Breeding Waterbird Populations Have Declined in South San Francisco Bay: An Assessment Over Two Decades

In south San Francisco Bay, former salt ponds now managed as wildlife habitat support large populations of breeding waterbirds. In 2006, the South Bay Salt Pond Restoration Project began the process of converting 50% to 90% of these managed pond habitats into tidal marsh. We compared American Avocet ( Recurvirostra americana ) and Black-necked Stilt ( Himantopus mexicanus ) abundance in south San Francisco Bay before (2001) and after approximately 1,300 ha of managed ponds were breached to tidal action to begin tidal marsh restoration (2019). Over the 18-year period, American Avocet abundance declined 13.5% (2,765 in 2001 vs. 2,391 in 2019), and Black-necked Stilt abundance declined 30.0% (1,184 in 2001 vs. 828 in 2019). Forster’s Tern ( Sterna forsteri ) abundance was 2,675 birds in 2019. In 2019, managed ponds accounted for only 25.8% of suitable habitats, yet contained 53.9%, 38.6%, and 65.6% American Avocet, Black-necked Stilt, and Forster’s Tern observations, respectively. Conversely, tidal marsh and tidal mudflats accounted for 42.9% of suitable habitats, yet contained only 18.4%, 10.3%, and 19.8% of American Avocet, Black-necked Stilt, and Forster’s Tern observations, respectively. Using a separate nest-monitoring data set, we found that for nesting may help to limit further declines in breeding waterbird populations.


ABSTRACT
In south San Francisco Bay, former salt ponds now managed as wildlife habitat support large populations of breeding waterbirds. In 2006, the South Bay Salt Pond Restoration Project began the process of converting 50% to 90% of these managed pond habitats into tidal marsh. We compared American Avocet (Recurvirostra americana) and Black-necked Stilt (Himantopus mexicanus) abundance in south San Francisco Bay before (2001) and after approximately 1,300 ha of managed ponds were breached to tidal action to begin tidal marsh restoration (2019). Over the 18-year period, American Avocet abundance declined 13.5% (2,765 in 2001 vs. 2,391 in 2019), and Black-necked Stilt abundance declined 30. 0% (1,184 in 2001 vs. 828 in 2019). Forster's Tern (Sterna forsteri) abundance was 2,675 birds in 2019. In 2019, managed ponds accounted for only 25.8% of suitable habitats, yet contained 53.9%, 38.6%, and 65.6% American Avocet, Blacknecked Stilt, and Forster's Tern observations, respectively. Conversely, tidal marsh and tidal mudflats accounted for 42.9% of suitable habitats, yet contained only 18.4%, 10.3%, and 19.8% of American Avocet, Black-necked Stilt, and Forster's Tern observations, respectively. Using a separate nest-monitoring data set, we found that nest abundance in south San Francisco Bay declined for all three species from 2005-2019. Average annual nest abundance during 2017-2019 declined 53%, 71%, and 36%, for American Avocets, Backnecked Stilts, andForster's Terns, respectively, compared to 2005-2007. Loss of island nesting habitat as a result of tidal marsh conversion and an increasing population of predatory California Gulls (Larus californicus) are two potential causes of these declines. All three species established nesting colonies on newly constructed islands within remaining managed ponds; however, these new colonies did not make up for the steep declines observed at other historical nesting sites. For future wetland restoration, retaining more managed ponds that contain islands suitable RESEARCH Breeding Waterbird Populations Have Declined in South San Francisco Bay: An Assessment Over Two Decades INTRODUCTION San Francisco Bay is one of the most important sites for waterbirds along the Pacific Flyway, supporting more than 1 million waterbirds annually (Page et al. 1999;Takekawa et al. 2001;Stenzel et al. 2002;Warnock et al. 2002). Often recognized for its importance to staging and wintering waterbirds, San Francisco Bay also supports large numbers of breeding shorebirds and terns. Over the last few decades, San Francisco Bay has supported the largest breeding populations of American Avocets (Recurvirostra americana) and Black-necked Stilts (Himantopus mexicanus) and nearly 30% of the breeding population of Forster's Terns (Sterna forsteri) along the Pacific coast (Stenzel et al. 2002;Rintoul et al. 2003;Strong et al. 2004;McNicholl et al. 2020). Once uncommon breeders in San Francisco Bay (Sibley 1952;Gill 1977;Rintoul et al. 2003), these three species benefitted from the conversion of approximately 14,000 ha of tidal marsh in south San Francisco Bay to commercial salt ponds between the 1860s through the 1950s (Goals Project 1999). As part of the South Bay Salt Pond (SBSP) Restoration Project, many of these salt ponds have since been transferred to government ownership, have been taken out of salt production, and are now managed as pond habitat for wildlife or are being converted to tidal marsh. In addition to providing abundant invertebrate and fish food resources (Takekawa et al. 2006), these former salt ponds provide nesting habitat in the form of islands and levees. More than 90% of Forster's Tern nests, >70% of American Avocet nests, and ~20% of Black-necked Stilt nests in south San Francisco Bay occur on islands within managed ponds (Strong et al. 2004;Ackerman and Herzog 2012;Hartman et al. 2016a;Ackerman et al. 2020).
