When a typical jumper skips: itineraries and staging habitats used by Red Knots ( Calidris canutus piersmai ) migrating between northwest Australia and the New Siberian Islands

The ecological reasons for variation in avian migration, with some populations migrating across thousands of kilometres between breeding and non-breeding areas with one or few refuelling stops, in contrast to others that stop more often, remain to be pinned down. Red Knots Calidris canutus are a textbook example of a shorebird species that makes long migrations with only a few stops. Recognizing that such behaviours are not necessarily species-speci ﬁ c but determined by ecological context, we here provide a description of the migrations migration, when birds often stopped along the way in southeast Asia and ‘ hugged ’ the coast of China, thus ﬂ ying an additional 1000 – 1500 km compared with the shortest possible (great circle route) ﬂ ights between NW Australia and the Yellow Sea. The birds staged longest in areas in northern China, along the shores of Bohai Bay and upper Liaodong Bay, where the bivalve Potamocorbula laevis , known as a particularly suitable food for Red Knots, was present. The use of multiple food-rich stopping sites during northward migration by piersmai is atypical among subspecies of Red Knots. Although piersmai apparently has the bene ﬁ t of multiple suitable stopping areas along the ﬂ yway, it is a subspecies in decline and their mortality away from the NW Australian non-breeding grounds has been elevated.

The ecological reasons for variation in avian migration, with some populations migrating across thousands of kilometres between breeding and non-breeding areas with one or few refuelling stops, in contrast to others that stop more often, remain to be pinned down. Red Knots Calidris canutus are a textbook example of a shorebird species that makes long migrations with only a few stops. Recognizing that such behaviours are not necessarily species-specific but determined by ecological context, we here provide a description of the migrations of a relatively recently described subspecies (piersmai). Based on data from tagging of Red Knots on the terminal non-breeding grounds in northwest Australia with 4.5-and 2.5-g solar-powered Platform Terminal Transmitters (PTTs) and 1.0-g geolocators, we obtained information on 19 route-records of 17 individuals, resulting in seven complete return migrations. We confirm published evidence that Red Knots of the piersmai subspecies migrate from NW Australia and breed on the New Siberian Islands in the Russian Arctic and that they stage along the coasts of southeastern Asia, especially in the northern Yellow Sea in China. Red Knots arrived on the tundra breeding grounds from 8 June onwards. Southward departures mainly occurred in the last week of July and the first week of August. We documented six non-stop flights of over c. 5000 km (with a maximum of 6500 km, lasting 6.6 days). Nevertheless, rather than staging at a single location for multiple weeks halfway during migration, piersmaiknots made several stops of up to a week. This was especially evident during northward migration, when birds often stopped along the way in southeast Asia and 'hugged' the coast of China, thus flying an additional 1000-1500 km compared with the shortest possible (great circle route) flights between NW Australia and the Yellow Sea. The birds staged longest in areas in northern China, along the shores of Bohai Bay and upper Liaodong Bay, where the bivalve Potamocorbula laevis, known as a particularly suitable food for Red Knots, was present. The use of multiple food-rich stopping sites during northward migration by piersmai is atypical among subspecies of Red Knots. Although piersmai apparently has the benefit of multiple suitable stopping areas along the flyway, it is a subspecies in decline and their mortality away from the NW Australian non-breeding grounds has been elevated.
There are good biological reasons for some birds to breed in one part of the planet and spend the rest of the year in another. The published research for these reasons now occupies a few metres of bookshelf, but the field has been ably summarized by Newton (2008). For example, long-distance migratory shorebirds that breed during the northern summer in the Arctic combine their reproductive activities on the tundra with long periods at softsediment seashores during the northern winter (or austral summer), the terminal non-breeding (or 'wintering') areas being found as far south as the sub-Antarctic (see generalizations in Piersma 1997, 2003, van de Kam et al. 2004, Conklin et al. 2017. Coastal shorebirds show a range of feeding specializations (Prater 1981). Among them, the Red Knots Calidris canutus possess a sensory system for the remote detection of hard objects in wet soft sediments , de Fouw et al. 2016. Red Knots make a living by probing for hard-shelled prey (usually bivalves) in intertidal soft-sediment flats (e.g. van Gils et al. 2006, Quaintenne et al. 2010, which is combined with visual hunting for surface-living arthropods on the tundra (Martin & Piersma 2009). Away from the tundra breeding grounds, using sensory attributes and prey types tolerant to foraging in dense flocks, they are highly social and often occur in large flocks (Piersma et al. 1993, Bijleveld et al. 2016, Oudman et al. 2018; this is also part of a strategy to avoid depredation by falcons (van den Hout et al. 2010). As suitable feeding habitats are rare and widely dispersed across the globe (e.g. van , the long migratory flights of Red Knots (Piersma & Davidson 1992, Shamoun-Baranes et al. 2010) may be considered a consequence of their ecological specialization.
