The effectiveness and cost efficiency of different predator exclosure designs to increase piping plover (Charadrius melodus) nest success and fledging rate in Alberta, Canada

An estimated 30% of the North American piping plover (Charadrius melodus) breeding population occurs in Canada, where it is Endangered. Predator exclosures are a common management tool across the species' range to increase nest success. Using data from 1998 to 2010 on 820 nests, collected as part of an ongoing management program in Alberta, we compare daily nest survival (DNS), numbers of chicks hatched and fledglings produced, and cost/chick among three treatments (large, medium, small exclosures) and natural nests. During the early period (1998–2001), when all three types of exclosures were applied, there were no significant differences in DNS between exclosed nests and natural nests. During the late period (2002–2010), when only small exclosures were applied, nests with small exclosures had a significantly higher DNS rate ( x¯ = 0.994, n = 594) compared with natural nests ( x¯ = 0.984, n = 88). Nests with small exclosures also hatched more chicks ( x¯ = 3.21, n = 598) and produced more fledglings ( x¯ = 1.17, n = 337) than natural nests ( x¯ = 1.73 chicks, n = 31; x¯ = 0.59 fledglings, n = 21) during the late period. However, considering only successful nests, the differences were not significant for either period, indicating no added benefit of exclosures beyond protecting the nest. The cost/chick rate was lowest using small exclosures (cylindrical, 40 by 60 cm), and this portable design is well‐suited for widespread application in the field. We demonstrate increased DNS for nests with small exclosures and a mean fledging rate close to the goal of 1.25 chicks/nest/year, although this did not increase Alberta's piping plover population beyond the timeframe of systematic exclosure application. Long‐term, cost‐efficient management tools that increase nest success and fledging rates should continue.