In 2001, Rintoul et al. (2003) conducted a comprehensive breeding survey to examine the abundance, distribution, and habitat use of American Avocets and Black-necked Stilts in San Francisco Bay south of the San Mateo Bridge. Through this effort, the populations of American Avocets and Black-necked Stilts in south San Francisco Bay were estimated to be 2,765 and 1,184 birds, respectively. Since 2001, south San Francisco Bay has experienced considerable landscape and management changes (http://www. southbayrestoration.org). First, most salt ponds were transferred to government management in 2003 and are now managed as ponds for wildlife habitat within the US Fish and Wildlife Service's Don Edwards San Francisco Bay National Wildlife Refuge and the California Department of Fish and Wildlife's Eden Landing Ecological Reserve. Second, the breeding population of California Gulls (Larus californicus), a key predator of waterbird eggs and chicks (Herring et al. 2011;Ackerman et al. 2014bAckerman et al. , 2014cTakekawa et al. 2015;Peterson et al. 2017), has increased from an estimated 16,998 gulls in 2001 to 45,026 gulls in 2019 (Burns et al. 2018;Tarjan and Burns 2019), likely exerting higher predation pressure on breeding waterbirds (Ackerman et al. 2014b(Ackerman et al. , 2014c. Finally, the SBSP Restoration Project is implementing a large-scale plan to convert 50% to 90% of former salt ponds into tidal marsh habitats within the next 50 years (USFWS and CDFG 2007). Since 2006, the SBSP Restoration Project has reconnected approximately 1,300 ha of former salt ponds to the bay, beginning the process of tidal marsh restoration. Restoring former salt ponds to tidal marsh is expected to benefit some plants and animals that depend on tidal marsh habitat, as well as buffer against storm surge, and potentially improve water quality and protect inland areas from sea level rise (USFWS and CDFG 2007;Goals Project 2015). However, the loss of managed pond habitat, and islands therein, may negatively affect thousands of waterbirds that rely on these productive wetland habitats for foraging, roosting, and nesting (Warnock and Takekawa 1995;Takekawa et al. 2001;Warnock et al. 2002;Strong et al. 2004;Hickey et al. 2007;Stralberg et al. 2009;Demers et al. 2010;Bluso-Demers et al. 2016;Hartman et al. 2016a To address this, the SBSP Restoration Project was designed to follow an adaptive management framework, whereby waterbird populations are monitored during each phase of managed pond conversion to tidal marsh, allowing managers to adjust plans to achieve the SBSP Restoration Project's goal of maintaining existing waterbird populations in south San Francisco Bay. To help achieve this goal, and to maximize waterbird use of reduced managed pond habitat, nesting islands were constructed within managed ponds during Phase 1 of the SBSP Restoration Project and the construction of additional islands are planned for future phases of the project. New islands have been constructed in the Ravenswood (30 islands at Pond SF2 in 2010) and Alviso (16 islands at Pond A16 in 2012) complexes of the Don Edwards San Francisco Bay National Wildlife Refuge, and at the Eden Landing Ecological Reserve (six islands at Ponds E12 and E13 in 2015; Figure A1). However, despite these many large-scale landscape changes, breeding waterbird population trends in south San Francisco Bay have not been investigated in the past decades.
