Area-restricted search in Magellanic Woodpeckers: importance of tree senescence, forest composition and open habitats

Large woodpecker species with stable territories, specialized diets and narrow habitat choices, such as the Magellanic Woodpecker (Campephilus magellanicus), are expected to adjust their movements based on the distribution of available resources within territories. Thus, Magellanic Woodpeckers should concentrate foraging activity in areas of higher quality, a behavior consistent with the area-restricted search (ARS) behavior. Although previous studies have deepened the understanding of how Magellanic woodpeckers select habitats, the assessment of ARS behavior would contribute to the identication of feeding areas important for their conservation and habitats hindering their movement. Methods We addressed the independent effects of tree senescence, forest succession, stand composition and open habitats on the adoption of area-restricted search (ARS) behavior in Magellanic Woodpeckers in a heterogeneous landscape dominated by southern beech forest in southern South America. Using GPS relocations from 24 woodpeckers, we estimated the First-Passage Time (FPT), a measure of the time individuals remain in a given area, with longer FPT values indicating the adoption of an ARS behavior. We determined the effects of habitat variables on FPT using a methodological framework based on Linear Mixed Effect models and a randomization procedure intended to reduce spatial autocorrelation arising from consecutive circles whose areas tend to be overlapped along trajectories.


Conclusions
The adoption of the ARS behavior in Magellanic Woodpecker is explained by forest composition, avoidance of open habitats and tree senescence. Our results suggest the conservation of Magellanic Woodpeckers in heterogeneous landscapes involve retention of senescent trees and restoring open areas in order to reduce edge habitat.

Background
Animals with stable territories, specialized diets and narrow habitat choices, such as some woodpecker species, are known to adjust their space use and movements based on the spatio-temporal heterogeneity of habitat resources [1,2]. Those animal species concentrate their movements in the habitats of higher foraging quality [3,4], [5,6,7], a behavior known as area-restricted search (ARS) that, in many cases, involves the use of spatial memory to return to suitable sites within territories [8,9]. Determining the areas where animals concentrate their movements and foraging activity not only contributes to understand decision-making behavior or foraging e ciency but also to the identi cation of feeding areas important for the conservation of vulnerable species [10,11,12,13]. The assessment of ARS behavior has been, for instance, carried out in marine species, including sharks [14], seals [15], whales [16], dolphins [17], sea turtles [18], pu ns [19], boobies [20] and albatross [21]. Although the ARS behavior has been tested in terrestrial species like carnivores [22], [13], small mammals [23], marsupials [24], lizards [25] and spiders [26], less attention has been paid to study it in woodpeckers. Determining habitat conditions under which forest specialized woodpeckers adopt an ARS behavior may contribute to sustainable landscape planning and forest management intended to increase habitat quality and connectivity.
The assessment of ner-scale movement patterns within territories provides a mechanistic basis for understanding the foraging decisions, which is especially evident in specialized animals [8]. In this sense, Hidden Markov Models (HMM) and State-space Models (SSM) have been used to distinguish between discrete modes of movement in woodpecker territories [27,28], where each behavioral mode is interpreted to represent a distinct behavior (e.g., foraging vs. traveling). Those discrete-time modelling approaches assume that, while moving across its home range, a woodpecker can either remain in a same mode or switch to a different mode between two successive time intervals [29]. Foraging woodpeckers, however, may stay for relatively long periods in the foraging mode when using forest stands or forest sites of high quality, as found in animals adopting an area-restricted search (ARS) strategy [30,31]. Although HMM and SSM capture the ne-grained step-by-step movement patterns, these approaches are usually di cult to implement and do not provide spatially explicit inference about the habitat area that animals use when moving. Alternative methodological approaches, such as First-Passage Time (FPT, [31] and residence time [32], provide statistical frameworks for the quanti cation of habitat areas where animals remain for more times, thus allowing for testing associations between ARS behavior and habitat conditions [33].
Indeed, FPT and residence time are useful approaches to evaluate habitat preferences and movement ecology of species that occupy spatially heterogeneous habitats [34,35,36].
In this study, we address the movement strategies adopted by Magellanic Woodpeckers (Campephilus magellanicus) in a heterogeneous forested landscape. Field observations suggest that Magellanic Woodpeckers move between neighboring trees by selecting and adjusting residence times in each visited tree based on its attributes, spending more time at trees that are highly decayed [37]. The tree decay-stage serves as a cue of habitat quality for woodpeckers, offering them with information about the presence of their prey inside trees, such as the larvae of long-horned beetles (Microplophorus magellanicus) [3,37,38,28,39]. Although movement decisions of Magellanic Woodpeckers have been shown to depend on the availability of senescent and dead trees [37,28], to date, no studies have determined how the ARS behavior of woodpeckers varies across an heterogenous landscape. Depending on forest disturbances, territories of Magellanic Woodpeckers may include forest stands that vary in tree species composition and age (old-growth vs. second-growth), but also include open habitats like prairies, bushlands, exotic beaver ponds, and bogs [40,41,39]. Woodpeckers may respond to such habitat heterogeneity by modifying their movement patterns when nding open habitats or intensifying their searching for prey in forest stands with more senescent trees. Here, we aim at examining the adoption of an ARS strategy across a heterogeneous forested landscape using the FPT approach, which allows for the identi cation of areas where woodpeckers behave as ARS foragers. We hypothesized that woodpeckers remain longer (i.e., higher FPT) in areas where individual trees are more senescent but also in areas dominated by oldgrowth forest with low representation of open habitats.

