Experienced observers and scientists aver that there have been significant reductions in the abundance of birds breeding in North America across the last half-century (Robbins et al. 1989, King and Rappole 2003, National Audubon Society 2004), perhaps owing, among other factors, to the influence of global atmospheric changes (Sillett et al. 2000, Walther et al. 2002, Parmesan and Yohe 2003, Root et al. 2003, Valiela and Bowen 2003, Anders and Post 2006) or to the loss of North American breeding habitats or Neotropical wintering habitats (Andren and Angelstam 1988, Temple and Cary 1988, Terborgh 1989, Böhning-Gaese et al. 1993, Brook et al. 2003, Johnson et al. 2005, Stratford and Robinson 2005). The loss of Neotropical wintering habitat has been given special attention because of the many reports, both in the scientific literature and in the popular press, of continuing alteration of tropical landscapes (Melillo et al. 1985, Woodwell et al. 1987, Sader and Joyce 1988, Houghton et al. 2000, Brook et al 2003, Hirsch et al. 2004, Marris 2005).

In this article we use the remarkably data-rich long-term (1966–present) record of breeding birds in the North American Breeding Bird Survey (BBS) to pursue three lines of inquiry: (1) to ascertain whether the abundance of eastern and central North American breeding birds indeed decreased during recent decades; (2) to see whether the abundance of birds that breed in North America was affected by the loss of habitat in the locations where the birds spend their winters; and (3) to determine whether changes in the abundance of North American breeders differed among birds that used different habitats.

The BBS data have been collected annually since 1966 in surveys staffed by skilled observers surveying set routes during the breeding season to record the abundance and identity of breeding birds. The data sets are massive: There are about 3700 routes defined in the BBS, and, on average, about 2900 of them are surveyed annually in the states and provinces of the United States and Canada. The routes were located so as to sample habitats that are representative of the regions. The survey of each 40-kilometer (km) route consists of 50 stops at 0.8-km intervals. Data on the numbers and species of breeding birds are recorded at each stop for 3 minutes, and all observations are added for the route; then all the route data are added for the state or province, and eventually reported (USGS 2005).

The BBS counts, repeated across years (and, when feasible, along the same routes) by qualified personnel, are intended to assess the relative abundance of breeding birds rather than to yield a complete count of them. Statistical issues arising from BBS counts are reviewed (Robbins et al. 1989, Böhning-Gaese et al. 1993, Peterjohn et al. 1995), and BBS abundance data have been used, albeit with some criticism, in previous papers (Böhning-Gaese et al. 1993, Keitt and Stanley 1998, Anders and Post 2006). Survey data such as those of the BBS have been subjected to a variety of normalizing procedures aimed at reducing bias and statistical irregularities. As appropriate as such corrections and modifications may be, it seems nearly impossible to fully correct for all potential statistical flaws (there has even been concern about the effects of the age of observers [and the consequent loss of aural acuity], interference by changing traffic frequency, sampling species that could be detected near roads, and more such variables). Tools for assessing the influence of all potential biasing variables are simply not available.

Most statistical problems arise at the level of the specific count. We avoided such problems as much as possible by pooling all of the observations made in all of the provinces and states within the Atlantic and Mississippi flyways. Pooling such large numbers of observations diminishes the impact of the inherent sampling problems to some degree. In addition, we focused on interpreting broad decadal trends (increases, decreases, or no change over the entire record) rather than attempting to interpret shorter-term interannual changes. Moreover, all normalizing corrections add variation as propagated error and produce transformed variables that might be less transparent than the straightforward “birds per kilometer of transect.” Since the effects of corrections in many cases were modest or variable, and given the difficulty of fully normalizing the data, we eschewed transformations and simply used pooled counts, keeping the data real and readily accessible. Although the BBS data may have biases and limitations, the BBS's long-term span, remarkably large number of observations, and regional coverage make these records attractive for comparative examination of long-term changes.

Previous studies have analyzed BBS data within restricted time periods, along selected routes, or in separate regions, or have focused on a single bird species or subset of the species recorded. We include the entire BBS data set for all data from the Atlantic and Mississippi flyways (figure 1), including all years, all routes, and all species.We therefore focus on the overall longer decadal trends, and use the entire data set to depict the changes in the avifauna as comprehensively as possible. This approach minimizes the influences of short-term interannual trends and of differences among specific bird species that we knew were present in the record.

Figure 1.

Approximate bounds of the Mississippi and Atlantic migratory flyways over the Americas. The dashed horizontal lines indicate the latitudinal ranges into which we stratified the North American Breeding Bird Survey (BBS) data. The inset graph shows the abundance of breeding birds recorded in the BBS from 1966 to 2004. Linear regressions were fit to BBS records from states and provinces that provided breeding areas for birds that used the Mississsippi and Atlantic flyways.

