Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2006) 15, 303–317 Blackwell Publishing Ltd

RESEARCH PAPER

A test of the mechanisms behind avian generalized individuals–area relationships Marco Pautasso* and Kevin J. Gaston

Biodiversity and Macroecology Group, Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK

ABSTRACT

Aim Questions related to abundances of organisms are central to ecological research. A priori, a scale independent estimation of abundances would be expected. However, we find estimates of numbers of bird individuals from all over the world to increase less than proportionately with increasing plot size. At the whole assemblage level, the pattern holds across biogeographical regions and habitats. The slope of the interspecific and, for the majority of species, the intraspecific individuals–area relationship is also significantly shallower than 1. The question arises as to which mechanisms cause these patterns. Location Global. Methods At the assemblage, interspecific and intraspecific levels, we tested three mechanisms that could be responsible for these patterns by comparing the slope of the individuals–plot area relationship for subsets of a database compiled from the literature. Spatial autocorrelation was controlled for. Results There was no evidence for an influence of plot area choice in order to sample a constant number of individuals. Evidence for higher survey efficiency was available only with increasing number of visits at the intraspecific level. Evidence for influences of habitat heterogeneity was present at the assemblage, interspecific and intraspecific levels. This mechanism can work only if small plots are delimited non-randomly in homogeneous habitat. Main conclusions Avian population size estimates without indication of the area over which they were obtained are of substantially less value than those coupled with that information. Ecologists planning to compare avian abundances between plots varying in some other factor of interest should minimize variations in their areas and/or account for them in data analyses. Population viability analyses, regional and global population size estimates, site prioritization and the scaling of ecosystem and species energy utilization need to address the plot area effect on assemblage and individual species abundances.

*Correspondence: Marco Pautasso,

Imperial College London, Wye Campus, High Street, Wye, Kent TN25 5AH, U5K E-mail: [email protected]

Keywords Birds, census technique, density, plot size, point counts, population sampling, scale dependence, species–area, survey methodology, territory mapping.

A vast number of abundance estimates have been generated for birds for an extremely diverse range of species and habitats (see, e.g. collations in Cramp et al., 1978–2000; Del Hoyo et al., 1992–2003). In so doing, a variety of issues related to abundance estimation have been identified, including sampling error associated with the short-term variability of assemblages

(e.g. Shields, 1977), variations in reliability between observations by different ornithologists (e.g. Best, 1975), variations in detectability between species (e.g. Järvinen & Väisänen, 1975) and different survey methodologies (e.g. Blondel, 1969; Emlen, 1971; Berthold, 1976). A further issue, with the potential to undermine substantially the intercomparability of abundance estimates has, however, received surprisingly meagre attention. This is the effect of plot area. Standard texts on the estimation of

© 2006 Blackwell Publishing Ltd www.blackwellpublishing.com/geb

DOI: 10.1111/j.1466-822x.2006.00222.x

INTRODUCTION

303

M. Pautasso and K. J. Gaston animal, and more specifically avian, abundances fail to assign to this the importance it appears to deserve (e.g. Bibby et al., 2000; Borchers et al., 2002; Gregory et al., 2004). Two different relationships between abundances and plot area have been distinguished (Gaston & Matter, 2002). The first, the patch individuals–area relationship (PIAR), refers to variations in plot area as a consequence of differences in the size of discrete habitat patches (as sampling units). Typically, it gives rise to a more than proportionate increase in the numbers of individuals with increasing area (e.g. Ambuel & Temple, 1983; Bellamy et al., 2000; Estades, 2001; Brotons et al., 2003; Crozier & Niemi, 2003; but see Cieslak & Dombrowski, 1993; Mayor & Schaefer, 2005), but is likely to act over only relatively narrow spatial scales and is influenced by the landscape matrix in which the patches are embedded. The second, the generalized individuals–area relationship (GIAR), is typically less than proportionate and much more general. Although Kendeigh (1944) noted that high abundances can be obtained when sampling plots of less extension than the area covered regularly by the birds counted, the effect was, to our knowledge, highlighted first in a paper that could have been seminal had it not been published in a low-impact ornithological journal and in German (Scherner, 1981). Scherner warned that it is objectionable not only to compare species numbers directly but also species abundances estimated from areas of differing size. He provided evidence for such a bias from breeding bird abundance data from the former Democratic Republic of Germany, where high densities were generally those obtained in small areas and low densities in large areas. A similar concern was expressed by James and Rathbun (1981). Subsequently, GIARs have been documented for birds on a number of occasions (e.g. Engstrom & James, 1981; Smallwood, 1995; and further references cited below) as well as for mammals (e.g. Schonewald-Cox et al., 1991; Smallwood, 1998; Smallwood & Schonewald, 1998; Silva et al., 2001; Prince & Silva, 2002; Schaefer & Mahoney, 2003; Mayor & Schaefer, 2005). The GIAR is of particular interest because abundance estimates are collated frequently from the literature to address a wide range of theoretical and applied issues in ecology and conservation biology. Typically this is conducted without reference to the areas from which those estimates were obtained, raising the possibility that results and conclusions may be seriously misleading (e.g. Smallwood, 1997; Blackburn et al., 1999). The severity of this problem depends on how general the GIAR is, and what causes it. Part of the difficulty in assessing the generality of the GIAR is that, although they have not been well differentiated in the past, there are three such relationships. The first concerns variation with area in the summed abundances of assemblages across all the species present on a plot (e.g. Solonen, 1996; Storch & Kotecky, 1999; Jimenez, 2000). We will call this the assemblage GIAR. The second is the interspecific GIAR, which concerns the relationship between the abundances of species and area, in which each species is represented by one or more data points (this has been best documented for mammals, e.g. Smallwood & Schonewald, 1996; Smallwood, 2001). The third relationship is the intraspecific GIAR, which concerns the relationship between abundance and area for a single species at multiple locations (e.g. 304

Village, 1984; Kostrzewa, 1988; Smallwood, 1995). In practice, of course, these relationships are likely to be tightly linked, but depending on the circumstances it is not necessarily the case that the observation of one leads inevitably to the others. Mechanisms Four principal mechanisms have been proposed to account for GIARs: plot area choice, survey efficiency, habitat heterogeneity and plot edge effects (Scherner, 1981; Wiens, 1986; Haila, 1988; Gaston et al., 1999; Nee & Cotgreave, 2002). Each of these can be used to explain assemblage, interspecific and/or intraspecific GIARs. First, numbers of individuals may increase less than proportionately with increasing area if smaller areas are chosen deliberately for study because they have higher densities. This might occur as a consequence of the publication of avian abundance estimates, at least in part, as a by-product of behavioural, physiological and environmental investigations, which may require a certain number of observations to be reached regardless of the survey effort needed. Hence, the prediction here is that a constant number of individuals is surveyed irrespective of the plot area. Secondly, numbers of individuals may increase less than proportionately with increasing plot area if survey efficiency declines with increasing plot area. In this case, numbers of bird individuals should increase with increasing area less rapidly in dense vegetation compared to open habitats, because bird detectability is likely to remain higher even over large areas in open habitats, whereas typically it declines rapidly in impenetrable forests. Similarly, if survey efficiency is responsible for the effect of area, then less reliable survey methodologies (point counts and line transects) should show a shallower relationship between number of individuals and area than territory mapping because, in the latter case, birds are commonly not counted from a certain distance but by walking over the plot and mapping the locations of individuals. For similar reasons, within data from territory mapping surveys, numbers of individuals from studies with a higher number of visits should show a steeper increase with increasing area than those from surveys with a smaller number of visits. Thirdly, numbers of individuals may increase less than proportionately with plot area due to increasing habitat heterogeneity with larger plot areas. Two ways have been suggested in which this might occur. In common with the other possible mechanisms mentioned thus far, this might be a result of survey methodology. If smaller plots are more likely than larger ones to be focused on individual habitat types with which a species or assemblage is associated principally or disproportionately, then habitat heterogeneity will tend to increase with plot area as greater expanses of unsuitable habitat are included (Gaston et al., 1999). Thus, the slope of the abundance–plot area relationship is expected to be shallower in plots of heterogeneous habitat compared to those of homogeneous habitat. Alternatively, Nee and Cotgreave (2002) have suggested that intraspecific GIARs may follow from the existence of species–area relationships. If increases in species numbers with increasing plot area

