#PAGE_PARAMS# #ADS_HEAD_SCRIPTS# #MICRODATA#

Detection and density of breeding marsh birds in Iowa wetlands


Authors: Rachel A. Vanausdall aff001;  Stephen J. Dinsmore aff001
Authors place of work: Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA, United States of America aff001
Published in the journal: PLoS ONE 15(1)
Category: Research Article
doi: https://doi.org/10.1371/journal.pone.0227825

Summary

Accounting for imperfect detection is an important process when obtaining estimates of density or abundance for breeding birds, and this is particularly true when researchers are monitoring birds to assess the success of restored wetlands. Due to the dramatic decline in areal cover and habitat quality, wetland restoration in the Prairie Pothole Region (PPR) is critically important to breeding birds. The Shallow Lakes Restoration Project (SLRP), a partnership between the Iowa Department of Natural Resources and Ducks Unlimited, Inc., aims to restore degraded shallow lakes throughout the Iowa PPR. We conducted unlimited-radius point counts with call-broadcast surveys for breeding marsh birds at 30 shallow lakes in various stages of restoration in 2016 and 2017. Our goals were to assess the impact of covariates on detection probability and estimate density of these species at non-restored, younger (1–5 years since restoration), and older (6–11 years since restoration) restorations. Detection probability ranged between 0.07 ± 0.009 (SE) for Red-winged Blackbird and 0.40 ± 0.09 (SE) for Common Yellowthroat. Percent cattail had a positive quadratic effect on detection probability for four species, with detection decreasing sharply as percent cattail increased and increasing slightly with 100% cattail cover. Wind speed negatively influenced the detection probability of Pied-billed Grebes but had a negative quadratic effect on the detection probability of Marsh Wrens. Both restored shallow lakes had greater densities of breeding Pied-billed Grebes, Marsh Wrens, and Yellow-headed Blackbirds than non-restored shallow lakes, but there was no significant difference between younger and older restorations. Including both habitat and environmental covariates on models for detection probability can improve the precision of estimates for density and should be considered when assessing bird populations pre- and post-restoration of shallow lakes.

Keywords:

Birds – Wetlands – Lakes – Wind – Clouds – Probability density – Marshes – Iowa

Introduction

Effectively monitoring wildlife to assess the impact of management or restoration can be critical to identifying population trends or the need for further management action. Typically, monitoring involves surveys repeated over space or time to get estimates of abundance or density. An important component of most wildlife surveys and obtaining estimates of density is the probability of detecting a species, as most surveys do not count every individual present [1,2]. Several sampling and analysis methods for estimating abundance or density when detection probability is less than one exist, and accounting for factors that influence detection probability can improve precisions of estimates [35]. This is particularly true for marsh birds, a group of birds that includes passerines and secretive marsh birds (i.e., rails, bitterns, grebes) and that can be difficult to monitor due to their elusive nature and the difficulty associated with surveying in wetland habitats [6,7], especially in large wetlands. Marsh birds are often considered wetland quality indicators [8] and some are experiencing declining populations [9,10], so obtaining reliable estimates of density can have important implications for wetland management and restoration.

Wetland restoration is a widely-used management tool for improving water quality and conditions for breeding marsh birds [1113], and this is particularly true in the Prairie Pothole Region (PPR). Of interest throughout parts of the PPR is the restoration and management of large semi-permanent or permanent wetlands, also known as shallow lakes (mean depth <1.5 m) [14,15]. Shallow lakes and other wetlands in the PPR are depressional and receive water inputs from precipitation and, in some cases, groundwater, which can lead to variable water levels within and between years [16,17]. Varying water levels lead to cyclical changes in vegetation, with the germination of emergent vegetation and annuals during drought years and the growth of submergent vegetation in wet years [18,19]. Additionally, variations in basin morphometry and water levels throughout a single wetland may create different degrees of vegetation density [20]. Breeding marsh birds rely on the emergent vegetation for shelter and building nests [2123], while aquatic vegetation can provide forage and habitat for prey items [24,25]. However, >50% of wetlands were drained due to agriculture and human settlement during the past 200 years [26,27], resulting in a loss of habitat for breeding marsh birds. Additionally, due to their large size, many remaining shallow lakes were never drained and degraded over time as a result of less hydrological variability [28], increased sediment loads [2931], and chemical and nutrient inputs [32]. Additionally, the invasion by planktivorous fishes increase bioturbation and decrease zooplankton, and collectively these factors increase phytoplankton biomass and decrease emergent and submersed vegetation growth [3335]. Ultimately, this highly turbid environment provides poor breeding habitat conditions for breeding marsh birds [14,36].

Efforts to restore shallow lakes tend to involve managing water levels to mimic the natural wet-dry cycle [12], along with removing planktivorous fish and nutrients [3739], in order to improve water and vegetation conditions for wildlife. Both the success of restorations and response of birds after restoration vary. Restoration success is usually based upon the long-term stability and diversity of submergent and emergent vegetation, along with the persistence of a clear-water state [40,41]. For example, Søngergaard et al. (2007) found that while there were several positive short-term effects from restoration efforts (e.g., fish removal, nutrient input reduction) throughout lakes in Europe, the majority of sites returned to a turbid state with little vegetation after 5–10 years. This could be due to a number of factors, including the return of planktivorous fish [38,42] and the release of phosphorous from the sediment [30]. Indeed, several species of birds respond positively within the first few years of restoration and species richness may increase with older restorations [43,44]. However, other studies have found that restored wetlands may not be functionally equivalent to reference wetlands (i.e., not recently degraded) and will have comparatively fewer species or waterbird use [45,46]. In the Midwest, for instance, restored shallow lakes and wetlands may be susceptible to invasion by aggressive or exotic plants given the appropriate conditions [4749], leading to a monoculture and a lack of open water. Breeding marsh birds have been shown to have a negative association with such plant species [8]. Monitoring restored wetlands throughout the restoration process and obtaining estimates of density while accounting for imperfect detection can help assess the potential impact of these restoration efforts on breeding marsh birds.

