We focused upon habitat use and home range size variation in an alpine population of the Black Woodpecker Dryocopus martius. Habitat selection analyses showed that the dominant local forest type, i.e. Mountain Pine Pinus mugo uncinata, was significantly avoided. Scots Pine Pinus sylvestris stands were positively selected, likely because of the abundance of ants; Beech Fagus sylvatica was chosen for nesting and roosting. A high level of flexibility in ranging behaviour and habitat selection, with remarkable among- and within-individual variability, was recorded. We assessed the reliability of census techniques by comparing results obtained by radio-telemetry and territory mapping, which showed that territory mapping failed to identify all pairs detected through radio-tracking. Several behavioural and ecological traits may bias count results irrespective of the methods used. Mapping methods are particularly ineffective because the Black Woodpecker is weakly territorial: in our study area home range overlap was rather high and no territorial defence was detected. Based on these results, the potential of the Black Woodpecker as ecological indicator is questioned.
Safeguarding forest biodiversity is an important branch of conservation biology, especially in Europe, where quality of the residual post-glacial forests has been severely reduced by anthropogenic activities (Thirgood 1989, McNeely 1994). Measuring forest biological diversity and determining appropriate policies of sustainable forest management are first steps in conservation efforts. Woodpeckers may be considered in this framework because they have already been proposed as indicators for forest biodiversity (Angelstam & Mikusiński 1994, Mikusiński & Angelstam 1998). To be an effective indicator species, patterns of distribution and abundance must reflect those of other taxa. This is the case, for instance, with the Lesser Spotted Woodpecker Picoides minor (Jansson 1998). In general, woodpeckers are adapted to habitat structures of old forests, and these very same structures (large and old trees, dead wood etc.) are likely important for a lot of other species. In Europe, this may be true for the Black Woodpecker Dryocopus martius, because it is the only woodpecker which creates breeding holes which other large hole nesters may use as well (Johnsson et al. 1993, Nilsson 1997). Although this species has never been considered as an indicator species per se, it is regarded, together with other woodpecker species, as a part of an indicator system for avian diversity (Mikusiński et al. 2001).
The Black Woodpecker usually favours tall and large trunks of many coniferous and broad-leaved trees forming extensive unbroken forests (Glutz & Bauer 1980, Cramp 1985). Habitat use is possibly related to its peculiar food requirements, especially carpenter ants (Formicidae) (Cuisin 1988, Ceugniet 1989, Pechacek & Krištin 1993, Rolstad & Rolstad 2000). The Black Woodpecker is considered a territorial species (Cuisin 1988), but it is also known that territorial boundaries are not clearly delineated (Glutz von Blotzheim & Bauer 1980) and that defence is concentrated at key sites like nest cavities (Cramp 1985).
In the light of the above, it could be affirmed that the Black Woodpecker is an ecologically relevant forest species which may be useful as part of an indicator system. However, to use a species as an indicator, local information on habitat use, home ranges, predators, energetics and food supply is needed because biological parameters may differ between areas. The ecology of the Black Woodpecker in the Alps, in particular, is poorly known.
The main aims of this radio-telemetry study were: i) to reveal habitat selection patterns, ii) to investigate home range variation over space and time, and iii) to assess the reliability of census methods used to count individuals. Accordingly, we studied habitat selection and home range spacing of a population of Black Woodpeckers inhabiting an Italian Alpine protected area characterized by extensive Mountain Pine woods, and assessed the reliability of census techniques by comparing results obtained by radio-telemetry and territory mapping. Finally, our findings are used to discuss the indicator potential of this species.