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances are generated by increasing habitat heterogeneity then, they argue, it is inevitable that numbers of individuals of particular species must also increase less than proportionately with plot area. A further mechanism for GIARs has been suggested to be the effect of plot edges. This is not the effect of habitat edge on bird abundances revealed in fragmentation studies, which results from environmental changes (e.g. Baker et al., 2002), but rather the effect of the study plot boundary. Smaller areas have a higher ratio of area at small distances from the study plot edge to the total area, compared to large plot areas. This may inflate abundances if surveys include individuals temporarily inside the plot but normally obtaining some or all of their resources from outside, but this argument can be reversed, i.e. a higher edge to area ratio can also mean that individuals living in an area being studied are missed because they are temporarily absent from that area (Nee & Cotgreave, 2002). In addition, interspecific GIARs may arise because investigators survey larger-bodied species over larger areas or using different techniques than for smaller species (Blackburn & Gaston, 1996; Smallwood et al., 1996). The relationship, or at least its precise form, may be artefactual or reflect genuine biological differences in the way in which organisms of different body size use space, depending on how good ecologists are at modifying survey methodology to address differences in this use of space (for discussion see Blackburn & Gaston, 1999; Johnson, 1999). In this paper, we test for the existence of a generalized individuals– area relationship at the assemblage, interspecific and intraspecific levels, employing a new global collation of avian abundance estimates. We then test the different mechanisms that have been proposed to give rise to these patterns. MATERIALS AND METHODS

We did not include plots in which assemblages had been disrupted recently by disturbance [addition of nest boxes, dieback, degradation, fire, fragmentation (fragmented plots pertain to PIAR, see Introduction), grazing, logging, predation, thinning, unusual weather, urbanization, windthrow]. These constraints left data for a total of 2711 plots (Fig. 1a). The number of species with at least one abundance entry in this subset is 3618, i.e. 37% of the species in the Sibley and Monroe (1990) world list of birds. The selected plots were surveyed mainly with territory mapping, the most reliable methodology for many species (63%); the rest were surveyed with line transects (22%) and point counts (15%). We recorded the number of visits or counts for each plot. Analysed plots span a latitudinal range between 54° S (Tierra del Fuego, Argentina) and 79° N (Ellesmere Island, Northwest Territories, Canada). We assigned them to a biogeographical region according to the classic modified diagram of Wallace’s realms (e.g. Watts, 1971; p. 246), with addition of the oceanian realm for islands of the Pacific Ocean. We also assigned sites to a habitat category, following the scheme of Klein Goldewijk (2001). Because of the global scale, this is a coarse grouping into 15 habitats. In turn, we grouped sites into an open habitat subset (cropland, pasture, tundra, grassland/steppe and hot desert) and a forest subset (wooded tundra, boreal forest, cool conifer forest, temperate mixed forest, temperate deciduous forest, warm mixed forest, tropical woodland, tropical forest), with the exception of scrubland and savannah. We added water, wetland (open habitats), wetland forest (forested habitat), urban area, plantation and mixed habitat, for a total of 21 habitats. We did not include plantations in the forest subset because of their artificiality. We retrieved body masses for 90% of the bird species with an abundance entry in complete surveys from existing compilations or published monographs. We did not include the remaining 10% of species in analyses of interspecific patterns of the individuals– area relationship controlling for body mass.

Data We searched the published literature for estimates of terrestrial bird abundances with an indication of study plot area, irrespective of the species, habitat, region or year to which they referred. We did not consider bird species lists without an estimation of the number of individuals, or bird counts that did not declare plot area. We obtained papers with systematic keyword searches in scientific databases, inspection of cited references in papers already found and scanning of the tables of contents of non-indexed journals. We restricted analyses presented here to abundances from studies aimed at a complete study of the avian assemblage in a definite area. Thus, we did not include abundances from studies of one or a few selected species. Similarly, we excluded counts carried out over variable distances. We expressed abundances as the number of individuals on a study plot. If surveys were repeated at the same site in multiple years, we included only data from one year chosen at random because of potential nonindependence. We excluded winter surveys at temperate latitudes. Analyses are thus restricted to intertropical surveys irrespective of the season and to breeding bird surveys at temperate latitudes.

Analyses We calculated total abundances of the assemblage (in number of individuals) for each plot by summing the abundances for the species with abundance estimates. These assemblage abundances do not include species recognized by the authors of particular studies as being present at such low numbers as to impede seriously their detection or a meaningful estimation of their abundance. To improve data consistency, we generally concentrated tests of mechanisms generating individuals–area relationships on forest plots. We calculated the interspecific individuals–area relationship for the 3427 species with at least one abundance entry in the database. We also documented intraspecific individuals–area relationships for the 100 species with the most abundance entries in the 742 forest plots surveyed with territory mapping. As a test of the habitat heterogeneity hypothesis we used species-specific slopes of the individuals–area relationship for the 54 species for which Imbeau et al. (2003) provide a categorization of their requirements for successional status (either early or mature forest = specialist; both successional stages = generalist) and patch use (either edge or interior = specialist; both edge and

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

305

M. Pautasso and K. J. Gaston

Figure 1 (a) Geographical distribution of survey plots analysed (Mercator projection) and (b) plot area (km2) as a function of survey year for the same set of surveys (n = 2711).

interior = generalist) and for which the individuals–area relationship in our data set is significant in controlling for spatial autocorrelation (see below). We performed a test of the suggestion by Nee and Cotgreave (2002) that the intraspecific GIAR follows from the species–area relationship following their methodology. We determined a species– area relationship for each species for the assemblages of which it was a part. We then used a two-tailed two sample t-test to compare the slope of the intraspecific individuals–area relationship predicted from this species–area relationship (obtained by subtracting the species–area slope from the assemblage individuals–area one for the same assemblages) with that observed. We obtained the standard error of the predicted slope from the square root of the sum of the single standard errors. Individuals–area relationships are between plot area and abundance (number of individuals), a variable of which plot area is not a component. A correlation between plot area and density (number of individuals/plot area) would, instead, be problematic. Although the claim that such a correlation is statistically invalid or spurious had been thought commonly to be a misconception (Prairie & Bird, 1989), Brett (2004) provides a way to gauge the magnitude of such potentially spurious correlations. Following his methodology, we found mean spurious coefficients of determination 306

(pooling data from all habitats and survey methodologies together; n random samples = 50) between assemblage density and plot area of 0.55 [standard deviation (SD) = 0.01] at the interspecific level (average values for abundance and plot area) of 0.46 (SD = 0.01) and for Fringilla coelebs (the species with most entries in the database) of 0.45 (SD = 0.02). We thus avoided using assemblage or species densities and restricted our analyses to individuals–area relationships per se, although the issue was framed in terms of negative density–area relationships in the published literature. As a test of the mechanisms explaining GIARs, we used interaction terms to assess the statistical significance of differences in the slopes of regression lines of the scaling of number of individuals with plot area coming from subsets of the data. Both variables were log-transformed prior to analyses to conform to the assumptions of statistical tests. We ran analyses in  8.2. We employed the same package to control for the effect of spatial autocorrelation with mixed models using an exponential covariance structure, after having ascertained its better fit (in terms of both Akaike’s and Bayesian information criteria) to the null model for assemblage and individual species abundances compared to spherical, Gaussian, linear, linear–logarithm and power structures. We added a nugget parameter to the two