We evaluated changes in breeding waterbird populations in south San Francisco Bay from 2001 to 2019 using two approaches. First, in 2019, we repeated the 2001 waterbird survey by Rintoul et al. (2003)  to present a detailed assessment of changes to these breeding waterbird populations in south San Francisco Bay over the past 2 decades. Additionally, as pond habitat availability decreases with tidal marsh restoration, those ponds that remain will be critical to meeting the SBSP Restoration Project's goal of maintaining breeding populations of waterbirds. Thus, identifying pond characteristics that support higher breeding waterbird abundance and reproductive success may help with the selection of ponds for tidal marsh conversion, as well as management of remaining ponds. We therefore evaluated the relationship between the abundance of American Avocets, Black-necked Stilts, and Forster's Terns and pond characteristics, including size, shape, location, and salinity during the 2019 survey. Data supporting this manuscript is available as a USGS Data Release (Hartman and Ackerman 2021).

Waterbird Population Survey
In 2019, we surveyed all accessible wetland habitat in San Francisco Bay south of the San Mateo Bridge (Figures 1-3). Importantly, the 2019 survey extent closely matched that of the original survey conducted in 2001 (Rintoul et al. 2003), allowing for direct comparison between years. We surveyed waterbirds over an 11-day period (May 14-24, 2019) equivalent to the 11-day period of the original 2001 survey (May 15-25, 2001;Rintoul et al. 2003). As with the original 2001 survey, we reduced the potential for doublecounting birds by having multiple teams survey simultaneously during a short period of time. We divided the south San Francisco Bay survey area into complexes, many of which were based on existing complexes within the Don Edwards San Francisco Bay National Wildlife Refuge and Eden Landing Ecological Reserve (hereafter Eden Landing; Figures 1-3). We then further sub-divided complexes into individual survey units ( Figures A1, A2), most of which could be completely surveyed in ≤1 day. Teams of 1 to 4 surveyors visited each unit, and by driving, walking, and/or boating, systematically surveyed the entire area using binoculars and 20x-60x spotting scopes. Individual observations consisted of solitary birds or groups of birds of the same species, that were close together (individuals within ~2-3 m of each other), in the same habitat, and engaged in the same behavior. For each observation, we recorded the species, the number of individuals, location within the survey unit (plotted on a printed map of the unit), their behavior, and the main and micro-habitat types (see below). Although the main focal species of this study were American Avocets, Black-necked Stilts, and Forster's Terns, we also recorded any Caspian Terns, Black Skimmers, Elegant Terns, and California Least Terns observed. In addition, we recorded any potential predator of waterbird eggs or chicks (e.g., corvids, raptors, coyotes) and also estimated the number of gulls (the vast majority of which were California Gulls) within the survey unit.
Recorded bird behaviors included feeding, roosting, walking, swimming, flushing, or flying, and those confirming or suggesting breeding activity (nest-building, incubating, brooding chicks, alarm calling, breeding display, or predator distraction display). Individuals observed in transit flying high over a survey unit (flyovers) were recorded for population estimates but were not included in any habitat analyses. We used the total number of individuals that exhibited breeding behaviors as a measure of the minimum number of breeding birds of each species. Although there were no data for 2001, average nest initiation dates obtained from nest-monitoring efforts across south San Francisco Bay (see Nest Abundance Monitoring section below) showed no significant trend in the timing of breeding for American Avocets, Black-necked Stilts, or Forster's Terns between 2005 and 2019 (JT Ackerman, MP Herzog, CA Hartman, unreferenced; see "Notes"). Thus, using the number of individuals exhibiting breeding behaviors to estimate the minimum number of breeding birds in south San Francisco Bay likely was comparable between 2001 and 2019. The mid-to late-May survey period coincided with the peak nesting time for American Avocets and Black-necked Stilts in south San Francisco Bay but was a few weeks before the peak nesting time for Forster's Terns (Ackerman and Herzog 2012).