The research was carried out from February 2000 to August 2003 in the Mont Avic Natural Park (Chalamy valley, in Aosta Valley) in the northwestern Italian Alps. The area (about 20 km2) was largely covered by coniferous forests ranging from 1100 to 2000 m a.s.l. The Mountain Pine Pinus mugo uncinata (Aeschimann et al. 2004) and, to a lesser extent, Scots Pine P. sylvestris were the dominant trees, occurring in mostly pure coniferous forests (47.3 and 26.1% of the forest area, respectively). The Mountain Pine, in particular, is not widespread in Italy and the forest of Mont Avic Natural Park (classified among the ‘seed national forests’) is the largest in the country (over 1100 ha). It creates dense forests with thick Alpen Rose Rhododendron ferrugineum and Blueberry Vaccinium myrtillus underbrush. Trees are characterized by straight, upright trunk and pyramid-shaped foliage from top to bottom, with low-growing branches (Fig. 1). Dwarfed and prostrate Mountain Pines were confined to higher altitudes (above 2000 m) and to a few deep gorges subject to avalanches. Another common conifer was Larch Larix decidua (pure Larch formations 1.3%; Larch-Mountain Pine formations 9.9%). Alpen Rose was particularly dense in Mountain Pine and LarchMountain Pine forests. Common broad-leaved trees were Beech (4.7%) and others like Chestnut Castanea sativa, Downy Oak Quercus pubescens, Birch Betula pendula and Aspen Populus tremula (together 10.8%). All forest types were spontaneous. Deciduous trees were dominant at low altitudes (below 1100 m; locally up to 1500 m), whereas coniferous trees mostly occurred at middle and high altitudes (1100–2000 m).
Snow depths, as a whole, were moderate because the Aosta Valley is a xeric area. Moreover, the ground of Mountain Pine stands below 1700– 1800 m was very often snow-free because of the high density foliage, and snow did not persist for a long time in Scots Pine sectors because they were usually south-oriented. Vegetation types and their distribution in the study area are shown in Fig. 2.
Woodpeckers were trapped in winter with long-handled dip nets set at night in front of the roosting hole. Most roosting cavities were already known because of previous studies (see below). Birds were banded and fitted with radio transmitters (4 g, i.e. less than 2% of the bird's body mass), which were glued and tied to the base of one of the two central tail feathers (Kenward 1978, 2001). Individuals were tracked with a portable receiver and a hand-held antenna, and locations were taken at intervals not shorter than 30 min while approaching the birds or by triangulation. Data were collected on 2–4 days each week until the loss of the tag for the late summer moult (on average 7.2 ± 0.39 months). A 100 × 100 m grid map was used.
During tracking specific efforts were made to detect territorial disputes. The so-called ‘static interactions’ between individuals were studied by calculating overlaps of home ranges of birds radio-tracked during the same period. To meet the assumption of independence between individuals, only the male's range of breeding pair was considered.
Summer detectability of the Black Woodpecker is rather poor, potentially leading to underestimates of numbers present (Majewski & Rolstad 1993). In our study area Black Woodpeckers were easily and regularly observed only from mid-February to April. Hence, pairs were counted by territory mapping from 1 March to 30 April 2002. We used this census method because it is a generally accepted way to identify areas or habitats of importance to species (Bibby et al. 2000). Moreover, territory distributions could be efficaciously matched with home range distributions obtained from radio tracking. Birds were recorded during 90 min-perday visits. We discarded the play-back option (Tjernberg et al. 1993, Fernandez & Azkona 1996) because when crowded individuals (as in our case) are attracted by taped calls, there is the risk of obtaining erroneous information about their actual distribution.
The study area had been systematically surveyed for Black Woodpecker holes in all forest types in previous years. During the present study a few additional holes were found. The status of each cavity (i.e. nest, roost, not used) was ascertained routinely.
We used Ranges VI software in the analysis of radio-tracking data (Kenward et al. 2002). We computed range size and overlap by the Kernel method (KER 95% of fixes for home ranges, 50% for core areas), given that the presence of disjunct home ranges and/or of multiple core-areas may be efficaciously pointed out by kernel estimator. The presence of stable home ranges was checked by incremental area plots, by sequential addition of fixes. In this case we used a minimum convex polygon for all locations (MCP 100%) since with KER the animal's range size estimate can diminish as locations are added sequentially (Kenward 2001). To avoid inadequate sampling, only stable ranges with more than 50 fixes collected during at least 4 different days in 2 consecutive months were analyzed to calculate home range size.
