Each spring approximately 600,000 Sandhill Cranes (Antigone canadensis) migrate through the central Great Plains from the wintering grounds in Texas, New Mexico, southeastern Arizona, and Mexico en route to their breeding grounds in Siberia, Alaska, and central and northern Canada (Kinzel et al. 2005, Krapu et al. 2011, Kruse and Dubovsky 2015, Dubovsky 2016). Nearly 80% of these cranes stage each spring along the Big Bend Reach of the Platte River in Nebraska, USA (Lewis 1976, Krapu et al. 1985, 2014, Pearse et al. 2017). While in this area, Sandhill Cranes forage during the day for waste grain in cornfields and for native sedges and macroinvertebrates in lowland meadows. At night they roost in shallow braided channels of the Platte River (Sparling and Krapu 1994, Pearse et al. 2017). This annual phenomenon is of critical conservation concern because it is one of the largest remaining mass wildlife migration events in the world (Johnsgard 2002).

The central Platte River is a steep, sand-bed braided river (Simons and Associates Inc. and URS Greiner Woodward Clyde 2000). As with most braided rivers, the central Platte is highly dynamic with a high width-to-depth ratio and a channel dissected by emergent bedforms at low discharges (Simons and Associates Inc. and URS Greiner Woodward Clyde 2000, Murphy et al. 2004). Sandhill Cranes roost in shallow water (<25 cm; 10 inches) that flows over submerged bedforms, and occasionally cranes will roost on aerially exposed sandbars (Norling et al. 1992, Kinzel et al. 2005) in areas with views unobstructed by dense vegetation (Folk and Tacha 1990). It is hypothesized that Sandhill Cranes roost in wide, shallow unvegetated channels because they rely heavily on eyesight to detect approaching predators (Armbruster and Farmer 1982, Krapu et al. 1984, Folk and Tacha 1990, Davis 2001, 2003, Pearse et al. 2017).

Given the importance of the central Platte River in the annual Sandhill Crane migration, ecologists have conducted numerous studies since the early 1980s to assess in-channel habitat suitability for roosting and to identify threats to the quantity and quality of that habitat (USFWS 1981, Faanes and LeValley 1993, Davis 2003, Pfeiffer and Currier 2005, Krapu et al. 2014, Pearse et al. 2017). Several of these studies were conducted using sophisticated physical/hydrodynamic modeling based on surveys of channel topography at roost locations (Prince 1990, Simons and Associates Inc. 1990, Kinzel et al. 2005). These investigations have primarily focused on assessing the channel widths and river discharges necessary to optimize the area of shallow water habitat in reaches that are suitably wide for roosting. Various researchers have proposed suitable channel width targets ranging from 50 m (164 feet) to ≥250 m (820 feet) and optimal river discharges ranging from 30 m³/s (cms; 1059 cfs) to 57 cms (2013 cfs; Currier and Eisel 1984, Platte River Management Joint Study Biology Workgroup 1990, Prince 1990, Kinzel et al. 2005). The high degree of variability in proposed width and discharge targets is likely due to the dynamic nature of the Platte River channel where bedforms and unvegetated channel width vary widely across the reach and at a single location through time. Furthermore, discharge can vary by up to 28 cms (1000 cfs) in a day during the spring migration season, substantially changing the quantity and spatial distribution of available roosting habitat (Kinzel et al. 2009).

Sandhill Cranes that use the Platte River during the spring migration season utilize a system that is far from static. Spatial shifts in roost locations that appear to correspond to flow variability are commonly noted during observational studies (J. Jenniges and M. Peyton personal communication) and during our systematic aerial monitoring surveys. For example, a large number of Sandhill Cranes have been observed roosting in fields and wet meadows under high discharge conditions when shallow-water habitat is limited (Davis 2001). These observations led us to hypothesize that the distribution of Sandhill Crane roost locations within and across years is a function of the interaction between available channel widths and discharge as it affects the distribution and availability of suitably shallow roosting habitat. The objective of this study was to test this hypothesis and to utilize the results to describe flexible flow targets that acknowledge the range of spatial and temporal variability present in this system. As such, we evaluated the influence that channel-width measures and flow per linear unit of channel width (i.e., average channel depth) have on roost-site selection by Sandhill Cranes. The results of our analysis could be used to help identify ideal flow conditions for all channel widths for Sandhill Cranes roosting on the central Platte River.

