The impact factors of first emergence time of V. sinensis
A total of 140 first emergence events of V. sinensis were recorded. There were no significant differences in first emergence time among the three consecutive years (Kruskal-Wallis test: H = 2.804, P = 0.246). In 49 records, V. sinensis started to emerge before sunset at dusk. Differences between sunset and first emergence ranged from −36 to 33 min (4.49 ± 13.23). Here, the negative values meant that first emergence events started before sunset.
Our analyses indicated that the best model describing the variation in first emergence time of V. sinensis used LIFE, LISS, and LS as the predictor variables (Table 1). Further, model averaging revealed that the three predictive variables were also significantly associated with the first emergence time ofV. sinensis (Table 2). Differences between first emergence time and sunset time were significantly and negatively associated with light intensity of first emergence (linear regression: t = −11.447,P < 0.001, Fig. 1b; curvilinear regression: t = 13.728, P < 0.001, Fig. 5a). In other words, the higher the light intensity at first emergence, the earlier the first emergence of V. sinensis. Moreover, differences between first emergence time and sunset time were significantly and positively associated with light intensity at sunset (t = 7.801, P < 0.001; Fig. 1a). Finally, significant differences in the first emergence time were detected between lactation and post-lactation (t = −4.779,P < 0.001; Fig. 1c). During the lactation period, V. sinensis departed earlier from the roost than during the post-lactation period (Fig. 1c). Specifically, 65.67% of first emergence events (44/67) started before sunset during the lactation period. In contrast, only 6.85% of first emergence events (5/73) were observed before sunset during the post-lactation period. A Chi-square test of independence showed that the first emergence events had a higher probability of occuring before sunset during the lactation period (χ 2 = 26.34,P = 2.86×10-7). Consistent with the results of the hierarchical partitioning analysis, the independent contributions of LIFE, LISS, and LS were 53.86%, 25.54%, and 14.17%, respectively (Fig. 4a).