In Iowa, published studies have estimated the detection probability and density of several marsh bird species across the state [5052], but none have attempted to obtain estimates particular to restored and pre-restored wetlands. The Shallow Lakes Restoration Project (SLRP) is focused on restoring shallow lakes throughout the Iowa PPR in order to improve conditions for wildlife. Since 2006 the Iowa Department of Natural Resources and Ducks Unlimited, Inc. have restored over 38 shallow lakes in the Iowa PPR through the SLRP. The goal of the SLRP is to improve water quality and the vegetation community to increase the establishment of diverse fish, bird, and invertebrate communities [15]. Since its implementation, these shallow lakes have shown improvements in water quality and vegetation structure [15]. However, as they age some shallow lakes appear to be showing either a decline in water quality and vegetation cover or a dramatic increase in the density of aggressive emergent plants [15,53], indicating a need for further management.

Our goals were to identify factors that influence detection probability of breeding marsh birds, including five species of marsh passerines and two species of secretive marsh birds, in restored and non-restored shallow lakes and to compare detection probabilities among these species. We evaluated the effects of five covariates, including percent cattail, cloud cover, observer, temperature, and wind speed, on detection probability of seven species. We also assessed the influence of call type for secretive marsh birds. We predicted percent cattail, observer, and wind speed would have the greatest influence on detection probability. When comparing estimates of detection probability, we hypothesized that marsh passerines would have higher detection probabilities than secretive marsh birds. Finally, we obtained detection-adjusted estimates of density based on the top models for detection probability and compared density estimates among three restoration states: non-restored, younger, and older.

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 [17,27,54,55]. Palustrine wetlands are typically vegetated, depressional wetlands with relatively shallow water depths, while lacustrine wetlands are deeper (>2.5 m) in some parts of the basin and mostly lacking vegetation [56]. In the U.S., emergent wetlands still cover about 20,000 km2, and in Iowa’s Des Moines Lobe region about 800 km2 is emergent wetland area (Fig 1) [27]. This area represents the southernmost extent of the Wisconsin glacial advance, which retreated from Iowa about 14,000 years ago [55,57].

Fig. 1.
Location of shallow lakes in the Iowa Prairie Pothole Region (gray outline) for breeding marsh bird surveys conducted in the summer, 2016 and 2017. Each black dot represents a site, and the 12 shaded counties are those that include the surveyed wetlands.

In this study, the term “shallow lake” describes both lacustrine and palustrine wetlands [56] that are large and on average <1.5 m deep. The term “restored” refers to palustrine wetlands that were once severely degraded and subsequently restored by manipulating the hydrology to improve water quality and vegetation. Prior to restoration these shallow lakes could be classified as lacustrine wetlands, as they were mostly lacking vegetation [56]. Restored shallow lakes were passively restored (i.e., no seed additions), and they were drained using an existing outlet or recently installed structures to begin the restoration process. Infrastructure, such as water control structures, channels, pipelines, or pumping systems, were installed into nearly all shallow lakes to manage water levels. Fish exclusion structures were also installed at wetlands to prevent the invasion of rough fish, such as black bullhead (Ameiurus melas) and common carp (Cyprinis carpio). 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. Thus, the age of restored shallow lakes ranged from 1 to 11 years.

Site selection

To examine how the breeding bird communities differed across wetlands in different restoration states, 19 restored sites were randomly chosen based on years since restoration and their large size (>20 ha) spanning the period from 1 to 11 years post-restoration (S1 Appendix). We also chose 11 non-restored shallow lakes to examine pre-restoration bird use of shallow lakes. One of these shallow lakes was in the early stages of restoration in 2017 and, therefore, considered to be a restored site in that year. Sites were located in 12 Iowa counties: Buena Vista, Calhoun, Cerro Gordo, Clay, Dickinson, Emmet, Hancock, Palo Alto, Pocahontas, Winnebago, Worth, and Wright. A total of 30 shallow lakes were surveyed in the spring of 2016 and 2017. For further analyses, we grouped the shallow lakes into one of three groups (hereafter referred to as restoration states): non-restored, younger (1–5 years since restoration), and older (6–11 years since restoration). Several studies have examined the changes in the vegetation community of restored sites at different ages [19,58,59], and others have found that restorations considered to be “older” may differ biologically from “younger” sites [60,61].

Bird surveys

We conducted unlimited-radius 10-min point counts throughout each wetland [62]. Points were situated randomly in wetlands, and the number of points depended on the size of the wetland. Two points were placed in wetlands with areas ranging from 10.1–20 ha, three points in sites 20.1–30 ha, four points in sites 20.1–40 ha, and five points in sites >40 ha [63]. Points were situated >400 m apart to avoid double-counting individual birds [7]. We surveyed sets of points twice during each year to account for any seasonal variation in the vocalization rates of species [64,65]. Point counts were conducted between a half-hour before sunrise and up to four hours after sunrise. We did not survey in rainy conditions or when winds exceeded 20 km/h [7]. We recorded several environmental covariates at the start of each survey, and they included temperature (°C), wind speed (km/h), and cloud cover. We estimated cloud cover using four categories: (0) few or no clouds, (1) partly cloudy, (2) overcast, and (3) fog. We only surveyed in foggy conditions when visibility exceeded 0.25 miles. Upon initial detection, the observer recorded the exact distance (m) to the bird using a laser range-finder (Nikon 8397 Aculon AL11 Laser Rangefinder, Nikon, Melville, New York).

To improve the detection of secretive marsh birds (e.g., rails, bitterns), we incorporated call-broadcasts into the survey period according to methods described by the North American Marsh Bird Monitoring Protocol [6,7]. These species are often difficult to detect using traditional survey techniques [64]. The first 5 min of each survey was a passive period (i.e., no recordings) followed by 5 min of recorded calls. Each minute corresponded to a species; the first 30 s included a recording of the particular species, followed by 30 s of silence. The sequence of calls we used, from first to last, was Least Bittern, Sora (Porzana carolina), Virginia Rail (Rallus limicola), King Rail (Rallus elegans), and Pied-billed Grebe (Podilymbus podiceps) [7]. Except for the King Rail, these are regular breeders in Iowa. In previous call-broadcast surveys conducted in Iowa [63], the King Rail did not have many detections, but Sora and Virginia Rail tended to respond to King Rail calls as readily as they responded to intraspecific calls. We used an MP3 player (SanDisk Sansa Clip 1GB, SanDisk Corporation, Milpitas, California) attached to portable speakers (JBL Flip 3, Harman International Industries, Inc., Stamford, Connecticut) and broadcast at 90 dB from a distance of 1 m in front of the speakers [7]. The speaker faced the interior of the wetland and was 0.5 m from the ground or water surface [63]. Our methods for recording detections of secretive marsh birds were the same as for marsh passerines, but we also recorded the call types of detected secretive marsh birds. Secretive marsh birds use a variety of calls throughout the breeding period [66], and there is evidence that detection probability may differ among the different call types [65]. The North American Marsh Bird Monitoring Protocol lists some of the call types given by each species, and we assigned each detection to a call type based on these protocols [6,7].

