Oecologia

, 159:435

Native bird breeding in a chronosequence of revegetated sites

Authors

  • Katherine Selwood
    • Australian Centre for Biodiversity, School of Biological SciencesMonash University
    • Australian Centre for Biodiversity, School of Biological SciencesMonash University
  • James R. Thomson
    • Australian Centre for Biodiversity, School of Biological SciencesMonash University
Conservation Ecology - Original Paper

DOI: 10.1007/s00442-008-1221-9

Cite this article as:
Selwood, K., Mac Nally, R. & Thomson, J.R. Oecologia (2009) 159: 435. doi:10.1007/s00442-008-1221-9

Abstract

Restoration of degraded landscapes through replantings of native vegetation has been proceeding in response to habitat loss and fragmentation and plummeting biodiversity. Little is known about whether the investments in ecological restoration have resulted in biodiversity benefits. We evaluated the potential of restored sites to support populations by assessing bird breeding activity. We surveyed 21 revegetated sites of various ages (9–111 years) in the box–ironbark region of Victoria, Australia. Sites differed in landscape context, patch features and in-site characteristics. The latter, including whether sites were grazed, amounts of fallen timber and numbers of remnant trees, were most important in affecting overall bird breeding activity. Patch-configuration (e.g., shape, area) was of secondary importance. Landscape context appeared to have little effect on bird breeding except for one species. While these results suggest that in-site habitat structure is the predominant driver, we caution against dismissing the importance of patch characteristics and landscape context for two reasons. First, the available sites covered a relatively small range of areas (<54 ha), and we could not provide a broad range of landscape-contextual contrasts given that we could only use existing plantings. Second, much of the breeding activity was by bird species known to be tolerant of smaller woodland areas or of the open countryside. We show that there is very little breeding activity in replantings by species that have declined dramatically in rank abundance between large ‘reference’ areas and fragmented landscapes. It seems likely that most replantings provide habitat configurations unsuited for dealing with declines of species most vulnerable to habitat loss and fragmentation.

Keywords

Box–ironbark forestsEvidence for breedingHabitat structureLandscape contextSoutheastern Australia

Introduction

Habitat clearance and fragmentation is a global problem that has resulted in large-scale habitat change and species declines and losses (Sekercioglu et al. 2004; Watling and Donnelly 2006; Lindenmayer et al. 2008). There has been much interest in habitat restoration in many parts of the world, especially over the last 20 year (Jordan et al. 1988; Hobbs and Norton 1996). For example in the UK, the Farm Woodland Scheme provides financial incentives for farmers to restore native vegetation (Green 1993). One of the primary objectives of restoration activities has been rehabilitation of land and water resources damaged by excessive land clearing, poor farming practices and over-grazing (Lambeck 1997).

In Australia, many landscapes have been cleared or heavily modified for agriculture, forestry, urbanisation and grazing since European settlement in the early 1800s (Saunders and Hobbs 1995; Radford et al. 2005; Fleishman and Mac Nally 2007). Land and water degradation has resulted, and there have been many adverse effects on biodiversity (Ford et al.2001; Bennett et al. 2006; Vesk and Mac Nally 2006). These stressors have contributed to local and regional extinctions of many species (Lunney et al. 1997; Ford et al. 2001).

Government bodies, industry and private landholders in Australia increasingly have become involved in restoration activities (Fitzsimons and Wescott 2001; Olsen et al. 2005). Plantings for carbon sequestration and salinity control have become high-profile topics, so there are potential opportunities to ‘piggy back’ biodiversity benefits onto these actions (Ryan 1999). Replantings also need to be integrated into strategic schemes, such as proposed ‘biolink’ programs (Brereton et al. 1995).

However, there have been some recommendations for improving benefits of restoration activities for biodiversity. Recher (2004) recommended establishment of corridors, minimisation of edges, removal of disturbances, such as grazing, and maintenance or placement of fallen wood. Majer et al. (2001) encouraged addition of logs and fallen timber, use of species-rich collections of regional trees, shrubs and herbs, and addition of nest boxes. However, assertions that replantings necessarily result in suitable habitat for native wildlife largely are untested (Block et al. 2001). Now that ‘biodiversity targets’ are more often being incorporated into restoration plans, it is increasingly important that goals for biodiversity be set and monitored (Hobbs 2003).

