Slow and Steady: Gradual Drawdown of Private Wetlands Supports Shorebirds During Northbound Migration

Prior to the mid-1800s, the Central Valley of California contained more than 1.6 million ha of wetland habitat (Frayer et al. 1989). Despite the loss of over 90% of this natural wetland habitat, the Central Valley continues to support millions of migratory waterbirds in the region's managed wetlands and compatible agricultural lands. Each year, nearly five million waterfowl use this area in winter, September through March (Olson 2019), and over 500,000 shorebirds rely on the Central Valley during spring migration from April through May (Shuford et al. 1998).

Over 30 species comprise these half a million shorebirds, including plovers (Charadriidae), dowitchers and sandpipers (Scolopacidae), and stilts and avocets (Recurvirostridae; Shuford et al. 1998). Water depth is an important driver of habitat use for shorebirds; bill and leg length determine the range of water depths used by each species (Elphick and Oring 1998; Isola et al. 2000). Sandpipers select for wetlands with more shallow water and mudflat and less vegetation (Skagen and Knopf 1994; Gillespie and Fontaine 2017). These conditions allow access to invertebrates in the water column and in the soft, bottom substrates by species with the shortest bills and legs (Isola et al. 2000; Taft et al. 2002; Strum et al. 2013). In the Central Valley, resources are available within a landscape matrix dominated by private agricultural lands including corn and rice fields that farmers flood postharvest (Stralberg et al. 2011; Dybala et al. 2017).

Migration is energetically taxing and requires access to consistent and reliable resources and habitats. Shorebird abundance and richness are positively associated with the density of nearby wetlands at stopover sites (Albanese and Davis 2015) and in the Central Valley, managed wetlands are one of the most reliable and important sources of flooded habitat (Reiter et al. 2018; Shuford et al. 2019). Because the primary objective of managed wetlands is provisioning waterfowl habitat, landowners flood most seasonal managed wetlands beginning in October and drain them in February and March (Dybala et al. 2017) before the peak of northbound shorebird migration in mid-April. The dewatering of seasonal wetlands prior to this time reduces the amount of available habitat and related food resources that are available to shorebirds, creating habitat shortfall (Dybala et al. 2017; Schaffer-Smith et al. 2017). As a result, the Central Valley Joint Venture (CVJV), a congressionally designated consortium of state and federal agencies and conservation nonprofits responsible for setting habitat objectives for migratory shorebirds and other waterbirds in the Central Valley, created a habitat objective of an additional 4,692 ha of shorebird habitat on the landscape in spring (mid-March–late April; Dybala et al. 2017).

The Grasslands Ecological Area (GEA; Figure 1) contains 38% of the Central Valley's privately owned managed wetlands (CVJV 2006). These wetlands provide an opportunity to work with private landowners in one of the most important wetland habitats in California toward a common goal of conserving and creating shorebird habitat during a period when it is scarce on the landscape. Together, we developed and implemented a practice to create and extend the availability of shallow-water habitat by delaying and slowing the drawdown of water on seasonal wetlands over 4 to 6 wk (gradual drawdown [GD]). Our goal was to determine if seasonal wetlands managed with a GD would provide more habitat for migratory shorebirds than is usually available and thus contribute to the overall habitat objective for nonbreeding shorebirds in the Central Valley. To evaluate our success, we monitored shorebird use of both traditionally managed (TM) wetlands and wetlands implementing GD during northbound migration in the springs of 2015–2017. If successful, we hypothesized that wetlands with GD would support migratory shorebirds more often (a higher proportion of sites with detections) and in greater abundance than TM wetlands during spring migration. We also hypothesized that GD of wetlands would support a greater diversity of migratory shorebird species during spring migration.

