Complex postbreeding molt strategies in a songbird migrating along the East Asian Flyway, the Pallas’s Grasshopper Warbler Locustella certhiola

Abstract Molt strategies have received relatively little attention in current ornithology, and knowledge concerning the evolution, variability and extent of molt is sparse in many bird species. This is especially true for East Asian Locustella species where assumptions on molt patterns are based on incomplete information. We provide evidence indicating a complex postbreeding molt strategy and variable molt extent among the Pallas's Grasshopper Warbler Locustella certhiola, based on data from six ringing sites situated along its flyway from the breeding grounds to the wintering areas. Detailed study revealed for the first time that in most individuals wing feather molt proceeds from the center both toward the body and the wing‐tip, a molt pattern known as divergent molt (which is rare among Palearctic passerines). In the Russian Far East, where both breeding birds and passage migrants occur, a third of the adult birds were molting in late summer. In Central Siberia, at the northwestern limit of its distribution, adult individuals commenced their primary molt partly divergently and partly with unknown sequence. During migration in Mongolia, only descendantly (i.e., from the body toward the wing‐tip) molting birds were observed, while further south in Korea, Hong Kong, and Thailand the proportion of potential eccentric and divergent feather renewal was not identifiable since the renewed feathers were already fully grown as expected. We found an increase in the mean number of molted primaries during the progress of the autumn migration. Moderate body mass levels and low‐fat and muscle scores were observed in molting adult birds, without any remarkable increase in the later season. According to optimality models, we suggest that an extremely short season of high food abundance in tall grass habitats and a largely overland route allow autumn migration with low fuel loads combined with molt migration in at least a part of the population. This study highlights the importance of further studying molt strategy as well as stopover behavior decisions and the trade‐offs among migratory birds that are now facing a panoply of anthropogenic threats along their flyways.