Fig. 1.

Platte River channels within the study area between Elm Creek and Chapman, Nebraska, USA, with the locations of cities included for reference.

Methods

Identifying Sandhill Crane Group Locations

To determine distribution patterns of Sandhill Crane roosts, aerial surveys of Platte River channels between Chapman and Elm Creek, Nebraska, USA (Fig. 1), were conducted annually by personnel of the Platte River Whooping Crane Maintenance Trust. Surveys were conducted via a fixed-wing single-engine airplane (generally a Cessna 172) once every week from mid-February to mid-April, as appropriate weather conditions and aircraft availability permitted. This 120-km stretch of river was flown 6 to 9 times per year at approximately 135 km/h and at an elevation of 200 m above ground. The surveys were initiated 15–25 min prior to sunrise and took approximately 1 h to complete. Survey flights were generally flown from east to west; however, flights were flown from west to east later in the seasons when the highest concentration of birds shifted from east to west (Krapu et al. 2014). One observer was responsible for estimating the size of each group of Sandhill Cranes by first counting a portion of each roost, then counting spatial replicates of that area within the roost as a whole (Gregory et al. 2004, Bowman 2014). A second researcher collected the latitude and longitude over the center of each Sandhill Crane roost using a handheld Global Positioning System (Garmin^® eTrex, 72, or 72H handheld GPS, Olathe, KS) and recorded count estimates obtained by the observer. Individual roosts were defined as a group of cranes that were separated from another group by more than 100 m (Iverson et al. 1987).

Fig. 2.

Examples of how (a) total unvegetated channel width (TUCW) and (b) maximum unvegetated channel width (MUCW) were delineated.

Validating Variables Included in Analyses

To validate results of the best model, we randomly partitioned the data sets of use locations and corresponding available locations for small, medium, and large groups into training data sets (2/3 of the choice sets) and test data sets (1/3 of the choice sets). We used training data to develop parameter estimates for best models and a comparison of available and use locations of the test data set to understand the reliability of a binary response (use/available) model. Predicted values of available locations within the test data set were scaled to the number of use locations in the test data set. These were then binned into 20 percentile categories and compared to the number of test-data-set use locations in each bin. Predicted values were summed to calculate the number of expected use locations in each bin, which were then compared to the actual sum of use locations in each bin with a linear regression model in order to identify the reliability of the model based on the closeness of the slope-relationship of 1. This method was repeated 1000 times to develop the average slope and 95% confidence intervals of model fit. As defined by Howlin et al. (2004), a “good” model had an average 95% confidence interval that incorporated 1 and not zero. An “adequate” model had an average 95% confidence interval that did not incorporate 1 or zero. If the average slope-relationship had a 95% confidence interval spanning zero, the model was deemed “poor.”

Table 1.

Average and standard deviation (SD) of habitat variables evaluated in a use-availability discrete-choice analysis of Sandhill Crane (Antigone canadensis) roost locations on the central Platte River, 2007–2017. Variables include maximum unvegetated channel width of the channel where the use or available locations were (MUCW), total unvegetated channel width of the channel where the use or available locations were (TUCW), and discharge per linear meter of total unvegetated width of all channels (DISCH).

Table 2.

Akaike information criterion (AIC) model selection results of our use-availability roost-site selection analysis by crane group size. Variables included maximum unvegetated channel width of the channel where the use or available locations were (MUCW), total unvegetated channel width of the channel where the use or available locations were (TUCW), and flow per meter of total unvegetated channel width of all channels (DISCH).

Results

We documented 2642 Sandhill Crane roost sites and 52,697 associated available locations within our study area between 2007 and 2017. Roost size averaged about 3500 individuals with a median of 1000 cranes (range 1–84,330 cranes). The average maximum unvegetated and total unvegetated channel widths were wider at roost sites of large crane groups than at roost sites of medium and small crane groups (Table 1). However, flow per meter of total unvegetated channel width was similar among crane group sizes (Table 1).

