Proximity of breeding and foraging areas affects foraging effort of a crepuscular, insectivorous bird

When complementary resources are required for an optimal life cycle, most animals need to move between different habitats. However, the level of connectivity between resources can vary and, hence, influence individuals’ behaviour. We show that landscape composition and configuration affect the connectivity between breeding (heathlands) and foraging habitats (extensively-grazed grasslands) of the European Nightjar (Caprimulgus europaeus), a crepuscular insectivorous bird. On a daily basis, nightjars connect breeding and foraging sites by rapidly crossing unsuitable habitats in order to exploit a higher prey biomass in foraging sites. However, low availability of foraging habitat near breeding sites and clustered landscapes greatly increase foraging distance. Birds occupying these sub-optimal breeding areas compensate for longer travels by increasing foraging duration, and their physiology shows increased stress levels. All findings suggest that landscape heterogeneity can affect population dynamics of nightjars. Therefore, we recommend an integrated management approach for this EU-protected bird species.

time in foraging their fitness can reduce 6,16,20,28,29 which could be reflected in biomarkers of oxidative stress [30][31][32] . In this study, we aim to further disentangle the foraging behaviour of nightjars in Flanders, using GPS technology, refined habitat maps and measures of plasma anti-oxidants. We tested whether 1) both measures of landscape heterogeneity: i.e. habitat composition, and habitat configuration, influence nightjars' foraging distance (see sections on "Landscape heterogeneity"); 2) landscape heterogeneity affects nightjars' foraging behaviour (i.e. foraging distance, foraging time and habitat specific travel speed; see sections on "Foraging ecology"); 3) nightjars' specific choice of foraging habitat could be explained by habitat-dependant food availability patterns (see sections on "Foraging economy") and 4) landscape heterogeneity could induce increased stress levels when birds occupy sub-optimal breeding areas (see sections on "Landscape economy").  Fig. 1). The sites are located approximately 15 km apart (Fig. 1) and were selected because they are different as to their configuration and composition of functional habitats (Fig. 2). Functional habitats are described following the criteria of Evens et al. 5 : 1) breeding habitat (dry heathland, inland dunes and forest clearings); 2) roosting habitats (pine stands); 3) foraging habitat (extensively-cultivated farmlands, oak and poplar stands, recreational areas, wet heathlands, swamps and riverine valleys); and 4) unsuitable habitat (urbanised areas, intensively-cultivated farmlands and anthropogenic water  Habitat composition was calculated from the amount of habitat that was available for each initial foraging flight (average foraging distance).

GPS
SCiEnTifiC RePoRtS | (2018) 8:3008 | DOI:10.1038/s41598-018-21321-0 bodies). Most breeding habitats are included as Special Protected Areas in Natura 2000 (75-80% Bosland, 100% Meeuwen-Gruitrode and 100% Mechelse Heide). The study sites are different as to their configuration of breeding habitats (fragmented vs. continuous) and as to area of unsuitable habitats that separate breeding sites from foraging sites (high vs. low; Fig. 2). The centre of Bosland consists of a mosaic landscape (2500 ha) with coniferous trees (82%), deciduous trees (14%) and fragmented heathlands (4%). In Bosland, large areas with potential foraging habitat are located in proximity to breeding areas 5 . The military area of Meeuwen-Gruitrode and the Mechelse Heide hold two of the most important continuous heathlands in Flanders (circa 1000 ha and 800 ha respectively). Potential foraging habitats in Meeuwen-Gruitrode are found adjacent to the breeding habitats and consist of remnant fragments within agricultural land. Larger areas with potential foraging sites are found in the Mechelse Heide, but pine forests, industrial areas and cities separate them from breeding sites.