| INTRODUCTION
The piping plover (Charadrius melodus) is a ground-nesting, migratory shorebird endemic to North America (Elliott-Smith & Haig, 2020). They spend the winter along the southeastern Atlantic seaboard and the coast of the Gulf of Mexico, including some Caribbean islands, migrating north in the spring to breed (Elliott- Smith & Haig, 2020). An estimated 30% of the North American breeding population of piping plovers occurs in Canada (Alberta Piping Plover Recovery Team, 2010). In Canada, the species was designated as Threatened in 1978 and then upgraded to Endangered in 1985 by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) (COSEWIC, 2003;Goossen et al., 2002). Two subspecies (Atlantic piping plover, C. m. melodus and interior piping plover, C. m. circumcinctus) were each listed as Endangered under Schedule 1 of Canada's Species At Risk Act in 2003 (Government of Canada, 2021). The piping plover was first listed in the United States in 1985: Endangered in the Great Lakes (Midwest) region and Threatened elsewhere throughout its breeding range (U.S. Fish and Wildlife Service, 2021). The province of Alberta, Canada, listed the piping plover as Endangered under Alberta's Wildlife Act in 1987 and it remains Endangered today (Alberta Queen's Printer, 2020;Prescott, 1997). An International Piping Plover Census has been conducted every 5 years since 1991 to better understand population numbers and trends across the species' range (Elliott-Smith et al., n.d.). During the early years of the census program, several North American piping plover populations showed continued declines (Haig et al., 2005), including Alberta where the population dropped to 150 individuals in 2001 (Prescott, 2001). Subsequent censuses in 2006 (274 individuals) and 2011 (244 individuals) showed a short-term increase in Alberta (Elliott-Smith et al., 2015). The most recent (2016) census identified a low of 123 piping plovers in Alberta, despite a rangewide increase in the population (Elliott-Smith et al., n. d.). Alberta's population of piping plovers has shown a declining trend since around 2008.
The continued decline of piping plover populations across their range is due to low nest productivity as a result of local threats, which may include predation and damage to nesting habitat from flooding, vegetation encroachment, property development, recreational vehicles, and livestock grazing (Environment Canada, 2006;Espie et al., 1998;Flemming et al., 1988;Patterson et al., 1991;Root & Ryan, 2004). In the Northern Great Plains, and particularly in Alberta, predation of eggs from mammalian and avian predators has been identified as a major limiting factor to piping plover reproductive success (Environment Canada, 2009;Ivan & Murphy, 2005;Kruse et al., 2002;Murphy et al., 2000;Prindiville Gaines & Ryan, 1988;Richardson, 1999;Westworth et al., 2004;Whyte, 1985). Predator removal, electric fencing, and predator exclosures have been proposed as methods to reduce nest predation on piping plovers (Schmelzeisen et al., 2004). A common management action for piping plovers across their distribution has been the use of predator exclosures to increase nest success (see table 2 in Gratto-Trevor & Abbott, 2011), though some studies have identified increased nest abandonment and adult mortalities associated with exclosures (Barber et al., 2010;Cohen et al., 2016;Murphy, Michaud, et al., 2003). Predator exclosures are wire mesh cages placed over the eggs to protect them from predators, while still allowing the adults to move freely. Previous studies in Alberta have shown that exclosures significantly reduce nest predation (Heckbert & Cantelon, 1996;Richardson, 1999). Similar success stories have been reported in other Northern Great Plains jurisdictions (Johnson & Oring, 2002;Kruse et al., 2002;Larson et al., 2002;Melvin et al., 1992;Murphy, Greenwood, et al., 2003;Rimmer & Deblinger, 1990). However, Johnson and Oring (2002) cautioned that not all exclosure designs are created equally. They recommended monitoring and evaluating the effectiveness of different exclosures under a rigorous study design including the use of control (natural) nests. In addition, while it has been shown that exclosures effectively reduce nest predation, their actual benefit to the population (i.e., fledglings produced) is largely unknown (Smith et al., 2011).
In Alberta, the use of predator exclosures to improve piping plover nest success began in 1994 (Richardson, 1999). Between 1998 and 2010, Alberta Conservation Association (ACA), in partnership with the Government of Alberta (GOA), systematically applied exclosures to as many piping plover nests as possible across the Alberta range; after 2010, the GOA continued applying them opportunistically at a much smaller scale. Modifications were made to the design of the exclosures over the years, which afforded us the opportunity to evaluate exclosure effectiveness at increasing nest success and assess the cost efficiency of the different designs. As a management-dependent species (Environment Canada, 2006), evaluating the effectiveness of piping plover management tools in the context of cost efficiency is extremely valuable.
Historically, there have been few studies that integrated economic costs into conservation planning and achievements, but this is slowly changing (Busch & Cullen, 2008;Engeman et al., 2002Engeman et al., , 2003Lindsay et al., 2005;Shwiff et al., 2005). For example, Hecht and Melvin (2008) estimated that inflation-adjusted expenditures for the Atlantic coast piping plover increased by 51% between 1993 and 2002 (from US $2.28 million to $3.44 million), but overall annual per-pair expenditures declined by 4% with paid-staff effort being similar between years. Using a stochastic population simulation model, Larson et al. (2003) determined a minimum cost of US $1-$11 million over 50 years was required to achieve a 20% chance of population stabilization for the piping plover. Underfunded species recovery programs (Buxton et al., 2020;McCarthy et al., 2012) place a premium on ensuring recovery actions are cost-effective while returning maximum biological value for dollars invested (Myers et al., 2000;Naidoo et al., 2006).
In 2001, the Alberta Piping Plover Recovery Team was formed and drafted the first recovery plan for Alberta (2002)(2003)(2004), with subsequent editions in 2005 and 2010 (Alberta Piping Plover Recovery Team, 2010). All three provincial recovery plans and Canada's Recovery Strategy (Environment Canada, 2006) describe a variety of actions necessary to achieve the recovery goal of a well-distributed, long-term average provincial population of 300 adult birds, no net loss of breeding habitat in the province, and a median fledging rate of 1.25 chicks/pair/year (Larson et al., 2002). Four of the nine principles guiding the recovery of piping plovers in Alberta are relevant to the research presented in this paper and include using an adaptive management approach to maximize effective tools with immediate benefits to piping plovers.
In this paper, we compared nest success (daily nest survival; DNS), number of chicks hatched, number of fledglings produced, and the cost of producing a piping plover chick among three treatment types (large, medium, and small predator exclosures) and a control (natural nest with no exclosure) for nests monitored over 13 years in Alberta, Canada. Treatments were not applied equally across the study period, so to reduce the potential influence of temporal bias we analyzed our data in two time periods (early [1998-2001], late [2002][2003][2004][2005][2006][2007][2008][2009][2010]). It has been inferred that exclosures not only increase nest success but also benefit the chicks and/or increase fledging success (Larson et al., 2002). This added benefit has not been tested in Canada (but see Anteau et al., 2021); therefore, we compared the number of chicks and fledglings produced per successful nest for each treatment type and natural nests. In addition, we compared the fledging rates achieved using exclosures to see if exclosing nests allowed us to reach the Alberta recovery goal of a median 1.25 fledglings/pair/year (Alberta Piping Plover Recovery Team, 2010;Environment Canada, 2006). We also assessed if the cost to produce a chick was lower for exclosed nests and therefore justified the added effort and cost to install and remove exclosures. We predicted that all three exclosure treatments function similarly to protect nests, resulting in no difference in nest success (DNS) among treatments, but that the exclosed nests would have a higher DNS than natural nests. The presence of humans (e.g., Environment Canada, 2006;Flemming et al., 1988;Patterson et al., 1991;Prescott, 1997) and cattle (e.g., Environment Canada, 2006;Prescott, 1997;Prindiville Gaines & Ryan, 1988) have been shown to affect nest survival; therefore, we predicted that DNS would be lower for nests in habitats with cattle and human disturbance. We predicted that fledging rate (i.e., the number of fledglings produced per nest) would be higher for exclosed nests than for natural nests but fledging success (i.e., the number of chicks that fledge per egg hatched) would be the same for all successful nests. If exclosed nests were more successful, we predicted they would fledge enough chicks to reach the Alberta recovery goal of median 1.25 fledglings/pair. Lastly, we predicted that the cost to produce a chick would be similar across exclosure types but lower than the cost to produce a chick from natural nests, justifying the use of exclosures. Using a long-term and large database of nests, this paper demonstrates the value in scientifically evaluating management practices under an adaptive management program, to ensure cost-efficient benefits for species at risk are achieved.

| Piping plover management area
Alberta's piping plover management program focused on naturally alkaline waterbodies with open, gravelly beaches in the Parkland Natural Region of east-central Alberta (Natural Regions Committee, 2006), where most of the province's piping plovers are known to breed ( Figure 1). Nesting typically occurs on gravel substrates within relatively wide, sparsely vegetated beaches. The availability of suitable nesting substrates depends on variations in water level; periodic high-water events restrict beach width and temporarily limit the availability of nesting habitat until water levels recede and expose gravel deposits (Alberta Piping Plover Recovery Team, 2010). Piping plovers move around from year to year based on fluctuating water levels and available habitat. In any given year, they occur on 20-25 lakes in Alberta, with the majority on 3-6 lakes (Alberta Piping Plover Recovery Team, 2010). The lakes vary in size from smaller than 1 km 2 to approximately 78 km 2 . The lakes used in our analysis had no habitat management techniques employed.