Despite extensive knowledge of geographical variation and migratory connectivity of Red Knot populations worldwide (Piersma & Davison 1992, Tomkovich 1992), a new subspecies of Red Knot was described as recently as 2001(C. c. piersmai, Tomkovich 2001. Evidence for their migration route included five re-sightings between November 1995 and September 1996 in NW Australia of a single Red Knot that was individually colourmarked on 10 July 1994 at Faddeyevski Island, New Siberian Islands group, Russia (Lindstr€ om et al. 1999). Also, biometric data and plumage observations of Red Knots in Roebuck Bay, NW Australia (e.g. Verhoeven et al. 2016), were all consistent with the idea that many piersmai spend the austral summer in NW Australia. Some appear to migrate as far south as New Zealand (Tomkovich & Riegen 2000. The observation that Red Knots departed on northward migration from NW Australia late into May, led Battley et al. (2005) to predict: (1) the use of high-quality shellfish food at potential staging areas along the Yellow Sea, (2) a window of about 3 weeks of potential fuelling time in Asia and (3) arrivals on the New Siberian Island breeding grounds in early June. The prediction of high food quality and abundance in the Yellow Sea was confirmed by Yang et al. (2013Yang et al. ( , 2016 for Red Knots staging in Luannan County, Hebei Province, Bohai Sea (see also , Yang et al. 2011, and see yearly field reports at http://globalflywaynetwork.com.au/bohai-bay/re ports-and-papers/). However, a capture-resight analysis of individually marked Red Knots showed that large numbers stage at this site for just 5-9 days (Lok et al. 2019), implying that there are other stopping sites en route. Does this mean that, in contrast to what Red Knots have been shown to do elsewhere in the world , Piersma 2007), in the East Asian-Australasian Flyway they do not necessarily 'long-jump', i.e. make a single refuelling stop during the migration from wintering to breeding areas (Piersma 1987)? Do they make multiple stops as in the 'skipper' strategy and, if so, where are the additional stopping areas located? Can such areas be characterized in terms of food availability?
To answer these questions, we applied an approach that combined tracking of individual Red Knots during migration with in situ sampling of benthic prey at potential stopping sites. In 2011-2019 we tracked the migratory routines of Knots caught at Roebuck Bay and Eighty Mile Beach, NW Australia, using three different tracking devices. Complementary to the 2018 tracking effort, in March-May 2018 we conducted surveys at several potential stopping sites for shorebirds and sampled macro-benthic bird food along the coast of China , Peng et al. 2021. Here we provide a detailed description of the seasonal migration of Red Knots from NW Australia, assessing the timing of migration, the lengths of non-stop flights, the locations and numbers of stopping sites used, as well as examining the possible food resources at these coastal stopping sites during northward and southward migration. With the knowledge that a small bivalve, Potamocorbula laevis, is the key highquality prey of Red Knots at a staging area in China (Yang et al. 2013), we focused on comparing densities of P. laevis between sites that the tracked Red Knots did and did not visit along the Chinese coast.

Satellite tracking
This study is part of an international collaborative long-term effort by the Global Flyway Network and associated institutions to study the demography and migration ecology of several representative shorebird species along the East Asian-Australasian Flyway (e.g. , Chan et al. 2019b, Lok et al. 2019. Red Knots were captured using cannonnets at the northern beaches of Roebuck Bay, Broome (17°48 0 58″S, 122°17 0 60″E) and at Eighty Mile Beach (19°20 0 24″S, 121°24 0 36″E), both located in NW Australia (see Table S1 for an overview of all tracking efforts included in this study). After capture, birds were measured and weighed, and a small blood sample was taken for molecular sexing (van der Velde et al. 2017). Birds were aged based on plumage characteristics (see Rogers et al. 1990, Higgins & Davies 1996 for guidance) and adults (birds older than 2 years) were selected for tagging. Due to incomplete breeding plumages at the time of year the birds were captured, we were unable to confirm subspecies identity, but we should have been picking mostly piersmai as it outnumbers rogersi in NW Australia (see . All birds were marked with an Australian Bird and Bat Banding Scheme (ABBBS) metal band and a unique colour band combination allowing individual identification in the field. The tagging work was carried out under Regulation 17 permits SF 010074, SF010547 and 01-000057-2 issued by the West Australian Department of Biodiversity, Conservation and Attractions.