During our study we captured 48 nightjars (Bosland: 29; Meeuwen-Gruitrode: 8; NPHK: 11; Supplementary  Table T1) using ultra fine mist-nets (Ecotone, 12 × 3m) and tape lures. All birds were captured within their territories and marked with a unique alphanumeric ring from the Royal Belgian Institute of Natural Sciences (RBINS). We noted individual information such as sex (male or female), age (older than 1st calendar year (CY), 2nd CY, or > 2CYs), weight (to the nearest 0.1 gram) and wing length (measured to the nearest millimetre). We fitted a Pathtrack Ltd. nanoFix or Biotrack Ltd. PinPoint-40 GPS-logger dorsally between the wings with a full body harness made of Teflon ribbon (Bally Ribbon) 33 . Tags weighed less than 3% of the average weight of tagged birds (72.10 ± 8.68 g; for a list of tagged birds see Supplementary Table T1). GPS-tags were programmed to start logging 24-hours after deployment, from dusk until dawn. Biotrack Ltd. PinPoint V1 loggers were programmed to log a total of 40 locations, once every 15 minutes, in one night. The other tags were able to log a bird's movement from three up to six nights, depending on the sampling interval (3 or 4 minutes for Biotrack Ltd. PinPoint-40 V2 and Pathtrack nanoFix V1 & V2) and the type of vegetation cover, as shown by Forin-Wiart et al. 34 . We performed attempts to recapture the birds one week after deployment. Upon successful recapture, tracking data were downloaded and imported in Quantum GIS V2.12. Environmental data were derived from nearest (approximately ten kilometres) online weather stations in Hechtel (Bosland), Bree (Meeuwen-Gruitrode) and Maaseik (NPHK).
Foraging ecology. We followed seven steps to calculate foraging distance, flight speed and foraging time (see Supplementary Methods M1 for full details of these calculations). For each night and each bird we defined foraging events with initial foraging flights as a measure of foraging distance. Initial foraging flights were defined as the first flights of the night, starting from breeding/roosting sites towards the foraging habitats and subsequent flights performed by birds that were present for at least one hour at the breeding/roosting site after they performed previous foraging flights. We defined complete foraging tracks (see below) as movements that include the start at breeding/roosting site, flight towards foraging habitats, foraging, return flight to breeding site and arrival at breeding/roosting site. An individual can perform more than one complete foraging track per night.
Landscape heterogeneity. We created a tailored structural and functional habitat map 12 per study site. The structural landscape map is based on the Biological Value Map of Flanders 35 and we reclassified 7500 habitat types into 23 relevant habitat types, following the criteria of Evens et al. 5 . We then created the functional landscape map by grouping the 23 habitat types into four functional categories following the criteria of Evens et al. 5 .
For each foraging event (see above), we cut out four circular sub-maps from the structural and the functional habitat map, to delineate the area of available habitats and to derive measures of local landscape heterogeneity related to that foraging event 5 (Supplementary Methods M2). The centre of each circle was placed at the start position of the corresponding foraging event. In other words, we created two sub-maps (one from the structural and one from the functional habitat map) with a radius equal to the initial foraging distance and two sub-maps (one from the structural and one from the functional habitat map) with a radius equal to the average foraging distance (1650 m; calculated as the average foraging distance of all initial foraging flights for all individuals).
Finally, from the sub-maps, we quantified the percentage of available foraging habitat and habitat diversity (Shannon diversity index) as measures of habitat composition. To measure habitat configuration we classified the spatial arrangement of functional habitats into three categories using Moran's I 36 . Moran's I measures spatial autocorrelation and indicates whether the configuration of habitat types is 1) random (p ≥ 0.05) or non-random (p ≤ 0.05), with similar habitat types to be 2) clustered (z-score ≥ 1.96) or 3) dispersed (z-score ≤ −1.96). Furthermore, we calculated the average patch size of foraging habitats as a second measure of habitat configuration. We used the sub-maps with available functional habitats (average foraging distance) to calculate the average rea of each functional habitat type for each study area (Fig. 2).