| Nest searches and monitoring
We located nests annually (1998-2010) by surveying potential breeding habitat for piping plovers displaying territorial, courtship, or predator distraction behaviors (Cairns, 1982). Nests were located in mid-May through mid-July (as per Murphy et al., 1999) during egg-laying and incubation, and assigned into one of two categories: natural (no exclosure) or exclosed (predator exclosure installed). We installed most exclosures the same day as the nest was discovered. Active nests were not approached unless we were installing an exclosure.
We monitored most nests weekly, from 50 to 100 m away to avoid disturbing incubating adults and attracting potential predators to nests, as per the guidelines in Richardson (1999). We only approached nests when there was no sign of activity or when contents needed verification. We recorded nest fate as successful, failed, or unknown. Cause of failure was recorded for all failed nests. Based on proximate evidence at the nest site, we classified cause of failure to 1 of 8 categories: adult depredation (e.g., feathers or adult carcass present), nest depredation (e.g., canine digging under exclosure), predation (not specified), cattle (e.g., tracks nearby or trampled exclosure), human disturbance (e.g., all-terrain vehicle [ATV] activity or crushed exclosure), unexplained abandonment, weather (e.g., local heavy rain/hail event), or unknown. In some cases, we had photo evidence from wildlife cameras at the nest site that allowed us to classify nest failure as the most proximate cause; for example, a failed nest with ATVs in the photo was classified as human disturbance, even though it was possible it was abandoned for other reasons. We considered a nest to be successful if one or more eggs hatched (Johnson & Oring, 2002;Melvin et al., 1992;Murphy et al., 1999;Murphy, Greenwood, et al., 2003). Exclosures were removed during the visit that nest fate was determined if young were not at the nest; otherwise, they were removed at a subsequent visit and usually within a week of determining nest fate.
We determined the number of chicks hatched based on eggs that were "missing" from the nest bowl during the expected hatch period and the number of young seen in the area. We monitored fledging success for a subset of our nests. Most chicks were monitored weekly, first by searching the nest site following the hatching of the eggs, and subsequently in the area where broods were last observed. We banded almost all young from lakes during our program (700 + ) with a single band combination for all chicks on a given lake in a given year. As a result, many of the returning adults at nests were banded, allowing us to separate some nests/broods from others. In addition, some chicks were banded before others to allow us to differentiate them from each other in a given year. Unlike some jurisdictions, the piping plover lakes in F I G U R E 1 East-central Alberta piping plover predator exclosure management program area with major water bodies, water courses, and key urban centers.
Alberta are relatively small and habitat is widely dispersed. It is unusual to come across multiple nests of the same age in close proximity, so brood mixing was not a significant issue. We monitored almost all nests and broods at least weekly and had a very good estimate of the age of each nest, hatching dates, and age of the unfledged chicks. This allowed us to be very confident in the age of the broods and, ultimately, how many chicks in a given brood survived to fledge. If our confidence was low in the number of chicks that survived from a particular nest, we removed that nest from our chick and fledgling analysis. This could occur if we were unable to return to a nest site during the time in which we expected the birds to fledge (i.e., we were unsure if the young were preyed upon or flew away, a problem that would occur with marked broods as well), or if an unknown brood (i.e., from a missed nest of similar age) produced chicks in the same area at the same time as a known nests. Therefore, our fledging metrics are likely underestimated because of the difficulty in keeping track of mobile chicks; however, we expect this underestimation to be random across all nests monitored for fledging.

| Predator exclosure designs and installation
Over the 13 years of this program, we refined the design of our predator exclosures to make deployment more efficient, while maintaining effectiveness in preventing egg predation and being mindful of adult piping plover predation at exclosures (Figure 2a-d; Murphy, Michaud, et al., 2003). As such, nests were not randomly assigned to treatments or controls across years or lakes, and sample sizes were not intentionally regulated.
The first treatment type, hereafter referred to as "medium" exclosures ( Figure 2a), was used from 1998 to 2000 and was the same as those used by Richardson (1999). They were square-pyramidal in shape, made of four, 5 cm Â 5 cm stucco wire mesh panels with a bottom F I G U R E 2 Depiction of three predator exclosure designs and a piping plover natural nest used in Alberta, Canada, 1998Canada, -2010: medium exclosure (a), large exclosure (b), small exclosure (c), and a natural nest (d). Medium exclosures were square-pyramidal in shape, made of four, 5 cm Â 5 cm stucco wire mesh panels with a bottom width of 1.2 m, a top width of 60 cm, and a height of 1.2 m. To protect against aerial predators, the top was woven with sisal twine. The large exclosures were 1.2 m high and 3 m in diameter. Small exclosures were 40 cm high and 60 cm in diameter. Both large and small exclosures were cylindrical in shape and made of 5 cm Â 5 cm stucco wire. To deter predatory birds from perching on both large and small exclosures, the horizontal wire along the top of each exclosure design was removed to expose the vertical wires. The top of the large exclosures were covered with 2 cm Â 2 cm plastic bird netting, while the tops of the small exclosures were covered with the same stucco wire used for the sides. To provide stability to medium and large exclosures, 1.5-m lengths of rebar were woven through the wire and inserted into the substrate. The bottoms of each of the three exclosure designs were secured in place by inserting 25 cm nails, bent at the top, through the exclosure and into the substrate at even intervals around the perimeter. width of 1.2 m, a top width of 60 cm, and a height of 1.2 m. To protect against aerial predators, the top was woven with sisal twine. The second treatment type, hereafter referred to as "large" exclosures ( Figure 2b), was used from 1999-2001 and was similar to the exclosures used by Melvin et al. (1992). The large exclosures were 1.2 m high and 3 m in diameter. The third treatment type, hereafter referred to as "small" exclosures ( Figure 2c), was used from 2001 to 2010. Small exclosures were 40 cm high and 60 cm in diameter (Engley & Prescott, 2005). Both large and small exclosures were cylindrical in shape and made of 5 cm Â 5 cm stucco wire. To deter predatory birds from perching on both large and small exclosures, the horizontal wire along the top of each exclosure design was removed to expose the vertical wires. The top of the large exclosures were covered with 2 cm Â 2 cm plastic bird netting, while the tops of the small exclosures were covered with the same stucco wire used for the sides. To provide stability to medium and large exclosures, 1.5-m lengths of rebar were woven through the wire and inserted into the substrate. The bottoms of each of the three exclosure designs were secured in place by inserting 25 cm nails, bent at the top, through the exclosure, and into the substrate at even intervals around the perimeter. Our controls, hereafter referred to as "natural" nests ( Figure 2d), did not have exclosures applied. Every year of the program had some nests that were never exclosed. The installation of exclosures was approved by the GOA at the provincial level and through the Government of Canada at the federal level.
For each exclosure design, we generally followed the installation techniques outlined by Richardson (1999). We installed exclosures and secured them to the substrate as quickly as possible, and then monitored the nest for the resumption of incubation from a distance of at least 100 m.