In April 2011 we deployed 4.5-g solar-powered Platform Terminal Transmitters (PTTs) (Microwave Telemetry, Inc., Columbia, MD, USA) on 30 Red Knots by gluing the transmitters onto the back of the birds with superglue (Warnock & Warnock 1993). Despite using methods that were previously successfully used in temperate climates on the same species, on the basis of field observations of colour-ringed birds that were seen without tags, we conclude that most birds shed their PTTs before migration. Here we report on the migratory movements of the remaining three birds (see Table S1 for an overview). We faced a similar problem in March 2012 when we tagged another 15 birds using the same method, with all transmitters technically failing before northward migration started. Consequently, the individuals tagged in 2012 were excluded from the analyses presented here. The 4.5-g PTTs were on a duty cycle of 10 h on for transmitting and 48 h off.
Before the start of migration in 2017 we tagged 21 Red Knots (n = 2 in October 2016 and n = 19 in February-March 2017), and in February-March 2018 we tagged 18 Red Knots, with 2.5-g solarpowered Argos 3 PTTs (Microwave Telemetry). The 2.5-g PTTs were deployed using a body harness ) made of nylon-coated stainless steel jewellery wire (provided by Microwave Telemetry; in 2017) or 1-mm-thick Flyneema (a smoothly covered fishing line with a strong Dyneemaâ core; de Lijnenspecialist, Amsterdam, The Netherlands; in 2018). After transmitter deployment, to allow them to acclimatize to the transmitter and harness, the birds were kept in cages indoors and observed for a few hours up to 48 h. We then released them on the beach near their capture sites. All but three of the deployed PTTs stopped before departure from NW Australia in 2017 due to what we think was loss of the PTTs because of corrosion and breakage of the harness. Here we report on movements of the remaining three Red Knots for which we collected data on migration in 2017. This problem of harness breakage was resolved when 18 PTTs were deployed in February-March 2018, although some other problems remained. Two of these PTTs never transmitted locations, eight provided locations from the area of release but stopped transmitting before departure (four of these birds were later resighted in NW Australia or China), one transmitter stopped transmitting at the Chinese coast during northward migration and one operated too intermittently for a complete reconstruction of the carrier's itinerary. Analyses presented in this paper are based on the migrations of six birds captured and tracked in 2018, one of which gave us a repeat track in 2019 (Table S1).
The 2.5-g PTTs were not on a duty cycle, but rather transmitted whenever sufficiently charged. All PTTs (when on) transmitted signals every 60 s to the Argos satellite system. When signals from the PTTs were received by a satellite, the perceived Doppler shift in signal frequency of successive transmissions was used to estimate the position of the transmitter (CLS 2016). We used the hybrid filter option of the Douglas-Argos Filter set for a high rate of speed (130 kph) and a relaxed minimum redundant distance (10 km) suitable for summarizing long-distance flights (Douglas et al. 2012). During filtering, all standard-quality locations (i.e. location classes 3, 2 and 1; for details on Argos location classes see CLS 2016) were retained, whereas low-quality locations (i.e. location classes A, 0, B and Z) were retained only if they passed filter thresholds.
Following Chan et al. (2019b), we refer to the places where birds during migration spent time on the ground as 'stopping sites', with no distinction between 'stopover sites' and 'staging sites' (see Warnock 2010 for definitions). To identify potential migratory stops, we first assigned a status of stationary (groundspeed <5 km/h) or moving (>5 km/h) to each filtered location after each bird departed its terminal non-breeding site (Roebuck Bay or Eighty Mile Beach). We then defined stops as a cluster of at least three stationary locations within 20 km of each other, with the first and last recorded locations at the stopover being at least 2 h apart.
Using speed of movement, departure times from a stopping site were extrapolated over the intervening travel distance between the last location at a stop and the next location. Extrapolation used the speed from the last location at the stop to the next non-stationary (in-flight point) or the median of flight points of all flights recorded in the same latitudinal interval and migration direction, whichever was faster. If there was no recorded location in-flight, the migratory flight was assumed to have occurred over the interval between the last point of a stop and the first point of the next stop. Arrival times were extrapolated in the same way over the interval between the first recorded location of a stop and the previous location (in-flight or not). Arrival times in the Yellow Sea area (between latitudes 30°54 0 N and 42°30 0 N) and at the New Siberian Islands were defined as the estimated times of arrival at the first stop within these respective regions. Duration of time at stopping sites was calculated as the time between the estimated arrival and departure times. Movements between detected stops were assumed to be carried out non-stop. Flight distances were calculated by summing up intervening distances between locations along the flight.