Thus, the minimum number of breeding Forster's Terns was underestimated.
Each individual survey unit was categorized into one of several main habitat categories: (1) managed pond, (2) active salt pond, (3) sewage/ holding pond, (4) tidal marsh, (5) non-tidal marsh, (6) tidal mudflat, and (7) other wetland habitat (vernal pools, channels, and manmade waterways). In addition, we recorded the micro-habitat in the immediate vicinity of each individual observation. Habitat and micro-habitat definitions are presented in Table 1. Where data were available (recorded during nest monitoring or other pond surveys), ponds were classified as having salinity levels that were relatively low (<60 ppt), medium (60 ppt to 120 ppt), or high (>120 ppt). We also noted whether tidal habitat (marshes and mudflats) occurred within sloughs, within restored ponds (former salt ponds that were opened to tidal action), or along the margins of San Francisco Bay. Observations were instantaneous (Altmann 1974), in that we recorded the behavior, habitat, micro-habitat, and location when we first saw the bird(s), even if they subsequently changed. There were two exceptions: 1. If birds changed from a non-breeding behavior to a breeding behavior, we recorded both behaviors to ensure that we captured potential evidence of breeding.
2. If birds were originally observed in flight, then landed and began another behavior, such as feeding, we recorded the second behavior because it was specific to the habitat.
During surveys, we recorded the habitat types present, which were later used to ground-truth habitat data used for mapping and analysis (see below), and we noted on unit maps any areas that we were not able to survey (because of limitations of access or visibility).
We entered all spatial data (locations of observed birds, habitat mapping, and areas not surveyed) into a geographic information system (GIS; ArcMap 10.6.1; Environmental Research Systems VOLUME 19,ISSUE 3,ARTICLE 4 Institute, Redlands, California) for mapping and spatial analysis. Mapping of available habitat was based on a published data layer (San Francisco Estuary Institute 1998), updated based on current aerial imagery and our own habitat assessments made in the field during the 2019 survey. We also revised the habitat categories in the published data layer (San Francisco Estuary Institute 1998) to match the habitats we used for our field protocol (Table 1).  Figure A1). We accessed each colony nesting site weekly from April through August each year, marked individual nests, and followed the fate of each nest until failure or hatch (detailed methods available in Ackerman 2014c; Hartman et al. 2016b). From these visits, we estimated the number of nests of each species observed each year at each colony site. However, because of time and logistical constraints, and the large study area, we were unable to visit all of the smaller colony sites weekly during every year of the 15-year study. Thus, we preferentially targeted the largest and historically used colony sites based on previous colony monitoring efforts (Strong et al. 2004) for weekly nest monitoring. At sites where we did not monitor weekly, and sites in which the number of nests was far greater than could be monitored during weekly visits, we instead estimated the number of nests of each species in each year from one or more nest surveys conducted during the peak of the breeding season. These surveys consisted of either a walkthrough survey, where individual nests were counted, or observational surveys, where nests were counted from a surrounding levee using binoculars and spotting scopes (most nests are on islands that are sparsely vegetated; Ackerman et al. 2014c). We then summed the number of nests across sites to estimate the annual number of nests of each species across the study area.

