Study area
Shengjin Lake (30°15′–30°30′N, 116°55′–117°15′E) is located in the middle and lower Yangtze River floodplain, one of the main wintering grounds for Hooded Cranes in China (Fang et al. 2006; Cao et al. 2008). Our study site is located in the upper part of Shengjin Lake where most of the cranes gather and abundant aquatic food sources for the cranes are found (Cheng et al. 2009). At Shengjin Lake the traditional foods of the Hooded Crane are submerged macrophyte tubers (Barzen et al. 2009), but more recently the cranes have also been feeding on Potentilla supine, Ranunculus polii and the like. During the autumn and winter drawdown, mudflats are exposed and the waterbirds feed on Anodonta woodiana (Zhang and Lu 1999; Fox et al. 2008). As a consequence of wetland degradation and habitat loss, the carrying capacity of these wetlands has decreased and the traditional food sources of these wintering waterbirds have experienced a sharp decline in quantity (Li et al. 2015). The cranes have shifted habitat from mudflats to farmland, such as harvested paddy fields with rice grains, to look for supplementary sources of food (Yang et al. 2015).
Habitat survey
We selected six sampling sites covering the three habitats in the main foraging areas of the Hooded Crane in the winter of 2014/2015 (Fig. 1). Each sampling site measuring 150 m × 150 m, was chosen to estimate food density for the specific date of each observation of foraging effort and foraging success. We estimated daily food-density values for each plot from regression equations. Our cranes mainly exploit Anodonta woodiana, mollus and small amounts of tubers of wetland vegetation from the bare substrate of the lake bottom in the mudflats. Potentilla supina, Ranunculus polii and Polygonum criopolitanum with high digestibility were obtained from meadows and rice grains (Olyza sativa) from paddy fields (Zheng et al. 2015). To investigate food density, we sampled the plots three times, i.e., at the arrival of the cranes at the lake (during the fall from 23 September to 17 October 2014), in the middle of the winter (from 5 to 25 January 2015) and at the departure of the cranes from Shengjin Lake (in the spring from 1 to 12 April 2015). Quadrats measuring 0.50 m × 0.50 m were excavated to a depth of 0.30 m, below which tubers were considered inaccessible to foraging cranes (Lovvorn 1989; Richman and Lovvorn 2003). During the three over-wintering periods, a total of 381 samples from these quadrats were collected, of which 43, 62 and 30 in meadows, 25, 29 and 55 in paddy fields and 47, 60 and 30 in the mudflats from the early, middle to late wintering stages, respectively (Gillespie and Kronlund 1999). The locations of the fall quadrats were not resampled in the spring (Lewis et al. 2007).
Behavior sampling
We collected behavioral data from November 2014 to April 2015. Hooded Cranes were located during regular-route surveys and locations were recorded with a GPS. Routes were never repeated on the same day to avoid pseudo replication. Focal samplings were carried out using binoculars (8×) or a telescope (20–60×). We defined November and December as the early wintering period; January and February of the following year as the mid-wintering period and March and early April as the late wintering period (Zheng et al. 2015). At the start of each focal observation, the location, date, time of day (morning, noon, afternoon), habitat type (meadows, mudflats, paddy fields), winter stage (early winter, mid-winter, or late winter), age (adult or juvenile) and family size (including 0, 1 or 2 juveniles) were recorded. A digital voice recorder was used to record behavioral events for 20 min unless we lost sight of the individual focal bird. We concentrated on studying the foraging behavior of Hooded Cranes, specifically their behavior in searching for food, handling food and swallowing.
Statistical analyses
In total, 397 behavioral observations were used in the analyses of foraging efforts. Foraging effort is defined as the ratio of total amount of time spent searching for and processing food and the activity time budget. If the head and neck of a crane lift backwards twice in a quick high-frequency tic when foraging and they then swallow, this means they have had foraging success. The rate of foraging success is defined as the ratio of the amount of food captured and the time required for foraging. We fitted a set of candidate generalized mixed linear models (Littell et al. 2000) in SPSS 19. We used a quadratic relationship (date + date2) for the date (Date), where Date was scaled so that 1 November 2014 = 1 to account for well-established patterns in foraging efforts that vary nonlinearly over the winter (Guillemette 1998; Fischer and Griffin 2000; Systad et al. 2000). Age and family size were considered in combination and referred to as “individual effects” (Individual) (Li et al. 2013). Time of day, habitat and winter period were referred to together as “Environmental effects” (Environmental). Date, individual, environmental and food density (Density) were used as units for the construction of the model.
Candidate models were made up of all single-variable models and additive models. The null model contained only the intercept and had no interactions, because there is no strong ecological explanation. For all models, the number of parameters (k) included intercept and variance estimates. Covariance structures for repeated measures included one parameter (k) for all models except for the null model. The foraging success rate may have included repeated observations of an individual crane. We allowed individual cranes to be random factors in order to control for repeated measurements. In the final model, we only included the habitat types (meadow, mudflat, paddy field), age (adult, juvenile), family size (2, 3 or 4 members), time of the day (morning, noon, afternoon), winter stage (early winter, mid-winter, or late winter), date and density. We used the method of information theory to guide the selection of the model and Akaike’s Information Criterion (AIC) to calculate the value of each model. We used the relative importance of each AIC and their weight, w
i
(i = 1, … 16) for drawing inferences from each model. To calculate the relative importance of each explanatory variable in a candidate model, we added the weights of all explanatory variables contained in each candidate model (Kuwae et al. 2010).
We studied the responses of foraging behavior to changes of food density and disturbances by analyzing their rate of feeding success and feeding effort. According to the frequency and distance of disturbances observed during the survey, we divided the intensity of disturbances into three grades (Jiang et al. 2007; Yang et al. 2015). The data of density, disturbances, feeding success and foraging effort were tested for normality by using the one-sample Kolmogorov–Smirnov test. If the data followed a normal distribution we used a one-way ANOVA or a t test for our analyses; if the data did not follow a normal distribution we used the non-parametric Kruskal–Wallis H and the Mann–Whitney U tests. As well, the effects of food density and disturbances on the foraging behavior were analyzed in a generalized linear model.