Influence of Weather on the Eurasian Woodcock’s Breeding Display

: Counts of displaying male Eurasian woodcock ( Scolopax rusticola ) form the basis for breeding Eurasian woodcock surveys in many regions across Europe and provide the only practical method of assessing the species’ abundance. This paper investigates the effect that weather may have on the results of these surveys, principally considering its influence upon Eurasian woodcock display behavior and detectability by surveyors. We assessed data from an annual Eurasian woodcock survey conducted in the Britain during 2004–2015 and correlated them with a number of weather variables. This is supplemented by tracking data gathered from 19 male Eurasian woodcock to assess how weather might affect each individual’s decision to display. We found that counts of roding Eurasian woodcock were positively related to the amount of rainfall in the 2 weeks preceding the survey and negatively related to wind speed on the evenings that surveys were conducted. The likelihood that tagged male Eurasian woodcock displayed decreased in relation to wind speed and increased in relation to minimum air temperature. To guarantee that counts of displaying males provide a representative measure of abundance, we recommend that surveys consist of at ≥3 visits to each site within each year, that visits are spread as widely as possible across the peak displaying season, and that analyses are based on maximal counts rather than means to reduce the effects of surveys conducted in sub-optimal weather conditions.

The breeding display flight of the Eurasian woodcock (Scolopax rusticola) offers a unique opportunity to survey a species that is otherwise rarely observed. Eurasian woodcocks' cryptic plumage and behavior means that encounter rates during the breeding season are generally low and that traditional diurnal bird surveys, such as those used for other woodland birds (e.g., Hewson et al. 2007, Hewson andNoble 2009), are unrepresentative. The display flight, known as 'roding' , contains conspicuous visual and vocal signals that provide an effective means of assessing site occupancy and the relative abundance of Eurasian woodcock.
Display flights are performed over roding grounds ranging in size from 43 to 134 ha (Hirons 1980) and are accompanied by a continuous repeated 2-part call (Hoodless et al. 2008). The roding grounds of several males may overlap (Hirons 1980) and the behavior appears to be a signal for female mate selection rather than male-male territoriality (Hirons 1980). Males perform their displays at dawn and dusk between February and July, with roding behavior peaking during May and June (Hoodless et al. 2006).
Counts of roding Eurasian woodcock form the basis of breeding Eurasian woodcock surveys in several European regions, including Switzerland, France, the Azores, and western Russia (Estoppey 2001, Machado et al. 2008, Fokin and Blokhin 2013, where measures of Eurasian woodcock activity or 'roding intensity' are used as an indicator of Eurasian woodcock abundance. The Eurasian woodcock is a quarry species in most European countries (Lutz 2005), and these surveys can be used to guide more sustainable hunting practices, particularly where migrant and resident Eurasian woodcock populations overlap.
A national survey of Eurasian woodcock in Britain, for instance, uses a calibration equation to convert surveys of roding activity into estimates of population size (Hoodless et al. 2009), and found a 29% decline in the British breeding population between 2003 and 2013 (Heward et al. 2015). Roding surveys are the only effective way of detecting breeding population trends as alternative methods conducted in winter, such as hunting bag analyses (e.g. Seamans and Rau 2017), are confounded by the large numbers of migrant Eurasian woodcock that winter alongside British breeders (Hoodless et al. 2013). In Britain, voluntary guidelines for hunters were produced in an attempt to minimize the impact of hunting on the declining resident component of the wintering population (Game & Wildlife Conservation Trust 2017). Continued monitoring of the resident population will help gauge the effectiveness of these measures and, if necessary, could form the grounds for tighter control in the future.
The British breeding Eurasian woodcock survey was conducted at approximately 800 randomly selected sites in 2003 and 2013 (Heward et al. 2015). A non-random sub-sample of these sites, surveyed on an annual basis in the intervening years, demonstrated a yearly decline of 4.9% (Heward et al. 2015). However, these annual surveys were also characterized by a high degree of annual variation, and weather was suggested as a possible cause for these fluctuations. There is already evidence that roding activity can be influenced by weather conditions during the roding period, although Hoodless et al. (2006) suggest that these must be relatively severe before they have a significant effect. If more subtle variations in weather influence the roding behavior of Eurasian woodcock, these need to be considered in the design and interpretation of roding Eurasian woodcock surveys. Moon phase may also influence the behavior of crepuscular species, although this has been demonstrated more clearly for mammals (Prugh and Golden 2014) than for birds (Mills 1986). We propose 4 ways in which weather might influence the results of roding Eurasian woodcock surveys: 1) if weather has a direct impact on survival or productivity in the previous breeding season, 2) if the prevailing weather prior to a survey affects individual fitness and therefore the likelihood of display behavior, 3) if weather during the survey period affects the likelihood of display behavior, or 4) if weather during the survey period affects detectability by the surveyor.
We investigate the relationship between roding survey results and weather, using Eurasian woodcock survey data collected over a 12-year period. The influence that weather has on the display behavior of male Eurasian woodcock is assessed using the tracking data gathered from a sample of 19 male Eurasian woodcock tagged with archival GPS loggers.