| INTRODUC TI ON
Annual cycles of birds comprise a series of energy-and time-demanding phenological processes (life-history stages; Wingfield, 2005): namely breeding, migration, and molt. Feathers deteriorate through time due to abrasion and exposure to sunlight and have to be renewed periodically. Molt, with its particular nutritional demands in terms of protein mobilization, is more flexible in timing than other annually repeated events (Jenni & Winkler, 2020;Lindstrom & Piersma, 1993;Newton, 2009Newton, , 2011. Timing and pattern of molt in relation to migration varies, and flight feather replacement occurs either in breeding areas, wintering sites, or at stopover sites on migration. The annual schedule depends on species or individual migratory strategies (Carlisle et al., 2005;Ginn & Melville, 1983;Jenni & Winkler, 2020;Newton, 2009;Stresemann & Stresemann, 1966) and frequently differs also among different geographical populations of the same species (Hemborg et al., 2001;Remisiewicz, 2011).
In contrast to resident populations or birds that molt in winter quarters, migratory species that molt during late summer at high latitudes are less flexible regarding the timing of their molt and may undergo more rapid flight feather replacement before departure. But rapid molting results in openings in the wing and/or reduced feather quality, impeding flight efficiency, and increasing risk of predation (Butler, 2013;De La Hera et al., 2009;Hedenström & Sunada, 1999; Moller & Nielsen, 2018; Morrison et al., 2015;Rohwer et al., 2011;Serra et al., 2007;Swaddle & Winter, 1997; but see Bridge, 2011).
Furthermore, molt-related feather gaps in the wing or tail may increase energetic costs and simultaneously impair accumulation of energy reserves for migration. Molt speed and the sequence of feather renewal may affect the size and shape of molt-related feather gaps with possible impacts on flight metabolism and performance during the molt process. Therefore, the timing of molt and extensive refueling rarely coincides (Barta et al., 2007). However, in some cases fuel deposition rate or body mass increase is little affected by molt (e.g., Holmgren et al., 1993).
Among almost all Western Palearctic passerines, the basic sequence of the complete primary molt is uniformly descendant, beginning with the innermost primary (P) P1, and the renewal of secondaries (S) begins ascendantly with S1 (Ginn & Melville, 1983;Jenni & Winkler, 2020). The few long-distance migrants that molt twice per year may replace their primaries in an ascending, descending, or eccentric sequence. The latter pattern of eccentric primary molt does not start with P1, but somewhere in the center of the primaries or in the outer primaries (Jenni & Winkler, 2020). In some passerines, the starting point of primary molt may be shifted from P1 to a middle primary (divergent sequence: renewal proceeding simultaneously both inwards and outwards) or to the outermost P9 (ascendant sequence; Kulaszewicz & Jakubas, 2015;Neto & Gosler, 2006;Stein, 2012;Steiner, 1970;Thomas, 1977). However, divergent molt is still exceptional among Western Palearctic passerines, and the occurrence of this molt sequence might be correlated with accelerated molt speed (Kiat, 2017).
The need for rapid molt might be highest for species breeding in highly seasonal habitats, with limited time on the breeding grounds.
Strongest seasonal changes in climate are found around the Siberian "cold pole," with temperature ranges of more than 100°C. One of the latest arriving species breeding in Siberia is the Pallas´s Grasshopper Warbler Locustella certhiola, which stays only around two months on the breeding grounds (Bozo et al., 2019;Kennerley & Pearson, 2010;Sleptsov, 2018). Geolocator tracking revealed a very rapid migration in this species: A breeding bird from the Russian Far East covered a distance of almost 5,000 km in less than one month during autumn migration (Heim et al., 2020). These extreme time constraints might have led to the evolution of a peculiar molt strategy, but little information is available so far.
The Pallas's Grasshopper Warblers are known to undergo a complete prebreeding molt in which up to four primaries may be grown simultaneously during normal descendant molt in March-April, shortly before spring migration (Bub & Dorsch, 1995;Nisbet, 1967;Stresemann & Stresemann, 1972). On the other hand, Stresemann and Stresemann (1972) stated that adult Pallas's Grasshopper Warblers undergo a complete molt in autumn as well, and two individuals in active primary molt were found in November in Myanmar (Williamson, 1976). Kennerley and Pearson (2010) reported adult birds from Hong Kong in September which had three outer primaries presumably replaced before migration. Chernyshov (2017) reported an individual trapped on 25 August at Chany Lake in West Siberia with freshly molted primaries, tertials, rectrices, and body feathers. The westernmost breeding population may have the longest migration route of all populations and is supposed by Dementev and Gladkov (1954) to have a complete molt twice annually. In addition, Svensson (1992) mentioned also two birds from Turkestan in August undergoing an advanced active postbreeding molt and wondered whether adults "might undergo an extremely rapid complete (or nearly complete) molt in late summer-early autumn" (Svensson, 1992). Furthermore, four adults with initiated but arrested wing feather molt were caught during autumn migration in Beidaihe, China (Norevik et al., 2020). These observations, though based on few individuals, suggest the occurrence of two molt seasons in the Pallas´s Grasshopper Warbler, a (partial) postbreeding molt, and a complete prebreeding molt. strategy as well as stopover behavior decisions and the trade-offs among migratory birds that are now facing a panoply of anthropogenic threats along their flyways.

K E Y W O R D S
divergent primary molt, fueling, molt migration, molt strategy, postbreeding molt, stopover Here, we ask the question whether the extreme time constraints resulting from a very short stay on the breeding grounds and rapid long-distance migration have led to specific adaptations in the molt pattern of the Pallas´s Grasshopper Warbler. First, we analyze the extent and direction of primary molt and the frequency of observed molt patterns based on individuals caught at six sites situated along the migration route. Second, we analyze the correlation of postbreeding molt with body condition and fuel loads.