Statistical modeling of habitat selection indicated that aspects of channel width were important for all crane group sizes. Flow metrics were also important for medium and large crane groups, but not for small crane groups (≤500 individuals; Table 2). Maximum unvegetated channel width was an important predictor of small and large crane group roost locations. Selection ratios were maximized at 131 m for small crane groups and 275 m for large crane groups but were statistically similar from 26 m to 190 m and ≥150 m, respectively (Fig. 3A, D). For medium-sized crane groups, the selection ratio for TUCW was maximized at 267 m but was statistically similar from 69 m to 408 m (Fig. 3B). Flow per meter of total unvegetated channel width was an important predictor for medium and large crane groups. Selection ratios were maximized at the lowest flows observed for medium and large crane groups (0.0361 cms/m and 0.0382 cms/m of total unvegetated channel width of all channels, respectively). However, these results were statistically similar from 0.0361 cms/m to 0.1420 cms/m and 0.0382 cms/m to 0.1006 cms/m of total unvegetated channel width of all channels for medium and large groups, respectively (Fig. 3C, E). Model validation results indicated an adequate and good model fit for large and small groups, respectively, where the slope of the regression lines and 95% confidence intervals averaged 0.323 (95% CI 0.081–0.565) for large groups and 0.907 (95% CI 0.111–1.704) for small groups. However, model validation results for medium groups indicated an adequate to poor model fit where the slope of the regression line and 95% confidence interval averaged 0.205 (95% CI –0.055 to 0.465), so results for medium-sized groups should be interpreted with care. Arbitrary group size cutoffs are a potential reason for the reduced certainty in medium group size model fit. Groups of 500 Sandhill Cranes need less area of suitable depths to roost, whereas the larger groups of 5000 birds may select channels more like groups >5000 in size along the central Platte River and thus require wider channels that provide more area of suitable depth.

Discussion

The most important factor regulating the numbers of cranes using the central Platte River is roosting habitat availability (Krapu et al. 1984, Tacha et al. 1992, Faanes and LeValley 1993). Consequently, data on habitat availability and use are key components in understanding crane-habitat relationships and for developing sound conservation plans. Sandhill Crane distribution along the Big Bend of the Platte River is determined by the availability of adequate roosting habitat more than any other factor (Krapu et al. 1984, Sparling and Krapu 1994). Substantial changes in river characteristics, including changes in river flow, have been implicated in the cranes' abandonment of large portions of the Platte River in west central Nebraska (USFWS 1981, Faanes and LeValley 1993, Krapu et al. 2014). Changes in the spatial distribution of Sandhill Cranes along the Platte River have been associated with changes in river morphology (Krapu et al. 1982, Faanes and LeValley 1993). Reduced flows in the Platte River have allowed for the encroachment of willows (Salix spp.), cottonwood (Populus deltoides), eastern red cedar (Juniperus virginiana), and more recently phragmites (Phragmites australis) onto sandbars used by Sandhill Cranes for roosting (Lewis 1976, Williams 1978, Currier 1982, Norling et al. 1992). As woody and exotic encroachment have increased following the appropriation of 70% or more of the Platte River's flows, large western portions of the Big Bend region have been abandoned by spring-staging Sandhill Cranes as channels have become narrower and deeper (Krapu et al. 1982, Faanes and LeValley 1993, Currier 1997). Faanes and LeValley (1993) reported that the distribution of staging Sandhill Cranes shifted eastward along the Platte River from 1957 to 1989, decreasing where woody vegetation encroachment and channel narrowing were the greatest and increasing along reaches that remained relatively wide and unvegetated. This abandonment has resulted in higher concentrations of cranes in a smaller reach of river, increasing the potential for an epidemic outbreak of diseases (Vogel et al. 2013). Because of the increased risk of disease, active channel width has been artificially maintained and vegetation growth on sandbars has been mitigated over the last few decades by the efforts of environmental organizations such as the Platte Valley Weed Management Area, the USFWS Partners for Wildlife Program, the Platte River Whooping Crane Maintenance Trust, and the Platte River Recovery Implementation Program (Pfeiffer and Currier 2005). These groups currently manage in-channel vegetation in the central Platte River nearly every fall through use of heavy machinery and through airboat and aerial applications of herbicide.