We focus on birds, which are good indicators of ecological ‘health’ and, by being conspicuous, require relatively little effort to monitor compared with other fauna (Lindenmayer et al. 2003). Birds also are among the most mobile of fauna, making it more likely that birds experience and respond to landscape configuration (Thomson et al. 2007, in press; Mac Nally 2008). It is also important to determine whether birds colonise restored areas because birds perform key ecological services, such as pollination and seed dispersal, which are vital for successful ecological restoration and on-going sustainability (Lundberg and Moberg 2003; Sekercioglu et al. 2006; Davis et al. 2008).

Many studies have shown that birds will occupy restored areas in Australia (e.g., Nichols and Nichols 2003; Martin et al. 2004; Jansen 2005; Barrett et al. 2008) and elsewhere (e.g., Twedt et al. 2002). Fewer studies address the specific habitat and site characteristics that might affect avifaunas of restored areas (e.g., Barrett 2000; Twedt et al. 2002; Martin et al. 2004). There is only limited information on the usefulness of specific habitat components or landscape attributes (Bennett and Mac Nally 2004) for encouraging colonisation (although see Mac Nally et al., in press).

While certain species may be present in particular habitats, presence does not necessarily indicate that these areas sustain populations (Westphal et al. 2007). Measuring birds’ breeding is a shift from ‘standing crop’ measures (e.g., presence and abundance), which are commonly used (e.g., Green and Catterall 1998; Fisher 2001; Martin et al. 2004), to more critical ‘production’ measures (Mac Nally 2007a; Barrett et al. 2008). Mac Nally’s (2007a) method for quantifying birds’ breeding activities was designed to measure the usefulness of restored areas for supporting birds because it is based on evidence for inferring breeding success. While evidence of bird breeding alone is not sufficient to demonstrate persistence of populations, conducting full demographic and dispersal studies logistically is very demanding (e.g., Poiani 1993) and probably will never be done for whole assemblages. Here, we investigate the suitability of restored vegetation for bird breeding using Mac Nally’s (2007a) method. We assessed the relative suitability of restored sites for breeding birds in relation to a range of site, landscape and habitat characteristics.

Are the ‘right’ species breeding in replantings?

From extensive prior data for this region, we were able to assess the degree to which species of (regional) conservation concern are catered for by replanted vegetation. It is possible that species known to be tolerant of much-changed landscapes are using replantings, while declining species are not. Independent data are available by which we can rank species insofar as they are affected by habitat loss and fragmentation (Mac Nally 2007b). To do so, we calculated the difference between the rank abundance of species in the largest extant remnant patches (>10,000 ha; these are referred to as ‘reference’ forests) and compared these with the mean of ranks of abundance in remnants ≤25 ha (Mac Nally 2007b); these remnants are not the sites used in the current study. These rank changes were used in our analyses to determine whether there is a relationship between rank-abundance changes and propensity to breed in replantings.

Methods

Study area

The study area is the box-ironbark region of Victoria, Australia, and areas west of the Great Diving Range (Fig. 1), on the gentle slopes and hills between 150 and 400 m ASL (Mac Nally et al. 2000). The mean annual rainfall for the region is 400–700 mm; summers are hot and dry (Mac Nally et al.2000). In the last 160 years, the system has been extensively disturbed by gold mining, timber felling and clearing for agriculture (E.C.C. 2001). Only 17% of the original natural vegetation remains, and most is heavily degraded (E.C.C. 2001). Several locations in the region have now been set aside for restoration, including private and public land (E.C.C. 2001).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-008-1221-9/MediaObjects/442_2008_1221_Fig1_HTML.gif
Fig. 1

The location of the 21 study sites within the study region (D.S.E. 2007). The study region is within the box-iron bark region of Victoria and areas west of the Great Dividing Range, on sedimentary soils, within the 400–800-mm rainfall boundary

Site selection

We selected 21 sites from 80 replanted sites (Vesk et al. 2008b) for which extensive data are available (Fig. 1). Sites were chosen to span as wide a range of restored stand ages, and habitat and landscape characteristics as was available (Table 1).
Table 1

Definitions of predictor variables

Predictor variable

Definition

Range

Landscape context variables

Distance to large remnants

Distance to areas of remnant vegetation >2,000 ha in area (km)

0–35.7

Surrounding vegetation

Total area of vegetation in the surrounding landscape (ha) within a 5-km radius of the site (log10-transformed in models)

79–3,040

Planting characteristics

Age

Number of years since restoration took place (log10-transformed in models)

9–111

Area

Total area of replanted vegetation (ha) (log10-transformed in models)

0.75–54.6

Shape

Shape index (log10-transformed) of the restored site, where \( {\text{shape}} = {\text{perimeter}}/2\sqrt {\pi \times {\text{area}}} . \) Circles = 1, elongated sites have larger values

1.06–4.98

In-site characteristics

Number of remnant trees

The total number of remnant trees present in the restored site (ha−1).