Static interactions and territoriality were examined using range overlap analyses, the Ranges VI program producing a matrix of the percentage overlap of range A on B and B on A for any dyad of ranges. A mean overlap value was calculated by using all overlap percentages, with a sample size of 2k where k is the number of dyads.
Coefficients of variation (CVs) were computed to estimate among- and within-individual variability in home range size. CVs were calculated for each re-captured individual, as a measure of within-individual variation, and for the overall radio-marked sample (in this case only the first recorded home range, following first capture, was considered), as a measure of among-individual variation. The ratio of the CV for the overall sample to the mean CV of individuals gives a measure of home range size individuality, being higher than 1 when among-individual variation is greater than within-individual variation.
Up-to-date digital maps of the study area were placed at our disposal by the Mont Avic Natural Park. Land cover GIS data digitised from 1:10 000 aerial photographs were used to detect forest types (Morra di Cella & Cremonese 2004). Six major types were identified based upon tree species prevalence: Mountain Pine, mixed Larch-Mountain Pine (both with underbrush, mostly Alpen Rose), Scots Pine, Larch, Beech, and other broad-leaved trees. Structural features of the vegetation were obtained from field measurements within circular sampling random plots (radius of 10m). Structural variables were: tree density, i.e. the number of trees/ha with trunk diameter (measured at breast height) exceeding 12.5 cm, dominant tree height, i.e. the mean height of the 100 tallest trees in the plot and canopy cover, i.e. the canopy projection on the ground. Raster maps (10 × 10 m pixels) were treated by Arc View 8 to be imported by the Ranges VI software for habitat selection analyses.
Habitat availability was measured within the whole study area identified by estimating a KER100% boundary round all the locations used in all the ranges. Preliminary tests indicated that this level of scale gave more information than that regarding individual home range. A 50 m radius buffer round every individual location was considered, and habitat class proportions were calculated within the area.
To examine whether individuals differed from random use of habitats, we employed χ2-tests of differences between observed and expected number of locations. Bonferroni confidence intervals were fitted to test whether individuals used habitats by availability (Neu et al. 1974, Byers et al. 1984). The sign test was used to test the hypothesis that positive and negative selections for each habitat category were sampled from a population in which the two kinds of selections were present in equal proportions (Sokal & Rohlf 1995). In this context, ‘preferred’ is said of a habitat type that is used significantly more than expected from its availability. We also used the terms ‘positive selection’, ‘negative selection’ (habitat type that is used more or less than expected), and ‘no selection’ (habitat used proportionally to its availability). A main problem with Bonferroni confidence intervals is that locations are treated as independent locations and that the ‘unit-sum constraint’ is ignored (Aebischer et al. 1993, Kenward 2001). Hence, to address these problems, we also used compositional analysis, by which the use of each habitat is expressed relative to each of the other habitats. We first verified overall significant departure from random use of habitat with Wilk's Lambda. Then habitat types were ranked according to relative use with regard to each bird (individual ranking) and the overall sample (general ranking). In the case of general ranking, Student's t was used to test whether the preference differed significantly from zero for each habitat pair (Aebischer et al. 1993).
To compare Bonferroni and compositional analyses, we checked whether the habitats which were classified as positively selected by the former method were classed as high-ranking by the latter.
A total of 18 woodpeckers were captured (11 males and 7 females); 10 individuals were re-captured and radio-tracked for more than one year (6 were recaptured once, 4 twice). We treated locations of the same bird in different years as independent samples. Admittedly, individual site fidelity was a source of pseudo-replication. To overcome this problem, in keeping with Nikula et al. (2004), we checked whether locations of each re-captured individual overlapped in consecutive years: if overlap percentage was higher than 50%, only the home range with the highest number of fixes was used.