The GEA is located in the San Joaquin River basin of California (37°03′30″N 120°51′00″W; Figure 1). A Mediterranean climate with hot, dry summers and mild, wet winters characterizes this area. Precipitation varies from year to year with average annual rainfall of 23 cm, two-thirds of which falls between October and April (WRCC 2018), influencing the amount and quality of habitat on the landscape. The GEA encompasses nearly 73,000 ha of managed wetlands, vernal pools, riparian habitat, rangeland, grassland, and farmland (Rahilly et al. 2010), including over 21,000 ha of private wetlands and 6,300 ha of public wetlands (CVJV 2006). The GEA is a Wetland of International Importance (Ramsar 2019) and a site of International Importance to shorebirds, supporting at least 100,000 shorebirds annually (WHSRN 1992).

Managed wetlands in the GEA are flooded using annual water supplies allocated to the region as part of the Central Valley Project Improvement Act and delivered via a network of canals. Wetlands consist of leveed units in which managers use water control structures at inlets and outlets to adjust water levels. Management of most of the private wetlands in the GEA are aims to provide habitat to waterfowl, produce forage for waterfowl, and contribute to recreational hunting opportunities in winter; these private wetlands are hereafter referred to as “duck clubs.” Water allocations for flooding and irrigating wetlands for these purposes can be restricted in drought years.

We sought to determine if seasonal wetlands managed with a delayed and GD would provide more habitat for migratory shorebirds in the GEA than TM wetlands. To implement the GD practice, wetland owners participated in a reverse auction similar to Reynolds et al. (2017) where landowners submitted their expected cost to implement this practice as a bid. We selected bids to maximize the amount of habitat provided at the lowest cost. Each GD wetland unit began fully flooded on April 1, either through maintenance of winter flooding (most common) or reflooding after an earlier drawdown event. We then reduced water levels gradually over a 4- or 6-wk period such that flooded habitat was still available until at least April 29 or May 13, respectively. Landowners monitored water levels on a weekly basis to ensure compliance.

To understand how shorebird response to GD wetlands compared to what is typically available on the landscape and help us understand the contribution of an incentivized practice, we also studied TM wetlands adjacent or close to the GD wetlands on the same or adjacent duck club (Table 1). The speed of the drawdown on TM wetlands was more rapid, usually occurring over 1–3 wk. Habitat conditions of the TM wetlands varied, largely due to the variation in the start of each unit's drawdown, which could have happened anytime from February to April. This variation in timing of drawdown, and thus, available flooded habitat during our study, reflects the broad range of habitat conditions typically provided by seasonal wetlands in the GEA. While variation in physical attributes (e.g., vegetation type and density, topography, etc.) among wetlands in this study existed, there were no other physical differences between the GD and TM wetlands prescribed as part of this study.

We selected 7 duck clubs in 2015, 10 in 2016, and 9 in 2017 to participate in this study based on their willingness to submit a bid and the cost-effectiveness of that bid. Wetland units implementing GD were chosen by the landowners; 12 wetland units implemented GD in 2015, 28 in 2016, and 21 in 2017 (Table 1). For each wetland unit (both GD and TM) on the selected duck clubs, we generated random survey points, at least 400 m apart, along the unimproved roads adjacent to the wetland units using ArcMap version 10.2.2 (ESRI 2014). From each random point, we defined an area from the edge of each wetland unit, bounded by a 200-m fixed radius or the levees separating units (whichever was closer) as the survey area. We limited survey areas to a maximum distance of 200 m from the observer to ensure a high probability of detecting birds (Shuford et al. 2019). Many survey areas had poor visibility due to tall emergent wetland vegetation, usually cattails Typha spp. and tules Schoenoplectus acutus. On our initial visit, we estimated the proportion of the survey area that was visible (area that we could see and subsequently survey for birds). We then retained only those survey areas that had at least 20% visibility for inclusion in the remainder of the study. To estimate the area surveyed, we calculated the area (ha) of each survey unit using ArcMap, then multiplied that area by the proportion of the survey area that was visible as estimated during the initial visit (a factor that was unchanged over the course of the study). Some wetland units, in both GD and TM wetlands, had more than one survey area. Within the same wetland unit, and thus within the survey areas themselves, water depth and vegetation structure and abundance varied.

