Are whooping cranes destined for extinction? Climate change imperils recruitment and population growth

Summary Identifying climatic drivers of an animal population's vital rates and locating where they operate steers conservation efforts to optimize species recovery. The population growth of endangered whooping cranes (Grus americana) hinges on juvenile recruitment. Therefore, we identify climatic drivers (solar activity [sunspots] and weather) of whooping crane recruitment throughout the species’ life cycle (breeding, migration, wintering). Our method uses a repeated cross‐validated absolute shrinkage and selection operator approach to identify drivers of recruitment. We model effects of climate change on those drivers to predict whooping crane population growth given alternative scenarios of climate change and solar activity. Years with fewer sunspots indicated greater recruitment. Increased precipitation during autumn migration signified less recruitment. On the breeding grounds, fewer days below freezing during winter and more precipitation during breeding suggested less recruitment. We predicted whooping crane recruitment and population growth may fall below long‐term averages during all solar cycles when atmospheric CO2 concentration increases, as expected, to 500 ppm by 2050. Species recovery during a typical solar cycle with 500 ppm may require eight times longer than conditions without climate change and the chance of population decline increases to 31%. Although this whooping crane population is growing and may appear secure, long‐term threats imposed by climate change and increased solar activity may jeopardize its persistence. Weather on the breeding grounds likely affects recruitment through hydrological processes and predation risk, whereas precipitation during autumn migration may influence juvenile mortality. Mitigating threats or abating climate change should occur within ≈30 years or this wild population of whooping cranes may begin declining.


Summary
Identifying climatic drivers of an animal population's vital rates and locating where they operate steers conservation efforts to optimize species recovery. The population growth of endangered whooping cranes (Grus americana) hinges on juvenile recruitment. Therefore, we identify climatic drivers (solar activity [sunspots] and weather) of whooping crane recruitment throughout the species' life cycle (breeding, migration, wintering). Our method uses a repeated cross-validated absolute shrinkage and selection operator approach to identify drivers of recruitment. We model effects of climate change on those drivers to predict whooping crane population growth given alternative scenarios of climate change and solar activity. Years with fewer sunspots indicated greater recruitment. Increased precipitation during autumn migration signified less recruitment. On the breeding grounds, fewer days below freezing during winter and more precipitation during breeding suggested less recruitment. We predicted whooping crane recruitment and population growth may fall below long-term averages during all solar cycles when atmospheric CO 2 concentration increases, as expected, to 500 ppm by 2050. Species recovery during a typical solar cycle with 500 ppm may require eight times longer than conditions without climate change and the chance of population decline increases to 31%. Although this whooping crane population is growing and may appear secure, long-term threats imposed by climate change and increased solar activity may jeopardize its persistence. Weather on the breeding grounds likely affects recruitment through hydrological processes and predation risk, whereas precipitation during autumn migration may influence juvenile mortality. Mitigating threats or abating climate change should occur within ≈30 years or this wild population of whooping cranes may begin declining.

K E Y W O R D S
atmospheric CO 2 , boreal pond, decadal cycle, groundwater, LASSO, precipitation, reproduction, solar activity, sunspots