Geolocation
We deployed a total of 129 geolocators (n = 36, Intigeo-W65; Migrate Technology Ltd, Cambridge, UK) in March 2012 and February and June-July 2015 (n = 93, mk50773; Biotrack, Lotek Wireless Inc., Wareham, UK; see Table S1 for an overview). Geolocators were attached to a Darvic PVC leg-flag using Kevlar thread reinforced with Araldite resin cement (after Lisovski et al. 2016) attached to a leg of the Red Knots. The combined mass of flag and geolocator was c. 1 g.
Of the 36 geolocators attached in early 2012, two were retrieved in November 2013; one of these only contained reliable migratory information for 2012, while the other one yielded information for 2 years of tracking (for 2012 and 2013). Another tag was retrieved in February 2015 but did not contain reliable information. Of the 93 geolocators attached in February 2015, one was retrieved in September 2015 and another in February 2020. Only the former yielded reliable information of the return migration in 2015 (see Table S1 for an overview).
We analysed data with the template fit approach (Rakhimberdiev et al. 2015b) in the R package FlightR (Rakhimberdiev et al. 2017). For calibration, we used average coordinates of individual re-sightings during their non-breeding season in NW Australia (from early September to mid-April). The stops were defined by the probability cut-off value of 0.1. The geolocation tracks were consistent with the geographical description of tracks obtained by the PTTs. However, due to the coarse nature of the geolocation data, the representations of geolocation tracks added no novel geographical information when compared with the tracks obtained by the PTTs and are therefore not presented in this study (Rakhimberdiev et al. 2015b(Rakhimberdiev et al. , 2016. Due to the constant daylight conditions at the high Arctic breeding grounds of the Red Knots, solar geolocation is unsuitable to position the birds in these areas. Therefore, we only present the migration timing and latitudes of geolocator-tagged birds until a latitude of 42°N, i.e. the northern boundary of the Yellow Sea.

Benthic food sources along the coast of China
From early April to late May 2018, we sampled the preferred benthic food of Red Knots, P. laevis (Yang et al. 2013), at 18 intertidal flats along the coast of China known to have shorebirds utilizing them during migration (Chan et al. 2019b), extending from Dongliaodao, Guangdong Province, in the far south (20°49 0 30″N, 110°23 0 02″E) to Panjin, Liaoning Province, in the far north (40°45 0 47″N, 121°51 0 36″E; see Peng et al. 2021 for further details). At each site, we sampled macro-zoobenthos across gridlines (after Bijleveld et al. 2012). Depending on the local geography, sampling stations were elected to be 50, 125, 250 or 500 m apart to sample adequately the area from the coast to the low water line. A total of 838 sampling stations were visited by foot. At each station, one sediment core with a surface area of 0.019 m 2 was taken to a depth of 20 cm and washed over a 0.5-mm sieve. The sieved samples were frozen and stored before analysis in the laboratory where shellfish were identified to the species level, counted and their maximum shell length measured. In the site Huanghua (38°24 0 27″ N, 117°51 0 06″E), the soft mud made grid sampling by foot impossible. However, it is an important area for commercial harvesting of P. laevis, so observations (visual, touching mud surface) were made to estimate the density level of P. laevis there. We examined whether a benthic sampling site was visited by any tracked Red Knot during the northward migration of 2018. A site is defined as visited by a tracked individual when the average coordinates of the individual's stops are within 10 km of the centre point of a benthic sampling area.

Geography of the migrations
Of the three Red Knots departing from NW Australia in 2011 (Fig. 1a), one was last recorded during its first stop in northeast Kalimantan, Indonesia. A second Red Knot also made a stop there and then continued to the Chinese coast, making a landfall on the coast of Fujian province. It then tracked the coastline of China to arrive in Bohai Bay, after which we lost contact. The third Red Knot of 2011, like the previous two, also travelled north across Makassar Strait, i.e. keeping to the east of Kalimantan, made a stop in the Philippines and an onward flight to the Fujian coast, at which point we lost contact.
Of the three Red Knots which we tracked in 2017 (Fig. 1b), one bird took a rather north-westerly course on a non-stop flight to the southern coast of Vietnam. After 10 days, while the bird was still at this site, we lost contact. A second bird migrated north, over eastern Kalimantan, turning southeast at the northern tip for a stop of 2 weeks on the Kalimantan coast just southwest of the city of Sandakan, Malaysia (5°50 0 N, 118°07 0 E). It then continued migration to make landfall on the western-most coast of China close to the border with Vietnam. It continued east by tracking the Chinese coastline until we lost contact, probably in flight across south Fujian on 21 May. The third bird travelled north across Makassar Strait, i.e. keeping to the east of Kalimantan, and also made a turn when it reached the northern tip of Kalimantan, and then stopped on the coast northwest of Sandakan. We lost contact on 8 May while the bird was still at this site.