From April 1 through mid-May of each year, we conducted surveys of shorebirds in each survey area once per week during daylight hours. We counted all individual shorebirds and identified them to species with the exception of dowitchers Limnodromus spp. and in some cases, yellowlegs Tringa spp. and calidrine sandpipers Calidris spp. We scanned all survey areas for a minimum of 2 min and until we counted all birds present; there was no maximum length of time for completing a survey. We counted only birds that were using (foraging, roosting, etc.) or landed in the survey area; we did not count birds that flew over the survey area. We did not conduct counts in inclement weather (winds >40 km/h, fog, or heavy rain). We also estimated the percentage of the survey area that was covered in water (flooded) as an indication of shorebird habitat.

This study occurred during and after an extreme drought in California (2012–2016), thus the amount of flooded habitat on the landscape varied from year to year. We characterized the amount of open water on the landscape by including the water year information in our comparisons. In California, the water year is the 12-mo period from October 1 to September 30. We used California Department of Water Resources' water year index as calculated from the unimpaired runoff (million acre-feet) for the San Joaquin River basin to categorize the water year type of the San Joaquin Valley for the 3 y of the study (DWR 2018). The amount of precipitation recorded in a given water year determines regional water management decisions in the following spring, thus influencing the flooded habitat available on the landscape.

To understand the habitat made available through the GD practice, during birds surveys we also estimated the percentage of each wetland unit that was flooded. The data on percentage of flooding were not normally distributed and we used bootstrapping and the percentile method to estimate the 95% confidence intervals for the weekly mean shorebird density estimates (Manly 2007). Specifically, we used the 2.5 and 97.5 percentiles of the weekly means calculated for each of 10,000 bootstrap iterations (random resample with replacement) to compare the weekly mean percent flooded between GD and TM wetlands in each year of the study. We also calculated the percentage of the CVJV's nonbreeding shorebird habitat objective (Dybala et al. 2017) that implemention GD as part of this project met by dividing the total acres enrolled in the GD practice summed over all clubs by the overall habitat objective for spring. This estimate is an upper bound and does not account for variation in water depth or the presence of emergent vegetation, which would reduce the area available to shorebirds. We used R v.3.2.4 (R Core Team 2019) in RStudio 1.1.4 (RStudio Team 2019) for all statistical analyses, and specifically the ‘boot' package version 1.3-11 for bootstrapping analyses (Canty and Ripley 2017).

Our objective to understand how the GD practice can contribute to CVJV habitat objectives for nonbreeding migratory shorebirds (Dybala et al. 2017) led us to combine all relevant species for our comparisons between GD and TM wetlands. We pooled the detections for all Calidris sandpipers, dowitchers, yellowlegs, Numenius spp., spotted sandpiper Actitis macularius, and plovers (with the exception of killdeer Charadrius vociferus), hereafter “migratory shorebirds” (Table 2). Nonmigratory shorebirds (primarily stilts, avocets, and killdeer) were present in comparatively low abundances and we excluded them from these analyses.

To characterize differences in migratory shorebird use between GD and TM wetlands, we compared mean migratory shorebird density (birds/ha) for each week of the study to elucidate changes in abundance over time and separately for each of the 3 y. The number of survey areas in each wetland unit varied. To reduce autocorrelation from repeated visits to each wetland unit, we pooled all the survey areas for each wetland unit and we performed all analyses at the level of wetland unit. Due to the large number of surveys with zero shorebirds detected (70–76% of surveys each year) and nonnormal distribution of bird counts and subsequent bird density estimates, we again used bootstrapping with 10,000 iterations to estimate the 95% confidence intervals for the weekly mean shorebird density estimates. We calculated density estimates independently for each week of each year to assess within- and among-year variation. We considered non-overlapping 95% confidence intervals to provide evidence of a difference in bird use. As another method of assessing the effectiveness of the GD practice that total abundance influences less, we also calculated the percentage of individual GD and TM wetland units where we detected at least one migratory shorebird each week of each year. Finally, we compared the mean ± standard error (SE) species richness of migratory shorebirds between GD and TM wetlands across all 3 y combined for each week of the study. Species richness data were normally distributed and followed the same pattern year to year, thus we combined years.

