Biodiversity rate of change data.

We used the rates of change in the species richness and assemblage dissimilarity of North American breeding birds across 30 years (1990–2019) obtained by [42]. We obtained data from the North American Breeding Bird Survey (BBS, [47]. We selected BBS routes in the conterminous United States sampled once a year in the period 1990–2019 that were sampled for more than 24 years, eliminating those that were not sampled in two or more consecutive years. Then, we retained only one BBS route per 10 × 10 km map grid cell by randomly eliminating one of two sampling units closer than 15 km (‘spThin’ R package, [48]. As in [42], this analysis included a total of 571 BBS routes: 309 on the Eastern Temperate Forest, 95 on the Great Plains, 61 on the Northern Forests, 56 on the Northwestern Forested Mountains, and 50 on the North American Deserts. Fig 1 in [42] illustrates the geographic distribution of the 571 BBS routes included in the study. We used the same exclusion and retention criteria as in [42] including in the analysis a total of 245 bird species (Table S1 in [42]). We converted the species abundances reported in the BBS to presence-absence data. We calculated the rates of change of species richness, total assemblage dissimilarity and its components as in [42].

Climatic data.

The climatic data were obtained from the Parameter-elevation Regressions on Independent Slopes Model (PRISM) datasets for the years 1990–2019 (PRISM Climate Group 2014; [49]. For this study, we selected climatic variables known to limit the distribution of birds physiologically [50–52]: mean precipitation (mm) minimum temperature(°C), and maximum temperature (°C). Each year, we calculated the mean values of these variables for the breeding season (i.e., May 1st to July 31st) using the methodology described in Supporting Information (S1 File).

LULC data.

To calculate the rate of LULC change between 1990 and 2019, we utilized the “Modeled Historical Land Use and Land Cover for the Conterminous US: 1938-1992” [53] and the scenario A1B of the “Conterminous United States Land Cover Projections - 1992 to 2100” [54] data sets. A detailed description of these data sets and the rationale for the selection of scenario A1B is provided in the Supporting Information (S1 File). We grouped the annual layers and retained for the analysis six combined layers containing the proportion of LULC per map cell for: barren land, crop/pasture, forest (deciduous forest + evergreen forest + mixed forest), grassland/shrubland, urban, and wetland (herbaceous wetland + woody wetland). S1 Fig in S1 File shows the methodology we used to obtain the grouped annual LULC layers.

Elevation data.

We included elevation in the analysis to account for its potential effect on biodiversity gradients. We used the North American Elevation map at 1 km resolution [55] to extract the elevation of each bird assemblage. Elevation information is given in meters, with values smaller than/ at/ above zero indicating below/ at/ above sea level, respectively.

To work with all data sets in the same spatial projection, we re-projected the environmental and biodiversity data to the Lambert Azimuthal Equal Area projection. We joined the annual climatic and land conversion layers and extracted the environmental information for each bird assemblage geographic coordinates.

Predictive power of dissimilarity models and species richness models across geographical scales.

We present values of the amount of variance in the biodiversity metrics explained by the environmental predictors (R²) for each model and resolution in Table 3. There was some variability in predictive power of the different models.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted R² across buffer resolutions for all GAMMs.

https://doi.org/10.1371/journal.pone.0330153.t003

The predictive power of the models of species richness was consistently lower than that of the models for total assemblage dissimilarity (Table 3). Further, more environmental predictors were significantly associated with the rate of change of total assemblage dissimilarity than of species richness, with the exception of the model of species richness in the Northwestern Forested Mountains (Tables 1 and 2). In comparison to the aforementioned models, the predictive power of models for total assemblage dissimilarity in the Eastern Temperate Forest, Great Plains, and North American Deserts, and of models for species replacement and species loss, was lower (Table 3).

The results of the sensitivity analysis showed that the amount of variance in the four biodiversity metrics studied that was explained by the environmental predictors varied with the geographical scale of the analysis (Table 3). For instance, while the variance explained of total assemblage dissimilarity for the Northwestern Forested Mountains was 27% when the scale of the predictors was tabulated for 25 km sampling areas, the variance explained was 38% for 12.5 km and 47% for 50 km sampling areas.

Increase of within-ecoregion assemblage dissimilarity (biotic heterogenization).

High latitude and high elevation places showed strongest relationships between biodiversity metrics and environmental drivers. Rapid within-ecoregion heterogenization of avian assemblages was associated with fast changes in climate and LULC in ecoregions at high latitudes or elevations such as the Northern Forest and the Northwestern Forested Mountains. These results point to the susceptibility of species assemblages at higher latitudes and elevations to environmental change also documented for other taxa [2,4,24,38,60,61]. Species at the edges of their thermal or habitat tolerances – due to species-specific physiological limits and breadth of ecological niches – will be particularly vulnerable to the changes we documented here.

For the two ecoregions with strongest relationships between the rates of change in biodiversity metrics and the environment, there were different environmental drivers and different outcomes (i.e., more species loss versus more species replacement). Below, we explore these two ecoregions further. The rate of within-ecoregion heterogenization of avian assemblages was impacted differentially by environmental variables in the Northern Forest and the Northwestern Forested Mountains. In general, the biological mechanism through which the biodiversity of bird assemblages was changing in response to environmental changes was also different between the two ecoregions, with species assemblages becoming subsets of richer assemblages in the Northern Forest and species being replaced in the Northwestern Forested Mountains.