| Nest success (daily nest survival)
We used generalized linear mixed models (glmer) with a logistic-exposure link function and the "bobyqa" optimizer to estimate the daily nest survival rate (DNS). The logistic exposure model was used to account for differences in exposure periods for individual nests, based on when each nest was found and when each treated nest received an exclosure (Shaffer, 2004). For the small number of nests that did not receive their exclosure the same day they were found, we created a separate record in the database for the time the nest was in a natural state (all survived) and a separate record for the same nest following the installation of the exclosure until nest fate was determined. This allowed us to model the exposure days while the nest was in a natural state and while it was under an exclosure.
We classified nests with one or more eggs hatched as successful/surviving (coded as 1) and the remaining nests as failed (coded as 0) for our response variable. We considered nest age, estimated nest initiation day, treatment type, presence of cattle, and presence of humans (i.e., foot traffic, ATVs, industrial activity) as explanatory covariates. Nest age was the number of days prior to hatch; however, we report nest age as days since clutch initiation, as this is more intuitive. We standardized the initiation day of the nest to January 1 of each year. We standardized both nest age and nest initiation day by subtracting the mean and dividing by two standard deviations (Gelman, 2008). We calculated nest initiation dates several ways to get the most accurate date possible. For nests found without a full clutch, but for which a maximum clutch size was known, initiation dates were calculated based on the number of eggs when found then back-dating assuming eggs were laid on alternate days. For successful nests found with a full clutch, initiation dates were calculated by back-dating 34 days from the hatch date; hatch date was calculated based on estimated chick age or by taking the mid-point between the last two visits if no chicks were found but there was evidence the nest had hatched. For failed nests found with a full clutch, initiation dates were calculated as the mid-point between 35 (i.e., maximum incubation days) and twice the number of eggs (i.e., number of laying days). Before the analysis, we converted the estimated nest initiation date to day of the year (i.e., January 1 = 1) using the Gregorian calendar to account for leap years. We also assessed treatment type (small, medium, or large exclosure, or natural [no exclosure]) and two disturbance covariates (presence of cattle and humans). During each visit to a nest, we recorded the presence of domestic livestock (physical or sign [i.e., cow patties, tracks]) and the presence of human activity (physical or sign [e.g., ATV tracks]) on the beach. Our analysis treated the presence of cattle and human activity as categorical (yes/no), with yes indicating that during at least one visit we detected cattle or human activity.
As exclosure deployment was not random across years or on lakes, we included a random term for each in our mixed models. Including year and lake as random effects accounted for the spatiotemporal dynamics of predation (Anteau et al., 2021). Year was treated as a factor to account for the fact that nest success on a particular lake in one year did not influence nest success on the same lake in the next or subsequent years. To further account for the potential non-random effects of how exclosures were distributed, we divided our data into two time periods and ran our models separately for each dataset: the early period (1998)(1999)(2000)(2001) included natural nests (all years) and nests with small (2001), medium (1998)(1999)(2000), and large (1999)(2000)(2001) exclosures; the late period (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) included natural nests and nests with small exclosures in each year.
Prior to assessing support for competing models, we calculated variance inflation factor (VIF) values for each of the five covariates to assess issues of multicollinearity and considered values over 5.0 to be significant (Thompson et al., 2017). We used Akaike's Information Criteria (AIC) to evaluate the support among models, with ΔAIC <2.0 as the cut-off to compare competing top models (Burnham & Anderson, 2002). We used a best subsets regression approach (Grueber et al., 2011) using the dredge function in Program R version 3.5.1 (R Core Team, 2018) package MuMIn (Barton, 2018) to compare all possible combinations of covariates from the global model. If we determined that more than one model was competitive, we report the full model averaged β coefficients plus their respective 85% confidence interval (CI) (Grueber et al., 2011).
In the logistic exposure model, when categorical covariates are included in the model, the β coefficient (direction and size) for each level of the categorical covariate are only in relation to the reference category and do not indicate how the other levels compare to each other. As we were particularly interested in comparing the DNS rates for each of the treatment types, we used a multiple comparison (Tukey) to account for the proper alpha level when comparing proportions to estimate pairwise-effect sizes among levels of the categorical covariate using the emmeans package (Lenth, 2018) on the probability scale. Estimates of the probability were obtained for each category of the treatment covariate in the model, with the other covariates in the top model being averaged. Likelihood methods generate an estimate and standard error (SE) on the logit scale, with the 95% CIs calculated on the logit scale by taking the estimate ±1.96 SE. We then estimated the DNS probability as the anti-logit of the logexp (1) scale and the 95% CI bounds and report these values. We used the higher-ranked model that contained the treatment covariate from the suite of competing models in the multiple comparison. If treatment was not a covariate in the top model(s), we used a model containing only treatment as a fixed effect in the multiple comparisons.