0–5

Fallen timber

Total volume of fallen timber >10 cm radius [m3 (0.1 ha)−1]

0–0.9

Grazing status

Grazing status: 1 = grazed, 0 = not grazed

Total basal area

Total basal area of all replanted trees [m2 (0.1 ha)−1], log10-transformed

0.24–4.64

Average basal area

Basal area per tree (m2) averaged over 0.1 ha

0.01–1.54

MDS1

Multidimensional scaling dimension 1. Negatively correlated with percentage grass cover (Spearman = −0.82); positively correlated with the number of trees with bark fissures present (= 0.53), shrub density level (r = 0.61), and species richness of shrubs (r = 0.61)

MDS2

Multidimensional scaling dimension 2. Negatively correlated with the number of trees with bark splits present (r = − 0.66)

MDS3

Multidimensional scaling dimension 3. Positively correlated with the number of trees with bark fissures (r = 0.51), the number of grass tussocks (r = 0.63) and the percentage ground cover of tussock grasses (r = 0.68)

Site and habitat measurements

We used the geographic information system ArcGIS 8.3 (E.S.R.I. 2002) with the tree cover layer ‘TREE25’ (D.S.E. 1999) to measure landscape contextual attributes of sites. These included: area and perimeter of sites, total area of vegetation in the surrounding landscape within a 5-km radius, and distances to moderately sized remnants (>400 ha) and to larger remnants (>2,000 ha). We calculated a shape index for the sites using the formula: \( {\text{shape}} = {\text{perimeter}}/2\sqrt {\pi \times {\text{area}}} , \) for which circles have a value of 1 and more elongated sites having larger values (there is no maximum value). We also measured the area of native remnant vegetation abutting sites.

A 10 m × 100-m plot was used to measure habitat variables and vegetation characteristics at each site. Species richness of trees and shrubs was recorded, density of trees measured and the number of trees with hollows or bark-fissures were counted (Harper et al.2004). The density of shrubs was classed into three levels: (1) <10, (2) 10–100 and (3) >100. The volume of fallen timber on the ground (>10 cm diameter) was measured, and the numbers of large tree boughs (>10 cm diameter) and remnant trees were counted. Percentage ground covers of herbs, flat-leaved weeds, grass and tussock grasses were measured in four 1-m2 quadrats, and the number of grass tussocks (a typical, natural form of grass coverage in box-ironbark forests) counted, with the values averaged for the four quadrats. The percentage canopy cover and total basal area of trees were measured, and the average basal area per tree calculated. The age and grazing status (1 = present grazing; 0 = no present grazing) of each replanted site were noted.

Bird surveys

Six rounds of surveys were conducted over the peak breeding season, at regular intervals, from early September to mid-November 2006. Each round of surveys was completed within 2 week of the previous one to contend with potentially short intervals between hatching and fledging of some bird species. Time-scaled survey methods were used for data collection (Mac Nally et al. 2000). For smaller sites (<10 ha), we surveyed one randomly placed 0.5-ha plot, and for sites ≥10 ha, we surveyed two plots, data for which were averaged.

Bird surveys were conducted between sunrise and 12:00 Australian Eastern time, and in the 3 h before sunset. We avoided the warmest parts of the day because bird activity is low. Surveys were not conducted on hot or windy days or in wet weather conditions because these conditions considerably decrease bird activity. Sites were surveyed in restricted, randomised order to avoid systematic sampling biases that might be associated with weather changes or observer (only K. Selwood) fatigue. We used an area-search method (Dieni and Jones 2002), which involved systematically perusing each survey area for 30 min and noting bird species present, and breeding behaviours. Locations of nests on each plot were recorded using GPS for subsequent surveys.