On average, core areas covered about a third of whole home ranges (92.4 ± 10.9 ha vs. 316.3 ± 32.8 ha, n = 27). Among-individuals variation in home range size was rather high, standard deviation being greater than 50% of the mean, both for home ranges and core areas (CVs = 0.527 and 0.595, respectively). The ratio of the CV for the overall sample to the mean CV of individuals was 1.48 for home ranges and 1.39 for core areas.
Home ranges during the breeding period (15 March – 30 June) were smaller than during the non-breeding period (median values 232.8 and 298.0 ha, respectively, n = 20, P = 0.03, Wilcoxon matched pair test). In most instances (17 out of 20) breeding and non-breeding home ranges greatly overlapped, in two instances breeding ranges were totally encompassed in non-breeding, and in another they were separated (mean seasonal auto-overlap 70.4 ± 5.3%, n = 20). Breeding core areas were not significantly smaller than non-breeding core areas (P = 0.33, Wilcoxon matched pair test). In most instances (14 out of 20) breeding and non-breeding core areas overlapped to a high degree, in three instances breeding core areas were totally encompassed in non-breeding core areas and in three they were different (mean seasonal auto-overlap 53.2 ± 7.4%, n = 20).
Seasonal home range (but not core area) size significantly varied from year to year (Kruskal-Wallis test, breeding ranges: H3,23 = 10.33, P = 0.02; non-breeding ranges: H3,26 = 10.69, P = 0.01), being larger in 2003 and smaller in 2001. The structure of ranges was also variable: disjunct home ranges (in 3 birds the feeding area was about 3.7 km from the roosting site) and multiple core areas were found in several instances (9 and 5, respectively, out of 27; Fig. 3).
Analysis of all-year ranges of the recaptured individuals showed high site-fidelity. Mean auto-overlap between successive years was 71.0 ± 3.5% for home ranges and 61.4 ± 4.5% for core areas.
Overlap of home ranges was much broader than overlap of core areas (on average 27.1% vs. 8.3%, breeding period; 35.7 vs. 11.2, non-breeding period, Table 1). Non-breeding overlap was usually not significantly broader than breeding overlap. Overlap values significantly varied among years, whatever the period considered (KruskalWallis tests). No evident territorial interaction was detected.
Bonferroni confidence intervals analysis pointed out a significant avoidance of Mountain Pine and mixed-Larch-Mountain Pine stands and a significant preference for Scots Pine stands. In Beech the number of positive selections was not significantly higher than the number of negative ones. Birds bred and roosted in Beech, but fed in Scots Pine habitat (Table 2). Although this was the general pattern, individual choices were not necessarily consistent over time: 23.1% of birds (out of 26) completely modified their preference from breeding to non-breeding and 15% (out of 20, i.e. 10 birds for each period) from year to year.
Static interactions, showing the degree of overlap between individuals radio-tracked during the same period. Mean overlap values (percentages) are given. To meet the assumption of independence between individuals, of each breeding pair only the male's range was taken into account. See methods for computation details.
Selection of forest types pointed out through Bonferroni confidence interval and Compositional analyses. The number of home ranges considered is 29 for the breeding period and 26 for the non-breeding period.
Selection of forest structure traits (tree height, tree density and canopy cover) pointed out by Bonferroni confidence interval and Compositional analyses. Symbols and number of home ranges considered as in Table 2.
In compositional analysis, Mountain Pine and mixed Larch-Mountain Pine stands scored the highest number of low individual ranks (i.e. 0 and 1), whereas Beech and Scots Pine scored the highest number of high individual ranks (4 and 5); t-tests showed that there was a clear significant split between the four top ranking habitats (which were sometimes interchangeable in the ranking order) and the two bottom ranking habitats (Table 2). Bonferroni-preferred habitats mostly corresponded to the top-ranking habitats (29 out of 34 in both periods; Table 2).
Bonferroni confidence intervals indicated that forest stands characterized by short trees were significantly avoided, whereas those with taller trees were positively selected. Birds significantly avoided both low density and very high density stands. Low and medium canopy closure stands were also significantly avoided. This general picture emerged in the same way both in the breeding and in the non-breeding period, with only few exceptions (Table 3).