| Reproductive success
We used four metrics of reproductive success for the nests monitored: chick hatch rate (i.e., number of eggs hatched per nest [successful and failed]), chick hatch success (i.e., number of eggs hatched per successful nest), fledging rate (i.e., number of fledglings produced per nest [successful and failed]), and fledging success (i.e., number of chicks that fledged per egg hatched). Since our metrics are discrete data and we were only interested in the effect of treatment on our four metrics, we used generalized linear mixed models (glmmTMB) with either a Poisson or negative binomial distribution with treatment as the fixed effect and year and lake as random effects using the glmmTMB package (Brooks et al., 2017). We calculated the variance-mean ratio and considered values greater than 1 as having a negative binomial distribution and values at $1 as having a Poisson distribution. We used a multiple comparison (Tukey) test to account for the proper alpha level when comparing proportions to estimate pairwise-effect sizes among levels of the categorical covariate treatment using the emmeans package (Lenth, 2018) on the response scale.
We present data on chicks/nest in this paper rather than chicks/pair; it is worth noting that there is likely some polygamy in piping plovers in Alberta because it has been documented elsewhere (Halimubieke et al., 2020;Hunt et al., 2015). We considered a chick to be fledged if it reached 20 or more days of age after hatching. For our analysis of fledging rate, we used the subsample of chicks monitored from successful and failed (fledglings = 0) nests to determine if there was a difference in productivity per nest between the treatments. We then analyzed fledging success between treatments using only the subsample of chicks monitored from successful nests. Assessing differences in fledging success allowed us to control for the difference in nest success rates. We completed our analysis splitting the data into early (1998)(1999)(2000)(2001) and late (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) periods to be consistent with the DNS analysis.

| Cost efficiency
We used internal ACA annual financial reports to calculate a cost per chick rate for each treatment and natural nests. See Table S1 for details on the steps taken to calculate these rates. We included actual costs for salary, fuel, meals, accommodation, field camp expenses, etc. during the egg-laying and hatching period (1 May to 30 June), and estimated costs for building exclosures in 2021 CDN dollars (large exclosure = $100.00, medium = $57.00, small = $5.50). Exclosure costs were calculated with two main assumptions: materials were purchased in bulk so that many exclosures could be built at one time; each exclosure was used only for 1 year. We did not add any costs associated with material for natural nests. To standardize dollar expenditures between years we adjusted for inflation using the consumer-price-index-inflation calculator to 2021 Canadian dollars (Bank of Canada, 2021).
Our cost estimates are conservative, as in-kind support from partner agencies has not been included because of the difficulty with quantifying it. We acknowledge there is likely error associated with expenses spanning beyond the 1 May to 30 June timeframe (e.g., seasonal truck rental); however, these errors would be similar across all treatment types and years. The true cost of a natural nest would be zero dollars; however, to determine if natural nests were successful in producing chicks, there would be costs associated with finding and monitoring these nests. We only used data from lakes monitored by ACA staff to calculate the cost per visit. To calculate the cost per chick, we used the nests monitored as part of the DNS analysis because we were confident in the outcome T A B L E 1 Full model-averaged regression coefficients and 85% confidence interval for the covariates in the top models for the early (1998)(1999)(2000)(2001)  F I G U R E 3 Predicted daily nest survival (DNS) probability for piping plover nests relative to nest age (a) and estimated nest initiation day (day 114 = April 24, day 162 = June 11; b) during the early period (1998)(1999)(2000)(2001) in Alberta, Canada. DNS rates based on the model averaged regression coefficients from a logistic-exposure model where we held the covariates Human Presence and Cattle Presence constant at Yes, the other standardized covariate at its mean, and set exposure to 1. Gray-shaded area represents the egg-laying stage (0-6 days). A nest that had ≥1 egg hatch was considered successful. Note that the y-axis scale differs between the graphs.
of those nests and the number of chicks that each nest hatched. Some of the nests within the DNS database were monitored by staff from GOA; in these instances, we assumed that the cost to monitor a nest by ACA or GOA would be equivalent. If during a given year for a particular treatment type there were no chicks produced, we set the number of chicks to one to alleviate issues with dividing by zero when calculating cost per chick. We compared the mean cost per chick per treatment type across years using an ANOVA to see if applying one treatment type resulted in a different cost per chick. If we detected an effect of treatment type on cost per chick, we used the Tukey multiple comparison to see which treatments differed.

| RESULTS
Between 1998 and 2010, we found a total of 1079 piping plover nests. Of these 1079 nests, we had estimated initiation dates for 871 nests. Across all years, the mean nest initiation date was May 23 (SE = 0.38 days; n = 871) and the mean hatch date was June 26 (SE = 0.39 days, n = 807).
Of the 1079 nests, we had complete data (i.e., initiation and hatch dates, eggs hatched) for 820 nests from 28 lakes, which we used in our DNS analysis. For nests with small, medium, large, and no exclosures, 88%, 38%, 59%, and 46%, respectively, were successful (i.e., hatched ≥ one egg). Predominant causes of nest failure documented in the field were unexplained nest abandonment and nest depredation for small exclosures, adult depredation for medium exclosures, unexplained nest abandonment for large exclosures, and nest depredation for natural nests (Table S2).