Each of the water years in this study was categorized differently by the Department of Water Resources; 2015 was “critically dry,” 2016 was “dry,” and 2017 was “wet.” Over the 3 y of this study (2015–2017), we monitored 107 survey areas with GD and 73 survey areas with TM and completed 1,245 surveys (Table 1; Table S1, Supplemental Material). Managers implemented GD practice on 617 ha in 2015, 1,211 ha in 2016, and 1,163 ha in 2017. These acres represent up to 13, 26, and 25% of the CVJV short-term habitat objective for nonbreeding shorebirds in the Central Valley in spring (Dybala et al. 2017), respectively.

When we looked at how available shorebird habitat compared between GD and TM wetlands, we found that for the first 3 wk of April in all years, GD wetlands had a higher percentage of the wetland unit flooded compared to TM wetlands (Figure 2; Table S2, Supplemental Material). Percent flooded between GD and TM wetlands in the last week of April was similar in 2015, higher in GD wetlands but with overlapping confidence intervals in 2016, and higher in GD wetlands without overlapping confidence intervals in 2017. There was no evidence of a difference in percentage of flooding of GD and TM wetlands in May in 2015 and 2016, while percentage of flooding remained higher in GD wetlands during the first week of May in 2017 (Figure 2; Table S2). In 2015, four TM wetland units at two duck clubs received irrigations in late April, resulting in an increase in the percentage of the wetland unit that was flooded during the last week of April while GD wetlands continued to draw down over the same timeframe (Figure 2; Table S2). We also documented minor irrigation events in both 2016 and 2017 on three and two survey areas, respectively; however, these events did not increase the mean percentage of flooding compared to the previous week.

We observed a total of 17 shorebird species during this study and counted nearly 42,000 individuals. The most abundant species were dunlin Caldidris alpina, dowitchers Limnodromus spp., and western sandpiper Calidris mauri, and the most often encountered species were American avocet Recurvirostra americana, killdeer, and dunlin (Table 2). Ninety-six percent of shorebirds observed during the study were using GD wetlands. We detected migratory shorebirds on a greater proportion of GD compared to TM wetland units in all weeks of the study in all 3 y except the last week of May (Figure 2; Table S2, Supplemental Material). Average species richness across all years was higher in the GD wetlands for the entire survey period except the last week of May (Figure 3; Table S3, Supplemental Material) and the number of migratory shorebird species detected during any one survey ranged from zero to five.

Overall, we found higher migratory shorebird densities on GD wetlands compared to TM wetlands (Figure 2; Table S2, Supplemental Material). In 2015 and 2017, migratory shorebird density was higher on GD than TM wetlands with no or marginal confidence interval overlap during all weeks of April and the first week of May—the period in which migration was concentrated. We also found the maximum mean shorebird densities were much higher on GD wetlands than TM wetlands in those years (21× higher in 2015; 6× higher in 2017). Results from 2016 were less distinct, when GD wetlands had higher densities only in the first and third week of April, with a maximum mean density of only two times that of TM densities (Figure 2; Table S2).

Migratory shorebird densities varied by week and the timing of peak shorebird migration varied among years. In all years of the study, the highest mean migratory shorebird densities occurred from the second through the fourth week of April. There was a single peak in shorebird density in 2015 (36.5 birds/ha). In 2016, shorebird densities remained at high levels (>25 birds/ha) for the last 2 wk of April. There was no clear peak in 2017, though densities were slightly elevated for the last 3 wk of April (Figure 2; Table S2, Supplemental Material). In all 3 y of the study, shorebird densities were lowest at the end of the study period—the second and third weeks of May.