Because we were not handling birds, we were not required to have approval by the Iowa State University Institutional Animal Care and Use Committee. Our general bird work is authorized under USGS Bird Banding Permit #23285 to Stephen J. Dinsmore, but permits were not needed for this specific study.

Habitat surveys

After each survey at every point we recorded seven habitat covariates following methods suggested by [6] and similar to [67]. Within 50 m of each point, we visually estimated the percent cover of the dominant vegetation types, which tended to include cattail (Typha sp.), bulrush (Schoenoplectus sp.), river bulrush (Bulboschoenus fluviatilis), or reed canary grass (Phalaris arundinacea) to the nearest 5%. We also estimated the percent cover of open water. Other less dominant vegetation types were grouped into a single category and were not used for analysis. Together these values summed to 100%. For analyses, we used an average value for each habitat measurement for the two visits.

Statistical analysis

To obtain abundance and density estimates of breeding marsh birds, we used the Distance package [68] in Program R [69]. We pooled data across years to increase our sample size and because we had no a priori reason to suspect detection probability varied by year. For each species we used the same modeling approach to select the most appropriate model for estimating parameters of interest. To reduce error related to distance sampling we assigned raw distances to distance bins [70]. We did this by visually examining histograms of distances and assigned bins to avoid heaping of distances [70]. We first examined four models recommended by [70]: the hazard-rate model with a cosine expansion, the half-normal model with a Hermite polynomial expansion, the uniform model with a cosine expansion, and the uniform model with a simple polynomial expansion. We fit a null model for each of these functions. For these and all other models, we stratified by restoration state to get estimates of density for non-restored, younger, and older restorations. For determining covariates that influenced detection probability, we used the multiple covariate distance sampling engine (MCDS), and this limited the key functions to the hazard-rate and the half-normal functions [71,72]. We determined the most appropriate key function for this step by comparing Akaike’s Information Criterion corrected for small sample sizes (AICc) [73]. For marsh passerines, we examined the effect of five covariates: observer, percent cattail, cloud cover, temperature, and wind speed. There were three observers throughout the 2-year period. One observer (Observer 1) surveyed wetlands during both years, while the other two observers surveyed in either 2016 or 2017. These two observers received training prior to completing surveys, but they were equally less experienced than Observer 1. Thus, we included the novice observers as a single observer (Observer 2) in the analysis. We assessed quadratic effects of percent cattail, temperature, and wind speed and compared these to the corresponding main effects models. We only included the quadratic effect in the subsequent step if it performed better (ΔAICc <2) than the main effects model. For secretive marsh birds we also assessed the effect of call type. This included four call types for Pied-billed Grebe: (1) owhoop, (2) hyena, (3) both owhoop and hyena given within a 10-min period, and (4) other [6,7]. For Virginia Rail this included two categories: (1) grunt and (2) other. The other category included less common call types and also visual detections, which were much fewer than aural detections for both species. Finally, we examined all possible combinations of the covariates for each species and used the best competing model to report estimates of density for each restoration state and an overall detection probability. We determined models to be competitive if they were <2 ΔAICc of the top model and they were not more complex versions of the top model [74,75]. For competitive models we assessed beta coefficients of covariates and determined them to have a significant effect on detection probability if the 95% confidence intervals did not overlap zero. If observer or call type were included in competitive models we used the predict function (mrds package) [76] in Program R to produce estimates of detection probability for each category, with all other covariates held at their means. We used the top model for each species to estimate density for each restoration state. We determined densities to be significantly different among the restoration states if the 95% confidence intervals did not overlap.

Results

We conducted surveys at 131 points on 30 wetlands during the course of two breeding seasons. Survey effort was similar at non-restored sites (n = 94), younger restorations (n = 86), and older restorations (n = 82). Surveys occurred from 2 June to 15 July in 2016 and from 1 June to 20 July in 2017. We focused our analysis on seven species for which we had an adequate number of detections, and these included Virginia Rail, Pied-billed Grebe, Marsh Wren (Cistothorus palustris), Swamp Sparrow (Melospiza georgiana), Common Yellowthroat (Geothlypis trichas), Red-winged Blackbird (Agelaius phoeniceus), and Yellow-headed Blackbird (Xanthocephalus xanthocephalus; Table 1). Of the focal species, 1,645 total birds were detected in 2016 and 2,330 total birds in 2017. Red-winged Blackbird had the most detections across all years and surveys (n = 1,000), while Virginia Rail had the fewest detections (n = 140).

Tab. 1. Total number of detections and percent of points with detections of breeding marsh birds surveyed at shallow lakes in the Prairie Pothole Region of Iowa, summer 2016 and 2017.
Total number of detections and percent of points with detections of breeding marsh birds surveyed at shallow lakes in the Prairie Pothole Region of Iowa, summer 2016 and 2017.

The top model for Virginia Rail included the half-normal key function and the effect of percent cattail (Table 2). Percent cattail had a negative effect on detection probability, whereas the effect of temperature was not significant (Table 3). Overall detection probability was 0.12 ± 0.02 SE and density was 0.69 ± 0.15 SE birds/ha. Densities of Virginia Rail were highest in restored shallow lakes, with the greatest densities occurring in older restorations, but this was not a significant difference (Fig 2).

Densities (birds/ha) of breeding marsh birds across shallow lakes surveyed in the Prairie Pothole Region of Iowa, summer 2016 and 2017.
Fig. 2. Densities (birds/ha) of breeding marsh birds across shallow lakes surveyed in the Prairie Pothole Region of Iowa, summer 2016 and 2017.
Sites are divided into non-restored, younger (1–5 years since restoration), and older (6–11 years since restoration) shallow lakes. Vertical error bars represent 95% confidence intervals.
Tab. 2. Model selection results for detection probability for seven species of breeding marsh birds surveyed at shallow lakes in the Prairie Pothole Region of Iowa, summer 2016 and 2017.
Model selection results for detection probability for seven species of breeding marsh birds surveyed at shallow lakes in the Prairie Pothole Region of Iowa, summer 2016 and 2017.
Tab. 3. Beta coefficients (with standard errors) on the logit scale included in the most competitive models for breeding marsh birds surveyed at shallow lakes in the Prairie Pothole Region of Iowa, summer 2016 and 2017.
Beta coefficients (with standard errors) on the logit scale included in the most competitive models for breeding marsh birds surveyed at shallow lakes in the Prairie Pothole Region of Iowa, summer 2016 and 2017.