Methods
We conducted our study in a forest landscape located on Navarino Island in the southernmost region of Chile (Fig. 1). The study landscape was covered by southern beech forest of Nothofagus betuloides, N. pumilio, and N. antartica (Fig. 1). Open habitats also were present in this landscape and included patches of shrublands, wetlands, peat bogs, meadows, and ponds, with the latter two being produced by the introduced beaver (Castor canadensis) ( Fig. 1; [41]). The cover of N. pumilio, N. betuloides, and N. antarctica was 29.2%, 29.5% and 6.4%, respectively (Table 1). Although forest stands in old-growth stage of succession covered 45.3% of the land, forest disturbances (e.g., logging and res) have resulted in some second-growth stands (20.2%). Shrublands and meadows (upland open areas) covered 10.2% of the study area and peat bogs and pond (lowland open areas) represented 6.5% of the landscape (Table 1). We acquired GPS locations from 24 tagged male Magellanic Woodpeckers [39,28] using ATS G10 UltraLITE GPS Logger (Advanced Telemetry Systems, Inc.) devices, each attached to a very highfrequency transmitter (ATS model A2440, 2.3 g) for later recovery. GPS devices were placed on the back of adult male woodpeckers using a small amount of epoxy to six feathers. We chose adult males because males guide family groups by eliciting a dominant social behavior while moving across forest habitat [42,43]. The locations of woodpeckers were recorded every 5 min between 08:00 to 16:00 and during the 2014-2015 post-reproductive period (Jan to Mar).
We estimated the accuracy (measurement error) of GPS locations as 12.9 ± 2.9 m (mean ± SE), which was obtained by quantifying the Euclidean distances (m) between 12 different GPS measurements and the actual position of a reference point identi ed on an imagery-based map layer [28]. We calculated the overall speed for each woodpecker by dividing the total traveled distance (i.e. the sum of the distance between GPS relocations) by the total time of a given burst of continuous relocations.

