Figure 1: Site map showing where reproductive shoots were collected. In Lake Macquarie there is two thermally affected locations which are illustrated by red markers and a power station image. Seagrass mapping data sourced from NSW Department of Primary Industries.

Reproductive shoots were collected over two different reproductive seasons, one during 21/22 spring/summer and one during the 22/23 spring/summer. At thermally affected sites (Myuna Bay and Mannering Park in Lake Macquarie), reproductive shoots were collected late October, while at ambient locations reproductive shoots were collected late November. The difference in collection times was due to the likelihood of temperature affecting seed maturation times (Smith et al. 2016). Reproductive shoots were transported to the laboratory in a 30 L tub filled with seawater from the collection site. Shoots were then placed outdoors in a shaded area in aerated 50 L containers with natural seawater (34 ppt) sourced from Swansea Channel, the entrance to Lake Macquarie. After maturation (~4 weeks after collection) negatively buoyant seeds were siphoned from the bottom of the tanks and stored in autoclaved seawater (sourced from Swansea Channel) at 4 °C for a maximum of 2 days before being setup in an experiment. To simulate a freshwater pulse and induce germination, seeds were moved from storage, placed into a 30 ml sample jar of distilled water, stirred lightly for 5 minutes and then stored for 24 hours to assist in initiating germination (Stafford-Bell et al., 2016; Cumming et al. ¸ 2017).

To determine viability, seeds (n = 100) from each site except Gwandalan 21/22 (n = 50) were subject to a cut test prior to being placed into the germination experiment (Smith et al., 2016). The cut test is an efficient (both time and economically) test which involves slicing the seed in half and deems seeds which are firm and have a blue tinge to be viable (Borza et al., 2007; Smith et al., 2016) highlighted no difference between tetrazolium staining and a cut test in determining seed viability. However, to ensure the cut test was being completed correctly, tetrazolium staining was also performed on four sites (Myuna Bay, Mannering Park, Murrays Beach and Yarrawonga Park; n = 100) and compared by ANOVA which found no significant differences in viability between the two tests. Consequently, in the second year only the cut test was performed. Unfortunately, in Brisbane Water and Tuggerah Lakes tanks, aerator failures resulted in usable seeds from only one site from each estuary which were Saratoga and Pipeclay Point respectively. This meant that the original design of n = 3 control sites in two separate control estuaries was not accomplished, and because of no significant differences between estuaries, ambient sites from Tuggerah Lakes and Brisbane Water were pooled with ambient sites from Lake Macquarie.

For each site, 100 seeds were randomly allocated to five petri dishes (n = 20 per petri dish). However, insufficient seeds were collected at Gwandalan in 2021, so three replicates (n = 14) were placed in the 16 °C and 20 °C treatments only. Likewise, Saratoga Point only had sufficient seeds for 5 replicates (n = 20) in the 16 °C treatment. Each petri dish was filled with approximately 5-10 ml of sterilized seawater sourced from the entrance of the lake on a single sampling occasion (34 ppt) and fitted with an autoclaved sponge to reduce evaporation and wet-strengthened filter paper to ensure the seeds did not move for the duration of the experiment (Tol et al. 2021). Seeds were then placed into plastic containers blacked out with aluminum foil and randomly assigned to one of four (16 °C, 20 °C, 24 °C, 28 °C) temperature control cabinets which allowed no ambient light. Excess seeds from five of the sites were used in an additional five replicates (n = 20) to be held at 8 ppt salinity in the 16 °C and 28 °C treatment to investigate any effect of salinity. In the 16 °C treatment, this included all sites except for Gwandalan and Saratoga and in the 28 °C treatment it included Mannering Park, Myuna Bay and Pipeclay Point. Seeds were checked for germination every 2-3 days and evaporative loss topped up when required with saline solution (Cumming et al., 2017). Water was siphoned and changed fortnightly or earlier in the case of unusual evaporation to ensure salinity did not fluctuate. Temperature treatments and salinity were based on likely scenarios within the estuaries where the seeds were sourced. On conclusion of the first year of experiment, seeds were subjected to another freshwater pulse for 48 h and results recorded for separate analysis.

To determine the effect of seed location (Lake Macquarie – ambient and thermally affected; Brisbane Water – ambient, Tuggerah Lakes - ambient) on mean time to germination (MTG), cox models were fit to seed germination data using the R package ‘survival’ (Therneau, 2022) and significant differences identified using estimated marginal means with the ‘emmeans’ package with a Tukey adjustment (McNair et al., 2012; Tol et al. , 2022; Lenth, 2022). Because there were no differences detected between ambient treatments regardless of estuary, all seeds were grouped into either ‘ambient’ or ‘thermally affected’ treatments. Non-proportional hazards were checked using Kaplan-Meier survivorship functions and the model with the lowest Akaike information criterion (AIC) was used (McNair et al., 2012). Germination analyses did not include the freshwater pulse in the first year which concluded the experiment to avoid inflating the original germination rate.

To determine the influence of temperature treatments and location of seeds on seed germination, a multiple logistic regression model was fit to individual seed data. To ensure the model fit the data and to determine which covariates or interactions (Site, Location, Treatment ) significantly influenced the model, deviance tables were constructed on various models and compared using ANOVA with a Chi square test. Multi-collinearity was tested by calculating the variance inflation factors (Heiberger and Holland, 2015; Cumming et al., 2017). Factors which did not significantly influence the model or suffered from multicollinearity were identified and removed from the model (Site and the interaction of Location and Treatment). As in the cox model, pairwise differences were identified using estimated marginal means with the ‘emmeans’ package with no adjustment (Lenth et al., 2022). The MASS package (Venables and Ripley, 2002) was then used to calculate e^β and 95% confidence intervals to quantify the influence of location on seed germination.

To compare means of dead seeds on experiment completion and maximum mean germinations pre and pulse post and any interaction between location and pulse, data was first analysed for homogeneity of variance which was confirmed by plotting residual values vs fitted values and normality using a Q-Q plot. ANOVA was then used with a Tukey’s HSD post hoc to determine where significant differences occurred (Z ar, 2005).