Of the eight PTTs deployed in 2018 that returned migratory movements, transmission of one PTT stopped during northward migration at the Chinese coast (the bird indicated in purple in Fig. 1c). Another bird migrated only as far as western Kalimantan to return to NW Australia from there after the northern summer (the bird indicated in green in Fig. 1c,d). The other six made complete migrations to the New Siberian Islands (Table 1, Fig. 1c,d). In the case of the bird carrying transmitter 48950 (coloured white in Fig. 1c, d), in 2018 no signals were received after it left the Yellow Sea on northward migration until August; however, a complete track was obtained during the second season of migration in 2019.
As suggested by the tracks in Fig. 1, during the first leg of the northward migration after leaving NW Australia, rather than following the shortest northward flight route (i.e. a route close to a great circle route across Sulawesi to the Yellow Sea as illustrated by the yellow line in Fig. 1c,d), most Red Knots took a longer route by initially flying north-westwards. The birds reached the coast of China between Guangdong and Fujian, rather than in the Shanghai area where they would have arrived if they had flown from NW Australia to the Yellow Sea along a great circle route (Fig. 1c). Most birds continued to the Yellow Sea by closely tracking the coastline. This flight behaviour added 1000-1500 km to a great circle distance of 6500 km between NW Australia and northern Bohai Bay in the Yellow Sea.
All tracks obtained from Red Knots carrying PTTs confirm that during the first leg of migration to the Yellow Sea, Red Knots made one to four stops (Fig. 1). Once in the Yellow Sea, the birds stopped at one to four different sites (Fig. 1, Table 1). During northward migration, the last coastal stopping sites before the trans-continental flight towards the New Siberian Islands were in the northern part of the Yellow Sea, at either the coast of Bohai Bay, Hebei and Tianjin Municipality, or in upper Liaodong Bay, Liaoning (Fig. 1). The stops in 2018 of two tagged Red Knots in Luannan County in northern Bohai Bay, China, were confirmed with on-the-ground observations of the colour-ringed birds.
Flying from the Yellow Sea to the New Siberian Islands, all six birds tracked in 2018-2019 made one to five stops at inland sites or on the coastal tundra just before crossing the Laptev Strait. On the way back to the Yellow Sea, the birds also stopped at continental sites (n = 1-6), and one stopped in the Lena River Estuary (38°24 0 27″N, 117°51 0 06″E; see Fig. 1d). Thus, during the migrations across the thinly populated areas of northern China, Mongolia and eastern Russia, most tracked Red Knots spent some days at freshwater lakes and riverbanks. Birds often used lowland lake systems, but some stops were made at water bodies at altitudes of up to 1100 m in mountainous terrain.
During migration from the Yellow Sea to NW Australia, only bird 48950 (Table 1) made a single non-stop flight in both years during which it was tracked; the other birds made one to three stops in Taiwan, the Philippines, Malaysia and/or Indonesia during southward migration. Four of the six satellite-tagged Red Knots of the 2018 cohort demonstrated the ability to non-stop fly distances of approximately 5000 km or more (with a maximum of 6500 km, a continuous 6.6 days of flight; Table 1).

Figure 1. A summary of all migratory tracks recorded by PTTs in Red Knots marked in NW Australia between 2011 and 2019.
Tracks of the partial northward migrations from NW Australia of three individual Red Knots which were tracked with 4.5-g PTTs in 2011 (a); and of three Red Knots tracked with 2.5-g PTTs in 2017 (b). The lower two panels present the tracks of the northward migration (c) and the southward migration (d) of eight individual Red Knots tracked from NW Australia with 2.5-g PTTs in early 2018. Small dots indicate the filtered Argos locations used. The larger green dots represent all Red Knot stopping sites observed in all years, during northward and southward migrations combined. All sites are plotted in all four panels, with the numbers in panel (a) corresponding to additional information of the sites in Table S3. These stopping sites were calculated by means of grouping all individual stopping sites (see Methods for definition) within a 10-km radius. The yellow lines in (c) and (d) represent the shortest, great circle, routes between Roebuck Bay in NW Australia and Bohai Bay in the Yellow Sea and between the Yellow Sea and New Siberian Islands. With respect to the birds marked in early 2018, Table 1 Table 1) only migrated as far as westernmost Kalimantan, returning from there to NW Australia. For Red Knot 48953 (in purple; not in Table 1) transmissions stopped during northward migration at the Chinese coast.