REpEAT RODing EuRASiAn WOODCOCK SuRvEYS
Roding Eurasian woodcock surveys were conducted at woodland sites across Britain between 2004 and 2015. These followed a large-scale national Eurasian woodcock survey conducted in 2003 in which counts of roding males were made at 807 randomly-elected 1 x 1-km squares across England, Scotland, and Wales (Hoodless et al. 2009). Observers taking part in the 2003 survey were encouraged to continue their counts on a yearly basis, resulting in a sub-sample of sites, all of which were occupied by Eurasian woodcock, receiving repeated visits in subsequent years. These were supplemented by 2 additional sites conducted in non-random squares. Ideally, these repeat surveys would have been conducted at an entirely random selection of sites, but this was not possible without increasing the risk of reduced uptake or poor continuity of surveys.
Each survey site was visited up to 3 times per year between the start of May and the end of June, with each visit ≥1 week apart. Volunteers selected a suitable count point where roding could be observed, generally a woodland clearing or the intersection of 2 open trails where the absence of canopy cover allowed a clear view of the sky (British Trust for Ornithology 2013). The same count point was used in successive years unless the development of surrounding vegetation meant that the point had to be moved to a new, more suitable location nearby. Volunteers recorded each separate occasion that a Eurasian woodcock was seen or heard as a single 'registration' .
We obtained daily rainfall data from the Centre for Ecology and Hydrology's (CEH) CEH-GEAR dataset (Tanguy et al. 2016) and, for each survey, we calculated the total rainfall on the day of the survey ('survey day' d = 0), total rainfall for the week preceding the day of the survey ('survey week' = d -7 to d -1), and for the week before this ('previous week' = d -14 to d -8). For the same periods, we extracted mean and minimum temperature from the Met Office's UKCP09 dataset (Met Office 2017a). Both datasets are interpolated raster maps based on observations from national Met Office weather stations.
We sourced moon phase data from a moon phase generator available online (somacon.com 2011). This gives the date of each full moon and new moon within the study period. We then used these dates to assign surveys to 1 of 3 categories: if conducted on the night of a new moon or full moon, or within 3 days either side, they were classified as 'new' or 'full, ' respectively, otherwise, they were classified as 'transitional' .
We placed survey sites into 1 of 4 geographic regions based on their location: Scotland, Northern England, East Anglia and the Midlands, and South-east England. These regions derive from natural groupings in the distribution of repeat survey sites and follow logical geographic divisions ( Fig. 1). We extracted the total area of woodland within a 1-km radius buffer of the count point from the CEH's Land Cover 2007 dataset (Morton et al. 2011). Woodland area is known to have a strong effect on the number of Eurasian woodcock registrations recorded at a site (Heward et al. 2015), hence its inclusion in the analyses described below.
We analyzed data using a generalized linear mixed model (GLMM) with a Poisson distribution and a logarithmic link function. The analysis was conducted in R (R Development Core Team 2016) using the glmmPQL function within the Mass package (Venables and Ripley 2002). We specified the number of Eurasian woodcock registrations recorded on each survey as the dependent variable. The fixed effects were mean temperature on the survey day, during the survey week, and previous week; total rainfall on the survey day, during the survey week, and previous week; moon phase; cloud cover and wind during the survey as recorded by the surveyor; day of year; and wooded area. We specified site, nested within region then year (with year as a categorical variable), as a random effect. We did not include rain during the survey, as recorded by surveyors, as a factor given a very heavy bias towards surveys experiencing no rain at all (n = 838 / 886). We also re-ran this model substituting minimum temperature for the mean temperature and the natural log of the minimum temperature.