| Study sites
We compiled data on molt and body condition of Pallas´s Grasshopper Warblers from a total of six sites situated along its migration route (Table 1): from a breeding site at the northwestern limit of its distribution (Mirnoye, Central Siberia), at breeding TA B L E 1 Extent of postbreeding molt in wing and tail feather tracts of the Pallas´s Grasshopper Warblers at six sites along the migration route (PP = primaries, SS = secondaries, TT = tertials, TF = tail feathers).  Figure 3). Birds were caught with mist-nets without using playback.

| Data collection and selection
We included only adult birds, as juveniles do not molt their flight feathers in autumn (Bub & Dorsch, 1995;Svensson, 1992). Aging in autumn was based on yellowish underpart color and dark spotting on the upper breast and lower throat in juveniles (Svensson, 1992).
Body mass was taken to the nearest 0.1 g with an electronic scale.
Visible subcutaneous fat load in the tracheal pit and on the abdomen was estimated on a nine-point scale (0-8) according to Kaiser (1995).
Most detailed information on molt was collected at the study site in the Russian Far East. Here, the molt of remiges was recorded by attributing a score from zero (old feather) to five (completely grown new feather) to all flight feathers (Ginn & Melville, 1983). The total molt score (TMS) was determined for each bird by adding the molt scores of the individual nine long primaries and the six secondaries of the left wing, but excluding the very small outermost primary (P10), to give a maximum TMS of 45 for primaries and 30 for secondaries.
When molting birds (i.e., birds in active wing feather molt) were recaptured, molt score was again calculated and used to estimate molt progress per day, calculated as the difference in TMS divided by the number of days between capture and recapture. Due to a lack of data on feather mass in this species, it was assumed that the primary molt score was linearly related to time (but see Newton, 2009;Underhill & Zucchini, 1988).

| Statistical analysis
We compared the percentage of individuals showing postbreeding primary molt and the number of molted primaries as well as the direction of molt. In a second step, we modeled the number of molted primaries as an index for molt progress using linear mixedeffect models in R package lme4 (Bates et al., 2014). Julian day and latitude were fitted as fixed independent effects. Study site was included as random effect. We built two sets of candidate models, one including all birds trapped during the autumn migration period (15 July-25 September, based on the bird tracked by Heim et al., 2020) and one including only individuals trapped at stopover sites (excluding the likely wintering site in Thailand). We graphically examined if model assumptions were met using residual plots and evaluated model fit with the conditional R 2 and marginal R 2 (Nakagawa & Schielzeth, 2013). Fixed effects were tested with a likelihood ratio test.
To test whether flight feather molt is connected to body condition, we examined molting and nonmolting birds caught between 27 July and 14 September (a time when both molting and nonmolting birds co-occur) at the study site in the Russian Far East. The relationship between molt and body mass was estimated by linear regression using log-body mass. The correlation of molt (yes/no) with fat and muscle score was estimated using a Wilcoxon-Mann-Whitney test as both fat and muscle score were measured at ordinal scales and were not normally distributed.

| Postbreeding molt patterns
In Central Siberia, nearly two thirds of adult individuals caught in July and August (n = 26) had recently commenced their primary molt with the starting point at the center of the feather tract; therefore, the molt patterns of these 16 birds (62%) were unidentifiable (Table 1) in a divergent way (e.g., P4 > P5 > P3 > P6 > P2 > P7 > P1 > P8 > P 9), but since the inner primaries were nearly fully grown, the precise molt sequence in these individuals was unknown. One adult showed an eccentrically descendant molt pattern, that is, starting from the middle primaries and then outwards, while one individual had just started molt with P5 and P6 emerging from the sheath. Only 3 out of 16 birds had replaced all primaries (Table 1), that is, were completing primary molt except for the two outermost primaries (still growing, each with score 4). P5 and P6 were invariably molted, whereas 81% of the birds retained an unmolted P1 and 38% retained P2, too ( Figure 1c). The retained, unmolted primaries were generally located in the inner wing, regardless of whether the molt pattern was divergent or eccentrically descendant.
We found that 56% of the birds molted their tertials completely, most often following the sequence T7-T8-T9 (Table 1). Single secondaries were shed quite unsystematically and without any clear connection with the sequence and extent of primary molt. In 53% of the birds S6 was shed, descendantly followed by S5 (Figure 1c).