In agreement with the literature, medium and large Sandhill Crane groups selected areas with wider channel widths (USFWS 1981, Krapu et al. 1984, Norling et al. 1990, Sidle et al. 1993), while small groups selected channels with moderate widths. Nearly 100% of Sandhill Cranes observed in many studies of the central Platte River, including ours, roosted in river channels >50 m wide (Krapu et al. 1984, Norling et al. 1990, Sidle et al. 1993). Large Sandhill Crane groups used channels where maximum unvegetated widths averaged 307 m and selection for MUCW was maximized at 275 m. Selection for small groups averaged 135 m and their selection was maximized at 131 m TUCW. Medium-sized Sandhill Crane groups used channels where maximum unvegetated widths averaged 258 m and selection for TUCW was maximized at 267 m. Sidle et al. (1993) found that more than two-thirds of cranes roosted in river channels wider than 150 m. Krapu et al. (1984), Davis (2001, 2003), and Pearse et al. (2017) reported that most roosting Sandhill Cranes selected river channels with widths >150–200 m and avoided river channels with widths ≤100–150 m. Davis (2001, 2003) and Krapu et al. (1984) also reported that most roosting Sandhill Cranes were observed in the widest river channels and avoided narrow channels. In contrast, Norling et al. (1992) and Parrish et al. (2001) found that Sandhill Cranes selected channels with moderate widths and that large and small flocks were located in wide channels as well as narrow channels. However, the discrepancy in selection for various channel widths is likely related to the way that “channel width” was defined and estimated in each of these studies, the timing of flights (middle of the night vs. after twilight) if crane groups in narrower channels depart the river prior to being detected after twilight, or the flow during survey periods. Our findings indicate that the optimal channel width for roosting Sandhill Cranes largely depends on crane group size and water depth as a function of discharge and channel width (flow per unit width of channel).

Fig. 3.

Influence of variables included in the top model on predicted relative selection ratios, with 90% confidence intervals, for (A) small, (B, C) medium, and (D, E) large crane groups. Variables included maximum unvegetated channel width of the channel where the use or available locations were (MUCW), total unvegetated channel width of the channel where the use or available locations were (TUCW), and flow per meter of total unvegetated channel width of all channels (DISCH).

The availability of suitable roosting habitat for Sandhill Cranes can be maximized under certain flow conditions because factors such as water depth and velocity that are important in determining the suitability of an area for roost sites are related to flow (FERC 1994, Kinzel et al. 2009). Norling et al. (1992) and Kinzel et al. (2005) found that Sandhill Cranes generally roosted in shallow channels with velocities <70 cm/s. Because channel width has always been evaluated independently of channel depth, it is possible that narrow channel use is not limited so much by a requirement for wider channels but by deeper water that flows through these channels (Latka and Yahnke 1986, Folk and Tacha 1990). Differences in channel flow during study periods may explain why Norling et al. (1992) and Parrish et al. (2001) found that Sandhill Cranes selected moderately wide channels (50–200 m) and found no preference for the widest channels. Theoretically, the widest channels should be the most preferred in times of higher discharge when dry sandbars within wide reaches are temporarily submerged, creating shallow roosting habitat, whereas narrower, more incised channels may provide adequate roosting habitat during dryer periods when water in those channels is appropriately shallow. This may be especially true when narrower channels are in close proximity to high-value foraging locations such as wet meadows (Sparling and Krapu 1994). Flow per unit of channel width is an indirect measure of average channel depth and is an important metric to consider, one that has never been evaluated directly. Sandhill Cranes respond to changes in flow by adjusting where they roost within the channel. Davis (2001) indicated that Sandhill Cranes use narrow channels—those that are normally too deep for roosting—during periods of low flows and have been observed roosting in shallow water areas adjacent to islands, on top of cleared islands, and in grasslands adjacent to the river during periods of high flows. In March 2017, flows exceeded 100 cms (3550 cfs) and the majority of the Sandhill Cranes roosted on land, forgoing the river during a period of high winds that created white-capped waves in the main channel of the Platte River (Platt River Whooping Crane Maintenance Trust, Brice Krohn, Vice President, personal communication).