| INTRODUCTION
Identifying drivers of animal population demographics (i.e., associations between environmental conditions and vital rates) and locating where they operate is necessary for steering efforts to optimize wildlife species management and conservation (e.g., Caughley, 1994;Mills, 2007). Doing so establishes a foundation to evaluate how climate change may impact an animal's population growth and vulnerability (Brook et al., 2009). This information is especially critical for recovering species endangered with extinction (Akçakaya, Butchart, Watson, & Pearson, 2014). Herein, we focus on the population growth of endangered whooping cranes (Grus americana), which hinges on recruitment (Butler, Metzger, & Harris, 2014).
Given that the approximate decadal population cycle observed in many boreal species appears mediated through climate processes (Scott & Craine, 1993;Yan, Stenseth, Krebs, & Zhang, 2013), we examined whether and how global climate change may influence whooping crane demographics. Climatic changes may affect whooping crane recruitment through direct (i.e., increases in predator densities) or complex indirect interactions. For example, whooping cranes build nests in shallow lakes and ponds of the taiga to deter predation (Kuyt, 1995;Kuyt, Barry, & Johns, 1992;Timoney et al., 1997).
Climate change is reducing the amount of surface water and likely the depth and seasonal longevity of these ponds (Carroll et al., 2011;Smith et al., 2005). Fewer, shallower, and more ephemeral ponds failing to persist through the entire nesting period may increase nest predation, thereby reducing recruitment of juveniles (Timoney et al., 1997).
Other factors outside of the breeding grounds may also affect whooping crane recruitment as it represents a combination of several demographic parameters (e.g., breeding propensity, clutch size, hatch and fledge rates, and juvenile survival during migration). Reproduction by adults can be supported using stored resources (known as capital breeding) or via the uptake of resources concomitant to a reproductive attempt (known as income breeding; Stephens, Boyd, McNamara, & Houston, 2009). If a capital breeding strategy is employed by whooping cranes, weather conditions prior to their departure to the breeding grounds may affect breeding propensity and nest success.
Alternatively, juveniles may be influenced by conditions encountered during southern migration (e.g., inclement weather, food quality, and quantity) that could reduce their survival and subsequently affect their recruitment into the winter flock.
We examined environmental drivers of whooping crane recruitment by modeling it as a function of solar, weather, and hydrological factors spanning all portions of its life cycle (breeding, wintering, and migration). On the breeding grounds, these models examined associations of recruitment with the solar cycle and winter temperature and precipitation. For the wintering grounds, models included drought conditions and freshwater balance of estuaries, as these factors may limit food production . During the spring and fall migrations, models evaluated temperature, precipitation and drought, as each affects food availability and production, and inclement weather may increase juvenile mortality during southern migration (Kuyt, 1992).
We predicted the potential impact of climate change on whooping crane recruitment and population growth for future scenarios, by linking the primary drivers of recruitment with atmospheric CO 2 concentrations (elevated concentrations are indicative of climate warming) and the solar cycle. Scenarios of atmospheric CO 2 concentration consisted of the 2015 concentration (≈400 ppm) and the predicted 2050 concentration (≈500 ppm; Intergovernmental Panel on Climate Change [IPCC], 2014). The solar cycle scenarios were the modern grand maximum (1954)(1955)(1956)(1957)(1958)(1959)(1960)(1961)(1962)(1963)(1964), the Dalton minimum (1811-1821), and a typical cycle. This procedure reveals how climate change may affect the whooping crane population's dynamics through 2050, provides expectations for species recovery, and indicates where management intervention would likely be most effective.

| Breeding grounds, wintering grounds, and migratory route
Only one wild, migratory population of whooping cranes exists. This population is growing, yet small (≈340 birds), range restricted, F I G U R E 1 Migratory route of whooping cranes in North America. The breeding grounds are on and around Wood Buffalo National Park, Canada, and the wintering grounds are on and around Aransas National Wildlife Refuge, Texas, USA and therefore considered endangered with extinction (Butler et al., 2013CWS & USFWS, 2007 (Kuyt, 1992). Their migratory route continues north-northwest across Saskatchewan to Alberta and then north to Wood Buffalo National Park (Figure 1; Kuyt, 1992).
Whooping cranes begin arriving on their breeding grounds by late April, and their nests are initiated by mid-May (Kuyt, 1995). Whooping cranes build their nests in shallow boreal ponds or lakes by constructing a mound of mud and vegetation in open water or placing it among clumps of brush on natural islets (Kuyt, 1995;Kuyt et al., 1992;Timoney et al., 1997). Whooping cranes, along with the young of the year, depart the breeding grounds starting in mid-September and follow the same migratory route south to the Texas coast (Figure 1; Kuyt, 1992). During the autumn migration, central Saskatchewan appears to be a staging area in which whooping cranes congregate and forage on waste grain during late September to mid-October (Johns, Woodsworth, & Driver, 1997). Once whooping cranes depart the staging grounds in central Saskatchewan, southern migration across the USA can be as short as 1 week with delays occurring due to lowpressure weather systems (Kuyt, 1992).

