1. Wetlands, and the species that rely upon them, are under significant threat worldwide, with wetlands often being completely removed or drastically altered. Successful wetland management requires an understanding of the interactions between wetland species and the microhabitats they use. The use of microhabitats for thermoregulation in wetland species is poorly studied, though anthropogenic influence on wetlands can reduce the diversity of microhabitats and thus the thermoregulatory options for animals. At high ambient temperatures birds may use the water-logged wetland margins to help with thermoregulation, and are often observed roosting in the sitting position within this microhabitat. However, whether sitting on the wet substrate helps in thermoregulation is unknown.
2. In this study, we tested whether birds selectively use microhabitats across temperatures by conducting field observations of nine species of shorebirds. We use comparative analysis to determine whether birds roost more on wet substrate in the sitting posture, i.e. ‘wet-sitting’, at high ambient temperatures.
3. We found substrate type across the wetland margins to be important in shorebird thermoregulation, with the time spent sitting being significantly mediated by the substrate on which the bird roosted. Individuals tended to sit on bare, wet ground much more under high ambient temperatures compared with low ambient temperatures. Vegetation on the other hand was used similarly across temperatures, and likely does not provide the same thermoregulatory benefits.
4. By roosting on wet substrate at high ambient temperatures, birds may increase the potential for heat dissipation across the uninsulated legs, as water-logged wetland margins are known to remain cooler than the ambient temperature or vegetated microhabitats under hot climatic conditions.
5. Synthesis and applications. Wetland creation and management requires an understanding of the functional significance of such microhabitats, not only for foraging and breeding, but also for roosting. We demonstrate that managing wetland margins is likely important in minimising heat stress in birds, with our findings emphasising the importance of maintaining open spaces in habitat mosaics for birds to use for thermoregulation. The ability of wetland species to manage heat stress is becoming exceedingly important as they are threatened by both decreased wetland availability and increasing ambient temperatures under climate change.

Observations of shorebird behaviour

Behavioural observations of roosting shorebirds were conducted along the coastlines of Port Phillip and Westernport Bays, southern Victoria, Australia (Tables S1 Supplementary Material), during the austral spring and summer (October – February, 2013-14). These sites included naturally occuring wetlands, artificially created/restored wetland reserves (e.g. from former saltworks, or military bases) and coastal regions. They included several different wetland types, such as coastal saltmarshes, permanent freshwater lakes and intertidal flats (Fig. 1 and Table S1). Tidal regimes and habitats were broadly similar across study locations with both tidal and non-tidal sites, with varying degrees of vegetation density. The wetland and broad habitat classifications are summarised in Table S1 (Supplementary Material). All study sites were non-ornamental, and all had unformed margins, offering a mosaic of muddy substrates and low wetland vegetation. Across these habitats, we observed nine shorebird species from four families: sharp-tailed sandpiper Calidris acuminata (Horsfield 1821) (n = 42 individuals), curlew sandpiper Calidris ferruginea (Pontoppidan 1763) (n = 38), red-necked stint Calidris ruficollis (Pallas 1776) (n = 39), red-kneed dotterel Erythrogonys cinctus (Gould 1838) (n = 39), pied oystercatcher Haematopus longirostris (Vieillot 1817) (n = 43), black-winged stilt Himantopus himantopus (Linnaeus 1758) (n = 53), red-necked avocet Recurvirostra novaehollandiae (Vieillot 1816) (n = 41), common greenshank Tringa nebularia (Gunnerus 1767) (n = 38) and masked lapwing Vanellus miles (Boddaert 1783) (n = 42). These species have a large degree of overlap between in female and male size, and are not able to be sexed by eye. Therefore, sexes were combined in our analyses. During the austral summer, these species all forage and roost locally at the study sites; with some small scale movements between foraging areas and roosts (Rogers, Herrod, Menkhorst & Loyn. 2010, Nebel, Rogers Minton & Rogers 2013, Rogers, Loyn & Greer 2013). Variations in tidal height are likely captured evenly for each species across the data collected, due to the birds mostly roosting at high tide (Rogers et al. 2006b) and filming being undertaken across a five month period. We used roosting birds only, with no birds demonstrating breeding or courtship behaviours. Birds were filmed from a stationary vehicle, from a bird hide or by approaching slowly on foot (for birds ~30m or more away), waiting five minutes before filming commenced.