TRACKing MAlE EuRASiAn WOODCOCK WiTh gpS lOggERS
We captured Eurasian woodcock in mist-nets in 2 large woods (each ca. 10-14 km²), in Nottinghamshire, UK (Lat: 53.186 N Long: 1.093 W) during the breeding seasons of 2015-2016 (Heward et al. 2017). We tagged 19 male Eurasian (2015: n = 5, 2016: n = 15 (1 individual was tagged in both years)) with a 7.5-g tag combining an archival GPS logger (Lotek PinPoint 50, Newmarket, Ontario, Canada) and a VHF radio (Perdix Wildlife Supplies, Stratford-Upon-Avon, UK). For most captured Eurasian woodcock (68%), roding behavior immediately prior to capture indicated that the individual was a male. When we could not determine sex based on behavior, we identified males based on the appearance of the cloaca [as described for other Charadriiformes (Tomkovich andSoloviev 1996, Giunchi andPollonara 2007)], lack of a brood patch, or using biometric data (Ferrand and Gossmann 2009). We aged captured individuals as juvenile (first-breeding season) or adult (≥2 breeding seasons) based upon their wing molt (Ferrand and Gossmann 2009). We attached tags to the skin of the synsacrum and a small number of surrounding feathers using a gauze mount and cyanoacrylate glue. Tagging was conducted under a license issued by the British Trust for Ornithology.
GPS loggers began recording 5 days after tagging, allowing individuals to acclimatize to the tag. Locations  were recorded at 1-minute intervals during the typical dusk display period; beginning 15 minutes before sunset and continuing for 90 minutes. This schedule was repeated every subsequent evening for 1 to 6 days depending upon the rate of battery depletion. We recovered tags by recapturing the individual or by using the VHF radio signal to relocate tags once shed.
We obtained weather data from the UK Met Office's observation station at Watnall, UK (Lat: 53.005 N Long: 1.250 W; Met Office 2017b), approximately 20 km from the tracking locations. We extracted daily rainfall, daily minimum temperature, and daily mean wind speed for each evening on which tracking data were recorded.
We classified each Eurasian woodcock, for each evening it was tracked, as 'roding' or 'non-roding' based upon its tracking data; for most individuals the distinction between the 2 types of behavior was obvious. Tracking data from roding individuals showed a period of almost constant movement for ≥30 minutes during the core of the roding period. For non-roding individuals, movements were short in duration and direct, typically consisting of a single flight between a roost site and a presumed feeding area (Fig. 2).
We assessed the correlates of 'roding' and 'non-roding' behavior using a binomial GLMM with a logit link function in R's MASS package. The independent variables tested were total rainfall, minimum temperature, and mean wind speed on the day that tracking data were recorded, and the individual's age. We specified 'individual' as a random effect to account for differences among tagged individuals. As a comparison, we created a 'null' model in which age and day of year were the only explanatory variables.

RODing EuRASiAn WOODCOCK SuRvEY SiTES
In total, 65 survey sites received repeated visits (Fig. 1). Few sites were visited every year between 2004 and 2015, but on average each site was surveyed 5.0 different years during the 12-year period. We excluded from analysis sites that were visited in fewer than 3 of the 12 years. The average number of sites visited per year was 26.6. Because of their self-selected nature, the average number of contacts with Eurasian woodcock recorded at these 65 sites [mean registrations = 9.57 ± 0.26 (1 SE)] was higher than the national average [mean registrations = 7.45 ± 1.03 (1 SE) at occupied sites] recorded in the large-scale national survey of 2003.

RODing SuRvEYS AnD WEAThER
The multivariate analysis of Eurasian woodcock survey data revealed significant relationships with 3 weather variables: 2 measures of rainfall and the wind speed variable. There was a positive relationship with rainfall during both the survey week (glmmPQL: t 550 = 2.81, P = 0.005, 0.051 ± 0.018) and the previous week (glmmPQL: t 550 = 2.70, P =  0.007, 0.046 ± 0.017) but no relationship with total rainfall on the survey day.
No significant relationships were found between the number of Eurasian woodcock registrations recorded and the minimum temperature variables. When the model was re-run using the natural logs of minimum temperature, however, the temperature during the survey week was associated with the number of Eurasian woodcock registrations (glmmPQL: t 550 = 2.19, P = 0.029, 0.055 ± 0.024). In the model using mean, rather than minimum, temperatures we found no significant relationships with temperature, and relationships with other weather variables remained consistent with those in the model that included minimum temperature.
Moon phase, day of year, and total wooded area were not related to the number of Eurasian woodcock registrations, nor was cloud during the survey, but for the latter, P-values were close to statistical significance when comparing between cloud cover of 66-100% and 0-33% (glm-mPQL: t 550 = -1.90, P = 0.058, -0.083 ± 0.046).