| Aerial surveys
Annual, repeated aerial surveys of whooping cranes have been conducted by USFWS personnel on the wintering grounds between October and May since 1950 (Butler, Strobel, & Eichhorn, 2016;Strobel & Butler, 2014). The survey has been primarily conducted from a fixed-wing aircraft with transects spaced approximately 250-800 m apart (Butler et al., 2016). The number of surveys conducted each winter between December 1 and March 31 has ranged from 4 to 21 . This survey produced counts of total winter abundance and number of juvenile birds recruited plus estimates of annual mortality (Butler et al., 2013). averaged the months of November-March of each year. We obtained monthly freshwater balance (i.e., balance = inflow − evaporation + precipitation) for the Guadalupe and Mission-Aransas estuaries (Guthrie & Lu, 2010;Schoenbaechler, Guthrie, & Lu, 2011). For both estuaries, we summed the months of November-March of each year and scaled estimates to 1 million acre-feet. The Aransas-Wood Buffalo whooping crane population overwinters along both estuaries, so we also combined freshwater balance for them. As environmental conditions on the wintering grounds may affect whooping crane body condition and therefore reproduction once birds return to the breeding grounds (i.e., capital breeding; Stephens et al., 2009) We divided the migratory route into two primary areas: northern U.S. Great Plains and southern Saskatchewan prairies. We obtained monthly values for the Palmer drought indices for Nebraska Climatological Divisions 2-9, South Dakota Climatological Divisions 2-3 and 6-9, and North Dakota Climatological Divisions 1-2, 4-5, and 8-9 (Figure 1; Palmer, 1965;NCDC 2007), as indicators of climatic conditions on the northern U.S. Great Plains. For each of the Palmer indices, we averaged data from the Climatological Divisions across months for the spring migration period (i.e., April) and autumn migration period (i.e., October-November).

| Possible predictors of recruitment
We used the proportion of days in which the maximum temperature remained below freezing (i.e., ≤0°C), the proportion of days in which minimum temperature was below freezing, mean temperature, and the total precipitation obtained from three Global Historical Climatology Network (GHCN) stations (i.e., Regina University, CA004016640; Saskatoon Airport, CA004057120; and Muenster, CA004015440) located across the southern Saskatchewan prairies ( Figure 1). All three of these stations had a complete record from 1950-2010 and covered the southern and northern portions of the migratory route through the southern Saskatchewan prairies. Data were pooled across the three stations and months for the spring (i.e., April) and autumn migration periods (i.e., September-October).

Few long-term GHCN stations exist in the vicinity of Wood Buffalo
National Park and the whooping crane's breeding grounds. We obtained weather data (same measures used for southern Saskatchewan prairies) from the GHCN station at Fort Smith, Northwest Territories, Canada (CA002202200), the station closest to the breeding grounds ( Figure 1). For the breeding grounds, we also considered two periods as potentially important to recruitment: winter (November-March) and breeding (April-September). We considered the winter prior to arrival because surface water conditions in breeding ponds may be affected by winter weather (Yoshikawa & Hinzman, 2003).
The solar cycle has been implicated as a potential factor in the cyclic population dynamics of lynx, snowshoe hare, and other organisms (Klvana et al., 2004;Krebs et al., 2001;Marques, Short, et al., 2014;Selås et al., 2004). Sunspots are a long-term indicator of total solar irradiance (Krivova et al., 2010), so we obtained the annual sunspot number from the Royal Observatory of Belgium (SILSO World Data Center 2015). We used year as a predictor to investigate trends in recruitment.

| Addressing bias in recruitment estimates
We suspected recruitment estimates may have been biased high during years with fewer surveys (similar biases occurred with winter mortality estimates; Butler et al., 2014). We examined recruitment estimates for this potential bias by regressing recruitment and the number of surveys conducted during a winter between December 1 and March 31 (logistic regression). If unbiased, recruitment estimates should not be associated with the number of surveys. However, if an inverse relationship between recruitment and the number surveys exists, it indicates some adults are missed when the number of surveys is small. For example, detection of a whooping crane group was estimated as 55.8% (Strobel & Butler, 2014) which means the cumulative detection probably does not exceed 99% until six surveys are completed. Therefore, estimates of recruitment are likely biased high during years in which the number of surveys conducted was less than six.
We removed those potentially biased years and reexamined the association between recruitment and the number of surveys conducted.