Breeding behaviour scores

We used a scoring system based on the opinions of 25 expert Australian ornithologists (Mac Nally 2007a). Each score is based on the consensus rank importance of the behaviour in relation to breeding success (Table 2). Behaviours with high scores, such as observations of fledglings, or feeding of young in the nest, clearly indicate breeding success (Mac Nally 2007a). Behaviours with low scores, such as territorial and courtship behaviours, indicate that birds at least ‘view’ the site as potentially suitable for breeding, but it is not known whether this necessarily results in breeding output (Mac Nally 2007a). Only the highest observed scored behaviour was recorded for a given nest.
Table 2

Results of surveys of weights of evidence for breeding success based on viewpoints of 25 expert Australian ornithologists (Mac Nally 2007a)

Behaviour

Score

Feeding of young out of the nest

9.0

Young birds seen

9.0

Feeding of young in the nest

8.0

Presence of juveniles

7.5

Young birds heard

7.5

Adults carrying food

6.0

Adult on a nest

6.0

Current breeding season’s nest

5.0

Past breeding season’s nest

3.5

Gathering nest material

3.0

Courtship

2.0

Territorial behaviour

1.0

Male and female pairs

1.0

Response variables

The response variables for the analyses were:
  1. 1.

    Number of breeding species = the number of species showing breeding behaviour at a site over the entire study (i.e., at any time during the six surveys);

     
  2. 2.

    Total score = the total score of all breeding behaviours performed by all species at each site over the entire study. Only the maximum score recorded for each territory or nest for a given species was included. For example, if young birds were observed being fed at a nest that was scored in an earlier survey as an ‘adult on a nest’, then only the former behaviour (the most highly ranked) would be counted (see Table 2);

     
  3. 3.

    Minimum realized = the total score of fledged young observed at each site over the entire study (i.e., behaviours scored 9 in Table 2). This is a key indicator of actual recruitment to at least free-living stage.

     

We also calculated the measures of ‘total score’ and ‘minimum realized score’ for the top seven breeding species. These species contributed ca. 65% of the total breeding score of all sites and were: white-plumed honeyeater (see Appendix for Linnean binomials), Australian magpie, red wattlebird, red-rumped parrot, willie wagtail, fuscous honeyeater and white-winged chough.

Predictor variables

The landscape variables used as predictors for the response variables included: area of the replanted site; total area of vegetation in the surrounding landscape; shape index; distance to remnants of area >2,000 ha (Table 1). The area, shape index and area of surrounding vegetation had skewed distributions, so these were log10-transformed. Although both the distance to remnants of area >400 ha (Barrett et al. 1994) and to remnants of area >2,000 ha (Harwood and Mac Nally 2005) were measured, there was high correlation between the two (R2 = 0.99), so both were not included. While remnants of >400 ha are considered by some to be source remnants (Barrett et al. 1994; Major et al. 2001), Harwood and Mac Nally (2005) found that even remnants of ca. 2,000 ha had different avifaunas to areas of >10,000 ha of continuous vegetation (i.e., reference forests; see also Mac Nally 2007b). Therefore, we used the distance to remnants >2,000 ha as a possible predictor variable.

In-site variables, including the number of remnant trees, replanting age, volume of fallen timber, grazing status, average basal area of trees and total basal area of trees, were included as individual predictor variables at the in-site scale (Table 1). The age of replanting had a skewed distribution and so was log10-transformed. We combined other habitat predictor variables using non-metric multidimensional scaling (MDS), which produced three dimensions (MDS1, MDS2 and MDS3), which were used as compound predictors (see below).

Are the ‘right’ species breeding in replantings?

We averaged ranks for each native species from data for three reference-forests surveys, which we took to be our best indicator of the pre-fragmentation avifauna (described in Thomson et al. 2007). There were 99 species recorded in those surveys. For species not recorded in reference-forests surveys, but in surveys of areas ≤25 ha, we allocated an ‘effective rank’ of 100, meaning that any of those species were at most the 100th most abundant species prior to fragmentation. For remnants ≤25 ha in area, there were 104 species; we allocated effective ranks of 105 for species not recorded in these remnants but that were recorded in the reference-forests surveys. From these two sets of ranks, we could compute a rank-difference, indicating the degree to which a species’ abundance has been affected by habitat fragmentation. We computed the total breeding scores across all replantings for each species individually. We then correlated these scores with the rank-differences using Spearman’s statistic.

Statistical analyses

We used Bayesian model averaging (BMA) (Ellison 1996; Hoeting et al.1999; Wintle et al.2003) to model the birds’ responses to restored sites. BMA incorporates model-selection uncertainty into inference and prediction, producing more accurate predictions than methods that select a single best model (Thomson et al. 2007). We used ‘bic.glm’ in the ‘BMA’ package (Raftery et al. 2005) in R (R Development Core Team 2005). This function performs BMA for generalised linear models, using the ‘leaps and bounds’ algorithm (Furnival and Wilson 1974) and Bayesian information criterion (BIC) approximation to Bayes factors (Hoeting et al. 1999).