TAggED EuRASiAn WOODCOCK AnD WEAThER
The probability of Eurasian woodcock roding was related to minimum temperature and mean wind speed. As the daily minimum temperature increased, so did the likelihood of display behavior being observed (glmmPQL: t32 = 3.51, P = 0.001, 3.574 ± 1.016; Fig. 4a). For tagged individuals, roding became less likely as wind speed increased (glmmPQL: t32 = -2.25, P = 0.031, -1.711 ± 0.757; Fig. 4b). There was no significant effect of rainfall or Eurasian woodcock age. In the 'null' model, containing only age and day of year, neither variable was significantly related to probability or roding.

Discussion RAinfAll
The number of Eurasian woodcock registrations recorded during roding surveys was positively related to rainfall, but only when rainfall was measured over a week or more prior to the survey. Lower rainfall in spring or summer is likely to drive the soil-dwelling invertebrates on which Eurasian woodcock feed deeper into the soil (Gerard 1967, Hoodless andHirons 2007), and poorer foraging conditions may leave Eurasian woodcock with less energy to expend on display during dry spells. Dry summers have been demonstrated to have a negative effect on reproductive success in Eurasian woodcock populations (Guzmán and Arroyo 2015), but the potential effect on adult body condition remains unclear.
Had there been a significant relationship with rainfall during the survey, a negative association might be expected, assuming that heavy rain reduces the male's ability to see or hear receptive females and makes display behavior less profitable [as for tawny owls (Strix aluco): Lengagne and Slater 2002]. This was true of American woodcock (Scolopax minor), for which display activity was reduced by even light rain (Blankenship 1957). We found no relationship with rainfall on the day of the sur-Displaying Eurasian Woodcock & Weather · Heward et al. Figure 4. Predicted likelihood of tagged Eurasian woodcock roding relative to a) daily minimum temperature and b) mean wind speed, plotted using the 'effects' package (Fox 2003) in R. This accounts for the relative effect of the 3 other explanatory variables and 1 random effect included in the model. Dashed lines signify 1 SE. A B vey, but this was probably because surveys were not conducted during heavy or persistent rain, and most experienced no rainfall at all, reducing the degree of variation in our model. Our tracking data also showed no correlation with rain on the day of recording, although our daily rainfall variable does not identify whether rain fell during the logger's recording period. Whilst rain during the roding period might inhibit roding behavior, at other points during the day it might encourage roding. This was proposed for American woodcock (Blankenship 1957) and, based on personal observation, appears to be the case for Eurasian woodcock, perhaps because rainfall during the day provides the opportunity for diurnal feeding.

TEMpERATuRE
We found that display behavior in tagged individuals was positively related to minimum daily temperature. This could be the product of an underlying relationship with the time of year, if both temperature and roding intensity vary in a linear fashion over the course of the breeding season, but our analyses are focused only on the peak of the roding season when roding activity tends not to show a consistent linear trend. This is reflected in our 'null' model, which found no direct correlation between the number of Eurasian woodcock registrations and the day of the year.
Woodcock may be less likely to rode on cooler evenings or following cooler days because the cost of maintaining a constant body temperature is greater and the demand to feed, rather than display, is increased. Woodcock are able to survive in much harsher conditions during the winter than are typically experienced in spring, but during the breeding season males carry minimal fat reserves (Owen and Krohn 1973) and bear greater energetic costs associated with the maintenance of reproductive physiology (Vézina and Salvante 2010) and the roding flights themselves. Alternatively, temperature might be a natural stimulus to breeding behavior as documented in other bird species (Garson and Hunter 1973). Whatever the reason, the same effect of temperature on display behavior is recorded for several other species (Slagsvold 1977, O'Connor and Hicks 1980, Gottlander 1987, Santos et al. 2009), including American woodcock (Blankenship 1957, Goudy 1960.
Whilst this relationship with temperature was not immediately apparent in the roding survey data, it became clear when minimum temperature values were logged. This probably suggests that the relationship with temperature is non-linear and that high air temperatures can also have an inhibitive effect on roding owing to lower rainfall, high rates of evaporation, and dry soils.

WinD
During roding Eurasian woodcock surveys in which the observer rated the wind as a light or moderate breeze, fewer Eurasian woodcock were encountered than during evenings that were still. Increased environmental noise could impede detection by surveyors (Alldredge et al. 2007), but our tracking data showed that tagged Eurasian woodcock were less likely to rode on windier nights, supporting a change in Eurasian woodcock behavior rather than just detectability. Males may be discouraged from roding on windier nights because their roding call is less likely to be heard by potential mates. Increased wind speeds were shown to influence the results of other bird surveys (O'Connor andHicks 1980, Santos et al. 2009) including the Singing Grounds Survey (SGS) used to detect American Eurasian woodcock (Bergh and Andersen, this volume, Estimating density and effective area surveyed for American woodcock), and this has also been attributed to changes in Eurasian woodcock activity rather than just detection by surveyors (Blankenship 1957, Goudy 1960).