First-passage time
We examined the role of habitat heterogeneity on the movement of woodpeckers by estimating the rstpassage time (FPT) from GPS location data. FPT can be used as an indicator of foraging behavior along animal's tracks, with longer FPT areas being interpreted as evidence of ARS at ecologically proper spatial scales [31,44,45]. The rst-passage time is de ned as the time spent by an individual in circles of radius r centered on subsequent GPS positions along each trajectory (i.e., a movement path consisting of a sequence of GPS locations within the home range; Fig. 2A; [46]). As r increases, longer trajectory sections will be included in the circle [31]. We established the proper spatial scale of FPT analysis by searching the value of r that maximizes the relative variance [ ] of the FPT because the ability of the FPT to detect area restricted search (ARS) increases as the variance takes maximum values [31]. The ARS is a behavior characterized by slow and tortuous movements typically displayed by woodpeckers when selecting trees for foraging [2].
To estimate the FPT, we used the fpt function of the R package adehabitatLT. The last six woodpecker's data were excluded from further analysis because it was not possible to de ne a regular trajectory due to the different time lag between relocations (Fig. S1). The rst-passage time method is designed for trajectories with three or more relocations, so the trajectory data with less than three observations were eliminated. Because observations of six woodpeckers were discarded from the analysis, we estimated the rst-passage time (FPT) from the trajectories of 18 woodpeckers. We maximized for each trajectory and individual by estimating FPT over 50 different radii (r) in a range from 12 to 250 m corresponding, respectively, to the GPS accuracy and a quarter of the calculated net distance displacement of Magellanic Woodpeckers [32]. We determined the proper FPT scale, r value at which reached its maximum value, by examining plots of against r. From plot examination, a set of 36 trajectories were selected out of 62 trajectories (Fig. S2). When we did not observe a maximum value of , we assumed that the path traced by a woodpecker was random and did not represent a movement pattern including different behavioral modes, as shown by the eighth trajectory of woodpecker 3 (Fig. S2).
However, we considered proper FPT scales for trajectories where a local maximum was observed, for example, the rst trajectory of woodpecker 4 (Fig. S2). Based on the proper FPT scales, we de ned two spatial scales to quantify properties of the habitat used by woodpeckers (see below), including site (habitat-patch) scale, de ned by the area of each FPT circle ( Fig. 2A), and home-range scale, de ned by the area comprising the union of all the FPT circles along the trajectory (Fig. 2B).

Habitat variables
We quanti ed habitat variables at the site and home-range scales using a high-resolution (0.50 m) multispectral image from the WorldView-2 sensor (2014). We created geographic information system (GIS) layers of the main categories of habitat types and tree species in the study site as well as a remote sensing index of tree senescence (  (Fig. 2). Fourth, PSRI values were averaged over all FPT circles, considering those values as estimates of tree senescence at the home-range level (Fig. 2). Tree senescence estimated at the site and home-range scales were not collinear (see below; Table S1), hence those variables were included in the same statistical models. All tree species located at the site and home-range levels (i.e. all trees across the trajectories) were interpreted as the foraging habitat quality at the home-range scale [37].

Statistical modelling
We used a methodological framework based on Linear Mixed Effect models (LMM) to determine the effects of habitat variables on FPT (Fig. 3). This approach was intended to reduce spatial autocorrelation arising from consecutive circles whose areas tend to be overlapped along trajectories ( [31]; Fig. 3A). To analyze independent FPT data, we performed a randomization procedure by randomly selecting subsets of not-overlapping circles for each trajectory (Fig. 3). This procedure was repeated 1000 times, which resulted in 1000 sets of trajectories, each contained independent data later used in LME analyses (Fig. 3). LME were tted to different datasets, which precluded comparing models with a criterion derived from likelihood functions, such as the Akaike Information Criteria (AIC; [47]). Thus, for each of the 1000 sampled datasets, we used the RMark R package to compute model-averaged coe cients based on the AIC weights [48]. The AICc weight quanti es the probability that a given model is the best among a set of candidate models [48,49]. The distributions of model-averaged coe cients were assumed to represent the effects of predictors. The mean and 95% con dence intervals of those distributions were used to interpret the signi cance of the coe cients. Model averaging was carried out on a set of 93 models including all possible combination of variables derived from a global model. The global model was built to assess the independent effects of habitat variables. Thus, we checked collinearity with the variance in ation factor (VIF ; Table S1), where a VIF > 10 indicates lack of orthogonality between variables [50]. From this analysis, we detected that the percent cover of Nothofagus species and that of old-growth forest were collinear, so we discarded the latter from analyses (Table S1). Therefore, the global model

Results
The overall speed and radius of FPT circles differed among individuals and trajectories, with speed ranging from 0.  Fig. 4). Conversely, FPT was higher in stands and home ranges where tree senescence (PRSI) was higher at the site and home-range levels ( Table 2; Fig. 4).

