Birds – Grassland

About this Indicator

The primary goals of the analysis were to understand the likely trajectory of 28 common species, provide a benchmark against which managers can measure future changes, guide current and future management, and suggest additional research needed to fill data gaps. The 28 species we analyzed were divided into four groups (with unequal numbers of species in each group) based on the type of vegetation the species are typically found in: riparian, grassland, oak woodland, or shrubland – the latter with an emphasis on birds in chaparral shrublands. We call these groupings “guilds,” and we define them as those species that exploit the same resource—in this case, a vegetation or habitat type. 

Why is this resource included? 

Birds are important to both ecosystems and human health. They provide a wide variety of services, including devouring pests, pollinating flowers, dispersing seeds, scavenging carrion, cycling nutrients, and modifying the environment in ways that benefit other species (Whelan et al. 2015). The San Francisco Bay Area (hereafter referred to as “Bay Area”) is a destination for many bird-watchers—a hobby that can have significant positive impacts on both local and national economies (Carver 2013). Therefore, resource management to promote bird community health is likely to have multiple benefits to non-target species and communities, as well as to promote human well-being. 

Birds are highly visible across both urban and natural landscapes, making them a readily observed and frequently monitored natural resource. Considerable public interest in bird watching has also resulted in community science observations through the Cornell Lab of Ornithology’s eBird platform, among others (see the Metrics in Detail section for more information on the eBird dataset). The great interest in birds on the part of scientists and community members has led to a detailed understanding of the life histories (e.g., nesting habits, migration, feeding) of many common species. 

Because of the relative wealth of data and their widespread distribution across ecosystems, birds are recognized as indicators of ecological change (Carignan and Villard 2002). Indicator species are defined as a species or a “group of species whose population trends, when taken together . . . cast light on trends [within vegetation types] and act as a surrogate for ecosystem health” (Gregory et al. 2005). Considering a group of indicator species, such as a bird guild, allows for multiple observations of health, with specific life-history differences providing clues to ecological conditions that are causing population changes. For example, a recent analysis found that over the last half-century, 57% of North America bird species have declined, with a net loss of 2.9 billion birds (Rosenberg et al. 2019). Because the declines were occurring across so many types of birds, the weight of the evidence pointed to widespread anthropogenic changes causing an alarming gradual decline in many species. Rosenberg et al. (2019) showed that 38 families of birds (e.g., larks, thrushes, wood warblers, finches, blackbirds, American sparrows) have lost more than 50 million individuals each. Even non-native species (e.g., starlings, Old World sparrows), which were assumed to be robust based on their initial population expansions upon introduction into North America in the 1800s, showed similarly drastic declines. Shorebirds and waterbirds have also evidenced serious declines, probably due to the 50% loss of North America’s wetlands. The management implications of a decline in non-native birds versus a decline in native birds are quite different, however. Whereas a decline in non-native species might be desirable because it makes resources available for native birds, a decline in native birds suggests an urgent need to identify causal agents and implement conservation efforts. By considering sub-groups (guilds) within the larger category of birds, we can make these more detailed and nuanced management decisions. 

Desired Condition and Trend 

Given the substantial alteration in the natural landscape of Network partner lands over the last century, this assessment does not use historical abundance as a realistic surrogate for desired condition. Instead, the desired condition is to have the trend (as measured in 2010–2020) in the presence/absence and abundance of each bird species remain unchanged or to improve. We also calculated a summary evaluation of individual species trajectories for all species within the guilds: riparian, grassland, oak woodland, and shrubland. The desired condition is that all species within a guild have trends that are unchanged or improving. 

Current Condition and Trend 

Globally, all species we analyzed except loggerhead shrike are ranked “least concern” by the International Union for Conservation of Nature (IUCN). Although the loggerhead shrike is ranked as “not threatened” by the IUCN, it has experienced declines across most of its geographic range (Smallwood and Smallwood 2021). The U.S. Fish and Wildlife Service (USFWS), taking a more cautionary approach, identifies “birds of conservation concern.” The belted kingfisher, grasshopper sparrow, northern harrier, oak titmouse, Nuttall’s woodpecker, wrentit, and California thrasher are all considered birds of conservation concern. Assessments at the state level (Shuford and Gardali 2008) identify the yellow warbler, grasshopper sparrow, northern harrier, and loggerhead shrike as California bird species of special concern. 