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances

Figure 2 Log–log frequency distribution of (a) plot area (km2) and (b) assemblage abundance (individuals). Plot area class −1.00 to 0.76 contains one plot, and assemblage abundance classes −1.25 to 1.01, 4.50–4.74, 5.25–5.49, 5.50–5.74 and 5.75 –5.99 also contain one plot, but the logarithmic scales of the y-axes cause these classes to appear empty on the figures.

parameters governing the converging process because of marked variations in the response variable at a small scale hindering convergence (Littell et al., 1996). We sought convergence with no more than 50 iterations of the mixed model at a significance level of 10−5. We controlled for spatial autocorrelation with absolute distances after reprojection of latitudes and longitudes of plots to Lambert conical coordinates for each biogeographical region. Generally, the evidence presented for the single mechanisms is affected by whether or not spatial autocorrelation is controlled for, highlighting the profound consequences of taking the relative positions of spatial data points into consideration (Legendre, 1993). Apart from the proportion of variance explained in regressions, here we concentrated on the more conservative outcomes controlling for spatial autocorrelation. RESULTS For the 2711 plots analysed, plot area varied by more than six orders of magnitude, from point counts of a few hundred square

metres in Indian tropical forests to line transects across hundreds of square kilometres in the Lake Constance region (Fig. 2a). The average plot area was 3.2 km2, the median was 0.13 km2. Altogether, the area surveyed totalled c. 8600 km2, which is approximately the size of Puerto Rico. Variations in plot area were not independent of a series of environmental factors. Even if the proportion of variance explained was not substantial, the area of plots from all habitats and all survey techniques was significantly negatively correlated with mean annual temperature [R2 = 0.01, log(km2) = 1.43 – 0.19 log(temp), slope standard error (SE) = 0.09, F1,2710 = 5.0, P = 0.03] and precipitation [R2 = 0.04, log(km2) = 1.02 – 0.60 log(prec), slope SE = 0.11, F1,2710 = 29, P < 0.0001]. There were also significant positive correlations of plot area with absolute latitude [R2 = 0.01, log(km 2) = −1.00 + 0.007 abs(lat), slope SE = 0.002, F1,2710 = 10, P = 0.001] and survey year [R2 = 0.00, log(km2) = −14.3 + 0.006 years, slope SE = 0.001, F1,2710 = 21, P < 0.0001; Fig. 1b], yet in these cases not only the proportion of variance explained but also the slopes were negligible.

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

307

M. Pautasso and K. J. Gaston consistent across biogeographical regions (Table 1), the only exception being data from the oceanian realm, where plots were bound to pertain more to the patch individuals–area relationship (see Introduction), as they were often surveyed on small islands with limited habitat extension. A positive squared plot area term was significant in the Australasian, Nearctic, Neotropical, Palearctic and Oriental realms, but only in the last case was the additional proportion of variance explained (12%) substantial. The slope of the increase of assemblage numbers of individuals with increasing plot area was significantly shallower than 1 in 16 of 21 habitats (Table 1). It was not significantly different from one in pasture, tropical woodland and hot deserts only. A positive squared plot area term was significant in four habitats (boreal and tropical forest, plantation, mixed habitat), but the additional proportion of variance explained was negligible. There was no significant effect of the range of plot areas on which habitats, or on how strongly habitats showed the pattern. The pattern, however,

Assemblage numbers of individuals also spanned six orders of magnitude, from less than one individual in urban surveys in Brisbane to several hundred thousand individuals in the Lake Constance surveys mentioned (Fig. 2b). The average assemblage abundance across all habitats was c. 1150 birds per survey, the median was 125 birds per survey.

Assemblage individuals–area relationships The logarithm of plot area explains roughly one-half of the variance in the logarithm of assemblage numbers of individuals, pooling data from all habitats and survey methodologies (Table 1; Fig. 3a). All other things being equal, for roughly every five orders of magnitude increase in plot area, assemblage abundance increased by four, i.e. by one order of magnitude less. A squared plot area term was not significant. The less than proportionate increase in numbers of individuals with plot area was

Table 1 Regressions of logarithmically transformed assemblage abundance (number of individuals) against logarithmically transformed plot area (km2), for different biogeographical regions and habitats; n = number of plots; min = smallest plot area (km2); max = largest plot area (km2); r 2 = proportion of variance explained in regression without controlling for spatial autocorrelation; intercept (a), slope (b), standard errors of slope (SE), F and P-values are given controlling for spatial autocorrelation; n.s., P ≥ 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.c. did not converge after 50 iterations (see Methods). S: slope significantly shallower than 1; P: proportional increase, i.e. slope not significantly different from 1; N: no increase, i.e. slope not significantly different from 0

All assemblages   Australasian Ethiopian Nearctic Neotropical Oceanian Oriental Palearctic  Cropland Pasture Water Tundra Wooded tundra Boreal forest Cool conifer forest Temp. mixed forest Temp. dec. forest Warm mixed forest Grassland Hot desert Scrubland Savannah Tropical woodland Tropical forest Wetland forest Wetland Plantation Mixed habitat Urban area

308

n

Min

Max

r2

a

b

SE

F

P

2711

0.0002

600.0

0.47

2.73

0.80

0.02

2584

****

S

275 112 844 179 25 32 1244

0.001 0.008 0.008 0.002 0.024 0.0002 0.002

207.0 54.5 83.2 14.8 0.5 0.2 600.0

0.45 0.68 0.40 0.46 0.79 0.25 0.45

2.78 2.80 2.64 2.98

0.84 0.78 0.78 0.80

0.05 0.08 0.03 0.07

242 102 739 139

S S S S

2.24 2.66

0.18 0.77

0.09 0.02

4.4 1184

**** **** **** **** n.c. * ****

109 51 61 33 10 93 152 198 505 96 90 12 293 25 76 58 116 128 161 106 347

0.009 0.025 0.004 0.045 0.040 0.040 0.015 0.006 0.008 0.003 0.007 0.062 0.001 0.034 0.005 0.0002 0.005 0.003 0.0002 0.053 0.006

19.1 5.4 3.9 43.0 1.0 104.3 4.0 6.4 5.1 13.0 83.2 0.7 46.6 54.5 41.2 9.5 17.0 20.0 3.8 600.0 485.0

0.48 0.77 0.15 0.65 0.19 0.69 0.32 0.54 0.60 0.59 0.63 0.78 0.37 0.77 0.38 0.60 0.68 0.75 0.29 0.22 0.57

2.52 2.55 2.90 1.94 2.21 2.55

0.58 1.10 0.59 0.50 0.23 0.78

0.07 0.09 0.15 0.08 0.17 0.05

70 162 15 39 1.9 203

S P S S N S

2.86 2.93 2.77 2.57 2.56 2.67 2.85 2.99 3.14 3.04 2.81 2.56

0.78 0.89 0.76 0.87 1.00 0.88 0.74 0.94 0.73 0.83 0.82 0.64

0.04 0.03 0.07 0.06 0.17 0.05 0.13 0.08 0.08 0.03 0.06 0.08

355 715 130 179 35 312 30 193 81 800 204 68

2.99

0.87

0.03

768

**** **** *** **** n.s. **** n.c. **** **** **** **** **** **** **** **** **** **** **** **** n.c. ****

S S

S S S S P S S P S S S S S

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances

Figure 3 The relationship between (a) log10 assemblage abundance (individuals) (dotted line shows slope of 1) and (b) log10 number of species at a site and log10 plot area (km2).

differed across habitats (Table 1). A general linear model of assemblage numbers of individuals as a function of area, habitat (categorical variable) and their interaction revealed that the latter was a highly significant term (P < 0.0001).

