB’, Magnusson et al. 2018) followed by a post hoc Tukey test (emmeans function ‘emmeans’, Lenth 2021). To normalize data, we converted nectar volumes using log10 and sugar concentrations were converted to proportion. We used nectar volume and sugar concentration as response variables and the interaction between cucurbit species, flower sex and time as predictive variables (emmeans function ‘emmeans’, Lenth 2021). Time was transformed to decimal time (time x 1440) for analysis. Because some pistillate and staminate flowers were sampled from the same plant, Plant ID was used as a random variable which was nested within the field site to control for site level variation. For nectar volume, we used a normal distribution and for sugar concentration we used a beta distribution

本数据集用于比较3个南瓜属（Cucurbita）姊妹物种的花形态计量性状、花蜜数量与质量。研究于墨西哥哈利斯科州、格雷罗州和米却肯州采集了雄花与雌花。

为明确南瓜属物种间花性状是否存在差异，本研究采用广义线性混合模型（generalized linear mixed model），通过R语言的lme4包`glmer`函数（Bates, 2010）以及stats包`glm`函数（R Core Team, 2020）对各花性状的大小进行比较，随后开展事后Tukey检验（emmeans包`emmeans`函数，Lenth, 2021）。为满足正态性要求，所有花性状测量值均经log10转换。

以传粉者吸引相关性状——花冠（CD、TL、CL、TD1、TD2、TD3；图1A、B）分别作为响应变量，以南瓜属物种与花性别的交互作为固定因子开展分析。以植物繁殖相关性状——雄蕊群（AD、StL、AL；图1C）、雌蕊群（SD、SL、PL、OL、OD；图1D）及蜜腺直径（NDm与NDf；图1C、D）分别作为响应变量，以南瓜属物种作为预测变量。将样地设置为随机效应以消除样地水平的变异。由于同一植株会被采集多朵花，因此将植株ID嵌套于样地随机效应中，以控制单株在同一样地内的个体水平变异。显著性通过III型Wald卡方检验确定（car包`Anova`函数，Fox & Weisberg, 2019）。

此外，为分析各花性状的个体变异，我们对每株植物随机选取的一朵花计算了变异系数（CV = s / x̅，其中s为样本标准差，x̅为样本均值）。所有花性状分析均在R程序（版本4.0.1）中完成（R Core Team, 2020）。

为分析花蜜体积与糖浓度在南瓜属物种、花性别及时间梯度下的差异，本研究采用广义线性混合模型（花蜜体积：lme4包`glmer`函数，Bates, 2010；花蜜糖浓度：glmmTMB包`glmmTMB`函数，Magnusson et al., 2018），随后开展事后Tukey检验（emmeans包`emmeans`函数，Lenth, 2021）。为实现数据正态化，对花蜜体积进行log10转换，将糖浓度转换为比例值。以花蜜体积与糖浓度作为响应变量，以南瓜属物种、花性别与时间的交互作为预测变量。分析前将时间转换为十进制时间（时间 × 1440）。由于部分雌花与雄花采自同一植株，因此将植株ID作为随机效应，并嵌套于样地中以控制样地水平的变异。花蜜体积分析采用正态分布，糖浓度分析采用β分布。