In spite of this habitat selection constancy over time at the population level, individual variability was considerable. On average, 41% of individuals totally changed their structural choices (e.g. from tall to short trees, from low to high density stands) from one period to the next, and 26.7% from year to year.
In compositional analysis, t-tests showed that in all analyses there was a clear split between the two (or three) top-ranking habitat types and the low-ranking types (Table 3). Correspondence between Bonferroni-preferred habitats and individual top-ranking types of compositional analysis was very good for canopy cover (25 out of 27 in the breeding period and 20 out of 22 in the nonbreeding) and trees density (24 out of 28; 23 out of 27), slightly less for dominant tree height (10 out of 17; 15 out of 23).
Radio-telemetry and direct observations at the nests showed the presence of at least 9 nesting pairs and 3 single birds (Fig. 4A). Censuses succeeded to identify and locate only 4 pairs (i.e. A, C, D and F) and failed to find and discriminate pairs nesting at close distance. Three close nesting pairs (G, H, and J) remained undetected because a single, spurious, unresolved cluster of points was mapped, so that a single territory was consequently assumed (Fig. 4B). Conversely, two nesting pairs with a large core area spatially shared the same area with non-breeding individuals (B and E), whereas the presence of four pairs was erroneously assumed on the ground of two distinct clusters of points within the home range of each of two nesting pairs (Fig. 4B). All in all, censuses were not accurate and resulted in both under- and overestimates.
Trees with cavities were not uniformly scattered over the study area; despite the broad expanse of Mountain Pine forests and the research effort we made, no cavity was found in sectors covered by this conifer (Fig. 2). Mostly Beeches and Scots Pines were used for nest building (respectively 62.1 and 32.7% of all nests found, n = 58). When tree availability was calculated from the proportion of the area covered, preference for the Beech was apparent (χ2 = 86.03, P < 0.001).
Bonferroni confidence intervals indicated that woodpeckers preferred Beech stands. As for forest structure, stands with high canopy closure (80–90%) and medium-high tree densities (1200– 1400 trees/ha) were significantly and positively selected. These same categories were classified as top-ranking by compositional analysis. Surveys conducted from 1991 to 2004 pointed out that several nest and roost cavities were continuously used over time (the longest record: 8 successive years), often by the same birds (4 successive years).
Three main results of the present research deserve to be discussed as regards the indicator potential of the Black Woodpecker for forest biodiversity and habitat management.
First, the classification of this woodpecker as specialist or generalist is ambivalent and should be used sparingly. According to Mikusiński et al. (2001), the Black Woodpecker can be considered a habitat generalist. At the same time, given its peculiar diet (Cuisin 1988, Ceugniet 1989, Pechacek & Krištin 1993, Rolstad & Rolstad 2000), it can be considered a specialist feeder. In our study area, for instance, woodpeckers are known to forage on ants of the genera Camponotus and Formica (unpubl. data). Ants are abundant and available in Scots Pine stands (pers. observ.), which were positively selected. In contrast, Mountain Pine sectors were significantly avoided, maybe because the presence of the thick underbrush hampered collection of food and trees (with dense foliage from top to bottom) were not suitable for excavating holes. It is worth mentioning that Mountain Pines in the Pyrenees, characterised by large straight trunks free of branches, are used as nesting trees (Ceugniet 1989). Beech was positively selected mostly for nesting and roosting. Forest stands with taller trees were positively selected, whereas those with low canopy closure were avoided. The conclusion is that our Black Woodpecker population seems rather flexible in the use of habitats, at the same time showing clear preferences for particular habitats, likely depending on feeding, nesting and roosting requirements. This is in keeping with other results which pointed out that the species is not absent from fragmented or managed forests, provided that tree composition and food supply are suitable (Tjernberg et al. 1993, Mikusiński 1997, Rolstad et al. 1998).