We filmed roosting birds with a Canon 60D DSLR camera using a Canon 100–400 mm lens (Canon Inc., Tokyo, Japan) or ‘digiscope’ (Swarovski ATM 80HD Spotting Scope, Swarovski Optik, Absam, Tyrol, Austria). These data were used to analyse roosting behaviour across species, with interspecific comparison of several postures presented in Ryeland et al. (2017, 2019). For the purpose of this paper, we aimed to understand whether birds used the sitting posture selectively across temperatures and roosting mediums. As such, birds were categorised to be either roosting in the sitting or standing posture (Fig. 2), noting whether they roosted on bare, water-logged wetland margins (generally consisting of sand or mud), in the shallow water on the edges of the water bodies or, on low, predominately dense vegetation such as saltmarsh. The substrate on which the bird was roosting (‘wet ground’, ‘water’ or ‘vegetation’) was noted during filming. Microclimate measurements of different substrates, including substrate (upward) radiation, downward (solar) radiation or wind speed at the height and location of the bird roosting, were unable to be recorded to provide detailed quantifications of subtrate microclimate. Gaining operative or microclimate measurment would cause disturbance to the bird, precluding behavioural observations. As such, we took measurements of ambient temperature and wind speed at the (nearby) location of the observer (5 – 580m). Our findings may therefore not be able to used for detection of exact thermal limits, but instead indicate the overall relationship between temperature, body posture and microhabitat. We took temperature and wind speed  parameters at 5-min intervals during each video bout using a Kestrel 3000 pocket weather meter (Nielsen-Kellerman Company, Boothwyn, PA, USA), and then averaging the values for each bout. Ambient temperature and wind speed has been shown in previous studies to effect the use of thermoregulatory behaviours (Anderson & Williams 2010, Javůrková et al. 2011, Yorzinski et al. 2018).

The complete filming parameter details are described in Ryeland et al. (2017). Filming bouts were a maximum duration of 25 min (17.28 ± 7.60; mean ± SD) with a minimum of six individuals of each species observed for each of six ambient temperature brackets (11 – 15, 16 – 20, 21 – 25, 26 – 30, 31 – 35 and ≥ 36 C). We included any bird that was considered to be ‘roosting’, classified by a condition of being immobile in standing (on one or both legs) or sitting position for > 120 s, regardless of what micohabitat the bird was in. The posture of the bird, sitting or standing, was recorded across each filming season, recording changes in postures throughout this period. If the bird was roosting in dense vegetation, or other areas where it could not easily be observed (e.g. behind a rock), the bird was excluded from the analyses as we could not determine roosting posture. The number and duration of video bouts across species is summarised in Table S2 Supplementary Material.

Video footage was scored giving the duration and frequency of roosting postures using iObserver’ app for iPad, Skware, 2011. In many cases, multiple birds were filmed within the same frame. Individuals from a video were choosen by first assigned individuals a number sequentially from left to right then top to bottom and then selecting individual numbers using a random number generator. Selecting individuals in this way limited the chance of overrepresentation of any specific posture (i.e. always choosing a bird that was mostly sitting). The total proportion of each video bout the bird was sitting was then calculated.