| Daily nest survival
Eighty-six of the 820 nests with complete data were considered natural until they received an exclosure following a period of time after they were initially found (X = 3.27 days, SE = 0.38, Min = 1 day, Max = 16 days). Therefore, we modeled DNS for 906 "nests." The particular lakes included in the DNS analysis varied annually and ranged from 7 lakes in 2001 to 19 in 2009 (  (1998)(1999)(2000)(2001). During modeling, we determined that none of the covariates displayed multi-collinearity (VIF values <1.5). There were four top models with ΔAIC <2.0 (Table S4). The next best model had ΔAIC = 2.20 and was not considered as a competitive model. The top models contained 64% of the model weight and contained varying combinations of four of the F I G U R E 4 Predicted daily nest survival (DNS) probability for nests with small exclosures and natural piping plover nests relative to nest age (a) and estimated nest initiation day (day 116 = April 26, day 176 = June 25; b) during the late period (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) in Alberta, Canada. DNS rates based on the model-averaged regression coefficients from a logistic-exposure model where we held the covariates Human Presence and Cattle Presence constant at Yes, the other standardized covariate at its mean, and set the exposure to 1. Gray-shaded area represents the egg-laying stage (0-6 days). A nest that had ≥1 egg hatch was considered successful. Note that the y-axis scale differs between the graphs. five covariates (Table S4). Treatment type was not in any of the top models. Table 1 provides the full model averaged regression coefficients plus their respective 85% CI. As expected, our multiple comparison of treatment type using just treatment type as the only fixed effect in a mixed model revealed that there was no difference in DNS between treatment types and natural nests ( Figure S1A). Overall, nests in the early period had a probability of surviving to day 35 of 78% ( Figure 3a). Nests with a later initiation day had a higher DNS probability than those with an earlier day (Figure 3b).
We modeled 8698 daily intervals representing 682 nests for the late period (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010). During modeling, we determined that none of the covariates displayed multicollinearity (VIF values <1.1). There were three top models with ΔAIC <2.0 (Table S4). The next best model had ΔAIC = 2.38 and was not considered as a competitive model. The top models contained 62% of the model weight and contained all five covariates. Table 1 provides the full model-averaged regression coefficients and their respective 85% CI. Using the top model that contained four covariates (holding the others at their mean) revealed that there was a significant difference in DNS between nests with small exclosures (x = 0.994, n = 594) and natural nests (x = 0.984, n = 88; Figure S1B). Overall, nests in the late period had a probability of surviving to day 35 of 92%, and nests with small exclosures had a higher probability of surviving (96%) than natural nests (89%) (Figure 4a). Nests with a later initiation day had a higher DNS probability than those with an earlier day (Figure 4b), though the 85% CI around the beta coefficient overlapped 0 indicating it is a non-informative covariate (Table 1).

| Number of chicks produced
The mean hatch rate ranged from 1.43 (± 0.19 SE) chicks/nest for medium exclosures to 3.21 (± 0.05 SE) F I G U R E 5 Mean and 95% confidence interval for the number of piping plover chicks hatched for all nests (a and b) and for nests that survived (c and d), for natural and treated (small, medium, or large exclosure) nests in Alberta, Canada, 1998Canada, -2010. Graphs in the left column represent nests during the early period (1998)(1999)(2000)(2001), while those in the right column represent nests during the late period (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010). N-values are the number of nests. Within each graph, similar letters above points indicate no significant difference based on multiple pair-wise comparisons. Results are based on a generalized linear mixed model where the fixed effect was treatment (natural; small, medium, or large exclosure) and the random effects were Year and Lake, except for nests that survived during the early period where only Year was used as a random effect to avoid singularity issues in our model. Note that the y-axis scale differs between the graphs. chicks/nest for small exclosures, across all years of the management program. A mean of 3.67 chicks hatched from successful nests across all years of the study. There was no difference in the mean hatch rate between the four treatments during the early period ( Figure 5a); however, there was a significant difference in hatch rates between nests with small exclosures (x = 3.21, n = 598) and natural nests (x = 1.73, n = 31) during the late period (Figure 5b), with small exclosure nests hatching 85% more chicks than natural nests. When we examined hatch success, there was no difference between the four treatments during the early period (Figure 5c) or between the two treatments during the late period (Figure 5d).

| Number of fledglings produced
The mean fledging rate ranged from 0.29 (± 0.11 SE) fledglings/nest for medium exclosures to 1.34 (± 0.08 SE) fledglings/nest for small exclosures across all years of the management program. A mean of 1.77 fledglings were produced from successful nests across all years of the management program. There was no difference in the mean fledging rate between the four treatments during the early period ( Figure 6a); however, there was a significant difference in fledging rates between nests with small exclosures (x = 1.17, n = 337) and natural nests (x = 0.59, n = 21) during the late period (Figure 6b), with small exclosures producing 98% more fledglings than natural nests. When we examined fledging success, there was no difference between the four treatments during the early period (Figure 6c) or between the two treatments during the late period (Figure 6d).