Mechanisms (i) Plot area choice. In the extreme case, the plot area choice mechanism predicts that the total number of individuals counted is constant across plot areas. However, the slope of the individuals– area log–log relationship, although generally shallower than 1, was not significantly different from 0 only in exceptional cases (stratifying analyses by habitat, in wooded tundra only; see Table 1). (ii) Survey efficiency. Counter to expectation from this mechanism, the slope of the individuals–area relationship was significantly shallower in open compared to forested habitats [open habitat: R2 = 0.54, log(ind) = 2.50 + 0.63 log(km2), slope SE = 0.04, F1,487 = 300, P < 0.0001; forests: R2 = 0.52, log(sumab) = 2.83

+ 0.81 log(km2), slope SE = 0.02, F1,1308 = 1578, P < 0.0001; differing slopes: F1,1795 = 5.7, P = 0.02]. Without controlling for spatial autocorrelation the slopes met the predicted direction of differences, but were not significantly different (F1,1795 = 0.6, P = 0.44). The slope of the individuals–area relationship from data obtained in forests with the most reliable survey technique was, counter to expectation, not significantly different from that obtained with less reliable methodologies [line transects and point counts: R2 = 0.51, log(sumab) = 2.86 + 0.85 log(km2), slope SE = 0.03, F1,521 = 787, P < 0.0001; territory mapping: R2 = 0.55, log(ind) = 2.82 + 0.77 log(km2), slope SE = 0.03, F1,780 = 771, P < 0.0001; differing slopes: F1,1301 = 0.7, P = 0.39]. Surveys with a higher number of visits (forests only; all survey techniques) did not show a significantly steeper increase in numbers of individuals with increasing area than those from surveys with a smaller number of visits [five or fewer visits: R2 = 0.58, log(ind) = 2.83 +0.80 log(km2), slope SE = 0.03, F1,486 = 691, P < 0.0001; more than five visits: R2 = 0.49, log(ind) = 2.86 + 0.83 log(km2), slope SE = 0.03, F1,741 = 843; P < 0.0001; differing slopes: F1,1227 = 1.4, P = 0.23].

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

309

M. Pautasso and K. J. Gaston

Figure 4 The relationship between (a) log10 average species abundance (individuals) at a site and log10 plot area (km2) and (b) the partial regression plot of log10 species abundance (individuals) as a function of log10 plot area (km2) controlling for log10 species mass (g) (dotted line shows slope of 1).

(iii) Habitat heterogeneity. We tested the habitat heterogeneity mechanism by comparing the slopes of the individuals–area relationship in forest plots of homogeneous habitat and in plots of mixed habitat (all survey techniques). As predicted, the slope of the individuals–area relationship was significantly shallower in mixed habitat compared to forests [mixed habitat: R2 = 0.22, log(ind) = 2.67 + 0.34 log(km2), slope SE = 0.06, F1,105 = 30, P < 0.0001; forests: R2 = 0.52, log(ind) = 2.83 + 0.81 log(km2), slope SE = 0.02, F1,1308 = 1579, P < 0.0001; differing slopes: F1,1413 = 6, P = 0.01].

Interspecific individuals–area relationships Unsurprisingly, the number of species in a plot increased with plot area, even though in isolation the effect was not strong, being confounded by many other sources of variation between plots [all habitats, all survey techniques: R2 = 0.05, log(spp.) = 1.43 + 0.18 log(km2), slope SE = 0.01, F1,2711 = 329, P < 0.0001; Fig. 3b]. Hence, all other things being equal, i.e. assuming that species turnover is constant, the increase in average number of individuals per species with increasing plot area has to be at least 310

as shallow as the one for assemblage GIARs, given that average species abundances are obtained by dividing assemblage abundance by the number of species on a plot. This was indeed the case [R2 = 0.48, log(avgind) = 1.30 + 0.62 log(km2), slope SE = 0.01, F1,2711 = 2201, P < 0.0001; Fig. 4a]. All other things being equal, for each increase in plot area by three orders of magnitude, the average species abundance increased only by approximately two, i.e. by one order of magnitude less. At the interspecific level, i.e. plotting the average number of individuals for every single species with an abundance entry as a function of the average plot area, the slope was highly significantly less than proportional [R2 = 0.43, log(avgind) = 1.21 + 0.70 log(km2), slope SE = 0.01, F1,3426 = 2572, P < 0.0001]. A squared plot area term was significant but did not explain much additional variance. Randomly sampling abundances and their associated plot areas (for those species with more than one abundance entry in the database, in order to remove multiple occurrences of the same species) again provided a less than proportionate relationship that was highly significant, with an even shallower slope [R2 = 0.23, log(ind) = 0.6 + 0.41 log(km2), slope SE = 0.01, F1,3426 = 1051, P < 0.0001; these were the results of the ninth random

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances sampling of 30; the average slope of the 30 regressions was 0.41, SE = 0.01]. A squared plot area term was significant but scarcely explained any additional variance. The interspecific individuals– area relationship remained less than proportionately increasing when controlling for differences in the body sizes of the species [ninth random sample: R2 = 0.24, log(ind) = 0.96 + 0.43 log(km2) − 0.17 log(mass), parameter estimates SE = 0.01 and 0.02, respectively, F2,3426 = 555, P < 0.0001 for both factors; Fig. 4b].

Mechanisms Following from the results above, to test mechanisms generating individuals–area relationships at the interspecific level, we employed one randomly sampled estimate of number of individuals and the corresponding plot area for species with more than one abundance entry. Therefore a species appeared in only one analysed set, even if it was surveyed in both forests and open habitats, with both line transects or point counts and territory mapping, in both forests and mixed habitat. Results of tests were robust to the particular set of random samples used. As before, testing of mechanisms focused principally on forest plots (see Methods). (i) Plot area choice. Although the slope of the interspecific individuals–area relationship was shallower than 1, there was still a significant interspecific increase in the number of individuals with increasing plot area (see above), counter to expectations of the plot area choice mechanism in its extreme form. (ii) Survey efficiency. Counter to expectations, the slope of the interspecific individuals–area relationship (all survey methods) was not shallower in forests compared to open habitats [open habitats: R2 = 0.21, log(ind) = 0.62 + 0.39 log(km2), slope SE = 0.02, F1,935 = 235, P < 0.0001; forests: R2 = 0.20, log(ind) = 0.70 + 0.42 log(km2), slope SE = 0.02, F1,1928 = 495, P < 0.0001; differing slopes: F1,2863 = 0.7, P = 0.39]. Counter to expectations, the slope of the interspecific individuals– area relationship for forest species was significantly steeper for species surveyed with point counts and line transects compared to territory mapped species [line transects and point counts: R2 = 0.27, log(ind) = 0.71 + 0.49 log(km2), slope SE = 0.03, F1,1008 = 382, P < 0.0001; territory mapping: R2 = 0.10, log(ind) = 0.66 + 0.30 log(km2), slope SE = 0.03, F1,659 = 76, P < 0.0001; differing slopes: F1,1668 = 170, P < 0.0001]. Counter to expectations the slope of the interspecific individuals– area relationship (forests, all survey methods) was not significantly shallower for species surveyed with a lower number of visits [five or fewer visits: R2 = 0.20, log(ind) = 0.79 + 0.39 log(km2), slope SE = 0.03, F1,774 = 194, P < 0.0001; more than five visits: R2 = 0.15, log(ind) = 0.63 + 0.44 log(km2), slope SE = 0.04, F1,811 = 144, P < 0.0001; differing slopes: F1,1585 = 1.3, P = 0.26]. (iii) Habitat heterogeneity. As expected, the slope of the interspecific individuals–area relationship for species surveyed in forest plots of homogeneous habitat was significantly steeper from the one of species surveyed in plots of mixed habitat (all

survey techniques) [mixed habitat R = 0.41, log(ind) = 1.08 + 0.09 log(km2), slope SE = 0.06, F1,250 = 1.9, P = 0.17; forests: R2 = 0.20, log(ind) = 0.70 + 0.42 log(km2), slope SE = 0.02, F1,1938 = 495, P < 0.0001; differing slopes: F1,2188 = 37, P < 0.0001].