Statistical Analysis: Waterbird Survey
To examine changes in abundance and distribution of birds over time, we used the raw data from the 2001 survey reported in Rintoul et al. (2003), which we re-processed using our modified behavior and habitat definitions (therefore the final numbers for 2001 presented in this paper differ slightly from those presented in their 2003 publication). The 2001 habitat data layer used by Rintoul et al. (2003)  Ponds (especially managed ponds and salt ponds) make up a large proportion of the habitat available to waterbirds in south San Francisco Bay. We used generalized linear models in R (R Core Team 2014) to look more specifically at landscape-level factors that might influence the distribution and abundance of American Avocets, Black-necked Stilts, and Forster's Terns in pond habitat. Specifically, we tested the effects of pond type (active salt ponds or managed ponds), pond area, pond shape, pond distance to San Francisco Bay (measured to the nearest pond edge), tide level, salinity, gull abundance, and whether the pond contained islands or not. Mean tide level (m) data were obtained from the Redwood City Station (https://tidesandcurrents.noaa.gov/). Tide level was included to account for the fact that as tide level increases, more tidal mudflat and tidal marsh would be unavailable, potentially increasing use of non-tidal pond habitat. For each pond, we used the mean tide level recorded at the beginning of the survey, unless it took more than 4 hours to completely survey a pond, in which case we used an average of the mean tide level at the beginning of the survey and the end of the survey. Ponds were categorized as low salinity (< 60 ppt), medium salinity (60 ppt-120 ppt), or high salinity (>120 ppt) using existing US Geological Survey and San Francisco Bay Bird Observatory salinity monitoring data. Salinity categories for some active salt ponds were provided by Cargill, Inc. We quantified pond shape using a pond shape index: , where a larger pond shape index indicates a pond with more shoreline relative to the pond's size (McGarigal 2014 (Zuur et al. 2009). We considered using a zero-inflated negative binomial model (many survey units contained zero birds), but good visibility during this survey (areas with poor visibility were excluded) and conspicuous study species meant that zeroes were likely to represent a genuine absence of birds, rather than an issue of detectability. Moreover, preliminary model comparisons showed no improvement in model fit when using zeroinflated negative binomial models vs. standard negative binomial models. We used a pseudo R 2 [1 -(residual deviance / null deviance)] to assess the fit of each model compared to the null model.
American Avocet (χ 9 2 = 1,174.9, p < 0.001, Figure 1) and Black-necked Stilt (χ 9 2 = 496.5, p < 0.001, Figure 2 (Table 3A). There also was less area of moderate to high densities (25 to > 300 birds km -2 ) in 2019 compared to 2001, most notably in southern Eden Landing and Newark (Figures 4 and 5). American Avocet and Blacknecked Stilt densities have decreased in southern Eden Landing and Newark, and increased in northern Eden Landing, and parts of Moffett, Mowry, Warm Springs, and Ravenswood ( Figure 6). Densities of both species decreased sharply in Alviso Pond A16 and the northern part of New Chicago Marsh but increased in the southern part of New Chicago Marsh ( Figure 6).     (Figure 7). Corvids, gulls, and raptors were observed throughout the study area, although the greatest concentrations of gulls were observed around California Gull nesting colonies in Alviso, Mowry, Newark, and Palo Alto (Figure 7). water), we did not consider them as a separate main habitat. Nevertheless, we evaluated how breeding waterbirds were using sloughs relative to their availability (10.5% of the suitable habitat) and found that sloughs accounted for 7.3% of American Avocets, 4.6% of Black-necked Stilts, and 6.2% of Forster's Terns observed. Restored ponds, which are in transition to tidal marsh and tidal mudflat, comprised 9.4% of available habitat, and contained 2.6% of American Avocets, 2.7% of Black-necked Stilts, and 5.0% of Forster's Terns observed.
In 2019, American Avocets (χ 4 2 = 881.8, p < 0.001), Black-necked Stilts (χ 4 2 = 2,546.5, p < 0.001), and Forster's Terns (χ 4 2 = 609.7, p < 0.001) were not randomly distributed among habitat types (Table 3B). American Avocet, Black-necked Stilt, and Forster's Tern abundances were all greater in ponds (includes managed ponds, salt ponds, and sewage/holding ponds) and non-tidal marshes, and lower in tidal marshes and tidal mudflats than would be expected based on availability (Table 3B). Black-necked Stilts showed a much stronger preference for non-tidal marshes (contributing 91% of the overall χ 2 value) than American Avocets (41% of overall χ 2 value) and Forster's Terns (10% of overall χ 2 value), and a weaker preference for ponds. When only birds engaged in breeding behaviors were considered, results were similar, but all three species exhibited similar strength in preference for nontidal marsh, and only American Avocets and Forster's Terns showed a preference for ponds.