For Pied-billed Grebe, the most competitive model included the half-normal key function with the covariates wind speed, observer, and a quadratic effect of cattail (Table 2). Wind speed had a negative effect on detection probability (Table 3). Observer had a significant effect on detection probability, with a greater detection probability for Observer 1 than for Observer 2. With all other variables held at their means, detection probability was 0.34 and 0.26 for Observer 1 and Observer 2, respectively. Percent cattail had a positive quadratic effect on detection probability but this quadratic component was not significant. Call type and temperature were also included in other competitive models. The most frequently used call was the owhoop call (294 detections), followed by the hyena (78 detections), other (66 detections), and both the owhoop and hyena (46 detections). Detection probability was greatest for the owhoop call, with a detection probability of 0.36 when all other variables were held constant, and 0.24, 0.27, and 0.28 for the hyena, both hyena and owhoop, and other call types, respectively. In competing models, temperature tended to have a positive effect on detection probability but this was not a significant relationship. Overall detection probability for Pied-billed Grebe was 0.30 ± 0.02 SE and density was 0.44 ± 0.06 SE birds/ha. Densities were greatest in restored shallow lakes, but there was no difference between younger and older restorations (Fig 2).

For Marsh Wrens, the most competitive model included the hazard-rate key function and a quadratic effect of percent cattail and a quadratic effect of wind speed (Table 2). Detection probability mostly declined with percent cattail, but there was as slight increase as percent cattail approached 100% (Table 3). Wind speed showed a negative quadratic effect, with detection probability peaking around 9 kph. Other models within 2 ΔAICc of the top model were more complex versions of the top model so we did not consider them to be competitive. Overall detection probability was 0.15 ± 0.02 SE birds/ha and density was 4.24 ± 0.59 SE birds/ha. Densities of Marsh Wrens were significantly greater in restored sites than in non-restored sites (Fig 2).

Due to a limited number of detections for Swamp Sparrows, we only assessed models with no more than three covariates for this species to avoid overfitting. The single most competitive model for this species included the hazard-rate key function and a linear effect of percent cattail (Table 2), which had a positive effect on detection probability (Table 2). Overall detection probability was 0.12 ± 0.08 SE, while overall density was 0.93 ± 0.62 SE birds/ha. There was no a significant difference among densities in the restoration states (Fig 2).

The most competitive model for Common Yellowthroat included the hazard-rate key function along with the covariates temperature, cloud cover, and wind speed (Table 2). Temperature and wind speed did not show significant effects, whereas cloud cover had a negative effect on detection probability (Table 3). Percent cattail was also included in a competing model but did not show a significant effect on detection probability. Overall detection probability was 0.25 ± 0.05 SE and density was 0.40 ± 0.09 SE birds/ha. There was no significant difference among densities in the restoration states (Fig 2)

The top model for Red-winged Blackbird included the hazard-rate key function and a quadratic effect of percent cattail, along with temperature and wind speed (Table 2). Percent cattail showed a positive quadratic effect, with detection probability highest at 100% cover and lowest at about 50% cover (Table 3). Temperature and wind speed did not show significant effects on detection probability. Cloud cover was also included in two other competitive models and showed a negative effect with detection probability. Overall detection probability was 0.07 ± 0.009 SE and density was 3.46 ± 0.53 SE birds/ha. Density tended to be greater in older restorations, but this was not significantly greater than younger restorations or non-restored sites (Fig 2).

The most competitive model for detection of Yellow-headed Blackbird included a model with the hazard-rate key function and quadratic effects of percent cattail and temperature (Table 2). Percent cattail showed a positive quadratic effect, with detection probability decreasing sharply between 0 and 60% cattail cover and increasing slightly up to 100% cattail cover (Table 3). Temperature did not have a significant effect on detection. All competitive models included the quadratic effect of percent cattail, and a negative quadratic effect of wind speed was also included in a competing model. Overall detection probability for this species was 0.10 ± 0.02 SE. Densities of Yellow-headed Blackbird were highest in older restorations and lowest in younger restorations (Fig 2), with an overall density of 5.07 ± 1.00 SE birds/ha.

Discussion

Our findings emphasize the importance of accounting for factors that may influence detection probability when estimating density or abundance. Furthermore, this is one of the few studies that has examined breeding birds at restored, large, shallow lakes in the PPR (but see [37,7779]). At least four variables significantly influenced the detection probability of breeding birds in these restorations, which will be important to consider when completing future monitoring efforts [1]. And as with other studies [44,80,81], restoration seemed to improve breeding bird densities when compared to non-restored sites, specifically for species such as Pied-billed Grebe, Marsh Wren, and Yellow-headed Blackbird.

Detection probability varies among avian species, and this is particularly true for breeding marsh birds [52,82,83]. We expected cryptic species such as the Virginia Rail to have the lowest detection probabilities, while more vociferous species such as Yellow-headed Blackbirds and Marsh Wrens would have higher detection probabilities. Indeed, the detection probability of Virginia Rail was relatively low but similar to estimates found in another study in this region [51]. We estimated a higher detection probability for Pied-billed Grebe. This species appears to be more aggressive than Virginia Rails during the breeding period and tended to vocalize more frequently, so this pattern was not surprising. The marsh passerines tended to have higher detection probabilities, with Common Yellowthroat having the highest detection probability at 0.40. While this species was not as abundant as the Yellow-headed Blackbird or Marsh Wren, other studies have shown that Common Yellowthroats generally have a high probability of detection and availability and found similar estimates of this parameter [8486]. In contrast, we found lower detection probabilities for Yellow-headed Blackbirds and Marsh Wrens, two species that prefer deep wetlands with dense vegetation [67]. While we did not record call types of these species, these species also make a variety of calls, and they can change as the breeding season progresses [52]. Yellow-headed Blackbirds, for example, tend to sing less once young have fledged, so during this phase the probability of detecting this species given it was present likely declined [52].