In this study, the GD of private duck clubs in the GEA provided over twice the amount of shorebird habitat and supported shorebirds densities from 2 to 21 times more than TM wetlands during peak migration, a time of habitat shortfall as identified by the CVJV (Dybala et al. 2017). These benefits were realized by delaying and slowing the rate of drawdown of seasonal wetlands, thus extending the timeframe during which a range of shallow-water habitat was available to migrant shorebirds. Our results demonstrate that collaborations with private landowners to implement small changes in spring water management can have potentially large impacts on conservation outcomes.

The GD practice contributed up to one-quarter of the habitat objectives for nonbreeding shorebirds for the entire Central Valley during spring migration (Dybala et al. 2017). In this highly managed landscape, there are opportunities to scale up this or similar practices to provide additional habitat where and when it is needed most. Landowners can apply this practice to other flooded land cover types and manage them to provide habitat and resources to shorebirds and other waterbirds; land cover types can include private and public wetlands, flooded cropland (e.g. Strum et al. 2013; Sesser et al. 2016), floodwater holding basins, and groundwater recharge basins. In addition, managers could implement this practice or one that creates similar habitat during other times of limited habitat such as southbound migration (Golet et al. 2018). Our study only assesses the response of shorebirds to additional shallow-water habitat; however, the quality of this habitat is also an important factor. Further research on the available food resources and subsequent benefits to body condition and survival for migratory shorebirds would be greatly informative.

We detected just over half (17 species, 56%) of the 30 shorebird species known to migrate regularly through the Central Valley. This is lower than Shuford et al. (1998), who conducted extensive surveys of the entire Central Valley, but comparable to studies with a more localized geographic focus such as the recent work in the rice fields of the Sacramento Valley (Strum et al. 2013; Sesser et al. 2016; Golet et al. 2018) and other studies conducted in the GEA (Taft et al. 2002; Rahilly et al. 2010). Of those 17 species, we detected most on both GD and TM wetlands but mean species richness was higher on GD wetlands for all but the last week of the study. During peak migration, the 4 wk in April, mean species richness on GD wetlands was 2.1 to 3.6 times greater than on TM wetlands.

Sandpipers and dowitchers, which comprised 78% of all migratory shorebirds counted in this study, largely drove patterns in migratory shorebird density. Calidris spp. sandpipers are small and numerous and comprised the largest proportion of northbound migrants in other studies conducted in the Central Valley (Shuford et al. 1998; Sesser et al. 2016; Golet et al. 2018). Despite their apparent abundance in this study, western sandpipers, dunlin, and long-billed dowitchers are species of moderate conservation concern due to climate change (USSCP 2016). An additional 5 of the 13 migratory species in this study are of highest conservation concern (USSCP 2016), reinforcing the importance of providing habitat for shorebirds during migration.

Shorebirds are highly mobile species and tend to aggregate in flocks. Thus, means used to compare practices were affected by high variation in flock size. This, coupled with our modest sample sizes, led us to employ bootstrapping methods to estimate variance as used in other studies with similar data structure (Strum et al. 2013; Sesser et al. 2018). The differences in the mean and variance detected between GD and TM wetlands in multiple years ranging from critically dry to wet provides evidence that gradual drawdown is an effective approach to provide shorebird habitat for northbound migrants. The higher density of shorebirds using GD wetlands as seen in this study also reflects a pattern seen in other studies conducted in the GEA (Taft et al. 2002; Rahilly et al. 2010). Even when shorebird densities overlapped, such as in the second week of April in 2016 when we detected one flock of over 1,000 shorebirds in a TM wetland, we detected shorebirds on GD wetlands more often than on TM wetlands.