Intraspecific individuals–area relationships Provided that there is a scale-independent species turnover, the numbers of individuals for particular species will tend to increase at least as slowly with increasing plot area as do assemblage abundances. This was the case. Eighty-one of the 100 species with the most abundance entries in the database (all habitats, all survey methodologies) showed a relationship between number of individuals and plot area on a log–log scale with a slope significantly shallower than 1 (Fig. 5a). The average slope of the relationship for the species for which the mixed model converged (81, i.e. all species for which the mixed model converged showed the pattern) was 0.52 (SD = 0.12). Restricting analyses to data from forested habitats and territory mapping, 75 of the 100 species with the most abundance entries in that data subset showed a relationship between number of individuals and plot area on a log–log scale with a slope significantly shallower than 1 (two examples are given in Fig. 5b,c). The average slope of the individuals– area relationship for the 81 species for which the mixed model converged was 0.59 (SD = 0.19). Using data from all habitats and survey methodologies, a positive squared plot area term was significant in 37 species, but the additional proportion of variance explained was negligible (mean = 0.02, SD = 0.04). The magnitude of the intraspecific individuals–area pattern can be gauged by comparing the coefficient of variation of single species numbers of individuals (the SD of single species abundances divided by their average abundance) obtained from a series of surveys at the same site through time (we chose the one plot with the most repetitions in different years, and at least five repetitions, which were found for 71 of the 100 species with the most abundance entries in territory mapped forests), with the coefficient of variation of the abundances for the same species from the plots used to assess the GIARs. The two subsets differed significantly (: F1,141 = 110, P < 0.0001): year-to-year coefficients of variation of single species numbers of individuals at the same plot tended to be much lower (mean = 0.47, SD = 0.21) than coefficients of variation of numbers of individuals of the same species from plots differing in area (mean = 3.27, SD = 2.23).

Mechanisms As before, testing of mechanisms for intraspecific individuals– area relationships focused principally on forest plots (see Methods) and, unless stated otherwise, because of limitations on available data, on plots surveyed with all methodologies. (i) Plot area choice. Evidence for an intraspecific plot area choice mechanism was absent, as none of the 100 species with the most abundance entries in all habitats showed a slope of the individuals–area relationship that was not significantly different from 0 (see above).

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

311

M. Pautasso and K. J. Gaston

Figure 5 (a) The frequency distribution of the slopes of the individuals–area relationship for the 81 species [of 100 species with more abundance entries in the database (all habitats, all survey methodologies)] for which the mixed model converged; the relationship (territory mapped forest plots) between (b) log10 Colaptes auratus abundance (number of individuals) and (c) log10 Garrulus glandarius abundance (number of individuals) and log10 plot area (km2) (dotted line shows slope of 1); (d) slopes of the predicted intraspecific individuals–area relationship (obtained by subtracting the slope of the species–area relationship from the slope of the assemblage individuals–area relationship) as a function of those of the observed intraspecific individuals–area relationship for the set of 100 species with the most abundance entries in territory mapped forest plots for which all these relationships are significant in mixed models (n = 81).

(ii) Survey efficiency. Only one species (Turdus migratorius) showed an individuals–area relationship in forests that was significantly shallower than that in open habitats, as expected according to the survey efficiency mechanism. Counter to expectations, 28 species (Dendrocopos major, Dryocopus pileatus, Buteo buteo, Contopus virens, Myarchus crinitus, Vireo olivaceus, Catharus fuscescens, Turdus philomelos, Muscicapa striata, Erithacus rubecula, Phoenicurus phoenicurus, Dumetella carolinensis, Sitta carolinensis, Thryothorus ludovicianus, Troglodytes troglodytes, Polioptila caerulea, Parus montanus, P. atricapillus, Regulus regulus, Phylloscopus trochilus, P. collybita, Sylvia borin, Anthus trivialis, Prunella modularis, Fringilla coelebs, Dendroica coronata, Pheucticus ludovicianus, Passerina cyanea) showed a significantly steeper individuals–area relationship in forests compared to open habitats. Because many of these 28 species are forest specialists, this pattern could be interpreted as evidence for the habitat heterogeneity hypothesis. Forest specialist species may perceive open habitats as more heterogeneous than forests, and their individuals–area relationship would thus be shallower in open habitats compared to forests. Evidence for an intraspecific survey efficiency mechanism from different survey methodologies was limited, as only seven species 312

(Dendrocopos major, Buteo buteo, Vireo solitarius, Garrulus glandarius, Turdus merula, Pyrrhula pyrrhula, Dendroica coronata) showed a significantly shallower individuals–area relationship for data from point counts and line transects compared to data from territory mapped plots. More evidence is available for the survey efficiency mechanism from information on variation in numbers of visits. Here, 22 species (Dendrocopos major, Buteo buteo, Garrulus glandarius, Catharus ustulatus, Turdus merula, T. philomelos, Erithacus rubecula, Sturnus vulgaris, Certhia familiaris, C. americana, Troglodytes troglodytes, Phylloscopus trochilus, P. sibilatrix, Sylvia borin, S. communis, Prunella modularis, Fringilla coelebs, Pyrrhula pyrrhula, Coccothraustes coccothraustes, Emberiza citrinella, Spizella passerina, Dendroica virens) showed a significantly shallower individuals–area relationship when counted or mapped with five or less visits than when the number of visits was greater than five. (iii) Habitat heterogeneity. Only one species (Turdus merula) showed a significantly shallower slope of the individuals–area relationship in mixed habitat compared to forests of homogeneous type, as would be expected if habitat heterogeneity were driving

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances the less than proportionate individuals–area relationship. Moreover, counter to expectations, 13 species (Columba palumbus, Buteo buteo, Garrulus glandarius, Corvus corone, Turdus pilaris, Certhia familiaris, Parus palustris, P. ater, P. major, Regulus regulus, R. ignicapillus, Prunella modularis, Coccothraustes coccothraustes) showed a significantly steeper slope of the individuals–area relationship in mixed habitat compared to forests. The slopes of the individuals–area relationship for specialist species are expected to be shallower than those of generalist species, because for the latter the suitability of a given habitat is less heterogeneous than for the former. We compared the slopes of the individuals–area relationship for generalist vs. specialist species surveyed with territory mapping in forest habitats. Slopes of interior-edge specialist species (n = 31; average = 0.56; SD = 0.21) did not differ from those of generalist species (n = 23; average = 0.58; SD = 0.14) in a significant way (: F1,53 = 0.3, P = 0.61). As for successional status, specialist species (n = 42; average = 0.58; SD = 0.17) did not show steeper slopes than generalist species (n = 12; average = 0.54; SD = 0.24) in a significant way (: F1,53 = 0.3, P = 0.59). Following Nee and Cotgreave (2002), we compared the observed slopes of intraspecific individuals–area relationships with those predicted from the species–area relationships derived from the assemblages in which the respective species occurred. The two did not differ significantly (average P-value = 0.73; SD = 0.23) for any of the 81 of the 100 species with the most abundance entries in forest territory mapped plots for which all these relationships (assemblage individuals–area, species–area, single species individuals–area) were significant in mixed models. DISCUSSION The less than proportionate increase of avian assemblage and single species numbers of individuals with increases in plot area is such a widespread and strong pattern that it verges on the requirements for an ecological law (sensu Lawton, 1999). Remarkably, our analyses revealed that the less than proportionate relationship between numbers of individuals and plot area was sufficiently strong as to emerge despite the influence on avian abundances of variation in factors of more direct ecological and conservation interest, such as year, habitat, biogeographic region and the methodology employed to estimate numbers of individuals. Mechanisms As for other patterns in ecology, there was no necessity for a single mechanism to be responsible for the observed avian individuals–area relationships in any given case or across different cases. Although it might be argued that the generality of less than proportionate avian individuals–area relationships suggests a single repeated cause, it may also imply the existence of a number of different processes operating at different levels and scales (Smallwood, 1999). There was no evidence that a constant number of individuals was surveyed across areas of different size at the assemblage and