| Cost efficiency
An estimated $425,538.48 CDN dollars (2021-dollar values) was spent by ACA on field season (1 May-30 June) recovery efforts for piping plovers between 1998 F I G U R E 6 Mean and 95% confidence interval for the number of piping plover fledglings produced for all nests (a and b) and for nests that survived (c and d), for natural and treated (small, medium, or large exclosure) nests in Alberta, Canada, 1998Canada, -2010. Graphs in the left column represent nests during the early period (1998)(1999)(2000)(2001), while those in the right column represent nests during the late period (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010). N-values are the number of nests. Within each graph, similar letters above points indicate no significant difference based on multiple pair-wise comparisons. Results are based on a generalized linear mixed model where the fixed effect was treatment (natural; small, medium, or large exclosure) and the random effects were Year and Lake, except for nests that survived during the early period where only Year was used as a random effect to avoid singularity issues in our model. Note that the y-axis scale differs between the graphs. and 2010 (Table S5). The average cost per nest visit ranged from a low of $40.06 in 2000 to a high of $147.50 in 2001 (Table S5). Including costs associated with monitoring and materials for exclosures, the average price to produce one piping plover chick was $359.63 (SE = $80.49),$555.84 (SE = $401.91),and $182.87 (SE = $42.78),for nests with large,medium,or small exclosures,respectively,and $194.56 (SE = $34.24) for natural nests. The differences in the yearly cost per chick rates between the treatments were not statistically significant (F value = 2.40, p = .09, df = 3).