We used Gaussian errors with BMA to model the overall ‘total score’, overall ‘minimum realized score’, the ‘total score’ for the willie wagtail and both the ‘total’ and ‘minimum realized score’ for the white-plumed honeyeater, Australian magpie, red wattlebird, red-rumped parrot, fuscous honeyeater and white-winged chough. Poisson regression with Bayesian model averaging was used to model the number of species displaying breeding behaviours because the Poisson distribution is more appropriate for response variables consisting of small, non-negative numbers (McCullagh and Nelder 1989).

We used the posterior probability that a variable had a non-zero coefficient in the predictor model [Pr(inc)] as a measure of the influence of that variable on the response. Variables with high values of Pr(inc) contributed most to model fit, whereas predictor variables with low values of Pr(inc) were included only in less probable models (Thomson et al.2006). We considered predictor variables that had value of Pr(inc) >0.75 to be ‘key predictors’, i.e., important in predicting the response variable (Viallefont et al. 2001). Values of Pr(inc) < 0.50 showed no evidence of being useful predictors, and values 0.50–0.75 showed weaker evidence of being useful predictors (Viallefont et al.2001).

Hierarchical partitioning was performed on those predictor variables with Pr(inc) >0.75 to the determine relative importance of their independent effects on the response variable (Mac Nally 2000). Hierarchical partitioning is used to estimate the ‘independent’ explanatory power of each predictor variable with the response variable (Chevan and Sutherland 1991). We used the ‘hier.part’ package for R (Mac Nally and Walsh 2005). R2 was used as the goodness-of-fit measure. The term ‘%I’ denotes the measure of the independent contribution of a given predictor variable as a percentage of the total explained variance for the full model (Mac Nally and Walsh 2005).

Results

Habitat structure

The stress for the three-dimensional MDS of in-site habitat structure was 0.11. MDS1 was most highly correlated with grass cover, the number of trees with bark fissures present, shrub density and species richness of shrubs (Table 1). MDS2 was highly correlated with the number of trees with bark splits present, and MDS3 was most highly correlated with the number of trees with bark fissures, the percentage cover of tussock grasses and the number of grass tussocks (Table 1).

Number of breeding bird species

Forty species of native birds (see Appendix, Linnean names listed) were recorded undertaking breeding activities listed in Table 2. No potential predictor variables were identified as key predictors for the number of species showing breeding activity. There was weak evidence for shape having a positive influence on the number of breeding species — elongated sites may have had more breeding species (Table 3).
Table 3

Probability of a non-zero coefficient in the predictor model [Pr(inc)] determined by Bayesian model averaging