Second, individual habitat use and spatial behaviour are flexible as well. We pointed out high among- and within-individual variability in habitat selection and in home range size. The year-round home range size was comparable to home range size observed in Scandinavian forests, where sizes decreased with increasing proportion of young plantations in the landscape (Rolstad et al. 1998). Even the structure of home ranges was not constant because, in addition to the typical home range with a single core area, disjunct ranges with multiple core areas were also recorded. All in all, in spite of the consistent site-fidelity and individuality in the size of home range (as indicated from CV analyses), the overall degree of variability in spatial behaviour was rather high.
Third, widely accepted census techniques may give rise to unreliable results. We showed that censuses failed in identifying all the pairs inhabiting the area. This can be partly explained by the occurrence of several individuals at one site due to a concentration of trees suitable for nesting, which is not uncommon in this species (Lang & Rost 1990, Johnsson et al. 1993, Lange 1995, 1996). However, even in areas with a scattered distribution of woodpeckers, ecological and ethological traits of this species may contribute to invalidate census estimates. Both sexes share the same vocal repertoire and instrumental signals are often used for contact between partners. This may easily produce overestimates of breeding pairs; the occurrence of two interacting individuals is not necessarily a clue of the presence of two territories and, contrary to the rule, simultaneous registrations are not the key to reliable mapping. Analogously, the presence of two or more territories may be incorrectly scored when individuals with disjunct home ranges and/or multiple core areas are involved. However, the most striking hindrance to the use of mapping methods results from the limited territorial performances of this species. Cuisin (1988) reports that territories are fiercely defended and this seems to be consistent with the occurrence of regularly spaced territories in Sweden (Tjernberg et al. 1993). However, it has also been stated that territories may overlap with those of neighbours without leading to disputes, and that boundaries are not clearly delineated (Cramp 1985). We did not find any cue of territoriality despite the potential or agonistic behaviour induced by the clumped distribution of some birds. Contrary to true territorial woodpeckers (Bachmann & Pasinelli 2002), home ranges of neighbouring birds overlapped to a high degree and core areas overlapped as well, albeit to a lesser extent. Home range overlap was not significantly higher in the non-breeding than in the breeding period, again suggesting that territoriality was weak. Accordingly, in the Black Woodpecker the basic assumption underlying territory mapping is violated. It must be underlined that even the results obtained from non-mapping techniques may be biased by some of the previously discussed ethological and ecological traits of this species (movements between portions or cores areas within the same home range, vocalisations produced by both sexes).
Our results suggest therefore caution in considering the Black Woodpecker as a potential ecological indicator, at least in the Alps. Indicator species are those species whose ecological requirements guarantee the existence of particular environmental conditions (Campbell & Lack 1985). A flexible species as the Black Woodpecker may adapt to a variety of conditions, and the existence of particular environmental conditions is therefore not guaranteed. Unreliability of standard census methods means that bird density cannot be properly assessed. Moreover, while most of the other woodpecker species suffer from intensive forest management (Angelstam & Mikusiński 1994), the Black Woodpecker is less sensitive to man-induced structural changes in forests (Tjernberg et al. 1993, Mikusiński 1997, Rolstad et al. 1998). Hence, we cast doubts upon the convenience of the Black Woodpecker per se to monitor ecosystem health, in keeping with doubts risen for several other vertebrates proposed as indicators (Simberloff 1997, Hilty & Merenlender 2000, Carignan & Villard 2002).
This research was partly funded by the Mont Avic Natural Park and the Aosta Valley Autonomous Region (Assessorato Territorio Ambiente e Opere Pubbliche). E. Broglia and M. Gamba assisted us in the capture; M. Ferro and G. Rossi in banding. We wish to thank all the people who helped in the fieldwork; we are particularly indebted to M. Ortolan, V. Napoli, M. Guglielminotti, F. Navillod, C. Cerrato, M. Jacquand, B. Billiout, J. Lanave, B. Toutain, M. Chambon, J. Ruel-Gallay, G. Bontemps, E. Duchenne and G. Garcel. We also thank U. Morra di Cella and E. Cremonese for providing accurate GIS forest maps of the study area.