A number of studies track bird trends across different North American ecosystems. Apparent declines in species that use grassland, shrubland, and agricultural sites are particularly concerning; pesticide exposure and fragmentation have been hypothesized as potential drivers of these decreases (Stanton et al. 2018, Eng et al. 2019). From Mexico to Alberta, grassland habitat conversion has left less than 40% of what was present before European settlement. Even more dramatically, California’s native grasslands have also been reduced to less than 1% of what they once were (Wiley et al. 2019), due in part to invasion by non-native annual grasses. 

Rosenberg et al. (2019) analyzed avian population changes since 1970 across North America and found declines in all of the grassland and shrubland focal species considered here: loggerhead shrike, grasshopper sparrow, horned lark, savannah sparrow, white-tailed kite, western meadowlark, northern harrier, California thrasher, wrentit, and rufous-crowned sparrow, as well as declining populations for the following riparian and oak woodland focal species: Wilson’s warbler, belted kingfisher, tree swallow, song sparrow, yellow warbler, oak titmouse, and lark sparrow. 

A few studies have analyzed bird trends within the East Bay region. A riparian breeding bird assemblage was monitored from 1994 to 1998 and again from 2004 to 2008 at Coyote Hills Regional Preserve in Fremont, California (Riensche et al. 2010). Although this location is outside of the area of focus for this ecological heath assessment, it is instructive to note that this study detected a significant decline in the common yellowthroat, Wilson’s warbler, and song sparrow, with no apparent change in measured vegetation variables within the study plots. 

The preceding brief literature summary provides context for trends and conditions measured on lands within Network partner agencies’ boundaries. 

Stressors

Additional Resources

Other Metrics Considered but Not Included

• We considered minor modifications to the trend analyses, however, alternative analyses were consistent with results obtained without these changes, so we did not report them.
• We originally considered looking at trends in the larger East Bay region instead of solely within Network partner agency boundaries but found considerably more biases in where observations were made outside those boundaries. While beyond the scope of this analysis, an analysis of trends at different spatial scales would be interesting. Generally, we recommend considering different extents (e.g., local versus regional) and resolutions (e.g., alternative territory sizes to the one- and four-ha grid cells used here). Specifically, we suggest comparing eBird trends within Network partner lands boundaries to regional analyses of the Bay Area and beyond (i.e., California). Of particular interest is how trends within Network partner lands compare to regional species trends.
• We compared the trends from this report to an analysis of all North American birds (Rosenberg et al. 2019). While the relative ranking of our guild-level assessments were consistent with that study, we found that our single-species results suggested more stable populations.
• We briefly compared our results to those available through the eBird data portal (which does little bias correction). We could not find rigorous documentation of how the eBird portal data summaries were created and thus did not include them here.

Data, Management, and Supporting Information

eBird Data

We used eBird data to analyze trends from 2010–2020 across 28 species associated with four vegetation types: riparian, grassland, oak woodland, and shrubland. These data were collected by community scientists and reviewed by members of the Cornell Lab of Ornithology. (A full description of the eBird dataset, contributors, submission process, and quality control can be found at: ebird.org/about.)

Briefly, eBird observations, sometimes called “checklists,” can be contributed by anyone with an eBird account. They include:
• Where and when the observation took place (time, date, and duration of the observation).
• The type of observation that occurred (e.g., stationary, traveling, historical).
• A specialized protocol (e.g., the nocturnal flight protocol, or an incidental observation that records only a single species and not all the species at that time and location).
• How far the observers traveled as they observed birds (for traveling counts, which were the majority of observations).
• The number of observers.
• The number of species and individuals observed.
Unusual sightings (e.g., a very uncommon species for a given location or time of year) are double-checked by eBird reviewers to confirm the identification. If a sighting cannot be confirmed, it is filtered out before the eBird data are made publicly accessible.