Response variable

Age

Area

Shape

Distance to large remants

Surrounding Vegetation

Number of remnant trees

Fallen timber

Grazing status

Total basal area

Average basal area

MDS1

MDS2

MDS3

Number of breeding species

0.10

0.10

0.59

0.08

0.08

0.09

0.11

0.11

0.14

0.11

0.22

0.14

0.18

Total score

0.09

0.24

0.90

0.68

0.16

0.82

0.79

0.98

0.24

0.11

0.10

0.18

1.00

Minimum realized

0.09

0.27

0.41

0.09

0.09

0.28

0.70

0.49

0.15

0.18

0.07

0.09

0.14

Australian magpie min. realized

0.10

0.12

0.12

0.11

0.10

0.10

0.10

0.10

0.11

0.10

0.10

0.10

0.19

Australian magpie total

0.13

0.15

0.15

0.08

0.11

0.08

0.13

0.14

0.08

0.10

0.12

0.16

0.08

Fuscous honeyeater min. realized

0.10

0.16

0.10

0.08

0.08

0.12

0.25

0.09

0.16

1.0

0.08

0.14

0.09

Fuscous honeyeater total

0.96

0.24

0.09

0.22

0.15

0.16

0.67

0.09

0.35

1.0

0.12

0.09

0.62

Red-rumped parrot min. realized

0.10

0.68

0.21

0.21

0.12

0.16

0.45

0.21

0.45

0.09

0.14

0.09

0.09

Red-rumped parrot total

0.13

0.24

1.00

0.16

0.34

0.51

0.07

0.52

0.08

0.86

0.12

0.54

0.81

Red wattlebird min. realized

0.14

0.10

0.11

0.09

0.10

0.13

0.12

0.12

0.18

0.10

0.08

0.09

0.53

Red wattlebird total

0.10

0.79

0.14

0.11

0.99

0.78

0.08

0.09

0.13

0.19

0.08

0.11

0.24

White-plumed honeyeater min. realized

0.38

0.17

0.16

0.32

0.44

0.82

0.95

0.31

0.09

0.08

0.14

0.23

0.08

White-plumed honeyeater total

0.62

0.11

0.11

0.39

0.21

1.00

1.00

0.11

0.11

0.11

0.18

0.03

0.17

White-winged chough min. realized

0.37

0.50

0.09

0.10

0.12

0.10

1.00

0.14

0.16

0.09

0.22

0.38

0.09

White-winged chough total

0.11

0.10

0.10

0.08

0.08

0.14

0.49

0.13

0.23

0.11

0.17

0.13

0.10

Willie wagtail total

0.15

0.12

0.07

0.11

0.10

0.16

0.07

0.08

0.10

0.12

0.36

0.11

0.25

Response variables are in rows and predictor variables in columns. Predictors with bold values of Pr(inc) had values >0.75, and so were identified as key predictors

Total scores

Elongated patch shapes, more remnant trees and higher loads of fallen timber increased total breeding scores, while grazing depressed the total breeding score (Tables 3, 4). The negative coefficient for MDS3 implies that total breeding was depressed in sites with greater numbers of trees with bark fissures, higher percentage ground cover of tussock grasses and more grass tussocks.
Table 4

(a) Estimated coefficients of key predictors (P > 0.75) with standard errors of estimates. Coefficients are based on the weighted posterior mean coefficient of retained models produced by Bayesian model averaging; (b) percentage independent effects (%I) of key predictors (P > 0.75) on the response variables determined by hierarchical partitioning

 

Age

Area

Shape

Surrounding vegetation

Number of remnant trees

Fallen timber

Grazing status

Average basal area

MDS3

(a) Model coefficients

 Total score

18.58 ± 10.09

5.40 ± 3.70

29.70 ± 21.45

−25.90 ± 10.44

−34.14 ± 11.77

 Fuscous honeyeater min. realized

17.95 ± 0.95

 Fuscous honeyeater total

3.73 ± 1.69

18.71 ± 3.71

 Red wattlebird total

−2.44 ± 1.73

4.85 ± 1.69

1.78 ± 1.29

 Red-rumped parrot total

3.84 ± 1.15

−4.33 ± 2.68

−2.52 ± 1.69

 White-plumed honeyeater min. realized

2.07 ± 1.41

17.30 ± 8.13

 White-plumed honeyeater total

3.84 ± 1.28

29.41 ± 8.11

 White-winged chough min. realized

6.44 ± 1.34

(b) Independent explanatory variance %I

 Total score

29.3

5.5

17.2

24.4

23.7

 Fuscous honeyeater min. realized

100

 Fuscous honeyeater total

22.9

77.1

 Red wattlebird total

14.7

72.4

12.9

 Red-rumped parrot total

68.0

21.6

10.5

 White-plumed honeyeater min. realized

42.8

57.1

 White-plumed honeyeater total

40.7

59.3

 White-winged chough min. realized

100

Species-specific outcomes (Tables 3, 4): Factors affecting total breeding scores differed among species for which there were substantial data. Numbers of remnant trees were important for the red wattlebird and white-plumed honeyeater. Higher fallen timber loads increased breeding activity in the latter. Breeding activity by the red-rumped parrot increased with elongated replantings. Replanting age and average basal area were positive predictors of fuscous honeyeater breeding. The only landscape-contextual effect (vegetation in the surrounding landscape, radius 5 km) was for the red wattlebird.

Minimum realized scores (fledglings)

The volume of fallen timber may have had an influence on the minimum realized score of all species (Pr(inc) = 0.70). The minimum realized score for the white-plumed honeyeater was influenced by the number of remnant trees and the volume of fallen timber (Table 3). The volume of fallen timber had a higher independent effect than the number of remnant trees (Table 4b). Volume of fallen timber had a positive effect (Table 4a) and was the sole key predictor of the minimum realized score for the white-winged chough. Average basal area was the only key predictor for the minimum realized score for the fuscous honeyeater, with strong evidence of its predictive ability. Replanting age was not a key predictor (Table 3). No variables were useful predictors of the minimum realized score for the Australian magpie (Table 3). There was only weak evidence for the effect of area on the minimum realized score of the red-rumped parrot and for the effect of MDS3 on the minimum realized score for the red wattlebird (Table 3). Fledglings observed for each species are shown in the Appendix. The list is similar to those reported by Barrett et al. (2008).