| Fuel loads and effects of molt on body condition
Lowest fat scores were found at the breeding site in Central Siberia, at the breeding and stopover site in the Russian Far East and at the potential wintering site in Thailand ( scores (median score: 2) of molting and nonmolting birds differ significantly (W = 1,210, p = .57), though muscle scores (median score: 2) of nonmolters were significantly higher (W = 1,356.5, p = .01; Figure S2 and Tables S1-S3).

| Postbreeding molt
This is the first study providing direct evidence for a partial postbreeding molt of the Pallas's Grasshopper Warblers. With 36 days, the duration of the primary molt (outermost primary excluded) of birds caught in the Russian Far East took place in significantly less time than the 50-60 days estimated by Nisbet (1967) for prebreeding molters in spring. The accelerated partial postbreeding molt might be caused by the late time of nesting, the long migration distance, and associated time constraints due to decreasing food availability in seasonal habitats on their temperate breeding grounds (Ginn & Melville, 1983;Newton, 2009). The shorter molt duration might also stem from differences in the extent of molt, as many individuals do not finish their primary molt on the breeding grounds, or from the low sample size. We suggest that besides a premigratory molt of local breeders prior to autumn migration, a percentage of the  (Leu & Thompson, 2002) and has been linked to aridity of the breeding grounds (Pageau et al., 2020).  (Heim et al., 2020;Yong et al., 2015). Similar observations of molt bridging autumn migration have been made in other songbirds (Elkins & Etheridge, 1977;Herremans, 1990;Kiat et al., 2018;Nisbet & Medway, 1971;Schaub & Jenni, 2000b).
Based on our data, the Pallas's Grasshopper Warbler is the first eastern Palearctic passerine for which a divergent primary molt pattern is demonstrated (Figure 1b). Regular descendant sequence has been documented in this species too, although to a much smaller extent than divergent molt. Similar observations were made in Savi´s Warblers Locustella luscinoides in Europe (Thomas, 1977). The distribution curve of molted remiges (Figure 1c) was also similar to that of Savi's Warblers showing divergent wing molt (Thomas, 1977).
According to Kiat (2017), the benefit of a divergent molt strategy is that the wing gap comes about shortly after the start of the primary molt and is later split into two smaller gaps rather than one larger wing gap during the normal descendant molt sequence. This might allow divergent molters a higher molt speed and lower aerodynamic costs than in descendant molters (Kiat, 2017(Kiat, , 2018, but see Bridge, 2011).
Retaining the central secondaries, usually the least abraded of remaining wing feathers (Norman, 1986), shortens molt duration and presumably allows resources to be diverted preferentially to primary feather synthesis. The centers of primary and tertial molt, arranged around P5/6 and T8, are thought to have an adaptive function (Mester & Prünte, 1982). In the folded wing of a primarily reed foraging species, such as the Pallas´s Grasshopper Warbler, they might serve as protective shields against forced abrasion and sun exposure for the outermost pointed primaries and the central secondaries, respectively. Simultaneously, P5 and its adjacent feathers are exposed to intense wear, quasi representatively for the important outer wing feathers regarding aerodynamic and flight efficiency (Lockwood et al., 1998).
Interestingly, the unmolted primaries of the majority of adult