Previous work has examined the detection probability of marsh birds and how several factors influence detection of different species [23,50,52,66], and our study further highlights variables that should be considered when surveying for breeding marsh birds. Of particular concern is the experience of observers [87,88]. Providing extensive training in identifying species and determining distance from observer are critical to obtaining robust estimates of detection and abundance [7,89,90]. We found that observer influenced the detection probability of Pied-billed Grebes, with the more experienced observer (Observer 1) having a higher detection probability than the less experienced observers (Observer 2). While the less experienced observers received training on identifying secretive marsh birds, the variety of calls of these species appears to make them less likely to be detected by novice observers. However, call type did not appear in the best supported model for either Virginia Rail or Pied-billed Grebe, but it was included in a competitive model for Pied-billed Grebe. Few studies have examined the effects of call type on detection probability of breeding marsh birds (but see [65]), yet there is evidence that detectability varies among call types of secretive marsh birds [6,65]. The owhoop call, which tends to be used during courtship and territorial displays, had a higher detectability than the hyena call, which is used more for greeting, or other calls [6]. It could be that the owhoop call is more easily detected because it is elicited during more aggressive displays and given more frequently than the other call types.

We expected environmental and habitat variables to influence detection probability, but we found this to be true for only a few species. Wind speed, for example, tends to negatively influence detectability of some species, as windier conditions make hearing birds more difficult [51,91]. However, others have found no effect of wind speed on detectability or availability [92], and some have found a positive influence [93]. We found a negative trend for Pied-billed Grebe, but wind speed had a negative quadratic effect on the detectability of Marsh Wrens, with detectability peaking around 9 kph. While temperature did not show a significant effect for any species, this covariate tended to have a positive relationship with detection probability for the Common Yellowthroat, Red-winged Blackbird, and Yellow-headed Blackbird. Temperature has been shown to have a slight positive effect on some breeding marsh birds, as it may increase the activity of birds [94], but others have found little to no effect [51,95]. On the other hand, cloud cover showed a negative influence the detection of Common Yellowthroats. Other studies have found an influence of cloud cover on detection probability [51,64], but the reason is still unclear. Furthermore, while covariates related to weather are important, habitat may also influence the behavior of birds and affect their detection probability [96]. We found that the percent cover of cattail had a negative influence on the detection probability of Virginia Rails. While this species prefers nesting sites with dense, tall vegetation cover, such conditions can make detecting them more difficult due decreased visibility and noise coming from moving vegetation. On the other hand, percent cattail had a positive effect on detection probability for Swamp Sparrows and a positive quadratic effect on detection probability for Marsh Wren, Red-winged Blackbird, and Yellow-headed Blackbird. Although vegetation can decrease visibility, it also provides more perching locations as birds sing, which could further increase detection probability.

Accounting for the various covariates that influence detection probability allowed us to obtain reliable estimates of density at shallow lakes in three different restoration states when detection probability was less than one. For all species density tended to be greater in older restorations than in younger restorations, but we did not find significant differences between these two restoration states for any species. For three species (Pied-billed Grebe, Marsh Wren, and Yellow-headed Blackbird) densities were significantly greater in both older and younger restorations than in non-restored sites. Yellow-headed Blackbirds and Marsh Wrens, obligate wetland breeders, build nests using robust vegetation [16,9799]. Pied-billed Grebes also prefer to nest in wetlands with emergent vegetation and prefer some degree of interspersion between emergent vegetation and open water [21,22,67]. The diversity of emergent and submergent vegetation, along with the development of sedge meadow and wet prairie habitat near the edges, may provide habitat for a more diverse community of invertebrates and seeds, which make up significant portions of the diets for several species [24,97,98,100]. While we did not directly measure the presence of sedge meadow and wet prairie vegetation, as most of our points were located away from shallow lake edges, there did appear to be a greater presence and abundance of these habitats at restored sites. On the other hand, we did not find significant differences in density for Common Yellowthroat, Swamp Sparrow, and Red-winged Blackbird. These species are not considered to be obligate wetland breeders and will nest in other habitat types, which could explain this finding [101103].

Detection probability is an important parameter to consider when estimating the density of birds at wetland restorations. Several of these species can be difficult to detect given the training required to conduct surveys and the habitat that may influence an observer’s ability to survey affectively, and other environmental factors can also influence detection probability. Future monitoring should focus on minimizing the effects of environmental variables by standardizing protocols and on providing training to surveyors, particularly for identifying secretive marsh birds. We did not assess the influence of habitat variables on density or abundance, yet our findings showed that restored shallow lakes are home to increased densities of several breeding marsh birds.

Supporting information

S1 Appendix [xlsx]
Names, counties, and UTM coordinates of shallow lakes in the Iowa Prairie Pothole Region for breeding marsh birds, summer 2016 and 2017.


Zdroje

1. Mackenzie DI, Nichols JD, Lachman GB, Droege S, Andrew J, Langtimm CA. Estimating site occupancy rates when detection probabilities are less than one. Ecology. 2002; 83: 2248–2255.

2. Burnham KP, Anderson DR. Model selection and multimodal inference: A practical information-theoretic approach. 1st ed. New York, New York: Springer-Verlag; 2002.

3. Bas Y, Devictor V, Jean-Pierre M, Jiguet F. Accounting for weather and time-of-day parameters when analysing count data from monitoring programs. 2008; 3403–3416.

4. Gu W, Swihart RK. Absent or undetected? Effects of non-detection of species occurrence on wildlife–habitat models. 2004; 116: 195–203.

5. Burnham KP. Summarizing remarks: environmental influences. In: Studies in Avian Biology. 1981. p. 324–325.

6. Conway CJ. Standardized North American marsh bird monitoring protocols, version 2009–2. Wildlife Research Report #2009–02. U.S. Geological Survey, Arizona Cooperative Fish and Wildlife Research Unit. Tucson, Arizona; 2009.

7. Conway CJ. Standardized North American marsh bird monitoring protocol. Waterbirds. 2011; 34: 319–346.

8. Glisson WJ, Brady RS, Paulios AT, Jacobi SK, Larkin DJ. Sensitivity of secretive marsh birds to vegetation condition in natural and restored wetlands in Wisconsin. J Wildl Manage. 2015; 79: 1101–1116.