Are the ‘right’ species breeding in replantings?

Species-specific breeding activity was inversely correlated with rank differences in abundance (Fig. 2; Spearman rank correlation = −0.38, standard normal Z > 4.0). Thus, species with higher rank abundances in fragmented landscapes than in reference forests were much more active in breeding, while species declining in such landscapes engaged relatively little in breeding activity. There were no records of breeding activity by species declining by more than 20 ranks (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-008-1221-9/MediaObjects/442_2008_1221_Fig2_HTML.gif
Fig. 2

Summed (across all replantings) breeding activity of individual species of birds as a function of difference in rank abundance between small fragments and reference forests

Discussion

Our data suggest that by and large landscape contextual variables and patch characteristics had little influence on breeding activity; we will return to this point later. In-site habitat characteristics affect the ‘attractiveness’ of sites for recolonizing individuals (Doerr et al. 2006) as well as the availability of nesting and foraging resources for sustaining bird populations (Recher 2004). In-site resources considered important for avifaunal recolonization of restored sites included availability of fallen timber, species-rich plant assemblages and mature trees (Barrett 2000; Majer et al.2001; Recher 2004; Vesk and Mac Nally 2006). In-site characteristics had the greatest effect on the total breeding activity of birds in restored patches. In order of importance, grazing status, MDS3 (see Table 2), fallen timber and the number of remnant trees were influential on overall breeding activity. Planting age and average basal area affected the breeding activity and fledgling production of some species.

Grazing status had a negative influence on bird breeding. Several studies previously have reported grazing to have a negative effect on bird species abundance and breeding success (Luck 2003; Martin and Possingham 2005). Grazing may restrict food and habitat resources for ground-foraging insectivores by reducing and altering ground cover (Bromham et al. 1999; Luck 2003; Martin and Possingham 2005; McIvor et al. 2005) and affecting tree health (Davidson et al. 2007). Our study suggests that reduced grazing in restored sites would elevate bird breeding activity (see Majer et al. 2001), but results from a comparable study of bird breeding in river red gum replantings in northern Victoria indicated the opposite (Mac Nally et al., in press). We attribute the latter result to the dominance of breeding activity by small, ground-foraging species in the river red gum sites (e.g., superb fairy-wren Malurus cyaneus). Such species may be favoured by the more open habitats maintained by grazing. The usefulness of replantings clearly will depend on the species most responsive to the replanting activities and the conservation status of such species.

The volume of fallen timber positively influenced total breeding activity, the breeding activity of the white-plumed honeyeater and the number of fledglings produced by the white-plumed honeyeater and white-winged chough. Previous studies have found fallen timber to have a positive influence on ‘standing crop’ measures of avifauna (Mac Nally et al.2001; Lohr et al. 2002) and breeding success in small mammals (Mac Nally and Horrocks 2008). Mac Nally and Horrocks (2007) reported that the white-plumed honeyeater consistently increased in abundance following experimental increases of fallen timber loads. Fallen timber may provide shelter from predators for ground foraging birds, which constitute a large proportion of the avifauna of Australian temperate woodland (Antos and Bennett 2006). Fallen timber is associated with increased insect abundance and diversity (Grove 2002; Evans et al. 2003), which increases food resources for ground foragers, such as the white-winged chough, and canopy and bark foragers, such as the white-plumed honeyeater (Mac Nally 1994). Mac Nally et al. (2002) recommended ca. 40–50 t/ha of fallen timber is best for achieving beneficial biodiversity outcomes. As such, we suggest that these amounts be added to or maintained on restored sites to facilitate breeding.

Large, old trees are valuable habitat resources (Ford and Barrett 1995), and previous studies have found remnant trees to have a positive influence on ‘standing crop’ measures of avifauna (Ford and Barrett 1995; Barrett 2000). Remnant trees provide diverse and abundant invertebrate communities, nectar resources, overhanging branches for perching and nesting, and specialized resources, such as decorticating bark and hollows (Yahner 1982; Barrett et al. 1994; Ford and Barrett 1995; van der Ree and Bennett 2001; Recher 2004). Remnant trees also are important for encouraging bird breeding in restored areas. The number of remnant trees had a positive effect on total breeding activity and on the breeding activity of the red wattlebird and the white-plumed honeyeater, and fledgling production of the latter. Preservation of remnant trees should be considered as a high priority in restoration plans because the resources these provide take a long time to be replaced (Ford and Barrett 1995; Vesk et al. 2008b). If feasible, restoration activities should be prioritised to take place in locations where there are remnant trees.