| Modeling recruitment and survival with LASSO
We modeled recruitment (i.e., proportion of juvenile whooping cranes in the total population count for each winter) as a function of the predictor variables described above (Table 1), many of which were collinear, using logistic regression. To select the model, we used the least absolute shrinkage and selection operator (LASSO), a statistical regularization technique (i.e., procedure for balancing model fit and complexity) appropriate for model selection when a relatively large set of possibly collinear predictor variables are used (Gerber, Kendall, Hooten, Dubovsky, & Drewin, 2015;Tibshirani, 1996). Unlike other statistical regularization techniques (i.e., ridge regression), LASSO regression allows some effects to be exactly zero which results in a more easily interpretable model, because the number of predictors is reduced (Reineking & Schröder, 2006;Tibshirani, 1996). In LASSO regression, the balance between model fit and complexity is achieved through the parameter lambda.
We used the glmnet package (Friedman, Hastie, & Tibshirani, 2010;Hastie & Qian, 2015) in program R (R Core Team 2014) to conduct the analysis. We used the cv.glmnet function to preform tenfold cross-validation to select the optimal lambda (i.e., most regularized model whereby the mean cross-validated error is within one standard error of the minimum; Hastie & Qian, 2015). As the folds used in crossvalidation are selected randomly, we repeated the cross-validation procedure 1,000 times to characterize the uncertainty in lambda and the model coefficients. We also calculated the proportion of repeats in which a predictor variable was selected.
We conducted the repeated cross-validated LASSO procedure described above for survival as well. We estimated annual whooping where N t was the total abundance during winter t and Â t+1 was the number of after-hatch-year birds in the population during winter t + 1 .

| Modeling weather variables
The repeated cross-validated LASSO procedure identified the most important weather variables for predicting recruitment. We modeled those weather variables as functions of sunspot number and the natural log of mean annual atmospheric CO 2 concentration (μmol of CO 2 / mol [ppm]; Keeling et al., 2001). Simple linear or logistic regression was used depending on the distributional characteristics of each variable.
We used data from 1958-2014 for these analyses as measurement of the concentration of atmospheric CO 2 began in March 1958 (Keeling et al., 2001). By assuming a log-linear relationship with CO 2 concentration, we obtain conservative estimates describing the potential magnitude of change. Kuyt et al. (1992) measured the water depth of whooping crane nesting ponds during late May 1976-1991 and reported the annual mean. As surface water conditions in boreal ponds are affected by winter temperature and snow depth (Yoshikawa & Hinzman, 2003), we examined these relationships on the whooping crane's breeding grounds. We used simple linear regression to relate mean pond depth to sunspot number and the proportion of freezing days and precipitation on the breeding grounds during November-March. We also used logistic regression to examine the relationship between recruitment and pond water depth. This analysis helped explain associations between some of the weather variables and recruitment.

| Population growth, recruitment, survival, and environmental change
To predict the potential impact of climate change on whooping crane recruitment, we assessed the potential changes in climatic condition for various scenarios of atmospheric CO 2 concentration and sunspot number. First, we developed three solar cycle scenarios: one based on the cycle that contained the modern grand maximum (1954-1964; (IPCC, 2014). We converted predicted recruitment from a proportion to a ratio by proportion/(1-proportion). We estimated predicted population growth (̂) for each scenario as (Ŝ) (1 +R) where Ŝ is the geometric mean of annual survival and R is the predicted ratio of hatch-year to after-hatch-year birds for each scenario .
To characterize the uncertainty in predicted recruitment and population growth rate for the climate change scenarios, we implemented a nonparametric bootstrap procedure. Specifically, for each iteration of the repeated cross-validated LASSO procedure, we bootstrapped the models for each of the important weather variables and predicted the weather condition for each scenario. We bootstrapped whooping crane annual survival for each iteration of the repeated cross-validated LASSO procedure so predicted population growth would incorporate uncertainty and variability in survival and predicted recruitment.
If management options that directly manipulate recruitment are unavailable, management options may exist to manipulate annual survival. Therefore, we estimated the annual survival (Ŝ) required to avoid population decline or avoid population growth from falling below average (i.e., λ ≥ 1.035) for each scenario where Ŝ is 1∕(1 +R) .