9. Sauer AJR, Pardieck KL, Ziolkowski DJ, Smith AC, Hudson R, Sauer JR, et al. The first 50 years of the North American Breeding Bird Survey. Condor. 2017; 119: 576–593.

10. U.S. Fish and Wildlife Service. Birds of Conservation Concern 2008 [Internet]. United States Department of Interior, Fish and Wildlife Service, Division of Migratory Bird Management. Arlington, Virginia; 2008. Available from: http://www.fws.gov/migratorybirds/

11. Chow-Fraser P. Ecosystem response to changes in water level of Lake Ontario marshes: Lessons from the restoration of Cootes Paradise Marsh. Hydrobiologia. 2005; 539: 189–204.

12. Zohrer JJ. Prairie/wetland complex restorations in the Prairie Pothole Region of Iowa. In: Proceedings of the 17th North American Prairie Conference. 2001. p. 136–137.

13. Zedler JB. Progress in wetland restoration ecology. Trends Ecol Evol. 2000; 15: 402–407. doi: 10.1016/s0169-5347(00)01959-5 10998517

14. Anteau MJ, Afton AD. Amphipod densities and indices of wetland quality across the upper-Midwest, USA. Wetlands. 2008; 28: 184–196.

15. Geisthardt E, Gaustch J, Harland B. Shallow lakes summary report. Des Moines, Iowa; 2013. p. 65.

16. Weller MW, Spatcher CS. Role of habitat in the distribution and abundance of marsh birds. Ames, IA; 1965. p. 31.

17. Kantrud HA, Millar JB, van der Valk AG. Vegetation of wetlands of the Prairie Pothole Region. In: van der Valk AG, editor. Northern Prairie Wetlands. Ames, IA: Iowa State University Press; 1989. p. 132–187.

18. van der Valk AG, Davis CB. The seed banks of prairie glacial marshes. Can J Bot. 1976; 54: 1832–1838.

19. Galatowitsch SM, van der Valk AG. The vegetation of restored and natural prairie wetlands. Ecol Appl. 1996; 6: 102–112.

20. Stewart RE, Kantrud HA. Classification of natural ponds and lakes in the glaciated prairie region [Internet]. Resource Publication. Washington D.C.; 1971. Available from: https://pubs.er.usgs.gov/publication/rp92

21. Lor S, Malecki RA. Breeding ecology and nesting habitat associations of five marsh bird species in western New York. Waterbirds. 2006; 29: 427–436.

22. Rehm EM, Baldassarre GA. The influence of interspersion on marsh bird abundance in New York. Wilson J Ornithol. 2007; 119: 648–654.

23. Darrah AJ, Krementz DG. Occupancy and Habitat use of the Least Bittern and Pied-Billed Grebe in the Illinois and Upper Mississippi River Valleys. Waterbirds. 2010; 33: 367–375.

24. Voigts DK. Aquatic invertebrate abundance in relation to changing marsh vegetation. Am Midl Nat. 1976; 95: 313–322.

25. Murkin EJ, Murkin HR, Titman RD. Nektonic invertebrate abundance and distribution at the emergent vegetation-open water interface in the Delta Marsh, Manitoba, Canada. Wetlands. 1992; 12: 45–52.

26. Tiner RWJ. Wetlands of the United States: Current Status and Recent Trends. Natl Wetl Invent. 1984; 1–76.

27. Dahl TE. Status and trends of prairie wetlands in the United States 1997 to 2009. 2014; Available from: https://www.fws.gov/wetlands/Documents/Status-and-Trends-of-Prairie-Wetlands-in-the-United-States-1997-to-2009.pdf

28. McCauley LA, Anteau MJ, Van Der Burg MP, Wiltermuth MT. Land use and wetland drainage affect water levels and dynamics of remaining wetlands. Ecosphere. 2015; 6: 1–22.

29. Martin DB, Hartman WA. The effect of cultivation on sediment composition and deposition in prairie pothole wetlands. Water Air Soil Pollut. 1987; 34: 45–53.

30. Søndergaard M, Jensen JP, Jeppesen E. Role of sediment and internal loading of phosphorus in shallow lakes. Hydrobiologia. 2003; 506–509: 135–145.

31. Gleason RA, Euliss NH. Sedimentation of prairie wetlands. Gt Plains Res. 1998; 8: 97–112.

32. Neely RK, Baker JK. Nitrogen and phosphorous dynamics and the fate of agricultural runoff. In: van der Valk A., editor. Northern Prairie Wetlands. 1st ed. Ames, Iowa: Iowa State University Press; 1989. p. 92–131.

33. Jurik T, Wang S, van der Valk A. Effects of sediment load on seedling emergence from wetland seed banks. Wetlands. 1994; 14: 159–165.

34. Zimmer KD, Hanson MA, Butler MG. Factors influencing invertebrate communities in prairie wetlands: a multivariate approach. Can J Fish Aquat Sci. 2000; 57: 76–85.

35. Gleason RA, Euliss NH, Hubbard DE, Duffy WG. Effects of sediment load on emergence of aquatic invertebrates and plants from wetland soil egg and seed banks. Wetlands. 2003; 23: 26–34.

36. Fox AD, Balsby TJS, Jørgensen HE, Lauridsen TL, Jeppesen E, Søndergaard M, et al. Effects of lake restoration on breeding abundance of globally declining common pochard (Aythya ferina L.). Hydrobiologia. 2018; 830: 33–44.

37. Hanson MA, Butler MG. Responses of plankton, turbidity, and macrophytes to biomanipulation in a shalow prairie lake. Can J Fish Aquat Sci. 1994; 51: 1180–1188.

38. Zimmer KD, Hanson MA, Butler MG. Relationships among nutrients, phytoplankton, macrophytes, and fish in prairie wetlands. Can J Fish Aquat Sci. 2003; 60: 721–730.

39. Søndergaard M, Jeppesen E, Lauridsen TL, Skov C, Van Nes EH, Roijackers R, et al. Lake restoration: Successes, failures and long-term effects. J Appl Ecol. 2007; 44: 1095–1105.

40. Hanson MA, Herwig BR, Zimmer KD, Hansel-Welch N. Rehabilitation of shallow lakes: time to adjust expectations? Hydrobiologia. 2017; 787: 45–59.