| DISCUSSION
With almost one-third of the piping plover population breeding in Canada, our management actions play an important role in the species' population trends and need to be evaluated regularly. Furthermore, cost-effective options are necessary for the sustainability of recovery efforts for this management-dependent species (Environment Canada, 2006;Gratto-Trevor & Abbott, 2011). A common management action to reduce predation of eggs, a known limiting factor, has been the use of predator exclosures to improve nest success (Barber et al., 2010;Cohen et al., 2016;Kruse et al., 2002;Larson et al., 2002;Melvin et al., 1992;Murphy, Greenwood, et al., 2003;Rimmer & Deblinger, 1990). However, different exclosure designs need to be evaluated to ensure they are achieving their intended purpose of increasing nest success (Johnson & Oring, 2002;Larson et al., 2002). Our results demonstrate that two of our three predator exclosure designs did not have different DNS rates than natural nests and were ineffective. Only our small exclosures (40 cm high, 60 cm diameter, 5 cm Â 5 cm stucco wire on top and sides) had significantly higher DNS rates than natural nests, and is therefore our recommended exclosure design for increasing piping plover nest success. We also emphasize the practicality of these small exclosures; each field biologist can easily carry at least four at a time. For context, 88% of nests with small exclosures in our management program were successful (at least one egg hatched). This is higher than the 65% success rate of exclosed nests in a 22-year study of piping plovers in eastern Canada (Barber et al., 2010) and 68% averaged over 20 years of data for snowy plovers (Charadrius nivosus) in Oregon (Dinsmore et al., 2014). Our mean nest survival rate to 35 days for nests with small exclosures was higher, while that for natural nests (and nests with medium/large exclosures) was similar, compared with the rates reported by Cohen et al. (2016) for exclosed and natural piping plover nests on the Atlantic coast.
Of the three designs used during our management program, our small predator exclosures were the cheapest to produce and install. Although the cost per chick rate from nests with small exclosures was not significantly less than the other designs, our data and experience show that small exclosures can be deployed on a scale large enough to affect change in the population because they facilitated a fledging rate close to the goal of 1.25 chicks/ pair and the Alberta population generally increased during the years small exclosures were applied systematically (Alberta Piping Plover Recovery Team, 2010). The largest cost associated with small exclosures was the travel and salary associated with monitoring these nests because they generally survived longer. Our results show very high nest success and fledging rate by applying small exclosures. Taken together, this justifies our recommendation to use small exclosures and minimize the number of visits to nests to reduce monitoring costs and nest disturbance. Although adult predation was not a concern with our small exclosures (nor for the larger exclosures used in research conducted by Anteau et al., 2021), nests should still be monitored frequently enough to respond to unexpected adult predation episodes (Murphy, Michaud, et al., 2003).
The evolution of our exclosure design illustrates the need for adaptive management and continued monitoring to ensure deleterious effects of applied management do not hinder the recovery of an Endangered species (Johnson & Oring, 2002). During the initial years of our program (1998)(1999)(2000), we determined that medium-sized exclosures were having an unintended consequence of increased adult mortality (likely from merlins [Falco columbarius]; Michaud & Prescott, 1999), which resulted in lower DNS rates and a very low proportion of successful nests. We documented lower adult mortality at nests with large and small exclosures and at natural nests than we did at nests with medium exclosures (Table S2). Our adult mortality rates at medium-sized exclosures were greater than those previously reported (Murphy, Michaud, et al., 2003 [5% at exclosed nests, 0% for natural nests]; Barber et al., 2010 [6% at exclosed nests, 0.3% for natural nests]). We considered these adult loss rates to be unacceptable and changed the design of the exclosures to make them larger, believing this would address the problems. However, the use of large exclosures was also short-lived (1999-2001); they were labor intensive to deploy (also see Johnson & Oring, 2002), and cattle were attracted to them as scratching posts, possibly causing the large number of unexplained nest abandonments. In 2001, we changed the design once again to small exclosures, which were much easier to carry and install and virtually invisible on the beach, which we believed would address the cattle and adult predation issues. The sample size of small exclosures used in the early period likely was too small to truly test their effectiveness. As our late-period results indicate, only small exclosures resulted in a higher DNS rate and an increase in the hatch rate and fledging rate over that of natural nests.
Many jurisdictions promote the use of predator exclosures to assist in the recovery of piping plovers, under the assumption that they protect the eggs and increase fledgling numbers (Larson et al., 2003). However, there is a lack of knowledge as to whether exclosures have benefits beyond increasing DNS (but see Anteau et al., 2021); specifically, is there an increase in the number of chicks hatched and fledglings produced from exclosed nests (Johnson & Oring, 2002)? Removing the effect of nest success, our results indicate there is no difference in the number of chicks or fledglings produced per nest between the three types of exclosures and natural nests; if a nest is successful, the number of eggs that hatch and chicks that survive to 20 days (i.e., fledglings) are similar between natural and exclosed nests. Therefore, we found no added benefit of the exclosure beyond protecting the nest, likely because chicks leave the nest within several hours of hatching but remain within the general area (Barber et al., 2010;Wilcox, 1959). In comparison to long-term piping plover data from Prince Edward Island National Park, Canada (Barber et al., 2010), our data show higher hatching success but similar fledging success from successful exclosed nests. Even though exclosures do not appear to increase piping plover fledging success, their effectiveness at increasing chick hatching rate results in higher absolute numbers of chicks and consequently more fledglings, justifying our recommendation to use our small exclosure design. In fact, Alberta nests with small exclosures had a probability of surviving to day 35 of 0.96, while exclosed piping plover nests in a long-term experiment in the Northern Great Plains, United States, had cumulative survival of 0.73 (Anteau et al., 2021).
Our piping plover recovery program was focused on conservation and not designed with a research perspective (Richardson, 1999); nonetheless, the data we collected and our analysis methods allowed us to evaluate the effectiveness of exclosure designs. There are several things to note when interpreting our results. First, our sample sizes of nests with medium and large exclosures are small and lead to large confidence intervals. We felt it was important to include these results to inform conservation and allow the reader to interpret our results as they see fit. Secondly, exclosure deployment was not random or uniform across years and lakes leading to possible spatial and temporal biases (although we reduced the potential for temporal bias by splitting our data into two time periods and including random effects in a mixed model to assess daily nest survival); a true evaluation of exclosure effectiveness would have entailed a balanced study design (Richardson, 1999) that used all three exclosure types plus controls (natural nests) every year and evenly distributed across lakes. That was beyond the scope of our work; consequently, there are likely year effects (e.g., a late spring) embedded within the data that we could not account for. Thirdly, we have limited understanding of the effects of disturbance covariates (humans, cattle) on nest survival because they were only assessed at a presence/absence level; for example, one set of ATV tracks in plover habitat was considered the same as many sets of tracks, so we cannot determine if there is a difference in the effect on piping plovers. It is possible that all lakes included in this management program fell within a threshold level of disturbance making the true effect of this covariate undetectable. Lastly, the installation of an exclosure took more time (increasing with exclosure size) than initially visiting a natural nest. However, we feel the increased staff-time cost associated with exclosing a nest was balanced with subsequent visits to determine nest fate and fledgling success; exclosed nests were easy to find, whereas more time was required to find natural nests to determine their fate.
Simulation modeling has demonstrated that recovery goals for piping plovers could be reached in 30-50 years if reproductive rates of 1.24 fledglings per pair was achieved (Larson et al., 2003); more recent work (Cohen & Gratto-Trevor, 2011) suggests a productivity rate of 0.75 fledglings/pair for population recovery in prairie Canada. The Alberta Piping Plover Recovery Team (2010) and Canada's Recovery Strategy (Environment Canada, 2006) have adopted the higher fledging rate (rounded to 1.25) as a goal in the recovery of piping plovers. Although we assessed fledging rate per nest, we feel it is equivalent to fledging rate per pair. We interpret the goal of 1.25 fledglings per pair per year to represent all pairs (nests in our case), whether they successfully hatched chicks or not. Only our small exclosures came close to achieving this fledging rate, with a mean of 1.17 fledglings/nest, based on our data from 2002 to 2010. However, the Alberta recovery plan goal indicates a median value of 1.25 fledglings per nest and we were unable to calculate median fledglings/nest because we used the emmeans package to do the multiple comparisons. Nonetheless, we know that over 600 small exclosures were applied over 10 years (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) of our management program, and the Alberta piping plover population had an increasing trend throughout most of that time (Alberta Piping Plover Recovery Team, 2010; Figure 2). Since 2008, however, the piping plover population has had a declining trend in Alberta, likely exacerbated by the reduced number of small exclosures installed after 2010, as well as a large reduction in available nesting habitat due to high water levels in the southern part of the management area and vegetation encroachment in the north. Vegetation encroachment in plover habitat reduces the availability of gravel substrate for nesting.
Our research indicates that small predator exclosures (cylindrical, 40 by 60 cm), are cost-efficient and easily can be applied at a rate that increases piping plover nest success and achieves fledging rates close to the target of a median of 1.25 fledglings/nest, but should not be the only management tool employed or lead to unrealistic expectations of conservation success (Gratto-Trevor & Abbott, 2011). To achieve the goal of a long-term average population of 300 piping plovers throughout their historical range in Alberta, additional (and concurrent; e.g., Larson et al., 2003) management options, such as fencing out livestock and predator control, that protect feeding and breeding habitats and increase chick and fledgling survival rates need to continue (Cohen et al., 2016;Prescott et al., 2010).

AUTHOR CONTRIBUTIONS
Susan H. Peters conducted field work in some years of this recovery program and assisted with data analysis and writing the manuscript. Lance Engley is a member of the Alberta Piping Plover Recovery Team and the federal Prairie Piping Plover Recovery Team; he provided oversight to this recovery program, conducted field work each year, and assisted with production of the manuscript. Amanda Rezansoff conducted field work and assisted with data entry each year and edited the manuscript. David R. C. Prescott (retired) was the Chair of the Alberta Piping Plover Recovery Team and a member of the federal Prairie Piping Plover Recovery Team; he provided oversight to this recovery program as well as Alberta's overall piping plover management program, conducted field work each year, and edited the manuscript. Paul F. Jones analyzed the data and assisted with writing the manuscript. All authors provided critical feedback, edited the manuscript, and approved the final version.