Reduction in breeding bird abundance

Our central finding was that the total number of birds recorded as nesting in the eastern and central United States and Canada has steadily decreased since 1966 (figure 1 inset). The long-term trend in the number of nesting birds revealed decreases of about 12 percent and 18 percent in the Mississippi and Atlantic records, respectively, for the period 1966–2004. These losses took place at rates of 0.3 to 0.5 percent per year. The abundance of nesting birds was consistently higher in the Mississippi than in the Atlantic flyway, and there was a degree of year-to-year variation (figure 1).

We initially examined the possible association of changes in bird abundance with changes in mean temperature and precipitation for winter and summer over the eastern United States, using data from the NOAA Satellite and Information Service ( www.ncdc.noaa.gov/oa/climate/research/monitoring.html), for the years covered by BBS records. The year-to-year variation was probably a result of large-scale interannual climatic shifts (Sillett et al. 2000,Walther et al. 2002, Parmesan and Yohe 2003, Root et al. 2003, Valiela and Bowen 2003,Anders and Post 2006), although we found no evident correlation between (a) eastern North American temperatures and precipitation and (b) the year-to-year variations in bird abundance shown in figure 1. The interannual changes merit further study, but here we focus on the overall decadal-scale trends. For simplicity, and because of the relatively similar time courses, we pooled data for the two flyways. The decadal-scale losses of numbers of nesting birds that have taken place across large regions of North America over the more than 40 years of the BBS record pose a question: What are the possible causes?

The effect of overwintering range on the abundance of breeding birds

To discern whether the reduction in the numbers of birds that breed in eastern and central North America was associated with factors (particularly habitat losses) related to the latitudinal range where the birds either bred or spent the boreal winter, we partitioned the BBS data in terms of species that (a) were fully resident within the United States and Canada, or (b) migrated south during the boreal winter to more southern areas of the United States, to Mexico, to Mesoamerica and northern South America, to the Amazon basin and other areas east of the Andes, or to southern South America (figure 1). The idea here was that residents were exposed to conditions in the northern latitudes year-round, but migrants were exposed to the effects of spending part of the year in the United States and Canada as well as to the hazards of spending the boreal winter in the more southerly wintering range. If, as has been suggested, the loss of Neotropical habitats has been detrimental for Neotropical migrants, we might see different time courses of abundance among resident and migrant species, particularly those migrants that overwinter in the Neotropics, where extensive habitat loss has been reported. The species listed in the BBS records, and our classification of their migratory status, are included in table 1.

Table 1.

Species recorded in the North American Breeding Bird Survey, sorted by boreal winter area and habitat use.

Table 1.

(continued)

Table 1.

(continued)

Nesting bird abundance decreased most clearly and consistently in species that either resided (19 percent overall loss) or migrated within the United States and Canada (30 percent loss; figure 2). The slopes of the time trends for these two groups were similar (figure 2), suggesting that the effect of living in the United States and Canada part-time or year-round was about the same: Migrants did not incur relatively greater losses.

Figure 2.

Abundance of birds breeding in central and eastern North America that either were resident (black circles) or migrated farther south in the United States and Canada (white circles), to Mexico (black squares), to Mesoamerica and northern South America (white squares), to Amazonia (black triangles), or to southern South America (white triangles; cf. figure 1) during the boreal winter, 1966–2004. Numbers in parentheses on the right indicate the percentage of change, calculated as the difference between initial (1966) and final (2004) bird abundances using regression equations fitted to the entire data sets (for residents, F = 91.099, R² = 0.711, p < 0.001; for migrants within the United States and Canada, F = 208.166, p < 0.001, R² = 0.849; for migrants to Mexico, F = 1.837, p = 0.183, R² = 0.047; for migrants to Mesoamerica and northern South America, F = 115.026, p < 0.001, R² = 0.757; for migrants to Amazonia, F = 12.373, p < 0.001, R² = 0.251; and for migrants to southern South America, F = 16.146, p < 0.001, R² = 0.304). Where data scatter was such that R² was too low to be predictive (Prairie 1996), no numbers are included. Abbreviation: ns, not significant.

In contrast, the abundance of nesters that migrated farther south after the nesting season did not diminish. Species that migrated to Mexico, Amazonia, or southern South America showed no significant change, while migrants to Mesoamerica and northern South America increased by 20 percent across the later 20th century (figure 2). In all cases, there were multiyear increases and decreases over more than four decades, perhaps related to climatic shifts, but for present purposes the salient feature of figure 2 is the consistent loss of abundance of nesting species that spend their lives within the United States and Canada, compared with the increased abundance or lack of change in abundance of Neotropical migrants.