at the intraspecific level. Nonetheless, observers may still choose unconsciously to survey birds where they are more likely to be present at high numbers and, conversely, some surveys may suffer from the tendency to increase the area of study in case of low rates of bird encounters. Both biases may be present, even though not as strongly as to cause a constant number of individuals to be surveyed across areas. While researchers might choose small survey sites in regions with high abundance, it is unlikely that they are actively choosing sites with proportionately lower abundances over large areas. This could lead to an observed relationship between abundance area that could be described by a constraint triangle, with a flat lower edge generated by the fixed minimum sample size and an increasing upper edge generated by the natural increase in abundance with scale. This prediction, however, is problematic, because the variance of abundances is likely to increase with the absolute value of abundances anyway (e.g. Jimenez, 2000). Moreover, there was little qualitative evidence for such a constraint triangle to fit the observed scatterplot of abundances as a function of plot area. Evidence for a survey efficiency mechanism for individuals– area relationships was very limited. There was evidence only at the intraspecific level, and in only one of three different tests (limited number of visits). There was direct evidence at the whole assemblage and interspecific levels and indirect evidence at the intraspecific level for a habitat heterogeneity mechanism determining less than proportionate individuals–area relationships. These results fit with Nee and Cotgreave’s (2002) suggestion that the intraspecific individuals– area relationship is linked with the species–area relationship, and that because increases in species numbers with increasing area are generated by increasing habitat heterogeneity then the abundance of individual species also increases less than proportionately. As they found, the avian abundance data analysed here also revealed no significant difference between the observed slopes of intraspecific individuals–area relationships and those predicted from the associated species–area relationships. Unfortunately, however, this was a rather weak test, because the compounding of the standard errors of two slopes (for the species– area and assemblage individuals–area relationships) resulted in very large confidence intervals for the slopes of the predicted intraspecific individuals–area relationships (Fig. 5d). Even were this not so, it is also questionable how informative the test actually was. Both for Nee and Cotgreave’s (2002) data and for those analysed here, there were strong less than proportionate assemblage individuals–area relationships. It is difficult to provide a biological explanation for why this should be so. If land units of different area were sampled randomly across a region, then there should be a proportionate relationship between assemblage number of individuals and survey area. Less than proportionate assemblage individuals–area relationships are thus artefactual, because in order for them to arise as a consequence of habitat heterogeneity, smaller areas have to be biased towards homogeneous habitat. Finally, there was no evidence in the data analysed here that interspecific individuals–area relationships are generated by simply surveying larger-bodied species over larger areas. Although

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

313

M. Pautasso and K. J. Gaston Table 2 Summary of the evidence for the mechanisms explaining GIARs at the assemblage, interspecific and intraspecific levels Mechanism Data Level Assemblage Interspecific Intraspecific

Plot area choice

Survey efficiency (a, b, c)

Habitat heterogeneity

All habitats, all survey methods

See footnote

Forests vs. mixed habitat, all survey methods

None None None

None, none, none None, none, none None, none, strong

Strong Strong Some

(a) Forest vs. open habitats (all survey methods) (b) counts vs. mapping (forests) (c) less than five visits vs. at least five visits (forests, all survey methods).

body size explained some of the variation in species abundance, additional variation was explained by survey area (cf. Johnson, 1999). This was not a particular surprise in the present case because abundances were derived from studies of entire avian assemblages, and thus there was limited opportunity to alter survey area in response to the body sizes of individual constituent species. A different outcome might be achieved by collating densities from studies focused on individual species or small sets thereof. The failure to find unequivocal support for any single mechanism determining the form of generalized individuals–area relationships (Table 2) suggests three things. First, while pervasive, such effects may arise for different reasons in different circumstances at different levels, undermining attempts to pick mechanisms apart. Secondly, a posteriori it is difficult to separate mechanisms, which would be a concern for future employment of existing abundance data. Thirdly, additional processes may be at work that cannot be explored readily. For example, other population processes (metapopulation spatial dynamics, philopatry) can contribute to spatial aggregation of assemblages and individual species which, allied with biased sampling, could give rise to individuals–area relationships akin to those predicted by the habitat heterogeneity hypothesis but by other routes (Storch et al., 2003). Consequences The abundance at which different populations of the same species, different species, and assemblages of taxonomically related species, occur often provides essential information for conservation and ecological purposes. Abundance estimates enable intraand interspecific temporal analyses of population expansion, contraction, and threat (e.g. Bell & Merton, 2002; Dunn, 2002). They are the basis for quantitative spatial studies of species distributions (e.g. Stoffels et al., 2003) and estimation of population size at a geographic scale (e.g. BirdLife International/European Bird Census Council, 2000), which in turn are central to species and site prioritization and conservation monitoring (e.g. Brown et al., 1995; Underhill & Gibbons, 2002). Abundance values across space are indeed a prerequisite for population viability analyses (e.g. Beissinger & Westphal, 1998), for the spatially explicit modelling of species’ responses to environmental conditions (e.g. Mackey & Lindenmayer, 2001) and for investigations of the scaling of population and ecosystem energy use (e.g. 314

Ernest et al., 2003). Variations in abundance may also provide evidence of the overall health of an ecosystem, and have been employed widely to assess the effects of degradation, defoliation, dieback and other disturbances (e.g. Ford & Bell, 1981; Rabenold et al., 1998; Eeva et al., 2002; Bel’skii & Lyakhov, 2003; but see Van Horne, 1983; Vickery et al., 1992; Bock & Jones, 2004). In the face of the diverse uses of abundance estimates, it is important to realize that these values cannot be divorced from the areas over which they were determined (Gaston & Matter, 2002; Mayor & Schaefer, 2005). Organisms may have not just one but several abundances, depending on how large and homogeneous an area is chosen in their estimation (Haila, 1988). This study confirms that argument. Just as for estimates of species richness, it is problematic expressing different estimates of number of individuals as abundance figures if these were obtained from areas of differing size. This conclusion casts a shadow on earlier avian monitoring work. Analyses at local, regional or biogeographical scales may need a reassessment if they tried to establish the ecological significance of some environmental or management factors with data obtained from plot areas varying markedly in size. The same argument can be made for studies trying to explain differences in the abundances of different species based on the traits they exhibit, where abundances were obtained from differently sized plots. How easily these problems can be overcome in order to enable meaningful use of published data rests on the extents to which assemblages in different environments, or different species, exhibit similar individuals–area relationships. There is evidence that the slopes of assemblage individuals–area relationships are significantly different in different habitats. The shallowest but still significant slopes of the assemblage individuals–area relationship are found in tundra, water and cropland (0.50, 0.58, 0.59). The steeper ones (and not significantly different from one) are found in tropical woodland, hot desert and pasture (0.94, 1.00, 1.10, respectively). The extent to which the individuals–area relationship differs with variations in environmental factors will be difficult to determine if plot sizes used in different environments, or in studying different kinds of species, are different. Avian abundances do not scale proportionately with the area over which they were estimated, mainly because of increasing habitat heterogeneity with increasing plot size. The number of bird individuals cannot be expected to average linearly across areas so that the estimate of the population abundance will not vary if the size of the sample unit is changed (Dungan et al.,

