Wetlands in the Prairie Pothole Region (PPR) have experienced declines in areal cover, and many remaining wetlands are degraded. Recently, restoration of wetlands has been a key management strategy for waterbirds. In Iowa, the Department of Natural Resources and Ducks Unlimited, Inc. have restored 38 shallow lakes, which are large, mostly permanent wetlands. To assess the impact of habitat variables on migrant waterbird use (waterfowl, shorebirds, and secretive marsh birds), surveys were conducted at 19 shallow lakes restored 1 to 12 years prior to this study and at 11 soon-to-be-restored shallow lakes in Iowa during the spring of 2016-2018. A total of 713,338 waterbirds were identified in 82 species, and more waterbirds and species were detected at restored shallow lakes (582,148 waterbirds and 78 species) than at non-restored shallow lakes (130,895 waterbirds and 70 species). Greatest numbers of diving ducks and waterbird species occurred around 40-50% emergent cover. Water level negatively influenced dabbling ducks and shorebirds and positively influenced diving ducks and total waterbird abundance. Years-since-restoration positively influenced goose/swan abundance and secretive marsh bird abundance, and total wetland area had a positive effect on all groups. These findings emphasize how management strategies for restored shallow lakes should mimic the natural wet-dry cycle of prairie wetlands to provide benefits for all waterbirds.
Wetlands in the Prairie Pothole Region (PPR) of North America are important resources for birds throughout the year. They are critical to the foraging, nesting, and shelter requirements for migrant waterbirds, including waterfowl, wading birds, shorebirds, rails, and other wetland-associated species (Tiner 1984; Johnson et al. 1994; Anteau and Afton 2004). The PPR was once comprised of a vast network of poorly drained potholes and larger shallow lakes, interconnected and interspersed with grasslands that supported a variety of migrating birds (Bishop 1981; Tiner 1984; Van Meter and Basu 2015). Today, this area still provides habitat for more than 100 species of birds. During migration, wetlands in the PPR provide foraging habitat and allow migrants to gain energy and build nutrient reserves in between flights of up to several thousand kilometers (Skagen and Knopf 1993; Naugle et al. 2001; Murphy and Dinsmore 2014). For some species, the energy gained from seeds, plants, and macroinvertebates may have an important influence on their reproductive success (Krapu et al. 1981; Ankney et al. 1991; Jenni and Jenni-Eiermann 1998; Anteau and Afton 2004). Thus, adequate habitat available for migrating birds in this region is a key management concern, and may be a limiting factor for several species (Stafford et al. 2014).
Abundance of wetland and grassland areas across the PPR has declined dramatically since European settlement, and many of the wetlands that remain are highly degraded (Anteau and Afton 2008; Dahl 2014). Draining wetlands and converting grasslands for agriculture were widespread in the 19th and 20th Centuries, resulting in a nearly 50% reduction in wetland area throughout the PPR in the USA (Dahl 2014). Furthermore, many of the wetlands that remain have been dramatically altered due to watershed changes, increased benthivorous and planktivorous fish abundance (Hanson et al. 2005), increased sedimentation (Martin and Harman 1986; Euliss and Mushet 1999; Gleason et al. 2003), and excess nutrient loading (Neely and Baker 1989) and chemical drift (Main et al. 2014). Such factors can lead to wetlands lacking hydrological fluctuations, which may further lead to persistently turbid water and a loss of vegetation (Scheffer et al. 1993; Hanson and Riggs 1995; Zimmer et al. 2002). Wildlife species, including birds, rely on emergent and submersed aquatic plants for nesting and foraging, so the severe decline in quality wetland habitat has likely led to a decline in many species (Igl and Johnson 1997; Anteau and Afton 2008; Anteau et al. 2011).
To try to combat these issues, several management agencies are restoring wetlands and protecting existing wetlands (e.g., Brinson and Eckles 2011), and one restoration project is targeting Iowa shallow lakes, which generally describes a semi-permanent or permanent wetland with a mean water depth < 1.5 m (Cowardin et al. 1979; Geisthardt et al. 2013). Through a partnership between the Iowa Department of Natural Resources and Ducks Unlimited, Inc., the Shallow Lakes Restoration Project (SLRP) aims to restore degraded shallow lakes throughout the Des Moines Lobe region, which is the southernmost portion of the PPR and located in Iowa (Miller et al. 2009). Approximately 90% of Iowa's historical wetland area has been lost due to agriculture and other anthropogenic land uses (Dahl 1990; Miller et al. 2009; Van Meter and Basu 2015). The overall goal of the SLRP is to restore the hydrology of shallow lakes to improve water quality and provide habitat for wildlife. This involves mimicking the wet-dry cycle that characterizes wetlands in the PPR by managing water levels. As part of the natural process, during periods of drought and low water levels wetland plant seeds can germinate (Harris and Marshall 1963), and more nutrients become available as plant litter decomposes (Bärlocher et al. 1978). As water depth increases, emergent and submersed aquatic vegetation replace mudflat annuals, and if the basin is inundated long enough, vegetation may eventually die off (Harris and Marshall 1963; van der Valk and Davis 1976; van der Valk and Davis 1978). For the SLRP, managers install infrastructure, such as water control and fish exclusion structures, to control water levels, improve water quality, and mimic the wet-dry cycle. Ideally, complete drawdowns occur every 5 to 10 years, but the timing varies among individual shallow lakes and based on logistical factors. About 38 sites have been restored using this method since the implementation of the SLRP in 2006, and these shallow lakes have shown improvements in water quality and vegetation structure (Geisthardt et al. 2013).