Average basal area and replanting age were important breeding predictors for the insectivorous fuscous honeyeater. Larger and older trees support more diverse and abundant invertebrate assemblages due to the greater surface areas and diversity of microhabitats (van der Ree and Bennett 2001; Timewell and Mac Nally 2004), as well as more opportunities for species requiring different nesting heights (Vesk et al. 2008a). This highlights the important role that specific in-site habitat characteristics play for supporting viable populations of particular bird species.

Inferential limitations

Our results suggest that in-site factors were the primary controller of breeding activities. However, there are several key qualifications on this conclusion. First, the range of areas of replanting vegetation was not extensive in this study because, in general, replantings take place at a small scale. The largest site was just 54 ha, and most sites were <10 ha. As bird assemblages are known to significantly differ between sites of much larger scales than the present study (Mac Nally 2007b), a larger representation of site areas may have revealed different effects of area on bird breeding. We sought fruitlessly for replantings of larger areas, but these have not been undertaken yet. The only extensive plantings are for commercial forestry applications, but these usually consist of close ranks of non-endemic species, such as the Sydney blue gum (Eucalyptus saligna). We are cautious concerning the absence of very extensive (>1,000 ha) blocks of replanted vegetation that may have provided outcomes that misrepresent the potential biodiversity value of large-scale plantings of endemic trees.

Second, while our study included sites ranging from 9 to 111 years since restoration, which is a much longer time than most other reports (Martin et al.2004), the oldest sites still may not have provided adequate nest sites for hollow-dependent species. Therefore, the time period of restoration covered by our sites may not have allowed the detection of an increase with age in the use of restored sites for bird breeding, and particularly hollow nesters.

Third, there was no uniformity in the planting methods used at the study sites (Vesk et al. 2008b). Therefore, there were substantial variations that were unaccounted for among sites. While all sites were within the box-ironbark region, there was great variation in the tree species used for replanting. This may subsequently have affected resources available for bird breeding because of variations in growth rates, structural characteristics, etc. It is only in recent years that shrubs and tussocky grasses have been included in restoration replantings, so species reliant on these habitat resources have not been provided with critical resources and ought not be expected necessarily to occupy and breed in such sites (Barrett et al. 2008).

Fourth, our results suggest that small replantings are not contributing much to recruitment of species of most conservation concern based on their declines in abundances in increasingly fragmented and denuded landscapes. Much of the breeding activity was by species known to be tolerant of open country and limited areas of available woodland (e.g., white-plumed honeyeater, Australian magpie, red-rumped parrot; Mac Nally et al.2000). A significant observation is that breeding by one species, the fuscous honeyeater, which appears to decline greatly in more fragmented landscapes (Mac Nally 2007b), was associated strongly with replanting age, fallen timber and average size of trees. That is, this species breeds mostly in older, structurally mature habitats.

Focus for future restoration

Using production measures, such as breeding activity and breeding output, as indicators of success of replantings much improves our assessments of the biodiversity benefits of restored areas (Mac Nally 2007a). In the set of sites considered here, in-site characteristics had the largest influence on overall breeding activity. In-site habitat characteristics were the only important predictors of the breeding activity and/or fledgling production for the white-plumed honeyeater, fuscous honeyeater and white-winged chough. We have touched on the limitations of the data set, highlighting both the small range of areas and of ages of replantings. These limitations almost certainly occlude what might be substantially greater benefits of much larger, or perhaps more connected, replantings. One of the major caveats is that the species most in need of support probably are not well catered for by the current availability of replantings vis-à-vis areas and ages.

Acknowledgments

We gratefully acknowledge the support of the Australian Research Council through grants nos. DP0343898 and LP0560518. We thank Erica Fleishman, Andrew Bennett, Hugh Ford, Michael Craig, Jim Radford and Peter Vesk for comments, support and feedback on this project. Work was conducted under the specifications of the Monash University Animal Ethics Committee application BSCI/2002/11. This is contribution 120 from the Australian Centre for Biodiversity at Monash University.

Supplementary material

442_2008_1221_MOESM1_ESM.doc (66 kb)
Electronic supplementary material (DOC 66 kb)

Copyright information

© Springer-Verlag 2008