41. Hilt S, Alirangues Nuñez MM, Bakker ES, Blindow I, Davidson TA, Gillefalk M, et al. Response of submerged macrophyte communities to external and internal restoration measures in north temperate shallow lakes. Front Plant Sci. 2018; 9.

42. Hanson MA. Responses to food web manipulation: a field test in a shallow waterfowl lake. Hydrobiologia. 1994; 279/280: 457–466.

43. LaGrange TG, Dinsmore JJ. Pland and animal community responses to restored Iowa wetlands. Prairie Nat. 1989; 21: 39–48.

44. VanRees-Siewert KL, Dinsmore JJ. Influence of wetlands age on bird use of restored wetlands in Iowa. Wetlands. 1996; 16: 577–582.

45. Delphey PJ, Dinsmore JJ. Breeding bird communities of recently restored and natural prairie potholes. Wetlands. 1993; 13: 200–206.

46. Brown SC, Smith CR, Press A. Breeding season bird use of recently restored versus natural wetlands in New York. J Wildl Manage. 1998; 62: 1480–1491.

47. Mushet DM, Euliss NH, Shaffer TL. Floristic quality assessment of one natural and three restored wetland complexes in North Dakota, USA. Wetlands. 2002; 22: 126–138.

48. Mulhouse JM, Galatowitsch SM. Revegetation of prairie pothole wetlands in the mid-continental US: Twelve years post-reflooding. Plant Ecol. 2003; 169: 143–159.

49. Seabloom EW, Valk AG. Plant diversity, composition, and invasion of restored and natural prairie pothole wetlands: Implications for restoration. Wetlands. 2003; 23: 1–12.

50. Harms TM, Dinsmore SJ. Density, abundance, and habitat associations of the inland Swamp Sparrow (Melospiza georgiana georgiana) in Iowa. Wilson J Ornithol. 2015; 127: 670–677.

51. Harms TM, Dinsmore SJ. Density and Abundance of Secretive Marsh Birds in Iowa. Waterbirds. 2012; 35: 208–216.

52. Harms TM, Dinsmore SJ. Optimizing surveys for marsh songbirds: does timing matter? 2018; 89: 276–286.

53. Vanausdall RA, Dinsmore SJ. Impacts of shallow lake restoration on vegetation and breeding birds in Iowa. Wetlands. 2018

54. Bishop RA. Iowa’s wetlands. Proc Iowa Acad Sci. 1981; 88: 11–16.

55. Iowa Association of Naturalists. Iowa Wetlands. 2001. p. 24.

56. Cowardin LM, Carter V, Golet FC, LaRoe ET. Classification of wetlands and deepwater habitats of the United States. FGDC-STD-004-2013 Second Ed. 1979; 79. Available from: http://www.charttiff.com/pub/WetlandMaps/Cowardin.pdf

57. Miller BA, Crumpton WG, van der Valk AG. Spatial Distribution of Historical Wetland Classes on the Des Moines Lobe, Iowa. Wetlands. 2009; 29: 1146–1152.

58. Galatowitsch SM, van der Valk AG. Characteristics of recently restore wetlands in the prairie pothole region. Wetlands. 1996; 16: 75–83.

59. Aronson MFJ, Galatowitsch S. Long-term vegetation development of restored prairie pothole wetlands. Wetlands. 2008; 28: 883–895.

60. Badiou P, McDougal R, Pennock D, Clark B. Greenhouse gas emissions and carbon sequestration potential in restored wetlands of the Canadian prairie pothole region. Wetl Ecol Manag. 2011; 19: 237–256.

61. Bortolotti LE, Vinebrooke RD, St. Louis VL. Prairie wetland communities recover at different rates following hydrological restoration. Freshw Biol. 2016; 61: 1874–1890.

62. Ralph CJ, Sauer JR, Droege S. Monitoring bird populations by point counts. Gen. Tech. Rep. PSW-GTR-149. Albany, California; 1995.

63. Harms TM. Population ecology and monitoring of secretive marsh-birds in Iowa. Wildlife Ecology. 2011.

64. Gibbs JP, Melvin SM. Call-response surveys for monitoring breeding waterbirds. J Wildl Manage. 1993; 57: 27–34.

65. Conway CJ, Sulzman C, Raulston BE. Factors affecting detection probability of California Black Rails. J Wildl Manage. 2004; 68: 360–370.

66. Robertson EP, Olsen BJ. Density, sex, and nest stage affect rail broadcast survey results. J Wildl Manage. 2014; 78: 1293–1301.

67. Harms TM, Dinsmore SJ. Habitat associations of secretive marsh birds in Iowa. Wetlands. 2013; 33: 561–571.

68. Miller DL, Rexstad E, Thomas L, Marshall L, Laake JL. Distance sampling in R [Internet]. Journal of Statistical Software. 2019 [cited 2019 May 1]. Available from: http://doi.org/10.18637/jss.v089.i01

69. R Core Team. R: a language and enironment for statistical computing [Internet]. Vienna, Austria: R Foundation for Statistical Computing; 2018. Available from: http://www.rproject.org/

70. Buckland ST, Anderson DR, Burnham KP, Laake JL, Borchers DL, Thomas L. Introduction to distance sampling: Estimating abundance of biological populations. 1st ed. Oxford, U.K.: Oxford University Press; 2001.

71. Marques FFC, Buckland ST, Marques FFC, Buckland ST. Incorporating covariates into standard line transect analyses. Biometrics. 2003; 59: 924–935. doi: 10.1111/j.0006-341x.2003.00107.x 14969471

72. Marques FFC, Buckland ST. Covariate models for the detection function. In: Advanced Distance Sampling. Oxford, U.K.: Oxford University Press; 2004. p. 31–47.

73. Akaike H. Information theory and an extension of the maximum likelihood principle. In: Petran BN, Csaki F, editors. International symposium on information theory. Akad éiai Kiad ó Budapest, Hungary; 1973. p. 276–281.

74. Devries JH, Armstrong LM, MacFarlane RJ, Moats L, Thoroughgood PT. Waterfowl nesting in fall-seeded and spring-seeded cropland in Saskatchewan. J Wildl Manage. 2008; 72: 1790–1797.

75. Arnold TW. Uninformative parameters and model selection using Akaike’s Information Criterion. J Wildl Manage. 2010; 74: 1175–1178.