We found that suitable areas within the river channel for Sandhill Crane roosts are identified not only by unobstructed visibility but partially by water depth as well, which is similar to the results of a few other studies (Folk and Tacha 1990, Norling et al. 1992, Kinzel et al. 2009). Although Pearse et al. (2017) were unable to establish a link between their surrogate measure of water depth and Sandhill Crane roost-site selection, average maximum daily flows at Grand Island during the spring staging season were more than 50 cms (∼1750 cfs) higher during our study period than they were during the time frame of the Pearse et al. (2017) study (100 cms in 2007–2017 vs. 45 cms in 2001–2005). Similarly, average daily flows at Grand Island during the spring staging season were nearly 25 cms (900 cfs) higher during our study period than they were during the Pearse et al. (2017) study period (51 cms [1791 cfs] and 26 cms [921 cfs], respectively). The difference in how the flow-to-depth relationship was estimated and included within the model sets and the very different ranges of flow observed during our study and the Pearse et al. (2017) study likely account for the differences in conclusions between the 2 studies about the importance of flow-to-depth relationships to roost-site selection by Sandhill Cranes along the central Platte River. Kinzel et al. (2009) predicted the area of suitable depth for crane roosts as a nonlinear relationship of river stage in which moderate flows corresponded to the greatest roost area. Likewise, Farmer et al. (2005) reported varying relationships between river discharge and available habitat for roosting Whooping Cranes (Grus americana), depending partially on channel width along the Platte River.

Our results suggest that flows between 0.0361 cms/m and 0.1420 cms/m of total unvegetated channel width of all channels would maximize selection ratios for medium and large Sandhill Crane groups. This equates to a flow of approximately 13.65 cms (482 cfs; lowest flows observed) to 39.05 cms (1379 cfs) for channel widths of medium and large Sandhill Crane roosts (∼275 m); flows above these levels reduce habitat availability. Channels of these widths typically occur in areas that have been managed on a near-annual basis, such as the Rowe Sanctuary and properties of the Platte River Whooping Crane Maintenance Trust and the Platte River Recovery Implementation Program. Similarly, Kinzel et al. (2005) examined results at several stream flows and reported that the largest amount of available roosting habitat in channels that are generally 250 m wide occurs when the stream flow is about 35–40 cms (∼1200–1400 cfs). Prince (1990) and Simons and Associates Inc. (1990), who used a Geographic Information System and hydraulic modeling approach, also estimated that optimum flows for Sandhill Crane roosting habitat were between 30 cms and 40 cms. Based on several models originally developed by the Platte River Management Joint Study Biology Workgroup (1990), optimum flows for Sandhill Crane roosting habitat on the Platte River near Grand Island, Nebraska, were generally between 30 cms and 50 cms. Currier and Eisel (1984), however, concluded that marginal roosting habitat was available on the central Platte River at lower flows (i.e., ≤30 cms), while adequate and ideal habitat was provided at the highest flows (i.e., ≥57 cms) within a 150-m-wide channel required by Sandhill Cranes, but these assessments were based on qualitative judgements rather than on statistical analyses (Safina et al. 1989).

Our results further clarify the dynamics of Sandhill Crane roosting habitat in relation to channel width and water depth as a factor of discharge along the Big Bend of the Platte River. Integrating our research with the existing literature can allow a specific target range of flows based on the management objective of optimizing Sandhill Crane habitat during the spring migration period. Relatively high summer flows that maintain the ecological integrity of wet meadows, which are important foraging areas for Sandhill Cranes (Currier 1990, Sparling and Krapu 1994, Currier and Henszey 1996), also maintain fish spawning habitat during summer months (Poff et al. 1997) and reduce the need for mechanical maintenance of riverine systems by preventing woody and exotic plant establishment within the high banks (Williams 1978, Currier 1997, Johnson 1997). While our flow estimates are lower than most reported in the existing literature (Currier and Eisel 1984, Kinzel et al. 2005), our findings are best applied and understood in the context of the Platte River ecosystem as a whole. They should be interpreted in the context of extensive riverine management that has maintained river segments at historically more similar but artificially wide MUCWs, given current hydrological conditions (Williams 1978, Currier 1997, Pfeiffer and Currier 2005). It is also important to note that our analyses of unobstructed channel width measurements suggest that regardless of flow, wide unvegetated channel widths represent selected roosting habitat for both medium and large Sandhill Crane roosts, which make up the majority of Sandhill Cranes counted along the Big Bend Reach of the Platte River. Sustained flows during the summer months help maintain these wide-open channels. Reserving and retiming flows that are not necessary during Sandhill Crane and Whooping Crane migrations to meet the aforementioned objectives may provide tangible ecological benefits to both species and save conservation resources by reducing the need for mechanical intervention to maintain wide channels during and following dry summers.