Changes in temperature were associated with increased heterogenization in avian assemblages in these two ecoregions. Rapid changes in maximum temperature (Northern Forest) and minimum temperature (Northwestern Forested Mountains) were associated with rapid heterogenization of avian assemblages. Warming temperatures may have resulted in rapid increase in abundance and range areas of warm-dwelling species while cool-dwelling species decreased [62,63]. The heterogenizing effect of warming temperatures has been reported under current [24] and future [64] climate warming, in line with our results. These results likely indicate that some species were tracking climate change more rapidly or accurately than other species, as others have found [65,66]. The rate of biotic heterogenization may, however, not be high enough to keep the pace of the warming climate [67], since the stronger the climate warming, the more species’ responses may lag behind at range limits [68].

In the Northwestern Forested Mountains (but not the Northern Forest), increased water stress (i.e., declining precipitation) was associated with increased biotic heterogenization of avian assemblages. In the Northwestern Forested Mountains, rapid biotic heterogenization was also associated with rapidly drying locations. Increased water stress has been associated with range shifts to higher elevations, particularly when combined with changes in temperature [69]. Thus, species intolerant to low precipitation levels going rapidly locally extinct and species tolerant to low precipitation levels replacing them appear to be the mechanism behind the fast heterogenization of montane species assemblages undergoing decreases in precipitation. Decreases in precipitation have also been linked to population declines in European birds in mountainous regions due to phenological mismatches between resource availability and breeding season [70].

Rapid forest losses were generally associated with an accelerated rate of biotic heterogenization in the Northern Forests and the Northwestern Forested Mountains. However, the mechanisms behind this pattern were potentially different for each ecoregion. Rapid biotic heterogenization in the Northern Forests was likely driven by high rates of species loss (for example, loss of forest specialist species), while in Northwestern Forested Mountains, this process was possibly linked to high rates of species replacement (for example, replacement of forest specialists by species more tolerant to disturbed habitats). In the Northwestern Forested Mountains, the rate of change in total assemblage dissimilarity lagged behind the rate of change of forest loss, indicating that either extinction debt or buffering mechanisms may be operating here [4,71]. Although the consequences of forest loss on vertebrate biodiversity may be severe not only for already fragmented landscapes but also for relatively intact forested areas [72,73], the lagged response we found in the Northwestern Forested Mountains could be an indication of species assemblage resilience if critical habitat loss thresholds are not surpassed [14].

Decrease of within-ecoregion assemblage dissimilarity (biotic homogenization).

Here, we provide some context on how homogenization in this study might compare to homogenization in other studies. Because we are examining homogenization inside of ecoregions, it is not necessarily the case that the most generalized species will become the most common, as the ecoregion becomes more homogenized. It is a question of the scale of the analysis. For example, it could be that species typical of the ecoregion become more common within the ecoregion as homogenization progresses; these more common species might not be viewed as the most generalized, when comparing them to other species at a larger spatial scale. This is a nuance but is a result that is different from how other authors working at larger scales have conceptualized homogenization [4,61,74].

Rapid within-ecoregion homogenization of avian assemblages was mainly associated with fast changes in LULC. An exception was found in the Northwestern forested mountains, where fast increases in precipitation were associated with fast homogenization. Rapid increases in precipitation appear to be favoring ecoregion-specific species and slowing down their replacement by species tolerant of low precipitation levels, similar to results found by [62].

For Northern Forests and Northwestern Forested Mountains, the rapid change in LULC types less common than forest impacted the rate of change of total assemblage dissimilarity. For example, rapid increase of grassland/shrubland and wetland (within the Northern Forests) and rapid decreases in crop/pasture (within Northwestern Forested Mountains) were associated with accelerated rates of within-ecoregion biotic homogenization through the slowdown of species loss rates (Northern Forests) or declines in the species replacement rates (Northwestern Forested Mountains). Thus, rapid increases of natural LULC and declines of anthropogenic LULC appeared to be favoring habitat specialist species and slowing down their loss or replacement by disturbance-tolerant, widespread, generalist species. This finding highlights the importance of the heterogeneity of landscapes in maintaining biodiversity and providing resilience and stability of ecological processes against environmental change [75].

In the Eastern Temperate Forest and the Great Plains, the higher the elevation, the more similar the species assemblages within ecoregions. This occurred either because the species were less often replaced by others (Eastern Temperate Forests) or because they did not become a subset of richer species assemblages as fast as the species assemblages at lower elevation (Great Plains). These patterns may emerge from some species being restricted to certain areas and not colonizing whole gradients, likely due to environmental filtering linked to temperature and resource availability [76,77].

Effect of the rate of change of the environment on the components of assemblage dissimilarity.