Using data from areas surveyed in both 2001 and 2019, American Avocet habitat use changed significantly (χ 4 2 = 172.5, p < 0.001). There was a large increase in use of tidal mudflats (+ 74%) and tidal marshes (+ 155%) and decreases in use of ponds (-24%) and non-tidal marshes (-29%). This analysis does not take into account habitat availability, so some of these changes likely reflect habitat changes between surveys (e.g., ponds transformed into tidal habitat). However, the magnitude of the change in use was, with the exception of non-tidal marshes, greater than the magnitude of changes in habitat availability (tidal mudflats: + 3%, tidal marshes: + 43%, ponds: -10%, non-tidal marshes: -29%). Unlike American Avocets, Black-necked Stilts did not show a significant change in habitat use between 2001 and 2019 (χ 4 2 = 8.79, p = 0.067).
Micro-habitat use varied by species and main habitat type (Table A1) Forster's Terns were mostly observed over open water (60.1%) and roosting on wet bare ground (28.3%).

Factors Affecting Waterbird Use of Pond Habitat
The number of American Avocets, Black-necked Stilts, and Forster's Terns observed within each individual pond unit (n = 89) during the 2019 waterbird survey is provided in Table A2, and a map showing pond locations is provided in Figure A1. However, we lacked sufficient salinity data for nine ponds, and tide level data for an additional pond that was surveyed over multiple days, leaving 79 ponds available for this analysis. The American Avocet model (pseudo R 2 = 0.40) showed that avocet abundance was greater in managed ponds than in salt ponds (z = 2.21, p = 0.03), was greater in ponds of medium salinity (60-120 ppt) than low salinity (< 60 ppt; z = 2.40, p = 0.04) and high salinity (> 120 ppt; z = 3.81, p < 0.001), was greater in ponds where islands were present (z = 2.91, p < 0.01), and increased with increasing abundance of Black-necked Stilts (z = 2.22, p = 0.03). There was no effect of pond area (z = 1.78, p = 0.07), pond shape (z = 0.71, p = 0.48), distance to San Francisco Bay (z = 0.46, p = 0.65), tide level (z = 1.12, p = 0.26), gull abundance (z = 1.25, p = 0.21), or Forster's Tern abundance (z = 0.21, p = 0.83) on American Avocet abundance in ponds.
The Forster's Tern model (pseudo R 2 = 0.68) showed that tern abundance was higher in managed ponds than in salt ponds (z = 4.45, p < 0.001), increased with pond area (z = 3.14, p < 0.01), and decreased with gull abundance (z = 2.97, p < 0.01) and distance to San Francisco Bay (z = 2.73, p < 0.01). Forster's Tern abundance increased as salinity decreased (low salinity (< 60 ppt) > medium salinity (60-120 ppt): z = 2.91, p<0.01; medium salinity > high salinity (> 120 ppt): z = 2.51, p = 0.03). There was no effect of pond shape (z = 0.45, p = 0.65), tide level ( The number of major colonies has similarly declined over the study period. There were (mean ± SD) 12.4 ± 5. Nest abundance at many of the major colony sites exhibited declines between 2005 and 2019 (Table 4, Figure A3). Of 21 major American Avocet colony sites, eight exhibited significant declines, nine showed no significant change, and four exhibited significant increases (three of which were recently established between 2015 and 2019) over the 15-year study period. Of ten major Blacknecked Stilt colony sites, four exhibited significant declines, four showed no significant change, and two exhibited significant increases (both of which were recently established between 2015 and 2019) over the 15-year study period. Finally, of 11 major Forster's Tern colony sites, six exhibited significant declines, two showed no significant change, and three exhibited significant increases (two of which were recently established between 2015 and 2019) over the 15-year study period. Over the past 3 years (2017-2019), there have been no nests at five American Avocet, two Black-necked stilt, and four Forster's Tern colony sites that historically supported major colonies (Table 4).

DISCUSSION
Using two independent data sources, we found strong evidence that breeding American Avocet, Black-necked Stilt, and Forster's Tern populations in south San Francisco Bay have declined considerably over the past 2 decades. In May 2001, Rintoul et al. (2003) counted 2,765 American Avocets and 1,184 black-necked stilts in south San Francisco Bay. Using the same methods used in that 2001 survey, we counted 2,391 American Avocets and 828 Black-necked stilts in May 2019, representing declines of 13.5% and 30.0%, respectively. Considering only individuals exhibiting breeding behaviors, declines were even greater, with American Avocets decreasing 57.2% and Black-necked Stilts decreasing 52.0% in 2019.