When implementing these restoration techniques it is important to understand the habitat variables that may influence wildlife or limit populations and how those variables may change over the course of the restoration. The important factors that influence habitat selection by migrant waterbirds include foraging habitat availability, roost availability and safety, and disturbance (Myers et al. 1987; Sprague et al. 2008). Additionally, these factors can vary by species, within a species' annual life cycle, and based on changing environmental conditions (Beerens et al. 2011). Different phases of wetland vegetation and inundation can provide optimal habitat for a variety of avian species. As such, the migrating birds using these wetlands change as resources in the wetlands fluctuate. For example, the amount and density of emergent vegetation can be an important predictor of use by several species (Weller and Spatcher 1965; Fairbairn and Dinsmore 2001). During the initial drawdown period and for a couple of years following, exposure of the basin bottom and lower water levels will not be optimal habitat for species such as rails or diving ducks. On the other hand, shorebirds may forage on the mudflats during spring and fall migration, and dabbling ducks will utilize areas with shallow water and emergent vegetation as the basin gradually refills (Taft et al. 2002; Skagen et al. 2008). Other important variables that affect waterbird use include water depth (Colwell and Taft 2000; Taft et al. 2002), water level fluctuations (Dimalexis and Pyrovetsi 1997; Ntiamoa-Baidu et al. 1998; Taft et al. 2002), disturbances (Webb et al. 2010), and the surrounding landscape (Naugle et al. 1999; Froneman et al. 2001; Pearse et al. 2012). Thus, monitoring restored wetlands over time or wetlands in different stages of vegetation development will help yield a more accurate picture of bird use.
The SLRP has not formally monitored wetland bird communities on these restored shallow lakes in Iowa. Geisthardt et al. (2013) recorded birds at several restored and degraded shallow lakes, but these were not systematic surveys. Furthermore, there has not been an assessment on the habitat variables that influence waterbird use at these shallow lakes. Our objectives were to monitor waterbird use of shallow lakes restored and soon-to-be restored by the SLRP and determine habitat characteristics that influence use by migrating waterbirds. We conducted weekly surveys of both non-restored and restored shallow lakes for waterbirds. We hypothesized that restoration would positively improve waterbird use and that this would be reflected in a positive relationship between abundance or species richness and percent emergent vegetation cover. Additionally, we expected water level management to provide habitat for a variety of waterbirds at both low and high water levels. Results from our study will help inform future decisions for managing restored shallow lakes for waterbirds and should be particularly informative for regular monitoring of migrant waterbirds and managing drawdowns for a variety of species.
Methods
Study Area
The PPR covers about 700,000 km2 in the United States and Canada and is characterized by palustrine wetlands, often known as potholes, and lacustrine wetlands (Bishop 1981; Kantrud et al. 1989; IAN 2001; Dahl 2014). Emergent wetlands still cover about 20,000 km2 in the United States, and in the Des Moines Lobe region about 800 km2 is emergent wetland area (Dahl 2014). This area represents the southernmost extent of the PPR and the Wisconsin glacial advance, which retreated from Iowa about 14,000 years ago (IAN 2001; Miller et al. 2009).
In this study, the term “shallow lakes” refers to both palustrine and lacustrine wetlands (Cowardin et al. 1979) that are relatively large (>20 ha) and on average < 1 m deep. The term “restored” refers to palustrine wetlands that were severely degraded and subsequently restored by manipulating the hydrology to improve water quality and vegetation. These shallow lakes were passively restored (i.e., no seed additions), and they were drained using an existing outlet structure to begin the restoration process. Infrastructure, such as water control structures, water channels, pipelines, and fish exclusion structures, were installed in nearly all shallow lakes to manage water levels and exclude rough fish. Once the restoration process began, sites were refilled gradually over (ideally) a 2-year period to allow vegetation to reestablish. Likewise, the term “non-restored” refers to lacustrine wetlands that were not manipulated. Most of these shallow lakes were void of emergent vegetation and contained turbid water; some may be restored within the next few years. We considered the date of restoration to be the start of the drawdown, even if it was before completion of the water control structure.
Site Selection
To examine how shallow lakes in different restoration states influence migrant waterbird use, we randomly chose 19 restored sites based on age and their relatively large size (>20 ha) spanning the period from less than 1 to 12 years post restoration. We also chose 11 non-restored shallow lakes to examine pre-restoration bird use of shallow lakes. Wetlands were in 12 Iowa counties (Fig. 1; Appendix 1), and all wetlands were surveyed in the spring of 2016, 2017, and 2018.
Bird Surveys
In Iowa, spring migration of waterbirds generally begins around early to mid-March and continues through May (Kent and Dinsmore 1996), and we initiated surveys in conjunction with spring thaw. We focused on waterbirds including waterfowl (Anseriformes), loons (Gaviiformes), grebes (Podicipediformes), cormorants (Suliformes), pelicans (Pelicanformes), cranes and rails (Gruiformes), and shorebirds (Charadriiformes: Recurvirostridae, Charadriidae, Scolopacidae).