Pallas's Grasshopper Warblers in Central Siberia and the Russian
Far East appeared in pristine condition. Rather than "exceptional strength" (cf. Svensson, 1992) Our results indicate that intraspecific variation in molt sequence can neither be attributed to specific regions of the species' vast breeding range, nor be attributed to a subspecies. Based on our data from the Russian Far East, plasticity in postbreeding primary molt sequence cannot be easily related to season either, contrary to observations in another species with eccentric molt (Neto & Gosler, 2006).
It should be added that in spring, apart from four actively molting birds with irregularities in the molt sequence that cannot be allocated, 16 individuals caught during February-May 1995-2014 in Thailand followed the basic molt sequence of primaries (P. Round, unpublished data). These records indicate that partial postbreeding flight feather renewal in adult Pallas´s Grasshopper Warblers in general represents a separate molt that is arrested both within primaries and secondaries, followed by a second complete prebreeding molt in the nonbreeding area, as assumed for other Locustella species as well (Jenni & Winkler, 2020;Pearson & Backhurst, 1983;Svensson, 1992).
Tail feathers in the Pallas's Grasshopper Warbler are dropped simultaneously (Nisbet, 1967), as in other Acrocephalus and Locustella warblers (Kennerley & Pearson, 2010;Round & Rumsey, 2003). The specific tail and primary molt in the Pallas's Grasshopper Warbler might be supported by its feeding mainly in reeds and dense scrub, without having to cross wide gaps and thus saving energetic costs connected to movement with temporary wing and tail feather deficits. This behavior is shared by other Locustella species (Kennerley & Pearson, 2010) and might explain why the occurrence of divergent molt seems to be more widespread in this group.

| Molt migration and body condition
New production of flight and body feathers requires protein synthesis, diverting energy (Murphy & King, 1992), which cannot be allocated to fuel reserves in preparation for migration. Therefore, average fuel deposition rates and fat scores should be lower in molting birds (Merila, 1997;Remisiewicz et al., 2018;Rubolini et al., 2002;Schaub & Jenni, 2000a). Contrarily, we found constantly moderate body mass levels and low fat scores at our study site in the Russian Far East, with no significant differences in fat scores between molting and nonmolting birds, nor any remarkable increase later in the season in either group. Although comparatively more nonmolting birds had a maximum muscle score, flight muscle scores on average remained moderate throughout the season in both groups, indicating a low protein catabolism within molting birds, presumably because the demand for protein was easily met by dietary intake (Jenni-Eiermann & Jenni, 1996;Murphy & King, 1992;Podlaszczuk et al., 2017). The Pallas's Grasshopper Warblers are migrating largely over land without facing any major ecological barriers (Heim et al., 2020;Kennerley & Pearson, 2010). According to optimality models (Alerstam, 2011;Alerstam & Lindström, 1990), this would make longer nonstop flights redundant and would instead allow a nocturnal flight strategy with low fuel loads ( Figure 3) and daily refueling in suitable habitats. Neither muscle hypertrophy in preparation for migration (Lindström & Piersma, 1993;Lundgren & Kiessling, 1985) nor the need to catabolize flight muscle protein is evident. A similar migration strategy seems to be a frequently used pattern in nocturnal passerine migrants flying across continental Europe where widespread fueling opportunities exist (Bairlein, 1995;Delingat et al., 2006;Ozarowska, 2015;Schaub & Jenni, 2000a;Stepniewska et al., 2018). Recent studies found huge variation in fuel loads among migratory warblers in East Asia, with comparatively low fuel loads in the Pallas´s Grasshopper Warblers (Bozo et al., 2020;Sander et al., 2017Sander et al., , 2019 (Nisbet, 1967). Different origins/subspecies could also play a role, as northern/western populations are somewhat larger than southern/ eastern populations (Kennerley & Pearson, 2010).  (Newton, 2011).

ACK N OWLED G M ENTS
We thank all the volunteers for the assistance during the bird ring-

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data used for the models on molt progress are available on Dryad  & Lindström, A. (1990). Optimal bird migration: The relative importance of time, energy, and safety. In E. Gwinner (Ed.), Bird migration: Physiology and ecophysiology (pp. 331-351). Springer Verlag. Bairlein, F. (1995). European-African songbird migration network manual of field methods. Vogelwarte Helgoland.