Discussion
Our results suggest that adoption of an area-restricted search (ARS) strategy by Magellanic Woodpeckers is in uenced by the quality of foraging habitat (i.e., tree senescence as estimated by PSRI) at different scales, in addition to the cover of open habitats and tree stand composition. Furthermore, the positive effect of tree decay at the home-range scale on FPT suggests that woodpeckers adjust their movements based on the foraging quality of home ranges. Our previous observations have revealed Magellanic Woodpeckers switch to the transient mode (exploratory movement) when the trees located in the proximity of the individual are either, on average, poor in quality or highly variable in quality [28]. Moreover, the residence time of Magellanic Woodpeckers on individual trees is positively associated with the home range foraging quality [37], suggesting that individuals with more suitable territories adopt a more intensive prey searching, consistent with an ARS behavior. Woodpecker preferences for more senescent trees, as found in Magellanic Woodpeckers, have also been reported in other species of woodpeckers [1]. For instance, Red-cockaded Woodpeckers (Leuconotopicus borealis) intensify foraging activity when surrounded by a habitat of better quality [27] whereas Black-backed Woodpeckers (Picoides arcticus) have smaller home ranges in higher-quality habitat (i.e., habitats with more recently killed trees; [52]). However, habitat preferences of woodpeckers also can be in uenced by factors other than foraging habitat quality. For example, edgling Northern Flickers (Colaptes auratus) occupied habitats with greater densities of trees, presumably for protection against predators [53]). Some large woodpeckers, such as Pileated Woodpeckers (Dryocopus pileatus), spend much of their time defending their roosting and nesting sites [54]. Thus, woodpeckers eventually may remain longer in sites offering them with habitat resources different than those used when foraging. Our results provide insights into woodpecker's preferences for Nothofagaceae tree species composing old-growth forest ecosystems, as shown by the negative effect of N. antarctica cover on the residence time of woodpeckers. Previous studies indicate the remote-sensed characterization of tree senescence in N. antarctica differs from that of the other Nothofagus species [39]. These ndings may imply that woodpeckers foraging in an old-growth forest of N. antarctica face a distinctive, and possibly less abundant, assemblage of prey (i.e., saproxylic invertebrates) that is inherent to ecosystems with watersaturated soils like forest stands of N. antarctica. However, the response of Magellanic Woodpeckers to forests covered by N. antarctica requires further exploration.

Conclusions
The adoption of the ARS behavior in Magellanic Woodpecker is accounted for forest composition, avoidance of open habitats and tree senescence. This knowledge had the potential to help the conservation of Magellanic Woodpeckers in landscapes subject to high pressures from anthropogenic land-use change [55]. We recommended that conservation efforts should focus on protecting the remaining old-growth native forest, restoring open areas and reducing edge habitat, especially in protected areas surrounded by anthropogenic landscapes [55,56]. With these indicators, it might be possible to distinguish the more suitable habitats for woodpeckers to better guide the conservation efforts while using these management guidelines to conserve forest communities [57]. Taking actions to preserve the Magellanic Woodpeckers would also help to preserve other forest-dwelling species due to their ecological importance as primary cavity excavator of south Patagonian forest and even other possible important ecological roles. Our results also provided valuable insights into the implementation of sustainable management in the southernmost forests of the world. The existing national forest legislation considers traditional shelterwood cut as the harvesting system to be employed in the subpolar Nothofagus forest. However, shelterwood cutting is intended to retain the canopy cover rather than preserve deadwood, thus being of low value for saproxylic biota associated with these forest ecosystems. Therefore, our results pose the need to retain old-growth forest conditions at the spatial scales that are relevant for saproxylic species at the top of the food chain, as the case of Magellanic Woodpeckers.