ClOuD AnD MOOn phASE
We found no significant relationship between roding activity and moon phase. There are a few examples of moon phase influencing bird behavior, but these mostly relate to behavior that is truly nocturnal rather than crepuscular (Mills 1986, Dodd 1998, Mougeot and Bretagnolle 2000. Increased ambient light levels could reduce roding activity if associated with an increased predation risk (Mougeot andBretagnolle 2000, Prugh andGolden 2014), or lengthen or delay the roding period if they improve visual signaling between males and females. This would assume dusk light levels correspond to moon phase, which seem far more dependent upon cloud cover.
Although not confirmed by our analysis, there may be a correlation between cloud cover and roding activity, particularly because the P-values associated with cloud cover were only marginally greater than the 0.05 threshold. Light intensity was the main stimulus for the onset of display behavior in American Eurasian woodcock and starting times were earlier in relation to sunset on cloudy evenings (Duke 1966). It seems likely the same is true for Eurasian woodcock (Heward et al. In press), which would alter survey results if it meant a proportion of the roding activity fell outside of the 75-minute survey period.

Management Implications
It appears that the display behavior of Eurasian woodcock is affected by the weather and that this can, in turn, influence the results of roding surveys. Factors such as heavy rain and strong wind were already known to discourage roding (Hoodless et al. 2006) and as a result, the British breeding Eurasian woodcock survey methodology stipulates that counts should not be conducted in these conditions (British Trust for Ornithology 2013). The more comprehensive study presented here suggests that more subtle variation in survey weather can have a detectible effect on count data. Further tightening the conditions in which surveys should be conducted is a possible solution, but could result in criteria that are impractically restrictive and difficult to forecast, and could discourage survey participation. Each survey site, in each year, is usually visited on 3 separate occasions, and estimates of local density are derived from the maximum count rather than the mean. This approach should provide some buffer to the effects of sub-optimal weather providing at ≥1 count is conducted during suitable conditions. Our findings reiterate the need for all 3 visits to be completed and the exclusion of sites that do not receive them all [which has not always been the case previously (Heward et al. 2015)].
The effects of more protracted weather events, such as extended dry spells, are harder to predict. If these events affect male fitness, they are likely to affect roding behavior and, as suggested here, surveys based upon counts of displaying individuals. This is countered to some degree by the fact that repeated visits are made within the year and that each visit is separated by ≥1 week. It may be sensible to separate visits with a longer interval to ensure that counts are spread over a larger portion of the breeding season and reduce the influence of seasonal variation in weather.
When unfavorable weather conditions become particularly protracted or severe, they may affect the results of roding surveys by reducing recruitment and perhaps increasing adult mortality rather than merely reducing the likelihood of Eurasian woodcock displaying. Weather has been shown to influence breeding success (Selås 2006, Guzmán andArroyo 2015), and should therefore have an effect on roding surveys in the following year. How to separate the real demographic impacts of unfavorable weather from the behavioral responses of Eurasian woodcock based on their fitness is not currently clear. It would be possible to shed more light on these types of weather events by reanalyzing annual data to include weather variables based on the entire breeding season, not just those immediately prior to the survey. Such analyses could also incorporate measures of the weather conditions experienced in the previous breeding season and intervening winter.
There have only been 2 large-scale national surveys of breeding Eurasian woodcock in the UK to date (2003 and 2013) and these could be biased by the prevailing weather in the years they were conducted. All evidence points towards a decline in the British breeding Eurasian woodcock population that is real and long-term (Hewson and Noble 2009, Balmer et al. 2013, Heward et al. 2015, but its true scale will become clearer with each repeat of the national survey scheme as the longer time-series reduces the influence of weather in any single year. In the meantime, the complementary data provided by conducting repeat surveys at a subsample of the national survey's sites, such as those analyzed here, give some impression of annual variability in roding counts in response to weather. Repeat count data could also be used to test how successional habitat change at count points affects detection rates (Nelson and Andersen 2013). Continuing to account for weather in survey methodology, and improving the way in which this is done, will help strengthen the results of future surveys and management policies based upon them.