Foraging economy. We quantified the abundance of nocturnal insects in breeding, foraging and roosting habitats during three consecutive years (2011-2014, May-August) in Bosland. At least three times per week we used eight insect traps with 15watt UVA-lamps (attraction radius for photosensitive insects is approximately five meters 37 ) to catch nocturnal insects from dusk until dawn. Insects were trapped in a mixture of 70% ethanol, 30% water and a drop of detergent. In this study we focussed on moths (Lepidoptera), as these constitute the main diet of nightjars 38 . Specimens of moths were removed from the liquid at dawn, dried and wing lengths were measured to the nearest millimetre. To estimate biomass, we used wing length as a measure, as proposed by Garcia-Barros 39 .
Other insects were picked out and have been preserved for other, general biodiversity studies.
We collected blood samples from adult nightjars in all study sites during the breeding season of 2016. A maximum of 300 µl blood was extracted from the brachial vein a few minutes after the bird was captured. The blood was immediately centrifuged in the field and the plasma and red-blood cells were stored separately in liquid nitrogen. Within 24-hours the samples were stored at −80 °C until laboratory analysis, which occurred within five months after sampling. We used the -SHp test (Diacron International) to quantify the plasma concentration SCiEnTifiC RePoRtS | (2018) 8:3008 | DOI:10.1038/s41598-018-21321-0 of total thiols (e.g. albumin, lipoic acid and glutathione, expressed as micromolar of -SH groups), a biomarker for an individual's fitness 31,32 .
All methods were carried out in accordance with relevant guidelines and regulations as licenced by the Royal Belgian Institute for Natural Sciences and the Flemish Agency for Nature and Forest.

Statistical analysis. Landscape heterogeneity.
To investigate the effect of individual, environmental and landscape characteristics on foraging distance we fitted linear mixed models to our data (LMM 40 ). Model selection was based on a backwards selection procedure. Due to multicollinearity issues the selection process was initiated from two separate models. More specific, correlation was found between the following environmental variables (for more details concerning the variables, see below): the amount of available foraging habitat and Moran's I, the amount of available foraging habitat and the average size of foraging sites, habitat diversity and Moran's I, and habitat diversity and the size of foraging sites. Before variable elimination, model  Foraging ecology. Using LMMs, we also assessed possible differences between the three study sites regarding foraging distance (log-transformed; random effect: individual) and the area of suitable foraging habitat in proximity to breeding/roosting sites (square root-transformed; average foraging distance; random effect: individual). The effect of foraging distance on foraging duration (log-transformed) was analysed using an LMM (random effect: individual nested within location). Similarly, habitat-specific flight speeds were compared between unsuitable (locations with dense forest, agricultural and urban area) and suitable habitats (locations with heathland and grassland; binary explanatory variable: habitat suitability, random effect: individual).
Foraging economy. To evaluate the influence of habitat type and temperature on the observed insect biomass (log-transformed), another LMM was used (fixed effects: habitat type, temperature and year; random effect: trap number). Finally, we performed an ANOVA to assess whether the plasma concentration of thiols differed between the three research areas.
For all LMM's approximate F-tests were used to assess significance. Model assumptions for both LMM and ANOVA analysis were thoroughly checked, via plots of marginal and conditional residuals against fitted values to assess homoscedasticity and linearity. QQ plots of both types of residuals were used to evaluate normality. Correlation tests and variance inflation factor (VIF) investigations were used to detect problems with multicollinearity 40 . Post-hoc pairwise comparison p-values were adjusted via the Tukey-Kramer method to correct for inflation of Type I errors caused by multiple testing 42 . Lastly, for all hypothesis tests a significance level of 5% was used.
The datasets generated during the current study are available from the corresponding author on reasonable request.

Landscape heterogeneity.