Because breeding waterbirds spend a great deal of their daily time budget engaged in behaviors that do not indicate breeding activity (e.g., feeding; Gibson 1978;Robinson et al. 2020), the number of individuals exhibiting breeding behaviors underestimated the actual breeding population size for each species. Moreover, Forster's Tern peak nesting is a few weeks later than peak nesting of American Avocets and Black-necked Stilts (Ackerman and Herzog 2012), and therefore the timing necessary to survey for breeding avocets and stilts underestimated the Forster's Tern breeding population. Nevertheless, because the same methods were used during the 2001 and 2019 surveys, the relative change suggests that the number of breeding American Avocets and Black-necked Stilts in south San Francisco Bay has decreased markedly. Using a second data set of annual nest abundance monitoring, we confirmed that the breeding populations of American Avocets, Black-necked Stilts, and Forster's Terns    (Rintoul et al. 2003). However, unlike the original 2001 survey, we separated marshes into tidal and non-tidal marshes, and found that American Avocets, Blacknecked Stilts, and Forster's Terns used tidal marsh less than expected but non-tidal marsh more than expected based on availability (Table 3B). Much of this effect is from high bird use of New Chicago Marsh, a large non-tidal marsh in Alviso. Availability of diked, non-tidal marsh is low in south San Francisco Bay, because most diked wetlands are sparsely vegetated managed ponds. New Chicago Marsh is one of the most important diked, non-tidal marshes and supports large nesting colonies of all three species (Table 4). Although nest success is lower in New Chicago Marsh relative to islands in managed ponds, the dense vegetation cover within marshes provides preferred brood-rearing habitat for chicks (Ackerman et al. 2014c). Conversely, only 9.8% of all waterbird observations were in tidal mudflat, and only 7.0% were in tidal marsh, even though these two habitat types combined accounted for more than 42% of the available wetland habitat, indicating strong avoidance of these habitats by these breeding waterbird species in south San Francisco Bay. Comparisons between the 2001 and 2019 surveys show that use of tidal mudflats and tidal marshes increased as their availability increased, whereas use of total pond habitat (managed ponds, salt ponds, other ponds) and non-tidal marsh has decreased as the amount of these habitats has declined. However, we did not observe similar increases in the proportion of nests outside of pond habitat, indicating that even if birds were able to use tidal mudflats and tidal marsh for foraging and other needs, they lacked suitable habitat for nesting.
Numerous studies have demonstrated the importance of managed pond habitat in south San Francisco Bay to waterbirds during the breeding and non-breeding seasons (Warnock and Takekawa 1995;Takekawa et al. 2001;Warnock et al. 2002;Strong et al. 2004;Hickey et al. 2007;Ackerman et al. 2009;Demers et al. 2010;Bluso-Demers et al. 2016;Hartman et al. 2016a). For example, multiple telemetry studies have demonstrated preferential use of ponds and avoidance of tidal marsh by terns and shorebirds (Warnock and Takekawa 1995;Hickey et al. 2007;Ackerman et al. 2009;Demers et al. 2010;Bluso-Demers et al. 2016 waterbirds. Therefore, we examined what pond characteristics supported the highest waterbird abundance during the 2019 survey. Abundance of all three species was greater in managed ponds vs. salt ponds. American Avocet abundance was greater in ponds of medium salinity (60-120 ppt) and in ponds in which islands were present. Black-necked Stilt and Forster's Tern abundance increased with pond area. Forster's Tern abundance in ponds also decreased with distance to San Francisco Bay and as salinity increased.