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances 2002). We thus suggest three main recommendations. (1) Plot area should be declared when publishing abundances. This may allow subsequent collations of data to control for variations among studies (see also Mayer & Cameron, 2003). Furthermore, area variations among plots should be (2) minimized when planning survey work or (3) controlled for during data analyses. ACKNOWLEDGEMENTS We thank ornithologists from all over the world for their fieldwork. Thanks to M. Codesido, C. Delannoy, J.P. Duguay, K. Kujawa, J.-D. Lebreton, C. Moskát, J. Pearce-Higgins, H. Schmid, N.S. Sodhi, D. Storch, and J. Weiner for providing data, B. Beehler, J. Blondel, R.J. Dean, Y. Ezaki, E. Fernández-Juricic, D. Guthrie, A. Mack, E. Odell, M. Possingham, K. Remers Hanssen, T.S. Romdal and J.M. Thiollay for responding to queries, A. Beckerman, R. Davies, K. Evans, N. James, S. Ross and P. Warren for help with analyses and M. E. Dominguez Castro, A. Golovanova, S. Höschele, K. Maehara, I. Raminhos, P. Schneider and L. Scobczyk for help with foreign languages. Thanks to K. Evans, R. Freckleton, A. Holt, J. Kouki, S. Leach, S. Smallwood and D. Storch for comments on the manuscript. MP thanks his parents for financial support. REFERENCES Ambuel, B. & Temple, S.A. (1983) Area-dependent changes in the bird communities and vegetation of southern Wisconsin forests. Ecology, 64, 1057–1068. Baker, J., French, K. & Whelan, R.J. (2002) The edge effect and ecotonal species: bird communities across a natural edge in Southeastern Australia. Ecology, 83, 3048–3059. Beissinger, S.R. & Westphal, M.I. (1998) On the use of demographic models of population viability in endangered species management. Journal of Wildlife Management, 62, 821– 841. Bel’skii, E.A. & Lyakhov, A.G. (2003) Response of the avifauna to technogenic environmental pollution in the southern taiga zone of the Middle Urals. Russian Journal of Ecology, 34, 181– 187. Bell, B.D. & Merton, D.V. (2002) Critically endangered bird populations and their management. Conserving bird biodiversity. General principles and their applications (ed. by K. Norris and D.J. Pain), pp. 105–138. Cambridge University Press, Cambridge. Bellamy, P.E., Rothery, P., Hinsley, S.A. & Newton, I. (2000) Variation in the relationship between numbers of breeding pairs and woodland area for passerines in fragmented habitat. Ecography, 23, 130–138. Berthold, P. (1976) Methoden der Bestandeserfassung in der Ornithologie: Übersicht und kritische Betrachtung. Journal für Ornithologie, 117, 1–69. Best, L.B. (1975) Interpretational errors in the ‘mapping method’ as a census technique. The Auk, 92, 452–460. Bibby, C.J., Burgess, N.D., Hill, D.A. & Mustoe, S.H. (2000) Bird census techniques, 2nd edn. Academic Press, London.

BirdLife International/European Bird Census Council (2000) European bird populations: estimates and trends. BirdLife International, Cambridge. Blackburn, T.M. & Gaston, K.J. (1996) Abundance–body size relationships: the area you census tells you more. Oikos, 75, 303–309. Blackburn, T.M. & Gaston, K.J. (1999) Density, survey area, and the perfection (or otherwise) of ecologists. Oikos, 85, 570–573. Blackburn, T.M., Gaston, K.J., Quinn, R.M. & Gregory, R.D. (1999) Do local abundances of British birds change with proximity to range edge? Journal of Biogeography, 26, 493–505. Blondel, J. (1969) Méthodes de dénombrements des populations d’oiseaux. Problèmes d’écologie: l’echantillonage des peuplements animaux des milieux terrestres (ed. by M. Lamotte and F. Bourlière), pp. 97–151. Masson, Paris. Bock, C.E. & Jones, Z.F. (2004) Avian habitat evaluation: should counting birds count? Frontiers in Ecology and Environment, 2, 403–410. Borchers, D.L., Buckland, S.T. & Zucchini, W. (2002) Estimating animal abundances. Springer, London. Brett, M.T. (2004) When is a correlation between non-independent variables ‘spurious’? Oikos, 105, 647–656. Brotons, L., Mönkkönen, M. & Martin, J.L. (2003) Are fragments islands? Landscape context and density–area relationships in boreal forest birds. American Naturalist, 162, 343–357. Brown, J.H., Mehlman, D.W. & Stevens, G.C. (1995) Spatial variation in abundance. Ecology, 76, 2028–2043. Cieslak, M. & Dombrowski, A. (1993) The effect of forest size on breeding bird communities. Acta Ornithologica, 27, 97–111. Cramp, S., Simmons, K.E.L. & Perrins, C.M. (1978–2000) Handbook of the birds of Europe, the Middle East, and North Africa: the birds of the Western Palearctic, vols ii–ix. Oxford University Press, Oxford. Crozier, G.E. & Niemi, G.J. (2003) Using patch and landscape variables to model bird abundance in a naturally heterogeneous landscape. Canadian Journal of Zoology, 81, 441–452. Del Hoyo, J., Elliott, A. & Sargatal, J. (eds) (1992–2003) Handbooks of the birds of the world, vol. I. (Ostrich to Ducks) to vol. VIII (Broadbills to Tapaculos). Lynx Editions, Barcelona. Dungan, J.L., Perry, J.N., Dale, M.R.T., Legendre, P., CitronPousty, S., Fortin, M.J., Jakomulska, A., Miriti, M. & Rosenberg, M.S. (2002) A balanced view of scale in spatial statistical analysis. Ecography, 25, 626–640. Dunn, E.H. (2002) Using decline in bird populations to identify needs for conservation action. Conservation Biology, 16, 1632– 1637. Eeva, T., Koivunen, V. & Hakkarainen, H. (2002) Population densities of forest birds in a heavy metal pollution gradient. Avian Science, 2, 227–236. Emlen, J.T. (1971) Population densities of birds derived from transect counts. The Auk, 88, 323–342. Engstrom, R.T. & James, F.C. (1981) Plot size as a factor in winter bird-population studies. Condor, 83, 34–41. Ernest, S.K.M., Enquist, B.J., Brown, J.H., Charnov, E.L., Gillooly, J.F., Savage, V., White, E.P., Smith, F.A., Hadly, E.A., Haskell, J.P., Lyons, S.K., Maurer, B.A., Niklas, K.J. & Tiffney, B.