Habitat composition is significantly different between the three research areas (df N = 2, df D = 181, F = 38.98, p < 0.0001), with the availability of foraging habitat in proximity to breeding/roosting sites (expressed in average cover percentage) being significantly lower in NPHK (mean = 3 ± 4.5%, original scale) compared to Bosland (19.6 ± 6.7%; t = −8.82, p < 0.0001) and Meeuwen-Gruitrode (14.7 ± 1.3%; −4.23, p = 0.0001; Fig. 2). Foraging distance was modelled via two models. In the first model, it was shown that foraging distance increased when the amount of potential foraging habitat in proximity to breeding/roosting sites decreased (i.e. foraging habitat measured as available area of average foraging distance β = − . Besides, it is important to note that for both models no other individual (sex, age, wing length) or environmental variables (maximal daily temperature, rain, moon phase) were found to influence nightjars' foraging distance. Furthermore, it was necessary to correct for individual variation in all models, while variation within locations was shown to be negligible (Table 1).

Discussion
The use of GPS-loggers yielded detailed information on the movement of 30 nightjars within their respective home ranges. Tracking data allowed us to study foraging behaviour and actual movement paths of 210 foraging flights collected in three research areas. Our findings indicate that not only landscape composition, as already indicated by Evens et al. 5 , but also landscape configuration influences the connectivity between two complementary resources for nightjars: nest sites and foraging areas. On a daily basis, nightjars connect breeding and foraging sites by rapidly crossing unsuitable habitats in order to exploit a higher prey biomass in foraging sites. Increased stress levels were found in birds occupying large, attractive, yet sub-optimal breeding areas. These birds travel longer distances over unsuitable habitat and compensate for longer travel time by spending more time in foraging habitats (i.e. longer foraging time).
Landscape heterogeneity. Two measures of habitat composition influence nightjars' foraging distance in our case: the availability of potential foraging habitat and the habitat diversity. The amount of unsuitable habitat that separates breeding and foraging habitats differs between our study sites, which could play a role in altering foraging distances. In NPHK, for example, large pine forests surround the heathlands, reducing the availability of potential foraging habitat in the vicinity of nightjars' nesting sites (i.e. foraging habitat measured as available area of average foraging distance). Consequently, nightjars from NPHK are obliged to forage three to six times further away from their nesting sites than are those from Bosland or Meeuwen-Gruitrode. On a wider scale (i.e. foraging habitat measured as available area of initial foraging distance), however, the availability of potential foraging habitat does not influence nightjars' foraging distance. On the one hand, this could mean that habitat measurements at a larger spatial scale become irrelevant in relation to nightjars' foraging distance and that only habitat characteristics on a small spatial scale are useful to predict their foraging behaviour. On the other hand, our findings could indicate that nightjars require daily access to a specific amount of foraging habitat, for which some have to travel further comparted to others, due to their initial choice of breeding site. Evens et al. 5 already indicated differences in foraging distance between telemetry studies in the UK and Belgium. Our results suggest that this observation could be possibly explained by variation in landscape heterogeneity between study areas.  In degraded, homogenized landscapes, birds generally fly further to find food compared to those in diverse landscapes 6,28,43,44 . However, in our study we found that nightjars' foraging distance is longer in diverse landscapes. This can be explained by the high diversity in unsuitable and roosting habitat types in our specific case, and by the low diversity in foraging habitat types. Nightjars generally forage near specific landscape elements, such as scattered trees or hedgerows in extensively-grazed grasslands, ponds, oak shrubland and wet meadows 5,38,45 . These foraging habitats usually are relics in vast agricultural lands or pine forests, where landscape fragmentation and homogenisation is still ongoing. Higher food abundance in larger patches 46 could explain why nightjars fly further to find larger foraging habitats in our study. As such, ongoing homogenisation of landscapes and fragmentation of foraging habitats will continue to reduce food resources because fragmented foraging habitats and specific landscape elements are being further eliminated 43 , which will increase the foraging distance for nightjars.