Peak shorebird density varied within and among years similar to Taft et al. (2002), likely owing to the highly transient nature of migratory shorebirds. Peak mean density of shorebirds in 2017, a wet year, was 67% lower than the previous 2 y, which were dry and critically dry water years. We suspect that the greater amount of precipitation created additional flooded habitat on the landscape and migratory shorebirds dispersed as they exploited the additional available resources. We recommend that future studies examining wetland management also take into account the extent of open water on the surrounding landscape to provide greater context (Golet et al. 2018; Reiter et al. 2018). Despite the potential for rainfall to create additional habitat in TM wetlands in 2017, GD wetlands still supported 11 times more shorebirds during peak migration that year compared to TM wetlands.

Despite significant benefits for shorebirds, managers should consider potential tradeoffs when implementing this practice. Rahilly et al. (2010) highlighted the reduced production of swamp timothy Crypsis schoenoides, an important waterfowl food, in wetland units with a similar practice implemented. We did not quantify seed production; however, wetland managers and owners in this study did not observe reduced waterfowl use of wetland units following implementation of GD and were willing to continue implementation of the practice. To ensure the health of wetland units and continued forage production, we recommend rotating GD units every 2–3 y to ensure salt accumulation remains low and implementing GD on units where other waterfowl food crops are being grown that are more compatible with a later drawdown (e.g., watergrass Echinochloa crusgalli). Depending on the time of year and the distance to population center when implementing this or similar practice, it may be necessary to coordinate with local mosquito and vector control districts to ensure no risk to human health and safety from the creation of this habitat. Further, the creation of this habitat coincides with the onset of nesting for local migrants and resident species like American avocet and black-necked stilt that prefer to nest near shallowly flooded areas. As water levels draw down, nests can become more susceptible to depredation by terrestrial predators; we recommend continued surveillance and study of the impacts to breeding shorebirds.

Private and working lands are an important part of bird and wildlife habitat in the Central Valley of California. While permanently protected wetland refuges provide important habitat, migrants are also reliant on the availability of flooded habitat on agricultural land and privately managed wetlands (Reynolds et al. 2017) such as those in this study. Incentive programs supporting landowners to flood agricultural land provide a significant proportion, sometimes up to 100%, of the available flooded habitat in the Central Valley (Reiter et al. 2018). This, along with the success of our work on private wetlands highlights the importance of working with private landowners to enhance their habitat for birds and other wildlife. Our results demonstrate how this type of collaboration can benefit migratory shorebirds navigating a landscape impacted by drought and water uncertainty.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. Data collected during shorebird surveys of duck clubs in the Grasslands Ecological Area, California from April to May, 2015−2017.

Found at DOI: https://doi.org/10.3996/102019-JFWM-089.S1 (245 KB XLSX).

Table S2. Mean weekly migratory shorebird density and bootstrapped 95% confidence intervals in gradually drawn down and traditionally managed seasonal wetlands from surveys conducted in the Grasslands Ecological Area, California from April to May, 2015–2017.

Found at DOI: https://doi.org/10.3996/102019-JFWM-089.S2 (19 KB XLSX).

Table S3. Mean weekly migratory shorebird species richness across all years in gradually drawn down and traditionally managed seasonal wetlands from surveys conducted in the Grasslands Ecological Area, California from April to May, 2015–2017.

Found at DOI: https://doi.org/10.3996/102019-JFWM-089.S3 (18 KB XLSX).

We are grateful for the participation and support of the owners of the private duck clubs in the Grasslands Ecological Area. We thank Bob Nardi, Tim Poole, and Rick Ortega from the Grasslands Water District. We thank Billy Abbott, Julia Barfield, Amy Carter, Desiree Loggins, Karen Velas, and the late Valerie Calegari for field and logistical support. We would also like to thank Katie Andrews at The Nature Conservancy for informatics support and Nat Seavy and Matt Reiter for analytical support. Kristen Dybala provided helpful comments on earlier drafts of this paper. We thank the Associate Editor and reviewers who helped improve our original manuscript. This project was generously supported by the S.D. Bechtel, Jr., Foundation and the Packard Foundation. This is Point Blue Conservation Science contribution #2317

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.