The rate of change in the components of total dissimilarity was mainly related to changes in LULC, except for the Eastern Temperate Forest where precipitation was the most important predictor. In general, we found that the change in environmental predictors had opposite effects on the rates of change of species replacement and species loss in the Eastern Temperate Forest, Great Plains and North American Deserts. Opposite and/or independent responses of species replacement and species loss to predictor variables have been found for multiple taxa along elevational and latitudinal gradients [76,78,79], in agreement with what we found (S1-S3 Tables in S1 File). The pattern of decreased species replacement and increased species loss we identified in Eastern Temperate Forests and Great Plains is generally attributed to the presence of disturbance-tolerant or widespread species, and the local extinction of specialist or narrowly distributed species [80]. The pattern of increasing replacement rates and decreasing species loss rates with elevation found in North American Deserts is consistent with the upward range shifts of low-elevation species tracking their thermal niches [81].

Even small changes in the proportion of LULC can have large effects in relatively species-poor ecoregions. In North American Deserts, fast increases in barren land were associated with low rates of species loss, since likely the number of species present there was low to begin with due to the lack of vegetation characteristic of barren lands. Forest cover in this ecoregion is minimal, and its proportion is not changing significantly with time. Our models, however, detected that declines in forests in this ecoregion were associated with increased species turnover. Losing the habitat and resources that trees provide, the replacement of existing species by others better adapted to harsher environments is likely. In the Northern Forests, rapid increases in urban LULC (even if small in quantity) were related to lower rates of species replacement. In this case, we argue that, with the rapid habitat degradation that accompanies growing cities, there were fewer species to replace.

Effect of the rate of change of the environment on species richness.

We found significant relationships between species richness and the environmental variables for the Eastern Temperate Forest, the Northern Forests, and Northwestern Forested Mountains. In addition, the effects of climatic and LULC predictors on the rate of change of avian assemblage species richness, when significant, differed across ecoregions (similar to findings by [82]. We explore these relationships further, below.

Factors driving species richness were mainly changes in LULC. The only exception to this result was found in the Eastern Temperate Forests. There, fast increases in richness were associated with rapidly warming locations. This result is consistent with the northward range expansion of southerly-distributed, warm-affiliated species reported for eastern North America [62,83,84]. In contrast, other ecoregions saw increase in species richness driven by changes in LULC. In the Northern Forests, rapid increases in richness correlated with rapid increases in forests and urban LULC. An increase in richness with an increase in a particular LULC occurs when this change favors a particular group of species and immigration events exceed local extinction events [85]. Increases in forests may favor forest specialist birds with the increase in potential niches and habitat complexity. In contrast, rapid increases in urban LULC (even if small in quantity) may favor the increase of generalist, urban- or disturbance-tolerant species [86,87] in the avian assemblages of the Northern Forests. Decreases in crop/pasture were associated with fast increases in richness in the Northwestern Forested Mountains. The decrease in agricultural lands can potentially result in the increase of natural vegetation and the reduction of the effects of pesticide/fertilizer exposure, improving the quality of habitat for less disturbance-tolerant, late‐successional, and insectivore species [88–90].

Assemblage dissimilarity is better predicted by climate and/ or LULC than is species richness.

Our models performed reasonably well at predicting total assemblage dissimilarity for the northern and latitudinally more restricted ecoregions (Northern Forest and Northwestern Forested Mountains), though the better models were at the 25 km (Northern Forest) and 50 km (Northwestern Forested Mountains) scales. Models of the rate of change of species richness did not detect a significant relationship with the rate of change in the environment for the avian assemblages in the Great Plains and North American Deserts. In contrast, models of the rate of change of the three assemblage dissimilarity metrics detected significant relationship with at least one environmental predictor for all ecoregions. However, for some ecoregions, the models of the rates of change of the component of assemblage dissimilarity performed poorly since they explained a small percentage or none of the variance in the response variable. This may be an indication that the variation in the data was very small or that it would be explained by unmeasured variables (for example, [88], and the results for these models should be interpreted with caution. Interestingly, the variance in the response explained by the predictor variables was higher for the rate of change of total assemblage dissimilarity than for the rate of change of species richness. This is an indication that assemblage dissimilarity is more sensitive to changes in the environment than species richness, as other studies have also demonstrated [91,92]. Eventually, if the current rates of environmental change do not slow down, the effects of these changes will be reflected in the rate of change of species richness as well.

Interactions between the rates of change in climate and LULC.

Changes in climate and LULC can interact in their effects on biodiversity [25,26,93]. In this study, it was not possible to explore the effects of interactions among these two types of predictors due to data quantity limitations for some ecoregions. This restricts our findings to the additive effects of the rate of change in climate and LULC on the rate of change of biodiversity. Nevertheless, our results suggest that the effects of the rate of change in climate and LULC combined produce compositional changes in the bird assemblages in different directions. Rapid within-ecoregion heterogenization of species assemblages was driven by high rates of temperature change and rapid declines in precipitation, plus rapid declines in forests and grassland/shrubland. Homogenization within an ecoregion was faster in species assemblages experiencing rapid increases in precipitation, grassland/shrubland, and wetland or rapid declines in crop/pasture LULC. Consistent with our findings, previous studies have found that warming or drying climates combined with habitat loss or an increase in land use intensity severely affect species assemblages and population trends of insects and birds [25,26,93].