| Addressing bias in recruitment estimates
We found that recruitment was biased high for years with few surveys. Specifically, it was inversely related to the number of surveys conducted such that an additional survey produced a 3.6% decline in recruitment (β = −0.036, SE = 0.010, p < .001). This occurred because detection of a whooping crane group was 55.8% (Strobel & Butler, 2014) resulting in the cumulative detection probably not exceeding 99% until six surveys were completed. Therefore, we removed years in which the number of surveys conducted was fewer than six (n = 6). Once years with <6 surveys were removed, recruitment was no longer associated with the number of surveys conducted (β = −0.016, SE = 0.012, p = .119). Recruitment during years with ≥6 surveys ranged from 0 to 0.31 (geometric mean = 0.137, n = 55, SD = 0.074).

| Modeling recruitment and survival with LASSO
The repeated cross-validated LASSO procedure suggested four variables were important to predicting recruitment (selected by >95% of iterations;

| Modeling weather variables
The proportion of days that remained below freezing during winter on the breeding grounds (DX32.WBNP.1103) decreased as the sunspot number and natural log of atmospheric CO 2 concentration increased ( Modeled recruitment (n = 55) using LASSO regression and tenfold cross-validation to select the optimal model. This procedure was repeated 1,000 times and proportion of times a variable was selected was recorded. Acronyms are defined in Table 1.
T A B L E 2 Variables predicting recruitment of juvenile whooping cranes into the winter population

| Relationships with nesting pond depth
Mean water depth in nesting ponds ranged from 14.3 to 27.8 cm
Whooping crane population growth fluctuates inversely to the solar cycle, resulting in greater population growth during years of reduced solar activity (Figure 3). At the current atmospheric CO 2 concentration (≈400 ppm), years with negative growth are likely rare during a typical solar cycle (Figure 3). However, our model predicts years with negative growth become more common in the future (500 ppm CO 2 ), as 4 years of the typical solar cycle have negative population growth (i.e., λ < 1; Figure 3). Average recruitment is predicted to fall below the long-term average (0.137) in all the scenarios, except one: a solar cycle similar to the Dalton minimum with 400 ppm CO 2 (Figure 4).
Indeed, at 500 ppm CO 2 during a solar cycle similar to the modern grand maximum, mean recruitment lies below the amount required for population growth (Figure 4).
Regardless of the solar scenario, our model predicts reduced whooping crane population growth if atmospheric CO 2 concentration continues to increase ( Figure 5). Whooping crane population growth likely falls below the long-term average (3.5%) for all solar conditions examined with 500 ppm CO 2 ( Figure 5). Further, the population will likely decline by −0.44% (95% CI = −2.98% to 1.57%) if a solar cycle similar to the modern grand maximum occurs in the future ( Figure 5).
We found the past survival rate of 90.6% was likely sufficient to maintain a stable population (λ ≥ 1) for all but one scenario: a solar cycle similar to the modern grand maximum with 500 ppm CO 2 ( Figure 6). In this scenario, our model predicts annual survival would need to be increased to 91.0% (95% CI = 89.3%-93.4%; Figure 6). If the past population growth rate of 3.5% is to be maintained, then all scenarios with 500 ppm CO 2 would require increasing annual survival above its long-term mean ( Figure 6). For example, during a typical solar cycle with 500 ppm CO 2 , our model predicts annual survival would need to be increased to 93.4% (95% CI = 91.8%-95.7%) to maintain 3.5% population growth ( Figure 6).