Surveys were organized based on the “standardized search” approach described by Watson (2003) and utilized by Hopps (2012), where unit effort is measured in terms of the survey, rather than a fixed number of samples (e.g., point counts). Migrating birds can be more difficult to survey than breeding birds and often aggregate in large numbers that may be difficult to accurately count. Birds are also generally not as vocal during the migration season when compared to the breeding season (Wilson et al. 2000). The surveyor moved around the entire site in order to count all individuals and species present (Watson 2003; Aagard et al. 2015; Loges et al. 2015). Specifically, we utilized vantage points around the perimeter of each shallow lake to count waterbirds. These points were fixed throughout the season and were chosen to provide maximum visibility of each shallow lake. The maximum amount of time spent at each shallow lake depended on the size of the shallow lake basin. For sites ≤ 50 ha we surveyed for up to one hour and for sites > 50 ha we surveyed for up to two hours (Hopps 2012). This time included the time spent counting birds and not the time it took to drive from one point to another. Individuals were identified to species by sight and/or sound. Several secretive marsh birds (i.e., rails, bitterns, grebes) were primarily detected aurally. Although playback recordings are appropriate for estimating secretive marsh birds (Gibbs and Melvin 1993; Lor and Malecki 2002), we did not use this method due to logistical constraints and the fact that we did not do in-wetlands surveys. We counted waterbirds such as Wood Ducks (Aix sponsa), Great Blue Herons (Ardea herodias), and Double-crested Cormorants (Phalacrocorax auritus) that were perched in trees immediately adjacent to the shallow lake. We attempted to get complete counts for flocks of < 100 birds, but for larger flocks a “blocking” technique was used (Webb et al. 2010; Loges et al. 2015). We also included birds that left or entered the site during the duration of the survey. Surveys were conducted weekly, and the average time between surveys was 7.14 days. Surveys were conducted during daylight hours (sunrise to sunset), so three to four sites were generally surveyed in a single day. We maintained a consistent daily schedule for which sites were surveyed each day of the week, but we varied the time of day each site was surveyed each week to account for any intraday variation (O'Neal et al. 2008). Surveys were conducted on days with no precipitation and when winds were consistently < 20 km/hr (Loges et al. 2015).
Figure 1.
Location of shallow lakes in the Iowa Prairie Pothole Region (gray outline) for waterbird surveys conducted in the spring, 2016-2018. Each black dot represents a site, and the 12 shaded counties are those that include the surveyed wetlands.
![f01_135.jpg](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/f01_135.jpg)
We acknowledge that emergent vegetation negatively influenced our ability to see the entire wetland and, as a result, record all birds at some sites. As such, our estimates of abundance may be biased low, particularly at restored sites. However, we visually estimated the percent visibility of each site during every survey and all sites had at least 80% visibility. Furthermore, while most non-restored sites had 100% visibility, and we were confident we recorded all birds at those sites, we still detected a greater number of birds and species at restored sites than at non-restored sites. Thus, diversity and abundances recorded in this study are conservative estimates and improvements in visibility would likely increase the magnitude of these differences.
Habitat Variables
Several habitat variables at the site scale were measured and hypothesized to influence waterbird use, and we measured six variables at this scale. After each vantage point had been visited and the survey completed, we obtained a single estimate of the percent cover of the entire shallow lake that was emergent vegetation, open water, bare ground, or “other” vegetation. These were visually estimated to the nearest 5% and summed to 100%, and all observers received training on this method prior to the start of the seasons to ensure consistency. Both live and senesced vegetation were recorded, and emergent vegetation included vegetation between 2.5 cm and 6 m in height (Loges et al. 2015). Percent other included shrub-scrub and forested habitat (Cowardin et al. 1979), and these were included in a single category because they were uncommon.
Due to logistical constraints, we were not able to estimate the actual basin elevation of each shallow lake. Instead, we measured the weekly change in water level for each shallow lake. Water depth was monitored using water level gauges installed in the shallow lakes. For sites that did not have a gauge, we used other installed structures, such as concrete weirs and culverts, or installed metal posts to monitor weekly change. We defined water level as the weekly deviance (in cm) from the initial water level measurement.
We measured wetland area (ha) using ArcMap 10.3 (Environmental Systems Research Institute 2014). The border of each shallow lake's area was defined using both management area maps from the Iowa Department of Natural Resources and shapefiles from the National Wetlands Inventory (USFWS 2009). We delineated a fixed area that contained semi-permanent to permanent wetland types within the management property, which resulted in an area that represented the shallow lake basin. We acknowledge that this is not the total area inundated, which might be considered true wetland size. However, due to their large size and the intense practices used to manage the water levels, we were confident that the shallow lake basin was very similar to the total area inundated in most cases. This was not true during the drainage process, but drained shallow lakes still have a basin area that is relevant to waterbird use because instead of ponded water they have mudflats used by shorebirds (Murphy and Dinsmore 2013) and avoided by other groups such as diving ducks.