Besides habitat composition, habitat configuration also affects nightjars' foraging behaviour. Foraging distance is three times greater when functional habitats are clustered. In clustered landscapes, longer foraging distances can be explained by the nightjars having to cross larger distances across unsuitable habitat. In randomly-distributed landscapes, breeding and foraging habitats can be found on a much smaller spatial scale 9 . This implies that, when the availability of foraging habitat and habitat diversity is held constant, modification of landscape configuration also influences the connectivity between functional habitats 14 and can affect individual 17 , population 47 and ecosystem processes 48 .
Foraging ecology. Being highly mobile, nightjars show the ability to connect complementary habitats that are separated by 100 m up to seven kilometres. We observed that flight speed is habitat dependant, as nightjars cross agricultural land, forests and urban areas faster than they cross grasslands and heathlands. Generally, animals move faster between resources when the expected food availability in target patches is higher 49 , or when moving through less-preferred habitats 50,51 , especially when movement through these habitats is risky 52,53 or if resources are low 52,54 . We found that in the foraging habitats and pine forests of Bosland moth biomasses are two to three times higher than they are in the breeding habitats. In pine forests, however, nightjars do not forage because of visual limitations 38 . Further assuming that food availability is distributed evenly in similar foraging habitats, it seem likely that nightjars increase flight speed as a response to lower food resources in agricultural land and above dense forests. Opportunistic, supplementary feeding on-the-wing might explain the reduced flight speeds across grasslands and heathlands.
Foraging economy. We found that nightjars prolong foraging time when foraging distance increase.
Therefore, nightjars probably attempt to balance the costs of travelling against the benefits of energy acquisition to maximise net-energy gain (sensu Hedenstrom and Alerstam 55 ). Variation in landscape heterogeneity thus can have a profound impact on nightjars' daily energy expenditure as greater distances across unsuitable habitats are harder to cross 56 and higher flight speeds are also more energetically demanding 55 .
The allocation of more time and energy for foraging most likely also influences the fitness of nightjars through reduced reproductive success 6,16,20,28,57 and increased predation risk 29 . We found that plasma concentration of thiols, indicator for plasma antioxidant levels 58 , is significantly higher in birds from NPHK compared with those from Bosland. Nightjars from NPHK forage further and travel longer distances across unsuitable habitats, which is more energy demanding and may have led to elevated levels of oxidative stress. Birds occupying sub-optimal breeding habitats thus may experience a lower fitness as shown by the elevated antioxidant defences 31,32,59 . We can expect that reduced food availability 30,60 and food quality 61,62 also affects nightjars' oxidative status. Despite having shorter foraging distances, nightjars from Meeuwen-Gruitrode show intermediate levels of thiol concentrations. In contrast to the large, extensively-cultivated grasslands in Bosland and NPHK, foraging sites in Meeuwen-Gruitrode consist of small landscape elements in intensively-cultivated farmlands that possibly hold lower quality food 63 . Evidently, further work is needed to fully understand the role of plasma thiols in the way they are influenced by environmental stressors 58 and how they relate to survival and reproductive success in nightjars.

Conclusion
Our study shows that landscape heterogeneity can affect the connectivity between nightjars' functional habitats, influence their foraging behaviour and might also affect individuals' health and population processes. However, current conservation plans for nightjars, developed within the Natura 2000 framework, focus on the management of heathlands (i.e. breeding habitat) 23 and ignore the importance of key foraging habitats. Following our results we, therefore, conclude that (1) Natura 2000 objectives should be revised, (2) creation of new breeding grounds should be preceded by an assessment of landscape heterogeneity to minimize the distance between breeding and foraging sites and (3) restoration of known breeding grounds should also focus on creating/restoring foraging habitats in proximity to these breeding grounds.
Experiments on live vertebrates. The authors declare that all experiments have been performed under licenses of the Royal Belgian Institute for Natural Sciences (bird ringing licence) and the Flemish Agency Nature and Forest (GPS-tagging, blood sampling with Felasa B licence).