Timing of the migrations
In 2011 the four Red Knots departed from NW Australia between 30 April and 9 May (Fig. 2). In 2017 the three Red Knots departed from NW Australia between 21 April and 13 May. The geolocator tracks obtained from 2012 (n = 2) indicate that one bird departed from NW Australia in the last week of April and returned to NW Australia in the last week of August (047-2012). The other bird equipped with a geolocator in 2012 (022) departed from NW Australia in mid-May 2012. Its southward migration is unclear. The geolocator track obtained from the migratory season in 2013 (n = 1) also came from 047 (047-2013). It then departed from NW Australia slightly later than the previous year (c. 10 May) and it returned to NW Australia again in the last week of October. The geolocator obtained for the migratory season in 2015 (n = 1; P536) showed departure from NW Australia in the first week of May and a return to NW Australia around 20 August.
During seven migrations of six PTT-tagged Red Knots tracked in 2018-2019, birds departed from NW Australia between 24 April and 24 May (of which no fewer than four departed on 4 May, see Table 1, Fig. 2) and reached the breeding grounds on the New Siberian Islands between 8 and 29 June. During northward migration, Red Knots stayed on average 12.9 days in southeast Asia and southern China (5.5 days per site), and 13.8 days in the Yellow Sea (5.4 days per site). Departures from the Yellow Sea occurred between 2 and 23   Table S1. Stationary periods are coloured by the latitude at which they occurred according to the colour-gradient presented in the inset. The tracks are arranged from earliest to latest migratory departure from NW Australia.
June. Arrival on the New Siberian Island tundra breeding grounds occurred during 8-29 June. Three of the four early arriving birds (individuals which also left NW Australia relatively early, i.e. before 4 May) stayed long enough on the New Siberian Islands for a successful breeding season (54, 49 and 53 days, respectively). The first Red Knots departed from the New Siberian Islands on southward migration on 18 July (recorded in 2019) and between 27 July and 3 August in 2018. Arriving back in the Yellow Sea between 30 July and 18 August, Red Knots then staged here for a period twice as long as during their northward migration (22-37 days, average = 27.7 days). Departure from the Yellow Sea occurred between 27 August and 21 September. The earliest return to NW Australia occurred on 3 September and the latest on 26 October.
All three types of tracking devices yielded the same pattern of timing of Red Knots reaching different latitudes (Fig. 2), with no clear clustering of different devices with respect to either departure or arrival dates. The three geolocator tracks that yielded sufficient information regarding the entire migratory period (047, 022 and P536; Fig. 2) confirmed that staging in the northern Yellow Sea was much longer during southward than northward migration.
China coast: use of stopping sites in relation to benthic food Among the 19 shorebird stopping sites along the coastline of China where we surveyed macrozoobenthos in spring 2018, we found the bivalve P. laevis at 14 sites (Fig. 3, Table S2). Not all the sites where P. laevis was found were used by the tracked Red Knots but P. laevis was found at all seven sites where they did stop during northward migration (including the longest used sites in Bohai and Liaodong Bays). At the five sites for which we have benthos data but where tagged Red Knots did not make stops, no P. laevis were found. All P. laevis encountered were living in the top 5 cm of the sediment and were much smaller than 21 mm, i.e. perfectly harvestable by Red Knots.

DISCUSSION
In this study we confirm that Red Knots from their terminal non-breeding grounds in NW Australia stop in the Yellow Sea region of China (especially Bohai and Liaodong Bays) during both northward and southward migration seasons, and breed on the New Siberian Islands in Russia and arrive there in June. The fact that all individuals with records during the breeding season were on the New Siberian Islands, and an absence of clear outliers in the timing of migration patterns (Fig. 2), suggests that most, and probably all, the tagged Red Knots indeed belonged to the piersmai subspecies (Tomkovich 2001). In fact, the data are consistent with earlier inferences on the occurrence and distribution of piersmai (Tomkovich 2001, except that the birds were making more stops than anticipated, especially during northward migration. Although carrying a tag may come with timing delays or foreshortened flight ranges (Bodey et al. 2018), the similarity of the timing of migratory flights and the occurrence of multiple stops during northward migration by Red Knots carrying different tracking devices (4.5-and 2.5-g solar PTTs and 1-g geolocators; Fig. 2) indicate that the stopping behaviour along the migratory trajectories is not an effect of incremental impediments from the tracking devices. Although the smallest devices could have an impact, carrying the devices did not prevent the Red Knots from making non-stop flights of 5000 km (Conklin et al. 2017 used 5000 km as a threshold for 'long-jump' migratory flights).