Statistical Approach
We assessed the relationship of six habitat variables, years-since-restoration, and survey week with total waterbird abundance, species richness, and waterbird abundance grouped by foraging strategy and migration chronology (Kent and Dinsmore 1996; Webb et al. 2010). Because we did not conduct in-wetland surveys for waterbirds, and there was some concern about seeing all birds from the shallow lake perimeter, we chose to model abundance rather than density. The waterbird groups included geese/swans, dabbling ducks, diving ducks, and shorebirds. Additionally, we were interested in secretive marsh birds, a group identified as wetland quality indicators (Conway et al. 2009, 2011). This group included the following species: Pied-billed Grebe (Podilymbus podiceps), American Bittern (Botaurus lentiginosus), Least Bittern (Ixobrychus exilis), Virginia Rail (Rallus limicola), Sora (Porzana carolina), and Common Gallinule (Gallinula galeata). American Coot (Fulica americana) are also considered to be secretive marsh birds and use wetlands and bodies of water in a wide range of conditions (Brisbin and Mowbray 2002). We excluded coots from the secretive marsh bird group for analysis because of differences in behavior (e.g., flocking), their frequent use of open water, and due to their lower conservation priority. We were also most interested in the use of these wetlands by species more sensitive to wetland quality changes. Based on a literature review, we made predictions concerning the effects of habitat and survey week on these groups, and these directional hypotheses are shown in Table 1.
We used a linear mixed modeling approach to examine the relationship between the explanatory variables and either goose/swan abundance, dabbling duck abundance, diving duck abundance, secretive marsh bird abundance, shorebird abundance, waterbird abundance, or species richness. We used a square root transformation for groups that did not meet assumptions of normality (Zar 2010), which we assessed by visually inspecting plots of the residuals. However, models for three groups (goose/swan abundance, secretive marsh bird abundance, and shorebird abundance) did not improve with this transformation, so they were transformed with log10(x + 0.5) (Zar 2010). To avoid using highly correlated variables, we only used variables that had a variance inflation factor (VIF) < 2 (Zuur et al. 2010). When we encountered variables with a VIF > 2, we used the most biologically relevant variable. We scaled and centered all variables to improve model convergence (Gelman 2008). Due to missing data for water level, we used only 604 wetland surveys in the final models for all groups except shorebirds. Birds in the order Charadriiformes primarily migrate through the PPR beginning around mid-April (Skagen et al. 2008), so we only included surveys conducted after the third week of April for shorebirds to capture the main migratory period. We used site and year as random effects (Pinheiro and Bates 2000; Schabenberger and Pierce 2002; Zuur et al. 2009). Additionally, to address issues with heterogeneity, we incorporated a variance structure for the abundance models that allowed for a different variance for each survey week (Zuur et al. 2009).
We used program R (ver. 3.4.4.; R Core Team 2018) to initially examine, build, and select models using the “nlme” package (Pinheiro et al. 2017). We used a two-step approach for model selection. We were particularly interested in the effect of the amount of emergent vegetation and its interaction with survey week on abundance and species richness, but we first wanted to determine whether it should be included in the subsequent model selection step. We compared a model with all covariates, including a quadratic effect of emergent vegetation and its interaction with survey week, to subsets of this model that included 1) no emergent vegetation, 2) a linear effect of emergent vegetation, 3) a quadratic effect of emergent vegetation, 4) an interaction between a linear effect of emergent vegetation and survey week, and 5) an interaction between a quadratic effect of emergent vegetation and survey week. Using Akaike's Information Criterion (AIC; Akaiki 1973), we compared the most competitive model with models that included all possible combinations of the covariates in that top model. Similar to Devries et al. (2008), we only considered models to be competitive if they had a ΔAIC < 2, and if they were not more complex versions of the top model (Burnham and Anderson 2002). We did not report or make inferences from models in such cases (Arnold 2010). After initial model selection, we used program SAS (PROC MIXED; SAS Institute, Inc. 2008) to determine coefficients, F-statistics, and significance of covariates (P. Dixon, pers. commun.). We used program SAS in order to obtain P-values using the Satterthwaite approximation, which is not available in the “nlme” package in program R (Luke 2017). For reporting purposes we used the highest ranking models and the back-transformed values of the response variables and their 95% confidence intervals.
Table 1.
Hypothesized responses of local-scale habitat variables to waterbird abundance by taxonomic group and three community-level measurements for waterbird surveys conducted in the Prairie Pothole Region of Iowa, spring 2016 and 2017. An “L” represents a linear trend and “Q” represents a quadratic trend. In parentheses, “+” indicates a positive trend and “-” indicates a negative trend, and both symbols in parentheses indicate the direction of the relationship for a quadratic trend. A “0” means we did not expect the covariate to influence the response variable. Some of the sources used to justify the relationships are given.
![t01_135.gif](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/t01_135.gif)
Results
We conducted surveys from 20 March to 26 May 2016, 17 February to 25 May 2017, and 12 March to 31 May 2018. The earlier start in 2017 was due to the unusually early onset of spring migration. We completed a total of 35 surveys per shallow lake throughout the three years: 10 surveys in 2016, 13 surveys in 2017, and 12 surveys in 2018. Shallow lake area ranged from 19.52 ha to 470.13 ha, with a mean area of 121.00 ha. Percent emergent cover was variable across non-restored and restored shallow lakes (Table 2), with restored shallow lakes having greater emergent cover than non-restored sites. In 2016 and 2017, weekly water level remained relatively consistent but showed an increase on average in 2018 (Fig. 2).
