2.4 Data processes
All our filmed pictures are sufficiently stable to distinguish each behavioural state and long enough to meet at least 5 samples of vigilance/non-vigilance bouts for statistical effectiveness, even for our shortest video sampling of 2.2 min, as scan and inter-scan duration for black-necked crane are 4s and 20s on average (Li et al., 2017). So, all our data were taken into subsequent analysis. Vigilance behaviour was evaluated by both vigilance proportion and duration for each group member. Vigilance proportion was calculated as percentage of time spent on vigilance during the observation secession while vigilance duration as the average time span (seconds) of each vigilance bout (Li et al., 2017). Studies documented that group members could coordinate individual vigilance in order to increase group collective vigilance (Rasa, 1986; Bednekoff, 2015), and in small crane groups (Ge et al., 2011). We focused on the collective vigilance time pattern (time serials) of two adults in family groups, as no vigilance interactions between juveniles and adults were detected (Ge et al, 2011; Che et al., 2018).
We classified the areas where cranes occurred into low, intermediate and high disturbance levels according to tourist accessibility. Cranes distributed along the road to Dahaizi lake and Jigong mountain where nearly 90% tourist visited facing the highest disturbance, moderate disturbance in the areas around Tiaodunhe lake with less 30% public visitation, and lowest disturbance in the other areas. We considered the observer (data collector) as proxy of potential tourists, the distance from observer to cranes is another effect variable on vigilance. Only one observer approached crane without aggressive behaving e.g., shouting, chasing, means the lowest disturbance from tourist. Because inexperience always correlated with high predation vulnerability, a higher proportion of juveniles in groups usually indicated higher predation risk (Xu et al., 2013; Beauchamp, 2015). We considered our study objectives of three family typs with two adults and 0-2 juveniles as three levels of predation risk; and adults without juveniles endure the lowest predation risk while families with two juveniles have the highest predation risk.
We found that our data sets deviated from normality in a one-way Kolmogorov-Smirnov test, and arcsine square root transformation for vigilance proportion and logarithmic transformation for vigilance duration were subsequently conducted in order to get normalized data for parametric tests. Comparisons of vigilance difference between adults and juveniles and observed to expected collective vigilance of two adults were accomplished with t-test. One-way ANOVA was used to test both individual and collective vigilance difference of adults in three family types. We selected general linear model to distinguish effects of disturbance level (categorical variable of three levels), observer distance (continuous variable) and predation risk (categorical variable of three family types) on both individual and collective vigilance deviation of black-necked crane (Ge et al., 2011). Collective vigilance deviation was used to determine collective vigilance pattern (synchronization or coordination) by considering the deviation between expected (independent vigilance) and observed collective vigilance (Pays et al., 2007a, b). Expected collective vigilance was calculated with the equation of 1-[(1-p1 ) * (1-p2 )], where p represents vigilance proportion of two adults in a family (Pays et al., 2007a, b; Ge et al., 2011). Statistics were accomplished with IBM SPSS 20.0 software with a two-tailed significant level of p<0.05.