We have so far dealt with the total abundance of breeding birds. It seems useful to add species-specific information on abundance. To examine the time course of the many species recorded in the BBS, we fitted regressions of abundance for each species for each year, and used the calculated F_reg value as an approximate way to assess the significance of the changes in abundance (increases or decreases, noted by the sign of the slope of the regressions) between 1966 and 2004. Among US and Canada residents, 38 percent of the species increased significantly, while 21 percent decreased significantly in abundance (table 2). For Neotropical migrant species, 51 percent of the species increased significantly and 25 percent decreased significantly in abundance. It is not surprising, therefore, that previous reports show a variety of results, some claiming increases, some decreases, depending on the taxa included. Ratios of numbers of species that increased significantly relative to those that decreased significantly in abundance were somewhat higher for migrants (2.1) than for residents (1.8; table 2). We focus here on the contrasts among species that migrate different distances, but other traits may be important—for example, whether the species feed on other birds, mammals, invertebrates, leaves, fruits, or seeds—since climate change and habitat loss may differentially affect these groups of birds. Such features need further study.

Table 2.

Number of resident or migrant bird species with significant or nonsignificant increases or decreases in abundance in the breeding areas between 1966 and 2004.

The abundance of specific species in the BBS record varied: Almost half the species increased significantly in numbers, and less than a quarter decreased significantly (table 2), during the last half of the past century. There were substantial differences in the time trends among species with different migratory behavior. Of the species that reside year-round or migrate within the eastern United States and Canada, 38 percent increased significantly in abundance (including the mourning dove, house finch, Carolina wren, and Canada goose, among other prominent species), and only 21 percent decreased in abundance (the common grackle, red-winged blackbird, eastern meadowlark, house sparrow, and field sparrow, among others). Thus, increases in abundance were about twice as common as decreases for these species of birds.

The trends for US and Canadian species that migrate to more southern latitudes contrast with the trends for those that do not do so (table 2). More than half the species that migrate south of the United States and Canada increased significantly in numbers, and a quarter of the species decreased significantly in abundance (table 2). Surprisingly, therefore, conditions between 1966 and 2004 for Neotropical migrant species seemed somewhat better than, or at least not as problematic as, conditions for species resident in the United States and Canada.

In spite of the significant reductions in total abundance (figure 1), less than a quarter of the individual species in the BBS significantly decreased in abundance (table 2). This suggests that the considerable reductions in abundance seen in figure 1 (inset) may have largely resulted from losses incurred by a handful of relatively common species.

Effect of habitat preference on changes in the abundance of breeding birds

To determine whether reductions in abundance were disproportionate among species that preferred different terrestrial (open, edge, forest) or aquatic (open-water, coastal, wetland) habitats, we sorted BBS data accordingly (table 1). Sorting species of birds into a few simple habitat preference categories and migratory ranges involved some arbitrary assignments. Many species use a variety of habitats; in such cases, we assigned the species to the principal habitat used, on the basis of facts provided in Poole and Gill's (2006) excellent and comprehensive review of North American avifauna, supplemented by information from several other bird guidebooks and confirmed by personal experience.

Terrestrial species

There were notable differences in the decadal trends of abundances of resident and migrant bird species that used different terrestrial habitats (figure 3). Terrestrial species that were resident or migrated within the United States and Canada and preferred open habitats suffered a 50 percent decrease in abundance (figure 3). This group included species such as the eastern meadowlark, rock pigeon, common nighthawk, gray partridge, and greater prairie chicken. There were also considerable reductions (13 percent) in the abundance of species that preferred edge habitats (figure 3). Among the affected species were the blue jay, yellow-shafted flicker, and European starling. In striking contrast, the abundance of the few resident species that preferred forests—such as the brown-headed nuthatch, pine warbler, and pileated woodpecker—increased by about 63 percent (figure 3).

Figure 3.

Abundance of breeding birds in the North American Breeding Bird Survey (BBS) data, stratified into two classifications: first, species that were resident or migrated within the United States (black circles) or migrants that spent the boreal winter farther south (white circles); second, species that used open (top left), edge (center left), or forest (bottom left) terrestrial habitats, or used coastal (top right), open water (center right), or wetland (bottom right) habitats. Numbers in parentheses show the percentage change between the start and end of the time course, calculated by comparing the start and end values from regressions fitted to the data. Where data scatter was such that R² was too low to be predictive (Prairie 1996), no numbers are included. Abbreviation: ns, not significant.