We counted a total of 713,338 waterbirds and identified 82 species. More waterbirds and species were detected at restored shallow lakes, with 582,148 waterbirds and 78 species, than at non-restored shallow lakes, which had 130,895 waterbirds and 70 species. Of the restored shallow lakes, those that were restored six years prior to this study had the greatest number of waterbirds overall with 101,764 waterbirds. Shallow lakes restored five years prior to this study had the greatest number of waterbird species with 67 overall. Of the taxonomic groups of interest, geese and swans were the most frequently encountered (25.87%), followed by dabbling ducks (22.94%), and diving ducks (20.42%). Secretive marsh birds (0.54%) and shorebirds (0.25%) made up a small proportion of encounters. Waterfowl numbers tended to be highest early in the season, while secretive marsh birds and shorebirds gradually increased in numbers later in the season. Total waterbird abundance was mostly driven by waterfowl numbers and was greatest early in the season. Species richness peaked around late April and early May in 2016 and 2018, but peaked in late March in 2017.
Table 2.
Mean (SE) percent emergent vegetation cover and the total number of surveys conducted (n) in spring 2016-2018 at non-restored shallow lakes (19 lakes) and restored shallow lakes (11 lakes) in the Prairie Pothole Region of Iowa.
![t02_135.gif](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/t02_135.gif)
Figure 2.
Mean water level change (with 95% confidence intervals) for each year throughout the survey period at shallow lakes in the Iowa Prairie Pothole Region, 2016-2018. The dashed line represents the first measurement and reference point from which subsequent measurements were compared.
![f02_135.jpg](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/f02_135.jpg)
The highest-ranking model for goose and swan abundance included years-since-restoration and wetland area (Table 3). Percent emergent vegetation, a quadratic effect of percent emergent vegetation, and water level were also included in competitive models. Years-since-restoration (F1, 32 = 15.99, P < 0.01; Table 4; Fig. 3A) and wetland area (F1, 26.8 = 10.03, P < 0.01; Table 4) had significant positive effects on goose and swan abundance in the top model, but the other covariates did not influence abundance in competing models.
The highest-ranking models for dabbling ducks included years-since-restoration, wetland area, water level, and an interaction between survey week and percent emergent vegetation (Table 3). In the top model, water level (F1, 155 = 5.11, P = 0.03; Table 4) had a negative effect (Fig. 4A) and wetland area had a positive effect on dabbling duck abundance (F1, 26.7 = 4.83, P = 0.04; Table 4). The interaction between survey week and percent emergent vegetation cover was significant (F1, 48.8 = 2.42, P = 0.01). Percent emergent vegetation had a negative influence on abundance early and late in the season, but the effect was positive during most of the weeks in the middle of the season. Survey weeks that portray this contrasting interaction the most were the first week in March in 2017, when emergent vegetation had a negative influence on dabbling duck abundance, and the third and fourth week in March and the first week in April, when there was a positive effect on abundance (Fig. 4B). Years-since-restoration did not influence dabbling duck abundance.
Table 3.
Model selection results for linear mixed models of habitat effects on spring migrant waterbirds surveyed in the Iowa Prairie Pothole Region, 2016-2018. The covariates considered included years since restoration (YSR), shallow lake area (area), water level change (water change), a linear and quadratic effect of percent emergent vegetation cover (emergent), and an interaction between emergent and survey week (week). Week and year were included as fixed effects in every model but are not shown unless included in an interaction. Models with a ΔAIC < 2 are shown, along with their AIC, model weights (wi), and number of parameters (K).
![t03_135.gif](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/t03_135.gif)
For diving ducks, the highest-ranking models included years-since-restoration, wetland area, water level, and an interaction between survey week and a quadratic trend in percent emergent vegetation (Table 3). Wetland area (F1, 25.4 = 17.05, P < 0.01; Table 4) and water level (F1, 166 = 17.05, P = 0.03) had positive effects on diving duck abundance (Fig. 5A). The interaction between survey week and a quadratic trend in percent emergent vegetation cover had a significant influence on diving duck abundance (F12, 52 = 3.36, P < 0.01). Abundance tended to peak around 40% emergent vegetation cover. This effect was strongest during the second and fourth weeks in March, and it gradually weakened after the second week in April (Fig. 5B). Years-since-restoration did not affect diving duck abundance.
The highest-ranking models for secretive marsh bird abundance included wetland area and years-since-restoration (Table 3). Wetland area (F1, 23.8 = 11.18, P < 0.01; Table 4) and years-since-restoration (F1, 27.3 = 9.31, P = 0.01; Fig. 3B) positively influenced secretive marsh bird abundance.
The highest-ranking models for shorebird abundance included water level and wetland area (Table 3). The greatest shorebird numbers tended to occur at sites with low water levels and exposed mud, and water level had a negative relationship with shorebird abundance. However, this and other covariates did not have a significant influence on shorebird abundance.
For total waterbird abundance, the highest-ranking models included years-since-restoration, wetland area, water level, and a quadratic effect of percent emergent vegetation (Table 3). Wetland area (F1, 24.7 = 22.59, P < 0.01; Table 4) and water level (F1, 138 = 9.77, P < 0.01; Fig. 6A) positively influenced total waterbird abundance (Table 4). Percent emergent vegetation did not influence total waterbird abundance in the top model, but there was a significant influence in the second-best model (F1, 71.7 = 5.97, P = 0.02). Total waterbird abundance was predicted to peak around 60% emergent vegetation cover.
Table 4.