Data Filtering

The full eBird dataset can be downloaded with an eBird account, but due to its very large size, it requires filtering to be uploaded into the open-source statistical programming software R (www.r-project.org/). We provide statistical code for our analysis at https://github.com/erinconlisk/EBSNEcoHealth.

We downloaded all eBird observations from 2010–2020 that explicitly recorded each species’ presence or absence within Network partner land boundaries. We began with 2010 because, starting in that year, there was enough data in any one year to allow for a comparison across years. We chose to include only data from the Network partner land boundaries within the overall area of focus because outside these areas, there were biases towards increased sampling in urban areas and decreased sampling south of Mt. Hamilton. We did not split the trend analysis into subregions (i.e., East Bay Hills, Mt. Diablo Range, Mt. Hamilton) because there were not enough data from the Mt. Hamilton and Mt. Diablo Range subregions, due in part to lack of public access. We also checked for representative sampling across vegetation types and found good general agreement except in grasslands, which were undersampled. Because of this undersampling, we have lower confidence in our assessment of trends for grassland birds than for the other guilds.

We considered only breeding-season data and only species that breed in the East Bay because the number of breeding birds is the most relevant metric of population sustainability. By focusing on this critical season, we also limited variability in the data. For example, the number of individuals for year-round resident species could be biased by an influx of migrants/overwintering birds. Further, during breeding season, individuals are more likely to defend a territory and thus, more likely to be seen or heard and less likely to be moving around the landscape, where they could be double-counted. Nevertheless, we recognize that migratory species that leave the area, such as the yellow warbler, may be subject to stressors on their wintering grounds that affect their breeding numbers.

As was done in Johnston et al. (2021) and Fink et al. (2020), we imposed a grid on the landscape and averaged observations taken on the same day within each grid cell for each of the 66 days in the breeding season. We did this so that observations on a single popular day or in a single popular location would not overwhelm the data from other days and locations. We considered two grid resolutions—100 m (1 hectare [ha]) and 200 m (4 ha)—depending on the species’ territory size.

Data Gaps and Data Collection/Management Needs

The following additional information and data collection would be beneficial to a future assessment:

• There are considerably less eBird data available for the Mt. Diablo Range subregion and very little in the Mt. Hamilton subregion. These areas are also where most of the grassland habitat is, meaning data for that guild are particularly limited. We recommend focusing some Network partner agency resources on sampling in these regions, especially on lands with restricted public access.
• We also recommend that Network partner agencies try to motivate community scientists to visit areas in which they are allowed, and to upload their observations to eBird.
• In future analyses, we might consider other metrics to describe vegetation-type condition (e.g., species richness of eBird observations within Network partner lands through time).
• In addition to comparisons across locations, comparisons across specific species could be interesting. For example, Network partner agencies could identify species of special concern that could be compared to other species with similar vegetation-type preferences. Alternatively, species identified as declining regionally or across California could be compared to trends within Network partner agency boundaries.
• This analysis did not consider specific Network partner lands. Should sufficient eBird data become available across all Network partner lands, it would be advisable to compare trends among them to link trends with management actions or to identify habitat restoration opportunities.
• It would also be interesting to compare eBird observations within Network partner agency boundaries to the urban areas just beyond them. Such a comparison could show if species prefer Network lands to nearby urban areas. For example, Dettling et al. (2021) revealed the value of protected areas for safeguarding riparian avian populations.
• A power analysis could be performed on simulated data. This involves creating a simulated population trend with known change, variability across observations, and error. It could be performed with and without effort variables. Observations from this trend would be selected to recreate observed regional biases (e.g., more observations in East Bay Hills locations). Grid-based averaging would be performed on this simulated dataset. This analysis would look at the ability to detect trends based on bird species prevalence, magnitude of simulated population trend, and number of years of data. While we were unable to run a power analysis for this project, it is worth noting that for the Hawaiian red-tailed tropicbird, Seavy and Reynolds (2007) found a >90% chance of detecting a 50% decline over 10 years using real estimates of total population size. Although their analysis has some differences, it still gives us a degree of confidence that our approach would reveal a very big population change.
• A second power analysis could analyze the influence of increasing data through time. We explored both the influence of increasing numbers of grid cells with eBird observations across the landscape and increasing numbers of observations within each grid cell. We were concerned that with more sampling, there would be more observations of a given species, skewing their prevalence upward through time. We tried to account for this by subsampling the data so that each year contained the same number of observations; however, we found this did not change the results. Another approach would be to include the number of observations within a grid cell as a covariate in the model. We suggest doing this in the future.
• We recommend quantifying population objectives (e.g., ideal abundance levels) for indicator bird species (following Dybala et al. 2017) to complement trend analyses. This would form the basis of a more robust species-level condition assessment that could be incorporated into a vegetation-type condition assessment.
• Constructing a metric looking at species richness (a measure of the number of species present in each location) or species diversity (a metric of both richness and evenness in abundance across species) could be useful. Similar to species trends, considerable care would need to be taken to avoid strong biases in the data.
• Detecting trends across only ten years is difficult, but additional years of data will make this easier. We recommend against starting analyses earlier than 2010; less data are available the further back in time you go. (This could, however, be explored in the previously mentioned power analysis.)
• A peer-reviewed publication on the analytical approach would further validate and improve this methodology and make it accessible to other networks interested in using eBird data for ecological health assessments.