Birds did not make stops 'at their earliest convenience' (i.e. stopping at the first possible site in southern China), but rather flew up to 1000 km up the Chinese coast before making the first stop. Their northwestward, rather than northward great circle, bearings during departure from NW Australia are consistent with the visual onshore observations made as early as 1991 by Tulp et al. (1994) and Broome Bird Observatory data to 2021 (C. J. Hassell pers. obs.). This suggests that the departure directions in this study are similar to those in 1991, and that the Red Knots, by not stopping at the first suitable coastal site in southern China, were not running out of fuel upon arrival at the Chinese coast. In addition, individuals vary in their use of stops in Southeast Asia and southern China and this has consequences in the subsequent leg of the migration, i.e. the individuals which made more stops en route to the Yellow Sea stopped for fewer days in the Yellow Sea (recorded for six birds; Fig. 4). This variation in the ways that individuals distribute their fuelling over multiple areas would be an avenue for future investigation.
In view of the general capacity of Red Knots to migrate across large swaths of inhospitable terrain (as they do during the flights across the Asian landmass to and from the New Siberian Islands, Fig. 1; see e.g. Niles et al. 2010, Newstead et al. 2013, Kok et al. 2020 for similar feats in other  Table S2.
Red Knot subspecies including the 8100-km nonstop flight reported at https://whsrn.org/uncove ring-the-mysteries-of-red-knot-movements-on-thegulf-coast/#), the Chinese coastline is probably 'traced' for good navigational or ecological reasons, including the possibility that they take advantage of favourable winds (Tulp et al. 1994). Northward stops occurred at sites where our sampling of the intertidal feeding areas showed the presence of P. laevis (Fig. 3), a strongly preferred and highquality prey type for Red Knots (Yang et al. 2013 and the similarly molluscivorous and closely related Great Knot Calidris tenuirostris (Choi et al. 2017. During the 3week southward staging bouts in the Yellow Sea, the Red Knots, using the same areas as on northward migration, most likely again fuelled up on a diet of P. laevis. Although piersmai subspecies achieved non-stop flight distances comparable to those by other subspecies during northward migration from NW Australia to the Yellow Sea (Table 1), contrary to the other subspecies, piersmai behaved as 'skippers' rather than 'long-jumpers' (Piersma 1987). This pattern of making several short stops signals the presence of multiple suitable staging habitats along the east Asian coastline from Vietnam to the northern Yellow Sea. That this may have been going on for quite some time is suggested not only by the migratory departures from Roebuck Bay of shorebirds including Red Knots to the northwest (Tulp et al. 1994), but also by Crossland's (2009) observations of the presence of Red Knots in Sumatra in late March to mid-April 1997. Although these birds occur further west and earlier in the year compared with piersmai from NW Australia, these findings indicate that quite a number of sites are potentially suitable for staging piersmai. To help governments and conservation bodies to take appropriate steps towards their protection, we have listed all sites in Table S3, corresponding to the graphical listing in Fig. 1a.
This, then, invites the question of whether the recent reductions in the extent of suitable habitat in the Yellow Sea area (Murray et al. 2014) have contributed to the current pattern of stopping at multiple sites. Reductions in the extent of suitable intertidal habitat in northern Bohai Bay appear to have led to a concentration of staging Red Knots at the Luannan coast (Yang et al. 2011). Land claims for industry, port and city development, and aquaculture tend to start from the much-used upper parts of intertidal soft sediment systems, a pattern which would have augmented the reduction of the extent of suitable feeding area for Red Knots along the coast of China (Mu & Wilcove 2020). This suggests that the extent of suitable intertidal habitat for Red Knots was much larger two to three decades ago than now, and thus before the time of rapid intertidal losses due to land claims (Ma et al. 2014). Red Knots could have been 'hopping' along even more coastal sites back then. To complicate matters further, the suitability of the remaining Chinese coastal wetlands will be affected by shellfish aquaculture on mudflats, which in fact could have increased the range and densities of P. laevis (Peng et al. 2021), and the offshore fishery pressure on the epibenthic predators of small bivalves, such as shrimps and crabs. High fishing pressures may lead to a lack of epibenthic predation, which facilitate the late-winter settlement of P. laevis .
One of the benefits of making several shorter migratory flights, rather than a single long one, would be the cost reduction that comes from flying with, on average, smaller fat stores (Piersma  Table 1 for details).