Beta coefficients (SE) of covariate effects on spring migrant waterbirds surveyed in the Prairie Pothole Region of Iowa, 2016-2018. Estimates are from the top linear mixed model. An asterisk (*) indicates a significant (P < 0.05) effect. Survey week was included as a factor, so we showed beta estimates from three different weeks when we found a significant effect of emergent and week. Survey week is represented by the week and the month in which the survey took place (e.g., 1st for the first week of the month).
![t04_135.gif](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/t04_135.gif)
Finally, the highest-ranking models for species richness included wetland area, water level, and an interaction between survey week and a quadratic trend in percent emergent vegetation (Table 3). Wetland area had a positive effect on species richness (F1, 25.4 = 25.76, P < 0.01; Table 4). The interaction between percent emergent vegetation and survey week also influenced species richness (F12, 51.2 = 2.36, P = 0.02). This quadratic effect tended to be negative or nearly zero throughout the season. The strongest negative effect occurred during the second week in March, when abundance peaked around 40% emergent vegetation cover (Fig. 6B). However, during the last two weeks in April, the quadratic effect was positive, with species richness peaking at very low or very high emergent vegetation cover, but this was a weaker effect. Water level did not influence species richness.
Discussion
While much wetland restoration in the PPR has focused on smaller, temporary or seasonal wetlands, the restoration of large, lacustrine wetlands (i.e., shallow lakes) is also beneficial to waterbirds (Hanson and Butler 1994; Anteau and Afton 2009; Fox et al. 2018). This is particularly true for migrant waterbirds that require adequate stopover locations to rest, refuel, and acquire energy reserves for breeding (Krapu et al. 1995; LaMontagne et al. 2001; Alisauskas 2002). In fact, the decline in some waterbirds, such as the Lesser Scaup (Aythya affinis), may be attributed to the degradation of these shallow lakes and loss of invertebrate diversity (Anteau and Afton 2004, 2008, 2009). Due to their large size and the resulting availability of microhabitats, these sites have the potential to accommodate a greater variety of waterbirds, especially during migration (Hansson et al. 2010). Additionally, the practice of managing shallow lakes to mimic the natural hydrological, wet-dry cycle provides different habitats through time and supports a greater number of species and individuals than if they were to remain in an open, turbid state (Hargeby et al. 1994; Murkin and Caldwell 2000; Hansson et al. 2010). Restored sites host a variety of plant life, such as emergent vegetation, which can provide shelter for waterbirds, and submersed aquatic vegetation, which can provide forage and a substrate for invertebrates (Murkin and Caldwell 2000; Paszkowski and Tonn 2000).
Figure 3.
The predicted effect of years-since-restoration (n = 604 waterbird surveys) on goose/swan abundance (A) and secretive marsh bird (SMB) abundance (B) at shallow lakes in the Iowa Prairie Pothole Region, 2016-2018. The shaded ribbons represent 95% confidence intervals.
![f03_135.jpg](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/f03_135.jpg)
Figure 4.
The predicted effects of water level (A) (n = 604 waterbird surveys) and an interaction between survey week and emergent vegetation (B) on the abundance of dabbling ducks at shallow lakes in the Iowa Prairie Pothole Region, 2016-2018. The shaded ribbons represent 95% confidence intervals, but these are not shown for B to improve clarity of the relationship. The number of waterbird surveys conducted for each week differed, with 16 surveys for the first week in March 2017, 71 surveys each for the third week in March (all three years), and 81 surveys in the second week in April (all three years).
![f04_135.jpg](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/f04_135.jpg)
Figure 5.
The predicted effects of water level (A) (n = 604 waterbird surveys) and an interaction between survey week and a quadratic effect of emergent vegetation (B) (n = 90 waterbird surveys for each week) on the abundance of diving ducks at shallow lakes in the Iowa Prairie Pothole Region, 2016-2018. The shaded ribbons represent 95% confidence intervals, but these are not shown for B to improve clarity of the relationship. The number of waterbird surveys conducted for each week differed, with 20 surveys for the second week in March 2017, 77 surveys each for the fourth week in March (all three years), and 81 surveys in the second week in April (all three years).
![f05_135.jpg](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/f05_135.jpg)
Figure 6.
The predicted effects of water level on total waterbird abundance (A) (n = 604 waterbird surveys) and an interaction between survey week and a quadratic effect of emergent vegetation (n = 90 waterbird surveys for each week) on waterbird species richness (B) at shallow lakes in the Iowa Prairie Pothole Region, 2016-18. The shaded ribbons represent 95% confidence intervals, but these are not shown for B to improve clarity of the relationship. The number of waterbird surveys conducted for each week differed, with 20 surveys for the first week in March 2017 and 83 surveys each for the third and fourth weeks in April (all three years).