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

315

M. Pautasso and K. J. Gaston (2003) Thermodynamic and metabolic effects on the scaling of production and population energy use. Ecology Letters, 6, 990– 995. Estades, C.F. (2001) The effect of breeding-habitat patch size on bird population density. Landscape Ecology, 16, 161–173. Ford, H.A. & Bell, H. (1981) Density of birds in eucalypt woodland affected to varying degrees by dieback. Emu, 81, 202–208. Gaston, K.J., Blackburn, T.M. & Gregory, R.D. (1999) Does variation in census area confound density comparisons? Journal of Applied Ecology, 36, 191–204. Gaston, K.J. & Matter, S.F. (2002) Individuals–area relationships: comment. Ecology, 83, 288–293. Gregory, R.D., Gibbons, D.W. & Donald, P.F. (2004) Bird census and survey techniques. Bird ecology and conservation. A handbook of techniques (ed. by W.J. Sutherland, I. Newton and R.E. Green), pp. 17–55. Oxford University Press, Oxford. Haila, Y. (1988) Calculating and miscalculating density: the role of habitat geometry. Ornis Scandinavica, 19, 88–92. Imbeau, L., Drapeau, P. & Mönkkönen, M. (2003) Are forest birds categorised as edge species strictly associated with edges? Ecography, 26, 514 –520. James, F.C. & Rathbun, S. (1981) Rarefaction, relative abundance, and diversity of avian communities. The Auk, 98, 785–800. Järvinen, O. & Väisänen, R.A. (1975) Estimating relative densities of breeding birds by the line transect method. Oikos, 26, 316–322. Jimenez, J.E. (2000) Effect of sample size, plot size, and counting time on estimates of avian diversity and abundance in a Chilean rainforest. Journal of Field Ornithology, 71, 66 – 87. Johnson, C.N. (1999) Relationships between body size and population density of animals: the problem of the scaling of the study area in relation to body size. Oikos, 85, 565– 569. Kendeigh, S.C. (1944) Measurement of bird populations. Ecological Monographs, 14, 67–106. Klein Goldewijk, K. (2001) Estimating global land use change over the past 300 years: the HYDE database. Global and Biogeochemical Cycles, 15, 417–433. Kostrzewa, V.R. (1988) Die Dichte des Turmfalken (Falco tinnunculus) in Europa. Übersicht und kritische Betrachtung. Vogelwarte, 34, 216–224. Lawton, J.H. (1999) Are there general laws in ecology? Oikos, 84, 177–192. Legendre, P. (1993) Spatial autocorrelation — trouble or new paradigm. Ecology, 74, 1659–1673. Littell, R.C., Milliken, G.A., Stroup, W.W. & Wolfinger, R.D. (1996) SAS system for mixed models. SAS Institute Inc, Cary, NC. Mackey, B.G. & Lindenmayer, D.B. (2001) Towards a hierarchical framework for modelling the spatial distribution of animals. Journal of Biogeography, 28, 1147–1166. Mayer, A.L. & Cameron, G.N. (2003) Consideration of grain and extent in landscape studies of terrestrial vertebrate ecology. Landscape and Urban Planning, 65, 201–217. 316

Mayor, S.J. & Schaefer, J.A. (2005) The many faces of population density. Oecologia, in press. Nee, S. & Cotgreave, P. (2002) Does the species/area relationship account for the density/area relationship? Oikos, 99, 545– 551. Prairie, Y.T. & Bird, D.F. (1989) Some misconceptions about the spurious correlation problem in the ecological literature. Oecologia, 82, 285–288. Prince, M.E. & Silva, M. (2002) On the determinants of census area: implications for mammalian macroecological patterns. Global Ecology and Biogeography, 11, 401–409. Rabenold, K.N., Fauth, P.T., Goodner, B.W., Sadowski, J.A. & Parker, P.G. (1998) Response of avian communities to disturbance by an exotic insect in spruce-fir forests of the Southern Appalachians. Conservation Biology, 12, 177–189. Schaefer, J.A. & Mahoney, S.P. (2003) Spatial and temporal scaling of population density and animal movement: a power law approach. Ecoscience, 10, 496–501. Scherner, E.R. (1981) Die Flächengrösse als Fehlerquelle bei Brutvogel-Bestandsaufnahmen. Ökologie der Vögel, 3, 145– 175. Schonewald-Cox, C., Azari, R. & Blume, S. (1991) Scale, variable density, and conservation planning for mammalian carnivores. Conservation Biology, 5, 491–495. Shields, W.M. (1977) The effect of time of day on avian census results. The Auk, 94, 380–383. Sibley, C.G. & Monroe, B.L. (1990) Distribution and taxonomy of birds of the world. Yale University Press, New Haven. Silva, M., Brimacombe, M. & Downing, J.A. (2001) Effects of body mass, climate, geography, and census area on population density of terrestrial mammals. Global Ecology and Biogeography, 10, 469–485. Smallwood, K.S. (1995) Scaling Swainson’s hawk population density for assessing habitat-use across an agricultural landscape. Journal of Raptor Research, 29, 172–178. Smallwood, K.S. (1997) Interpreting puma (Puma concolor) density estimates for theory and management. Environmental Conservation, 24, 283–289. Smallwood, K.S. (1998) Patterns of black bear abundance. Transactions of the Western Section of the Wildlife Society, 34, 32–38. Smallwood, K.S. (1999) Scale domains of abundance amongst species of mammalian Carnivora. Environmental Conservation, 26, 102–111. Smallwood, K.S. (2001) The allometry of density within the space used by populations of Mammalian Carnivores. Canadian Journal of Zoology, 79, 1634–1640. Smallwood, K.S., Jones, G. & Schonewald, C. (1996) Spatial scaling of allometry for terrestrial, mammalian carnivores. Oecologia, 107, 588–594. Smallwood, K.S. & Schonewald, C. (1996) Scaling population density and spatial pattern for terrestrial, mammalian carnivores. Oecologia, 105, 329–335. Smallwood, K.S. & Schonewald, C. (1998) Study design and interpretation of mammalian carnivore density estimates. Oecologia, 113, 474–491.

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

The plot area effect on avian abundances Solonen, T. (1996) Patterns of variations in the structure of forest bird communities in southern Finland. Ornis Fennica, 73, 12– 26. Stoffels, R.J., Closs, G.P. & Burns, C.W. (2003) Multiple scales and the relationship between density and spatial aggregation in littoral zone communities. Oikos, 103, 81–92. Storch, D. & Kotecky, V. (1999) Structure of bird communities in the Czech Republic: the effect of area, census technique and habitat type. Folia Zoologica, 48, 265–277. Storch, D., Sizling, A.L. & Gaston, K.J. (2003) Geometry of the species–area relationship in central European birds: testing the mechanism. Journal of Animal Ecology, 72, 509–519. Underhill, L. & Gibbons, D. (2002) Mapping and monitoring bird populations. Conserving bird biodiversity. General principles and their applications (ed. by K. Norris and D.J. Pain), pp. 34–60. Cambridge University Press, Cambridge. Van Horne, B. (1983) Density as a misleading indicator of habitat quality. Journal of Wildlife Management, 47, 893–901. Vickery, P.D., Hunter, M.L. & Wells, J.V. (1992) Is density an indicator of breeding success? The Auk, 109, 706 –710.

Village, A. (1984) Problems in estimating kestrel breeding density. Bird Study, 31, 121–125. Watts, D. (1971) Principles of biogeography. McGraw-Hill, London. Wiens, J.A. (1986) Spatial scale and temporal variation in studies of shrubsteppe birds. Community ecology (ed. by J. Diamond and T.J. Case), pp. 154–172. Harper & Row, New York. BIOSKETCHES Marco Pautasso undertook a PhD in macroecology with Kevin Gaston and is also interested in forest and landscape pathology. Kevin J. Gaston has research interests in the fields of biodiversity, conservation biology and macroecology, with particular emphasis on the ecologies of rare organisms, the structure of geographical ranges and patterns in species richness. Editor: David Currie

Global Ecology and Biogeography, 15, 303–317 © 2006 Blackwell Publishing Ltd

317