| Environmental conditions and recruitment
Heightened solar activity increases the amount of radiation striking the Earth's surface (Krivova et al., 2010). This raises atmospheric temperatures and alters weather patterns thereby influencing plant growth, surface hydrology, and faunal responses (Klvana et al., 2004;Krebs et al., 2001;Marques, Short, et al., 2014;Stager, Ruzmaikin, Conway, Verburg, & Mason, 2007). Solar activity, as indexed by the sunspot number, had the greatest magnitude of effect on whooping crane recruitment (based on Standardized β ; Table 2), having ≥2.3 times more effect than any other variable. The mechanism(s) describing how sunspot activity affects whooping crane recruitment remains unknown. Unraveling similar correlations between sunspots and other cyclical biological phenomena within the boreal ecosystem, and identifying mechanism(s) that generate these observed relationships also remains elusive (Krebs et al., 2001;Sinclair & Gosline, 1997).
The amount of precipitation in the northern U.S. Great Plains during the autumn migration was the second most important predictor. This variable had ≥36% the influence of the remaining weather variables.
Recruitment declined as the amount of precipitation increased during autumn migration. We hypothesize that juveniles are more vulnerable to mortality or abandonment during inclement weather. As juveniles have greater energy expenditure during migration than adults (Rotics et al., 2016) and are inexperienced foragers, inclement weather may reduce foraging opportunities (Clark, 2009) T A B L E 3 Weather variables that were associated with recruitment of juvenile whooping cranes into the winter population were modeled as functions of atmospheric CO 2 concentration and/or sunspot number more vulnerable to collisions than adults (Brown & Drewien, 1995;Chavez-Ramirez & Wehtje, 2012;Ward & Anderson, 1992).
On the breeding grounds, recruitment increased with more days remaining below freezing during winter. We also observed that nesting pond depth decreased as the number of days remaining below freezing during winter decreased. The potential mechanism behind the association between recruitment and the number of freezing days during winter may be that fewer cold days (<0°C) reduces the amount of time the active groundwater layer remains frozen. Frozen earth retains pond water into the nesting period, with thinner layers more The solar cycle as indicated by sunspot number has varied over the last three centuries . The Dalton minimum occurred during 1811-1821 and modern grand maximum occurred during 1954-1964. We estimated the typical cycle as a sixth-order polynomial of the number of years since a cycle began F I G U R E 3 Predicted whooping crane population growth for two atmospheric CO 2 concentration scenarios during a typical solar cycle (horizontal line indicates 0% population growth, λ = 1) easily breached by the water trapped above (Yoshikawa & Hinzman, 2003). Thawing of the ground allows water to permeate the soil, resulting in reduced pond depth, size, and longevity. Presumably, adult whooping cranes nest within ponds to reduce predation on themselves, eggs and hatchlings Timoney et al., 1997). If ponds dry too quickly, nests may become more vulnerable to terrestrial predators, thereby lowering hatchling survival resulting in reduced recruitment.
F I G U R E 4 Predicted juvenile recruitment (ratio of hatch-year [HY] to after-hatch-year [AHY] birds) of whooping cranes during an 11-year solar cycle for two atmospheric CO 2 concentration scenarios. Top dotted line represents long-term mean recruitment (0.137) and bottom dotted line represents required recruitment (0.103) needed to maintain a stable population (λ ≥ 1) given the longterm mean annual survival of 90.6%. Points represent mean predictions and vertical bars 95% confidence intervals F I G U R E 5 Predicted whooping crane population growth during an 11-year solar cycle for two atmospheric CO 2 concentration scenarios. Top dotted line represents long-term mean population growth of 3.5%. Points represent mean predictions and vertical bars 95% confidence intervals. Only a solar cycle similar to the Dalton minimum with ≤400 ppm CO 2 will likely maintain population growth at or above the longterm mean of 3.5% Precipitation on the breeding grounds during winter, though much less influential than other weather variables, reduced recruitment.
Increased precipitation, likely in the form of snow, acts as an insulator maintaining soil warmth, reducing the thickness of frozen earth and thereby reducing water depth in nesting ponds (Yoshikawa & Hinzman, 2003). Corroborating this, we observed nesting pond depth declined as winter precipitation increased.
Increased precipitation during the breeding season on the breeding grounds was associated with reduced recruitment. We hypothesize that increased precipitation during incubation causes nest flooding and failure (Ivey & Dugger, 2008;Layne, 1983). Furthermore, increased rainfall is more apt to soak nesters and hatchlings, thereby increasing their risk of illness or death (Chavez-Ramirez & Wehtje, 2012;Layne, 1983).
Variables associated with recruitment primarily operate on the breeding grounds or during southern migration across the northern U.S. Great Plains. The variables operating in other locations were much less influential. For example, precipitation on the wintering grounds prior to spring migration and breeding was associated with increased recruitment. This suggests that whooping cranes do not exhibit a complete income breeding strategy (Stephens et al., 2009). Instead, breeding propensity or other reproductive parameters appear partially controlled by conditions prior to migration. However, precipitation on the wintering grounds was at least 69% less influential than the other weather variables already discussed (Table 2).