![f06_135.jpg](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/f06_135.jpg)
Several habitat variables have been associated with abundance, density, or richness of waterbirds (Fairbairn and Dinsmore 2001; Webb et al. 2010), and understanding these relationships is important when managing restored wetlands (Ma et al. 2010). In our study, shallow lake area was an influential variable for several waterbird groups, total waterbird abundance, and species richness. In all cases, area had a strong positive effect for geese and swans, diving ducks, secretive marsh birds, total waterbird abundance, and species richness. This relationship has been demonstrated previously for several waterfowl species and total waterfowl abundance (LaGrange and Dinsmore 1989; Colwell and Taft 2000; Webb et al. 2010) and for secretive marsh birds such as the Least Bittern and Pied-billed Grebe (Tozer et al. 2010; Baschuk et al. 2012; Harms and Dinsmore 2013; Monfils et al. 2014). During spring migration, waterfowl feed on a variety of food items, including waste corn, moist soil seeds, and invertebrates (Anteau and Afton 2008; Pearse et al. 2013; Tidwell et al. 2013). The propensity for certain forage varies by availability, species (Baldassarre and Bolen 1994; Tidwell et al. 2013), and breeding condition (Krapu et al. 1995; Pearse et al. 2011). Secretive marsh birds also feed on both plant and animal material during spring migration (Conway 19995; Muller and Storer 1999). With a larger area, these shallow lakes have a high degree of water level and vegetation complexity and are able to support various species with different nutritional requirements (Weller and Spatcher 1965; Kantrud and Stewart 1984).
At the same time, the positive association of area with several waterbird groups contrasts with other studies that found that a complex of several smaller wetlands supported a similar, if not greater, number of waterbirds and species than a single large wetland (Brown and Dinsmore 1986; Fairbairn and Dinsmore 2001). Indeed, small, seasonally flooded wetlands or a complex of both large and small wetlands are considered important for several species of waterbirds (Kantrud and Stewart 1977; Fairbairn and Dinsmore 2001; Naugle et al. 2001; Tozer et al. 2010). However, with an average size of 125 ha, all of our shallow lakes were much larger than the smaller wetlands considered by some of these studies, which examined wetlands as small as 1 ha or less (Brown and Dinsmore 1986). Therefore, we conclude that the shallow lakes in our study area may provide a greater variety of food items that can accommodate more individuals and species than complexes of small wetlands (Kantrud and Stewart 1984; Brown and Smith 1998).
Another important variable for some groups was the percent cover of emergent vegetation. We found emergent cover to be important for dabbling ducks, diving ducks, total waterbird abundance, and species richness. The importance of emergent vegetation has been well documented for waterbirds that use wetlands during migration and breeding (Weller and Spatcher 1965; VanRees-Siewert and Dinsmore 1996; Webb et al. 2010; Harms and Dinsmore 2013). Emergent vegetation provides shelter and a substrate for invertebrates (Murkin et al. 1992), and there is evidence that invertebrate abundance is greater when open water is interspersed with emergent vegetation (Voigts 1976). Indeed, waterbird abundance tends to be greatest when emergent vegetation and open water are present at a 1: 1 ratio (Weller and Spatcher 1965; Kaminski and Prince 1984; Webb et al. 2010). We found that total waterbird abundance and the abundance of diving ducks had a quadratic relationship with emergent vegetation, with abundance peaking around 40-60% vegetation throughout the season. An equal interspersion of open water and emergent vegetation likely provides diverse habitat for different diving duck species. For example, Ring-necked Ducks (Aythya collaris) tend to prefer to forage in habitat interspersed with vegetation, while Ruddy Ducks (Oxyura jamaicense) are more commonly found in open water areas (Bergan and Smith 1989). Additionally, some degree of open water improves visibility of potential predators, while emergent vegetation provides protection against other disturbances, such as weather and human disturbance. (Kaminski and Prince 1981; Austin et al. 2017).
Species richness also had a quadratic relationship with emergent vegetation with most weeks showing a negative quadratic effect. Similar to abundance, previous studies have found that species richness of waterbirds is highest in wetlands with 50% emergent vegetation cover (Weller and Spatcher 1965; Webb et al. 2010), and we found a similar relationship throughout most of the season. The last two weeks in April showed the opposite relationship with the highest numbers of species occurring at low and high emergent coverage. Although change in waterbird species composition was not reported for this study, this variability was likely due to the changing species composition throughout the migration period. For example, waterfowl species numbers were declining slightly in April and May, but species richness for shorebirds and secretive marsh birds was higher during this time period. Migrant shorebirds tend to prefer sites with little to no vegetation and open water (Velasquez 1992; Taft et al. 2002), while secretive marsh birds rely on areas with emergent vegetation (Harms and Dinsmore 2013; Blake-Bradshaw 2018; Fournier et al. 2018; Wilson et al. 2018). Such preferences could explain the change in the relationship between species richness and emergent vegetation.
On the other hand, dabbling ducks showed a linear relationship with emergent vegetation cover. This pattern was negative early in the season but became more positive in the mid- to late-season. Other studies have found that dabbling ducks tend to a have positive linear relationship with emergent vegetation cover (Stafford et al. 2007; Webb et al. 2010). Along with geese, the Mallard (Anas platyhynchos) and Northern Pintail (Anas acuta) are the earliest spring migrants in the PPR, and they often arrive along with the spring thaw (Drilling et al. 2002; Haukos et al. 2006). Indeed, these two species made up a large proportion of the waterbirds we counted early in the season, and they often appeared on thawed areas of mostly frozen shallow lakes. In such cases the presence of open water seemed more important than emergent vegetation, and this could explain this relationship early in the season.
