Monitoring
Biweekly field monitoring of seed germination was conducted over a year from April 2020 to May 2021. During each monitoring session, we recorded and labelled each seed germinated. Overall, over 90% of the geminated seeds were recorded within 1-3 months after seed sowing. Very few seed germination (< 5%) was observed beyond 6 months after seed sowing.
2.4 Statistical analysis
We used generalized linear mixed models (GLMM) to analyse the results of the seed coating and surface soil treatment experiments. These were performed using R version 3.6.3 with the ‘nlme’ package (Pinheiro et al., 2021). We used germination rate as the response variable in both experiments, while the three seed coating treatments (clay-dominant, biochar-dominant and control) and two surface soil treatment (biochar and control) were considered as fixed covariates respectively. In order to incorporate the dependency among observations in the same site and nested structure of our experiment, we used site and species as random effects. We did not include any interaction term due to small sample size. Moreover, due to the poor seed viability, some of the species such as Eurya nitida, Tetradium glabrifolium and Zanthoxylum avicennae failed to germinate in both nursery and field trials and were excluded from further analyses.
We separated the analysis into two parts: 1) across all species and 2) each species independently. For all species, we ran four models by including different covariates and selected the model with the lowest Akaike information criterion (AIC) value (Table 2). For the species models, we modelled the relationship between germination rate and treatment with site as a random factor. Model assumptions of normality and homogeneity of variances were verified by plotting model residuals. Post hoc analyses for pair-wise comparisons of means were undertaken using least square means using the lsmeans package (Russell, 2016).
Additionally, to test the effect of seed size and successional stage on germination rate of native seeds on landslide scars, data from the controls of the two experiments were pooled and analysed. We used linear mixed models to compare seed germination rate of different successional groups and seed sizes. All analyses were conducted using R version 3.6.3 (R Core Team, 2020) for GLMM.