In a previous study, we found that waterbird nest abundance and nest success in south San Francisco Bay was greater within ponds < 1 km from San Francisco Bay (nest abundance was also high among ponds > 4 km from San Francisco Bay), and that nest abundance decreased as the number of islands increased, such that ponds with < 5 islands had more waterbird nests than ponds with > 20 islands (Hartman et al. 2016a). Thus, managed ponds that are larger (> 50 ha), < 1 km from San Francisco Bay, of low to medium salinity, and with three to five islands are likely to support greater waterbird abundance and nest success. However, given the varied preferences in salinity and pond characteristics among waterbird species, reducing the total managed pond area to as little as 10% of what existed before restoration will present significant challenges in ensuring that conditions will be sufficient to maintain the diversity of breeding waterbird populations in south San Francisco Bay.
The increasing population of predatory California Gulls (Burns et al. 2018) is also likely to continue to negatively influence waterbird use of pond habitat. California Gulls account for more than 50% of American Avocet and Forster's Tern chick mortalities (Ackerman et al. 2014b(Ackerman et al. , 2014c and at least 13% of American Avocet nest failures (Herring et al. 2011). In this study, abundance of Black-necked Stilts and Forster's Terns within individual ponds decreased as the number of gulls in the pond increased. Previously, Ackerman et al. (2014b) found that Forster's Tern chick survival increased by 900% after the managed relocation of a nearby California Gull colony in the Alviso complex of south San Francisco Bay and that chick survival at waterbird nesting colonies was negatively related to a gull predation index, which incorporated both the number and distance of gulls from waterbird colonies. Thus, breeding waterbirds appear to prefer ponds and have higher productivity in ponds where California Gull presence is reduced, suggesting that managed ponds further away from California Gull nesting colonies are likely to support more breeding waterbirds. Moreover, as more managed ponds are converted to tidal marsh, breeding waterbird colonies may become concentrated into fewer remaining ponds, potentially facilitating predation by gulls on these colonies. Thus, management of breeding California Gulls may be necessary.

CONCLUSION
In conclusion, breeding waterbirds have decreased in abundance and have changed their distribution markedly in south San Francisco Bay over the past 2 decades. Two factors are likely causes for these declines. First, the number of California Gulls-major predator of Forster's Tern, American Avocet, and Black-necked Stilt chicks (Ackerman et al. 2014b(Ackerman et al. , 2014c)-has almost tripled from 16,998 gulls in 2001 to 45,026 in 2019 (Burns et al. 2018;Tarjan and Burns 2019). Second, the conversion of managed ponds to tidal marsh has resulted in the direct loss of historical island nesting habitat. Forster's Tern, and especially American Avocet nest abundance, began to decline after 2010 (Figure 8), coinciding with the loss of highly productive island nesting habitat in Alviso (Ponds A7 and A8) as a result of the opening of these managed ponds to muted tidal action to begin the process of tidal marsh restoration. The abandonment of highly productive nesting islands in Moffett Ponds A1 and A2W in 2017 from erosion that resulted from higher managed water levels further contributed to declines in nest abundance, particularly of Forster's Terns (Figure 8). In an effort to maintain breeding waterbird populations, the SBSP Restoration Project constructed new nesting islands to enhance four managed ponds (SF2, A16, E12, and E13) that will not be converted to tidal marsh. Forster's Tern and Caspian Tern nesting colonies have been established at two of these https://doi.org/10.15447/sfews.2021v19iss3art4 ponds, aided in part by social attraction efforts (Hartman et al. 2019;Hartman et al. 2020), and American Avocet have nested at all four enhanced ponds. In addition, Forster's Tern nest abundance has increased substantially at New Chicago Marsh where nest success is lower (Ackerman et al. 2014c), and new colonies have been established in Ravenswood Pond SF2 and the Redwood Shores Nob Hill Market area. However, increased nest abundance at these sites has not made up for the steep declines observed at other historically large colony sites (Table 4). As the SBSP Restoration Project continues to convert more managed ponds to tidal marsh, breeding waterbirds will lose additional nesting habitat, which may lead to further population declines. Retaining more managed ponds (particularly those that already support large numbers of nesting waterbirds), and ensuring that suitable nesting islands are provided, while reducing predation by California Gulls, may help to limit breeding waterbird population declines during future phases of the SBSP Restoration Project. Furthermore, regular monitoring of breeding waterbirds in south San Francisco Bay would help assess the effect of tidal marsh restoration, track population changes, and inform management decisions.