| Potential impacts of climate change on recruitment and population growth
Elevated concentrations of atmospheric CO 2 warm the Earth, reducing the proportion of days below freezing on the breeding grounds (i.e., increased temperature; Lemmen, Warren, Lacroix, & Bush, 2008) and increasing precipitation on the northern U.S. Great Plains during autumn (Melillo, Richmond, & Yohe, 2014). Given these relationships, we substituted the amount of atmospheric CO 2 for these variables to evaluate the potential effects of CO 2 concentration on recruitment.
Our model predicts recruitment declines as the magnitude of the solar cycle moves from the Dalton minimum to the modern grand maximum, or as atmospheric CO 2 concentrations increase from 400 to 500 ppm (Figure 4). The influence of sunspots and atmospheric CO 2 on recruitment is comparable, although sunspots have a greater effect.
For example, at 400 ppm CO 2 concentration, the difference in average population growth between the Dalton minimum (3.7% mean growth; Figure 5) and the modern grand maximum (1.2% mean growth; Figure 5) is 2.5% (95% CI mean difference = 1.3%-3.8%). Hence, a solar cycle similar to the grand maximum would result in ≈68% less population growth than a solar cycle similar to the Dalton minimum.
Alternatively, during a typical solar cycle, the difference in average population growth between 400 ppm (2.2% mean growth; Figure 5) and 500 ppm CO 2 concentration (0.4% mean growth; Figure 5) is 1.8% (95% CI mean difference = 0.3%-3.6%). Hence, as atmospheric CO 2 F I G U R E 6 Predicted whooping crane survival needed to maintain 0% population growth and 3.5% population growth (longterm mean) during an 11-year solar cycle for two atmospheric CO 2 concentration scenarios. Dotted line represents long-term mean annual survival of 90.6%. Points represent mean predictions and vertical bars 95% confidence intervals concentration increases to 500 ppm, population growth will likely be reduced by ≈80% compared to the current CO 2 concentration.
Only one scenario, a solar cycle similar to the Dalton minimum with 400 ppm CO 2 concentration, results in population growth commensurate with the long-term average (i.e., >3.5% growth; Figure 5).
This scenario results in no simulations indicating a population decline. All other scenarios portend a future with whooping crane population growth likely falling below the long-term average ( Figure 5).
Population growth will likely be reduced by half during a solar cycle similar to the Dalton minimum with 500 ppm CO 2 , suggesting downlisting and recovery may take two times longer. The most likely future scenario, a typical solar cycle with 500 ppm CO 2 , suggests population growth will likely be minimal (0.4% mean growth which is an eighth of the long-term average), with 31% of the simulations indicating a population decline. These results suggest downlisting and recovery may require eight times longer than conditions without climate change (Butler et al., 2013 Hence, our predictions about potential impacts of climate change on recruitment and population growth are tempered and reasonable.

| Management intervention to improve recruitment
Recruitment should remain at or above 0.103 to maintain stable to increasing population growth (λ ≥ 1; Figure 4). If the scenario of a solar cycle similar to the grand maximum with 500 ppm CO 2 occurs, recruitment may decline to 0.099 (95% CI = 0.071-0.120; Figure 4) and management intervention on the breeding grounds would likely have to result, on average, in an additional 0.5 juveniles per 100 adults to avoid population decline. To maintain long-term average population growth (3.5%), recruitment should remain at or above 0.137 (Figure 4).
The only scenario predicted to achieve this is 400 ppm CO 2 concentration during a solar cycle similar to the Dalton minimum (Figures 4   and 5). Other scenarios, such as a typical solar cycle or modern grand maximum with 500 ppm CO 2 concentration suggests average recruitment may decline to 0.108 (95% CI = 0.081-0.128) and 0.099 (95% CI = 0.071-0.120; Figure 4), respectively. Thus, any management intervention on the breeding grounds would likely have to result, on average, in an additional 3 to 4 juveniles recruited per 100 adults to maintain the long-term population growth.
While it has been suggested that management intervention on Wood Buffalo National Park is infeasible given its remoteness and lack of infrastructure (Wilson et al., 2016), these challenges would need to be overcome to recover whooping cranes. Without management intervention, the required amount of time to achieve population goals for downlisting and recovery likely becomes greatly extended or potentially impossible, given our model's predictions.
How climate change may manifest in the predator-prey cycle remains unclear. Lynx declines have been linked to climate warming (Yan et al., 2013), yet we do not know the ecological ramifications of this outcome, or climate effects on other predators. Other complex interactions associated with ponds drying and increased vulnerability to nest and fledgling predation could also decrease recruitment even if predator densities remained similar or declined. Furthermore, increased temperatures may result in predator populations becoming less cyclic and more generalized in prey selection (Scott & Craine, 1993).