Statistical analysis

To evaluate the relationship of ambient temperature and substrate on the use of the sitting posture across species, we employed a Bayesian phylogenetic generalised linear mixed modelling approach, implemented through the package MCMCglmm (Hadfield 2010) in R. This method controls for the statistical non-independence of related species by including species identity as a random effect. We used the phylogeny derived from the Global Phylogeny of Birds website – birdtree.org (Jetz, Thomas, Joy, Hartmann and Mooers 2012). We first created a major rule consesus phylogeny from 2000 downloaded putative phylogenies from the pseudo-posterior distribution of trees using the Hackett et al. (2008) ‘backbone’, constructed with Mesquite software (Maddison & Maddison 2010). Branch lengths were assigned using the Grafen (1989) algorithm whereby the depth of each node in the tree is equal to the number of daughter species derived from it. The constructed phylogeny is depicted in Table S3 Supplementary Material.

We used the proportion of time per observation period spent roosting sitting as the response variable, and ambient temperature and substrate (factor levels: vegetation, wet ground and water) as predictors, with mean wind speed and body morphometrics as covariates. We also included an interaction term between substrate and temperature to investigate whether the reponse to temperature differed depending on the subtrate on which the bird was roosting. We included body mass and tarsus length as morphometric covariates as they have been previously shown to effect the use of thermoregulatory behaviours (Ryeland et al. 2017, 2019). Although these measurements are correlated (r = 0.7), as this is the most appropraite method to control for body size effects when interested in the effect of tarsus length relative to body size, as we are here (see Freckleton 2002). Mean tarsus size (length from the inner bend of the tibiotarsal articulation to the base of the toes) and body mass was derived from the literature for each species (Marchant & Higgins 1993, Higgins & Davies 1996) using weighted species averages for live birds (both sexes combined) with Victorian captured birds used in preference (where data were available). Tarsus is the only leg measurement routinely measured in shorebirds (Baker, Dettmann, Scotney, Hardy, & Drynan 1997) and as such, tarsus length was used as a proxy for leg length, having been shown previously to predict for roosting postures (Pavlovic et al. 2019, Ryeland et al. 2019). Migratory shorebirds generally increase in mass over the austral summer in preparation for migration (i.e. red-necked stint, common greenshank, sharp-tailed and curlew sandpipers), which could result in an increased propensity for these species to sit at the end of our study period when they are heavier (Pavlovic et al. 2019). However, we could not take morphometrics from individual birds at the time of recording and as such, we use an avergae body mass across seasons. We also capture a wide range of temperatures both prior to summer (October, just after return from migration) and in late summer (February, just prior to migration) (Figure S4 Supplementary Material), and as such, have captured extremes in temperaures when the birds were at their lightest and heaviest.

As the time spent sitting is proportional to the length of the video bout (and restricted to values between 0 and 1), we modelled the response variable (proportion of time sitting) as a binomial response with logit-link function (family = “multinomial2” in MCMCglmm). This treats each minute of observation as a ‘presence’ (i.e. the bird was sitting) or ‘absence’ (i.e. the bird was standing) within a fixed number of Bernoulli trials (total minutes of observation). We standardised mean ambient temperaure and mean wind speed, and included ambient temperature with a polynomial term to reflect previous observed patterns of the use of the sitting posture in the mid-range temperatures (Ferns 1992). Video bout identity was included as a random effect in the model, as multiple birds were observed from particular filming sessions (see Ryeland et al. 2017) and the possibility exists that individuals in close proximity during video bouts may have influenced each other's behaviour (Beauchamp 2007, 2008).

For the phylogenetic GLMM we ran the model with 401,000 iterations with a burn-in of 1,000 and thinning interval of 400, using uninformative priors. A Bayesian posterior distribution was generated from a subsequent effect sample size of 1,000, with convergence checked by visual inspection of plots of parameter estimates. This model produced a Bayesian posterior mean estimate, with 95% credibility intervals for the distribution of estimates, and a probability estimate (pMCMC) which is two times the probability that the posterior mean is less than or greater than zero (whichever is the smaller). All R code used in analyses and the raw data are presented in Table S2 Supplementary Material.