How environmental stressors affect reproductive potential in a saltmarsh plant species Plantago maritima

Samples were collected during the first week of September 2016 in the three populations of Plantago maritima. We selected a set of 72 plots that differed in the types of management: sites lightly mowed – mowed once/twice a year (MowL), lightly grazed – sites avoided by cattle, rarely grazed with higher vegetation (GrazL), heavily grazed – sites preferred by cattle with short vegetation (GrazH), heavy trampled – sites along the pathways intensively used by tourists (TramH), old marks of wild boar rooting – uneven surface, but almost completely overgrown by plants (RootOld), fresh marks of rooting – uneven surface with a significant share of bare soil (RootNew). Infructescence density per one sq. metre was calculated on every plot. Additionally, 254 infructescences (Table 1) were collected randomly from the plots, stored in envelopes and air dried for two weeks. After drying, the length of each infructescence was measured with a resolution of 1 mm and the number of fruit capsules were counted. The average length of infructescences was multiplied by their number per 1 sq. m in order to estimate the potential for seed production per site (seed yeald). Germination test and growth experiment From each infructescence, a total of 30 randomly chosen seeds were weighed with a resolution of 1/100 mg. The seeds were subjected to stratification at 5 ˚C for 60 days, then planted in a growth chamber in germination tanks on filter paper. The germination regime was set with a 25 ˚C light(18h)/20˚C dark (8h). Light was supplied from warm white fluorescent tubes. The constant high moisture of the substrate was sustained with deionized water as it is known of Plantago maritima that its seeds germinate best in distilled water (Lotschert, 1970).The effects of germination – the number of emerging sprouts and their condition were checked daily for a period of 14 days as it was determined that this period was enough for an average seed to germinate completely, and after that period almost no new sprouts appeared. The seeds were considered to have germinated when the radicle and two cotyledons were present. Mean germination time (MGT) was calculated by using the equation: MGT = ∑ (n × d) /N, where n = number of seeds germinated on each day, d = number of days from the beginning of the test, and N = total number of seeds germinated at the termination of the experiment (Ellis & Roberts, 1981). After germination, a representative number of sprouts (Table 1) for every variant of management were grown in a separate pots filled with organic soil. After a period of two months, just before the plants started allocating material to reproductive organs, the specimens were harvested, separated into underground and aboveground parts, and their dry mass was measured. Statistical Analysis To test the influence of the environmental stressors on each of the response variables eight multivariate linear regression models were performed. The statistical significance level of all regression coefficients and models were established as α = 0.05. All of the independent variables were coded binary, in which "1" represents the influence of a particular stressor and "0" – the absence of a particular stressor. In some cases, the response variables were influenced simultaneously by more than one stressor (by two: in 39 cases by bough MowL and GrazH, in 4 cases - MowL and RootOld, in 4 cases - MowL * RootNew, in 4 cases - GrazH * RootOld, in 4 cases - GrazL * RootNew; by tree: in 4 cases - MowL * GrazL * RootNew) therefore all the stressors were tested simultaneously in one linear model. Depending on the skewness, the response data were normalized by logarithm log(x+1), square-root or square transformation (compare Table 2). To describe the importance of the environmental stressor influence, the percentage of each explained variation was calculated. Statistical analysis were performed in R (R Core Team, 2017), data visualisations were performed using ‘ggplot2’ (Wickham, 2016) and ‘ggridges’ packages (Wilke, 2018) in R studio (RStudio Team, 2015). The correlations between measured variables were tested with the Pearson’s coefficient.