| Management intervention to improve annual survival
If management intervention cannot improve recruitment, maintaining long-term average population growth (3.5%) likely requires increases in annual adult survival in all but one of the scenarios (solar cycle similar to the Dalton minimum with 400 ppm CO 2 ; Figure 6). For instance, the scenario of a typical solar cycle with 500 ppm CO 2 , annual adult survival should be increased from 90.6% (long-term geometric mean; SD = 0.060) to 93.4% (95% CI = 91.8%-95.7%; Figure 6) to maintain the long-term average population growth. At the current population of ≈340 whooping cranes, this is equivalent to, on average, 9 to 10 fewer adult mortalities per year. However, none of the variables we examined were important to predicting survival and many management actions are already underway to reduce adult mortality (e.g., signage to inform hunters, marking of power lines in strategic locations).
Past mortality from power line collisions and incidental or illegal harvest is rare (<1 bird/year succumbs to these mortality causes; Stehn & Haralson-Strobel, 2014). Thus, little gains in annual survival can currently be expected from these management actions.
Long-term annual adult survival is 90.6% and likely remains sufficient to maintain population stability (0% growth) across all solar cycle and CO 2 concentration scenarios except during a solar cycle similar to the grand maximum with 500 ppm CO 2 ( Figure 6). Given this scenario, annual survival needs to reach 91.0% (95% CI = 89.3%-93.4%; Figure 6) to avoid a population decline. In a scenario where a typical solar cycle occurs with 500 ppm CO 2 , annual adult survival could decline to 90.2% (95% CI = 88.7%-92.5%; Figure 6) and still avoid population decline. Hence, given a typical solar cycle with 500 ppm CO 2 and the current population of ≈340 whooping cranes, on average, 1 to 2 additional mortalities per year could result in a declining population.

| Synthesis
This whooping crane population will likely continue experiencing periodic and predictable years of low recruitment and growth, as sunspots form the primary driver of their approximate decadal population cycle. Climate change likely threatens this population by 2050 when CO 2 concentrations are expected to reach 500 ppm, as our model predicts population growth can become negative during any solar cycle (unlike 400 ppm CO 2 ; Figure 5). Hence, science has roughly three decades to reveal the onthe-ground mechanisms relating solar activity and atmospheric CO 2 with recruitment (i.e., proximate causes) and to design approaches for mitigating the adverse effects. Alternatively, other management actions could increase annual survival to avoid population declines. Optimistically, increasing recruitment and survival so the population continues growing requires only 1-4 additional recruits or survivors per year. Survivorship is the tougher to address, as many of the actions are continental in scope and multifaceted (e.g., power line marking, wind turbine curtailment, hunter education and law enforcement, etc.).
Recruitment affects whooping crane population growth most Wilson et al., 2016), and our model predictions show that climate change may imperil it. Hence, any perceptions that whooping cranes remain secure in Canada, and can remain free of active management, are false. If conserving whooping cranes in North America remains an important goal, then attention must center on understanding mechanisms limiting recruitment. There appears to be a connection between recruitment, breeding pond hydrology, and predation. Already, the numbers of boreal ponds and their temporal longevity during summer are declining (Carroll et al., 2011;Smith et al., 2005). Unless trends of increasing atmospheric CO 2 are abated, science must determine how to mitigate the likely reduction in recruitment by 2050. Future monitoring and research should focus on relationships of depth and seasonal longevity of nesting ponds with nest success and causes of nest failure. Additionally, fledgling mortality and its causes on the breeding grounds and during southern migration should be investigated. The results of this work could steer conservation actions toward long-term sustainability and protection of whooping cranes in North America. If such efforts prove unsuccessful, then the only wild population of this species must adapt or likely face extinction.

ACKNOWLEDGMENTS
Findings and conclusions in this manuscript are those of the authors and do not necessarily represent the views of USFWS. Use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. We thank Leandra N.
Taylor for helping design Figure 1.