1987), the lack of need for major 'organizational' changes to organs and body composition (Piersma 1998, Piersma et al. 1999) and the predation-related 'safety' gains from not having to fly with compromised manoeuvrability (van denHout et al. 2010). Equally, even in places where intertidal losses due to land claims have been small, variation such as reductions in food abundance may still occur (Zhang et al. 2018. Thus, the use, at least by piersmai, of a succession of several suitable sites should make them less susceptible to resource degradation (Piersma & Baker 2000, Iwamura et al. 2013, and includes the possible benefit that, by visiting multiple sites, Red Knots collect information on the quality of staging areas during migrations. This allows them, in subsequent migrations, to know where probable food resources are, should a site would be lost to industrialization or other factors.
The ecological reasons for the occurrence of several stops in freshwater habitats during the migrations from the Yellow Sea to New Siberian Island and back are not at all clear. On the way north, Red Knots often stopped on the Laptev Strait coast before crossing to the New Siberian Islands, perhaps to await suitable weather conditions before arrival on the tundra. However, some of the birds only made such onshore stops several weeks after conditions on the New Siberian Islands would have become suitable. Many of the Red Knots also used freshwater wetlands during their southward continental crossing. Do these Red Knots capitalize on unknown seasonal peaks in unknown freshwater invertebrates?

Conservation prospect
Even though, from 2004 to 2017, there was no significant change in the numbers of Red Knots in NW Australia (Rogers et al. 2020), the East Asian-Australasian Flyway populations of the piersmai and rogersi subspecies have been in decline for more than a decade (Conklin et al. 2014, Clemens et al. 2016, Studds et al. 2017. For a better understanding of the precise causes of such declines in the most threatened shorebird flyway in the world, the present description of migratory pathways and stopping sites of piersmai, which still has a population in the tens of thousands, opens up a system amenable to scientific study. We suggest that a combination of studies on local resources and staging, the use of tags to track individual birds lifelong, and analyses of times and places of death (Loonstra et al. 2019) will reveal the potential ecological reasons leading up to further declines or recovery (Rakhimberdiev et al. 2015a). In this way we remain vigilant in the hope that this contributes to science-based conservation outcomes which extend beyond the world of Red Knots.
We thank the many dedicated volunteers who participated in our satellite tracking fieldwork and coastal surveys in China from 2014 to 2017, and Broome Bird Observatory and the Australian Wader Studies Group (AWSG) for logistical support. The satellite tracking was funded by the Spinoza Premium 2014 awarded by the Netherlands Organization for Scientific Research (NWO) to T.P. and by the MAVA Foundation, Switzerland, with additional support from NWO TOP-project 'Shorebirds in Space' (854 11 004), with WWF-Netherlands and BirdLife Netherlands contributing to the work of C.J.H. at Broome. Some of the benthic surveys were funded by a KNAW China Exchange Programme grant. E.M.A.K. was supported by NIOZ and RUG. Y.C.C. was supported by private donors through the Ubbo Emmius Fund of the University of Groningen (fundraising by Tienke Koning and Wilfred Mohr), by the Spinoza Premium 2014 to T.P. and by the University of Groningen. H.B.P. was supported by the China Scholarship Council (201506100028). T.L.T. was supported by the Ecosystems Office of the U.S. Geological Survey. We are very grateful to Ken Gosbell and the late Clive Minton of the Australian Wader Studies Group for the joint instrumentation by geolocators in Broome, as well as their recapture; Ken Gosbell dealt with geolocator handling issues. We thank the editors, and David Melville and Danny Rogers as reviewers, for great constructive feedback. We thank Dick Visser for finalizing the figures. Thanks go to the Stoate family of Anna Plains Station for access and logistical support. We acknowledge the Yawuru People via the offices of Nyamba Buru Yawuru Limited for permission to catch birds on the shores of Roebuck Bay, traditional lands of the Yawuru people. We acknowledge the Karajarri and Nyangumarta people for permission to catch birds to be marked for this project on the shores of Eighty Mile Beach, traditional lands of the Karajarri and Nyangumarta people. Any use of trade, product or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Conflict of interest: P.W.H. is the founder and CEO of Microwave Telemetry, the producer of the PTTs used in this study. The authors declare no further conflict of interest.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Overview of tracking attempts included in this study. Table S2. List of intertidal sites along the coast of China with information on densities of Potamocorbula laevis and usage by satellite-tagged Red Knots. Table S3. Name list of stopping sites used by Red Knots in the East Asian-Australasian Flyway.