76. Laake JL, Borchers D, Thomas L, Bishop J. mrds: Mark-Recapture Distance Sampling. [Internet]. R package version 2.2.0; 2018 [cited 2019 May 1]. Available from: https://cran.r-project.org/package=mrds

77. Moreno-Ostos E, Paracuellos M, De Vicente I, Nevado JC, Cruz-Pizarro L. Response of waterbirds to alternating clear and turbid water phases in two shallow Mediterranean lakes. Aquat Ecol. 2008; 42: 701–706.

78. Epners CA, Bayley SE, Thompson JE, Tonn WM. Influence of fish assemblage and shallow lake productivity on waterfowl communities in the Boreal Transition Zone of western Canada. Freshw Biol. 2010; 55: 2265–2280.

79. Anteau MJ, Afton AD. Wetland use and feeding by lesser scaup during spring migration across the upper Midwest, USA. Wetlands. 2009; 29: 704–712.

80. Hickman S. Improvement of habitat quality for nesting and migrating birds at the Des Plaines River Wetlands Demonstration Project. Ecol Eng. 1994; 3: 485–494.

81. Ratti JT, Rocklage AM, Giudice JH, Garton EO, Daniel P. Comparison of avian communities on restored and natural wetlands in North and South Dakota. J Wildl Manage. 2001; 65: 676–684.

82. Harms TM, Dinsmore SJ. Influence of season and time of day on marsh bird detections. Wilson J Ornithol. 2014; 126: 30–38.

83. Conway CJ, Gibbs JP. Summary of intrinsic and extrinsic factors affecting detection probability of marsh birds. Wetlands. 2011; 31: 403–411.

84. Sauer JR, Blank PJ, Zipkin EF, Fallon JE, Fallon FW. Using multi-species occupancy models in structured decision making on managed lands. J Wildl Manage. 2013; 77: 117–127.

85. Tozer DC, Falconer CM, Bracey AM, Gnass Giese EE, Niemi GJ, Howe RW, et al. Influence of call broadcast timing within point counts and survey duration on detection probability of marsh breeding birds. Avian Conserv Ecol. 2017; 12: art8.

86. Twedt DJ, Wilson RR. Breeding birds in managed forests on public conservation lands in the Mississippi Alluvial Valley. For Ecol Manage. 2017; 384: 180–190.

87. Conway CJ, Simon JC. Comparison of detection probability associated with Burrowing Owl survey methods. J Wildl Manage. 2003; 67: 501–511.

88. Diefenbach DR, Brauning DW, Mattice JA. Variability in grassland bird counts related to observer differences and species detection rates. Auk. 2003; 120: 1168–1179.

89. Nadeau CP, Conway CJ, Smith BS, Thomas E, Nadeau CP, Conway CJ, et al. Maximizing detection probability of wetland-dependant birds during point-count surveys in northwestern Florida. Wilson J Ornithol. 2008; 120: 513–518.

90. Allredge MW, Simons TR, Pollock KH. A field evaluation of distance measurement error in auditory avian point count surveys. J Wildl Manage. 2007; 71: 2759–2766.

91. Holt RD, Butler MJ. Modeling audible detection of prairie grouse booming informs survey design. J Wildl Manage. 2019; 83: 638–645.

92. Tozer DC, Drake KL, Falconer CM. Modeling detection probability to improve marsh bird surveys in southern Canada and the Great Lakes states. Avian Conserv Ecol. 2016; 11.

93. Manning JA. Factors affecting detection probability of burrowing owls in southwest agroecosystem environments. J Wildl Manage. 2011; 75: 1558–1567.

94. Verner J. Time budget of the male Long-billed Marsh Wren during the breeding season. Condor. 2019; 67: 125–139.

95. Conway CJ, Gibbs JP. Factors influencing detection probability and the benefits of call broadcast surveys for monitoring marsh birds [Internet]. U.S. Geological Survey, Final Report, Patuxent Wildlife Research Center. Laurel, Maryland; 2001. Available from: http://ag.arizona.edu/research/azfwru/NationalMarshBird/downloads/technical_reports/Conway_and_Gibbs_2001_Report.pdf

96. Rush SA, Soehren EC, Stodola KW, Woodrey MS, Cooper RJ. Influence of Tidal Height on Detection of Breeding Marsh Birds Along the Northern Gulf of Mexico. Wilson J Ornithol. 2009; 121: 399–405.

97. Twedt DJ, Crawford RD. Yellow-headed Blackbird (Xanthocephalus xanthocephalus) [Internet]. The Birds of North America. 1995 [cited 2018 Jun 15]. Available from: https://birdsna.org/Species-Account/bna/species/yehbla/introduction

98. Kroodsma DE, Verner J. Marsh Wren (Cistothorus palustris) [Internet]. The Birds of North America. 2013 [cited 2018 Jun 15]. Available from: https://doi.org/10.2173/bna.308

99. Murkin HR, Murkin EJ, Ball JP. Avian habitat selection and Prarie wetland dynamics: A 10 year experiment. Ecol Appl. 1997; 7: 1144–1159.

100. Driver EA. Chironomid communities in samll prairie ponds: some characteriestics and controls. 1977; 7: 121–133.

101. Guzy MJ, Ritchison G. Common Yellowthroat (Geothlypis trichas) [Internet]. The Birds of North America. 1999 [cited 2019 May 1]. Available from: https://doi.org/10.2173/bna.448

102. Yasukawa K, Searcy WA. Red-winged Blackbird (Agelaius phoeniceus) [Internet]. The Birds of North America. 2019 [cited 2019 May 1]. Available from: https://doi.org/10.2173/bna.rewbla.02

103. Mowbray TB. Swamp Sparrow (Melospiza georgiana) [Internet]. The Birds of North America. 1997 [cited 2019 May 1]. Available from: https://doi.org/10.2173/bna.279


Článek vyšel v časopise

PLOS One


2020 Číslo 1
Nejčtenější tento týden
Nejčtenější v tomto čísle
Kurzy Podcasty Doporučená témata Časopisy
Přihlášení
Zapomenuté heslo

Zadejte e-mailovou adresu, se kterou jste vytvářel(a) účet, budou Vám na ni zaslány informace k nastavení nového hesla.

Přihlášení

Nemáte účet?  Registrujte se

#ADS_BOTTOM_SCRIPTS#