Seed dormancy and germination play an important roles in regulating germination timing and plant fitness in harsh environment, especially for fragile karst region. Broadening the knowledge regarding seed ecology of native plant species can improve local restoration management due to the increased biodiversity and success of seedling establishment, yet few studies have addressed this. We examined the dormancy types and germination responses to temperature regimes as well as light conditions of totally 27 that were dominant or common species inhabiting a degraded karst mountain from central Yunnan-Guizhou Plateau and assessed the relationships between seed traits and germination index using a partial least squares regression (PLSR). Approximately 48% of the investigated species exhibited physiological dormancy, 37% exhibited nondormancy, 7% exhibited morphophysiological dormancy, 4% exhibited morphological dormancy and 4% exhibited physical dormancy. Germination behaviors occurred in 16 of 27 species. The seed germination of 13 species strongly responded to seasonal temperature regimes, whereas 12 species were substantially affected by light conditions. About 94% species germinated to the maximum percent in warm temperature regimes (20/13 and 25/18 °C), while the remaining was in cool temperature regime (10/4 °C). The PLSR analysis indicated a significant positive correlation between seed mass and T50 m (time to 50% final germination), and a negative correlation between seed moisture content and percent germination, showing that small seeds, particularly those with a low moisture content, were more likely to have a greater germination capacity (including relative less germination time and higher germination percent) than large seeds in harsh karst environments. In general, these results strengthen the understanding of seed dormancy classification of dormant species and germination requirements of nondormant ones that representative of pioneer species of early stage of positive succession in seriously degraded sites and provide suggestion for species selection in local seed-based vegetation restoration.

To test the main effects (temperature and light conditions) and their interactions on the final percent germination, generalized linear models (GLMs) with a binomial error distribution and logit function were adopted. All models were simplified by using a stepwise backward selection of main factors and respective interaction terms until significant factors/interaction terms remained. A partial least squares regression (PLSR) with full cross-validation was used to relate the variation in a single-component response variable to the variation in multicomponent explanatory variables.