Such shifts in abundance in resident breeding birds may be linked to the substantial land-cover changes (Goss-Custard et al. 1994) taking place in North America during the period covered by the BBS. The expansion of urban sprawl at the expense of agricultural area, for example, may be associated with loss of birds of open habitats. We use “urban sprawl” here to describe a pattern of landscape cover that is not truly urbanization, but rather a complex mosaic of suburban, industrial, and other human-affected land covers that is proliferating across most of the developed world, and North America in particular. Detailed surveys of land-cover changes across the second half of the 20th century (figure 4) demonstrate that urban sprawl has recently expanded throughout eastern North America, at the expense of open (mainly agricultural) land covers and, to a more variable extent, of forest area (figure 4, top). Moreover, the extent of urban sprawl continues to increase (figure 4, bottom). The extent of this transition is impressive: About 14 and 32 percent, respectively, of the area of the states that fall within the Mississippi and Atlantic flyways falls into a land-use classification that has been defined as the “wildland–urban interface,” a land cover consisting of a reticulate mosaic of remnant natural and new residential parcels (Radeloff et al. 2005).

Figure 4.

Top: Frequency distributions of gains and losses of area covered by agricultural, forest, or urbanized land cover, 1982–1992, in 16 watersheds of the eastern United States. Bottom: Number of watersheds with different percentages of urbanized land cover within 51 eastern US watersheds. Each bar is divided into watersheds where the population within urbanized areas was increasing, decreasing, or not changing across the previous decade. Modified from Valiela (2006); data are from Van Breemen and colleagues (2002) and Dow and DeWalle (2000).

The shift to more wildland–urban interface should, however, have furnished considerably more edge habitat, but the BBS records show significantly decreased rather than increased abundance of North American edge-loving birds (figure 3). This suggests that the increased edge habitat undoubtedly provided by the remarkable proliferation of wild-land–urban interface is somehow inadequate as bird habitat, perhaps owing to overfragmentation or poor quality for bird use.

The increased abundance of resident forest species (figure 3) may be related to expansion of area of northern forests during much of the 20th century. The changes in forest area vary geographically (hence the wide scatter of the losses and gains of forest area seen in figure 4). Although we record mostly losses of forest area, forest cover increased in, for example, New England during much of the 20th century, as felled areas were reforested (Foster and Aber 2004). Such increases may have tapered off during the last 20 years or so (figure 3), as recovered forest area has been newly altered to “wild-land–urban” land covers.

The history of the abundance of terrestrial migrant species that spend the boreal winter south of the United States differed from those of bird species that spend their life within the United States and Canada; in addition, there were clear differences among migrants that prefer open, edge, or forest habitats (figure 3). There was about a 22 percent loss of abundance of migrant open-habitat species between 1966 and 2004 (figure 3). Because of the widespread conversion of Neotropical forest to pastures (Melillo et al. 1985, Woodwell et al. 1987, Sader and Joyce 1988, Houghton et al. 2000, Brook et al. 2003, Hirsch et al. 2004, Marris 2005), with the provision of extensive open areas, we conjecture that the decrease in migrant open-habitat birds might not be related to the loss of Neotropical habitat area, but rather may be a measure of the spreading footprint of loss of suitable open breeding habitats in the United States and Canada, an impact that may be spreading to affect the bird fauna of the entire Americas. We note that this seems to have occurred in the group of species that showed the greatest loss within the breeding areas (open-habitat birds), so it seems reasonable to find that this is the group that may show the first signs of the incipient global effects of the loss of North American breeding habitat.

In contrast to the trends for migrant open-habitat birds, there were no obvious changes in migrant edge species across the 38 years of BBS data (figure 3). For species that use edge habitats, the footprint of the loss or impoverishment of northern breeding habitats still seems to be constrained to their northern place of origin.

In further contrast to the time courses of migrants that prefer open and edge habitats, there was a clear increase in the numbers of forest-loving migrants, amounting to about 37 percent (figure 3). This increase suggests that the well-documented loss of forest in the Neotropics (Melillo et al. 1985, Woodwell et al. 1987, Sader and Joyce 1988, Houghton et al. 2000, Brook et al. 2003, Hirsch et al. 2004, Marris 2005) has not been paralleled by reductions in the total abundance of migrant forest species that nest in the United States and Canada. The increases in both the resident and the migrant forest species (figure 3) point out another instance of a possible spreading ecological footprint. The expansion of North American forest habitat area in the earlier 20th century might have altered—in this case positively—the abundance of the bird fauna of the Americas. It is unclear, however, how long this increase might continue, since in the breeding areas there is much conversion of forest to urban cover, and in the Neotropics the prediction is that in coming years agricultural expansion will consume about 40 percent of the Amazon forests (Silveira Soares-Filho et al. 2005). There is no evidence as yet (figure 3) that the reported losses of Neotropical forest area can be associated with lower total abundance of birds migrating to Neotropical forest habitats during the boreal winter. We do not know at what point migrant forest species might, as a group, begin to show negative effects from the continuing loss of suitable habitats.