Water level was also an influential predictor for several groups. Published studies have identified water depth as an important variable predicting waterbird use (Pöysa 1983; Colwell and Taft 2000; Isola et al. 2000). It can directly affect the ability of some birds to forage (Pöysa 1983) and influence the food items available (Murkin and Ross 2000). Managers can manipulate water levels, but water level changes in the PPR can still be highly variable and unpredictable within a given year (van der Valk 2005), especially since many of these shallow lakes are isolated and the main source of water input is precipitation (Winter 1989). In our study, dabbling duck abundance was negatively influenced by water level. This relationship has been demonstrated by other studies (Colwell and Taft 2000; Murkin and Caldwell 2000; Taft et al. 2002; Baschuk et al. 2012). Because dabbling ducks generally feed along the surface of the water or just a few centimeters below the surface, they are constrained by relatively shallow waters (< 25 cm; Colwell and Taft 2000). Contrastingly, diving ducks and total waterbird abundance showed a positive relationship with water level. Compared to dabbling ducks, diving ducks tend to forage for macroinvertebrates and tubers in deeper water (Afton et al. 1991; Colwell and Taft 2002). Additionally, nearly half of the observed waterbirds were diving ducks and American Coots, a species that can forage in relatively deep water during migration (Brisbin and Mowbray 2002). This group and species could have been driving the relationship between water level and total waterbird abundance.
We originally anticipated that several groups might be positively influenced by years-since-restoration, but only secretive marsh birds and geese and swans showed this relationship. Some of the earliest migrants in the spring in Iowa are geese and swans, not long after wetlands and ponds begin to thaw. The primary diet of midcontinent, migrant geese consists of agricultural items, such as waste corn and shoots of winter wheat (Krapu et al. 1995; Pearse et al. 2013). As such, wetlands appear to mostly be used for roosting by geese and less for foraging in the spring in the PPR (Krapu et al. 1995; Pearse et al. 2013). Thus, older restorations with increased emergent vegetation cover may provide better protection from predators or cover from weather than more open sites. On the other hand, older restorations tended to have more areas with thicker vegetation, and this could explain the positive relationship between secretive marsh birds and years-since-restoration. However, we did not find emergent vegetation to influence either geese and swans or secretive marsh birds, so there may be another habitat variable related to years-since-restoration that we did not measure that influences these groups. We found this to be particularly surprising for secretive marsh birds. This could partly be due to our sampling scheme, which may not have been appropriate for getting accurate counts of many secretive marsh birds. Future studies of migration for secretive marsh birds should use playback recordings, which have been shown to improve detection of species in this group (Gibbs and Melvin 1993; Lor and Malecki 2002).
Management Implications
This study highlights some of the habitat requirements of waterbirds migrating through the PPR in Iowa in spring, and the importance of restoring large shallow lakes by the SLRP to meet these requirements. There are two major components managers should consider when restoring large prairie wetlands and managing them for waterbirds. First, restoring wetlands that have lost major natural hydrological fluctuations will provide habitat for a greater number of birds and species during migration. Shallow lakes that remain in the lake phase of the wetland wet-dry cycle are the least productive when compared to wetlands with more emergent and submersed aquatic vegetation (Murkin and Caldwell 2000). After dewatering a shallow lake, the natural changes in the vegetation throughout the next few years provides habitat for different birds. Initially, the low water levels attract many shorebirds and some dabbling ducks (Taft et al. 2002). With gradual reflooding, conditions improve for waterfowl and other wetland-associated migrants. Second, additional periodic drawdowns could be implemented to further mimic the hydrological cycle. Ideally, restored shallow lakes will host emergent vegetation for a few years, then begin to die back due to persistent inundation and muskrat activity, which would prevent the “choking out” of wetlands by aggressive emergent vegetation (i.e., Typha sp.; Van der Valk and Davis 1976). At this time, another drawdown might be appropriate.
Acknowledgments
This project was funded by a grant from the U.S. Fish and Wildlife Service (Prairie Pothole Joint Venture) with additional support from the Iowa Department of Natural Resources. We thank Sarah Moodie, David Jabot, Jacob Newton, and Jason Newton for assistance with field work. We thank Brian J. Wilsey and Timothy W. Stewart for providing edits and advice on field methods and analyses. We also thank Mark Gulick, Todd Bishop, Doug Janke, Bryan Hellyer, Steve Wood-ruff, TJ Herrick, Clint Maddix, Chris LaRue, Karen Kinkead, Kevin Murphy, and Tyler Harms for help with planning and logistics. We thank Katherine Goode and Philip M. Dixon for assistance with statistical modeling. We thank two anonymous reviewers for constructive comments to improve the manuscript. We followed all applicable ethical guidelines for the use of birds in research, including those presented in the Ornithological Council's “Guidelines to the Use of Wild Birds in Research” (Fair et al. 2010). This paper is a product of the Iowa Agriculture and Home Economics Experiment Station, Ames, Iowa. Project No. IOW05438 is sponsored by Hatch Act and State of Iowa funds.
Literature Cited
Appendices
Appendix 1.
Names and counties of shallow lakes in the Iowa PPR surveyed for spring migrant waterbirds, 2016-2018 (restored lakes indicated by an asterisk). All properties are managed by the Iowa Department of Natural Resources (“WMA” represents wildlife management area). Locations and shapefiles of these properties can be found at https://geodata.iowa.gov/dataset/national-wetlands-inventory-iowa.
![tA01_135.gif](ContentImages/Journals/cowa/42/2/063.042.0201/graphic/WebImages/tA01_135.gif)