Animal capture and GPS-GSM deployment
From the East Asia Continental population, we captured 68 Greater White-fronted Geese between October and March on their wintering grounds at Poyang Lake (29° 07′ N, 116° 16′ E) and Chenyao Lake (in Anhui lakes, 30° 54′ N, 117° 40′ E) in the Yangtze River floodplain, China from 2013 to 2017. From the West Pacific population, a further 70 GWFG were caught in July and August during the moulting period in the Chaun (68° 53′ N, 170° 58′ E) and Indigirka Deltas (70° 45′ N, 151° 28′ E) in Russia during the summers of 2017 to 2019. From the North Sea-Baltic population, 18 GWFG were caught in their Dutch and German wintering areas and during flightless moult on Kolguev Island in Arctic Russia between 2013 and 2017.
Birds were fitted with a variety of telemetry devices (Druidtech, China 35 g mounted on neckbands, Hunan Global Messenger Technology Company, China, 26 g or 27 g, mounted on neckbands or using back packs, Ornitela, Lithuania, 38 g, mounted on neck bands, E-obs, Germany, 48 g, using back packs, Konstanz University, 35 g, mounted on neck bands, madebytheo, Netherlands, 38 g, mounted on neck bands) providing GPS positions to within 10 m accuracy via the Global System for Mobile Communications (GSM) and GPRS mobile telephone networks. These devices provided 1 to 288 GPS positions per day, depending on tag capacity and battery conditions (dependent on absorption of solar radiation by the in-built solar panel). Device failure, low battery power levels and signal loss often hindered the accumulation of regular and precise data, especially in winter when day length was short. We therefore only used tracks from which we obtained 1 to 288 fixes per day along the entire length of the migration routes between breeding and wintering areas. These left tracks of sufficient precision to compile 20 complete autumn migration tracks (3 in 2015, 8 in 2016, 5 in 2017 and 4 in 2018) and 23 complete spring migration tracks (3 in 2015, 9 in 2016, 5 in 2017 and 6 in 2018) for the East Asia Continental population; 57 complete autumn migration tracks (23 in 2017, 14 in 2018, and 20 in 2019) and 17 complete spring migration tracks (9 in 2018 and 8 in 2019) for the West Pacific population and 29 complete autumn migration tracks (4 in 2013, 3 in 2014, 3 in 2015, 9 in 2016, and 10 in 2017) and 24 complete spring migration tracks (4 in 2014, 3 in 2015, 7 in 2016, and 10 in 2017) for the North Sea-Baltic population (for full details of the individuals involved and their devices, see Additional file 1: Table S1).
To achieve segmentation of movement tracks and identify staging/stopover sites, we followed the methods of Wang et al. (2018) as follows. Firstly, we modified the first passage time method to achieve movement track segmentation using the modified methods of Lavielle (2005), Barraquand and Benhamou (2008), Le Corre et al. (2014) and Edelhoff et al. (2016). Secondly, we identified individual migration/stopover periods and sites by applying the net squared displacement and minimum convex polygon techniques of Mohr (1947) and Bunnefeld et al. (2010). See Wang et al. (2018) for full details. We defined the departure date as the date of the first position when the individual departed from wintering/moulting/breeding sites and was judged by the methods above to have acquired flight status. Arrival date was defined as the date of the first time when the individual was judged to have arrived at wintering or breeding sites, based on the methods above to qualify as non-flight status after a period of flight. A site where an individual bird spent over 2 days for resting and feeding during migration was defined as stopover site (Kölzsch et al. 2016).
Home range estimation
We defined the kernel areas used most frequently by geese during summer, winter and at staging/stopover sites as derived above. In those areas, then the land cover composition (habitat use) and the degree of nature conservation protected areas (degree of protection) within which these fell were derived.
We used Brownian Bridge Movement Models (BBMM) to calculate home range kernels. This method not only considers each activity center, but also the movement path of the animal, effectively avoiding the unused area between the activity patches as the animal’s home range (Bullard 1991; Horne et al. 2007). Compared with other models, the BBMM can more reasonably deal with the problems of spatial autocorrelation and unequal time intervals between fixes. In addition, the parameters within the BBMM model have ecological significance, taking the animal’s moving speed and measurement site errors into account (Bullard 1991; Powell 2000).
We applied the BBMM method to generate individual home ranges (50% Utilization Distribution) of each of the “stationary” segments of each track identified above, including the wintering, summering, and stopover periods during both migration seasons. BBMM was calculated using the “adehabitatHR” package (Calenge 2006) in R 3.6 (Team RDC 2017). To differentiate between nighttime and daytime habitat use, we separated the “stationary” segments by the local time of sunrise and sunset. “Daytime” was defined as the time of GPS points between one hour before sunrise and one hour after sunset and “nighttime” the time of GPS points between one hour after sunset and one hour before sunrise next day.
Analysis of habitat use
For all locations in China, we determined land use from the 2015 China Ecological Remote Sensing Survey and Evaluation Data Set (30 cm × 30 m accuracy) released by the Chinese Academy of Sciences and the Ministry of Ecology and Environment Department. This data set contains 6 types of Class I land type: Forest, Grassland, Cropland, Wetland, Artificial surfaces and “Other”. For the remaining locations outside of China, we used FROM-GLC10 2017 Data Set (10 m × 10 m accuracy) created by Tsinghua University (Gong et al. 2019). This system defines 10 types of Class I land Type: Forest, Grassland, Cropland, Water bodies, Wetland, Artificial surfaces, Permanent snow/ice, Shrubland, Bare substrate and Tundra.
We treated level I category 1 “Forest” for the purposes of this investigation including forest and shrubland, category 2 as “Grassland”, category 3 as “Cropland”, category 4 “Wetland” as natural wetland habitat including water bodies and wetland, category 5 as “Tundra”, then combined all other classes into an “Other” category. For the purposes of this analysis we have amalgamated the categories 2 and 3 and treated these as farmland, since most grassland habitats in the temperate zone are grazed pastoral systems.
Because the tracking data of the three populations were gathered at different time intervals, and the data quality was affected by the logger model, weather, battery conditions, and the behaviour of individual geese, the frequency of derived fixes was uneven (ranging from 1 to 288 GPS points per day). For this reason, it was decided that habitat use and degree of protected area coverage needed to be derived from the BBMM estimated home range delineation rather than of individual points. Thus, for each population, we extracted the land use type within each home range, and then calculated the percentage of land use cover shown in Table SX2. All above processes were extracted using ArcGIS 10.6 (ESRI 2013) and the calculations were run in R 3.6 (Team RDC 2017).
Conservation status at national level
To estimate the contribution of current conservation of the three GWFG populations by protected areas, we calculated the percentage of the home ranges which fell inside designated protected areas during each phase of the life cycle (summer, winter, and stopover periods during both spring and autumn migration episodes). The national nature reserve (NNRS) boundaries from the Resource and Environment Data Cloud Platform (http://www.resdc.cn/) were used to define the protected sites inside China (outside Russia), and the boundaries of the protected areas in all other areas were derived from the World Database on Protected Areas (https://www.protectedplanet.net/).
Add Comment