Past and Current Management

As prescribed by regulatory requirements and internal policies, Network partner agencies engage in best management practices and implement avoidance and minimization measures when engaging in landscape and infrastructure maintenance, construction, and development. For example, active vegetation management to reduce fuel loads and wildland fire risk is conducted to the extent practical outside of bird nesting season. If this work is conducted during bird nesting season, biological monitors perform bird-nesting surveys to identify locations of active bird nests. This information is used to establish disturbance-free buffers around the nests as well as timelines for when work can recommence once nesting over. Every effort is made to have biological monitors input all avian observations into eBird.

EBRPD actively engages in habitat restoration that benefits birds both within the area of focus as well as outside its boundaries. This includes volunteer native vegetation planting, habitat clean-up, invasive plant species removal, bird-box installations, and monitoring (Riensche 2008). It also includes grant- or bond-funded major restoration projects, such as the McCosker Sub-Area Creek Restoration and Recreational Improvements Project (https://ceqanet.opr.ca.gov/Project/2017062055), which is restoring more than 1.0 linear mile of highly incised creek bed and riparian habitat in Contra Costa County.

Network partners also have an extensive history of monitoring the focal species analyzed in this report, including area searches from 2000 to 2005 in the East Bay Hills, point counts established at 36 plots in grasslands throughout the area of focus from 2004 to 2011 (Gennet et al. 2017), and long-term riparian bird monitoring both within and outside of the area of focus (Riensche et al. 2010, 2014). Efforts to better understand the impacts of wind energy production on avifauna in the Altamont Pass Wind Resource Area are ongoing (Smallwood et al. 2009, Smallwood and Bell 2020, Smallwood et al. 2020, Smallwood and Smallwood 2021).

Potential Future Actions

• Compare eBird monitoring to focused monitoring using a standard protocol. There is a great need to validate the use of eBird data at the local level and to provide guidance on how to best use eBird data to capture local trends.
• Institute additional bird monitoring in East Bay grasslands and in more remote locations (i.e., not the East Bay Hills). This would augment existing eBird data.
• Repeat the 37 permanent breeding-season grassland point-count surveys on a rotational basis every five to seven years (Bartolome et al. 2013). This would help to more precisely determine the long-term persistence and diversity of grassland birds in relation to grazing management and grassland restoration.
• Create a database locating where management or disturbance has occurred. This could allow eBird data to be divided into treatment groups for analyses.
• Combine additional research and monitoring to make broad management suggestions that might benefit the species studied here.
• Focus future studies and actions on augmenting grassland bird habitat. Consistent with studies outside the Bay Area, we found that grassland birds were not doing as well as riparian, oak, and shrubland birds.

Key Literature and Data Sources

For additional information about this indicator including key literature and data sources see NatureCheck