Exploring the role of parental proximity in the maternal–neonate bond and parental investment in moose (
 Alces alces
 ) through postcapture movement dynamics

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2021 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd. 1Forest Wildlife Populations and Research Group, Minnesota Department of Natural Resources, Forest Lake, MN, USA 2Department of Fisheries, Wildlife, and Conservation Biology, University of Minnesota, Saint Paul, MN, USA 3Department of Geography, Environment, and Society, University of Minnesota, Minneapolis, MN, USA 4Wildlife Biometrics Unit, Section of Wildlife, Minnesota Department of Natural Resources, Forest Lake, MN, USA 5Department of Veterinary Population Medicine, University of Minnesota, Saint Paul, MN, USA 6Farmland Wildlife Populations and Research Group, Minnesota Department of Natural Resources, Madelia, MN, USA

The welfare of offspring relies not only on the dam's ability to provide a healthy internal preparturition environment, but on her proficiency at affording the neonate an ecologically appropriate and safe external environment following birth (Keech et al., 2000(Keech et al., , 2011Ralls et al., 1986;Severud et al., 2015;. Spatial proximity to its neonates is likely critical to the dam's success . The ever-changing natural environment poses many challenges to the dam's ability to provide for the basic needs (food, water, cover) of its offspring, whether related to habitat, or natural and anthropogenic disturbances associated with weather, fire, predation, or humans. Indeed, research is increasingly demonstrating the notion that human disturbance stimuli are analogous to predation risk for prey species, because responses to both "…divert time and energy from other fitness-enhancing activities…" including feeding and parental care (Gutzwiller et al., 1994;Lima, 1998;Lima & Dill, 1990;Steidl & Anthony, 2000). Specifically, studies have shown that variability of perceived risk (e.g., attack and capture probabilities) induces similar responses by prey; the greater the perceived risk, the stronger the response (Abrams, 1993;Frid & Dill, 2002;Hugie & Dill, 1994). Disturbance stimuli associated with humans approaching on foot may be indistinguishable by prey from true predatory stimuli (Frid & Dill, 2002). Behavioral responses by dams to such challenges may be quite variable, ranging from steadfast commitment to their young, including aggressive defense of them, to immediate or ultimate abandonment (Alexander et al., 1984;DelGiudice et al., 2015DelGiudice et al., , 2018Keech et al., 2011;Livezey, 1990;Ralls et al., 1986).
In a world of intensifying wildlife management, enhancing our understanding of influences of human disturbance is increasingly critical. Just as neonates are not passive recipients of maternal care, the strength of the maternal-neonate bond is not determined solely by the dam's behavior. Rather, interactive behavior between neonates and their dams helps to insure successful rearing through this most vulnerable period. Stringham (1974) reported that for semicaptive moose (Alces alces), a period of "intensive reciprocal stimulation" begins between a calf and its dam immediately after birth. Studies of captive and semicaptive moose indicated that the persistence of a short visual and vocal distance between the dams and their neonates is important to the interactive formation and strengthening of the maternal-neonate bond (Bogomolova et al., 1992;Bubenik, 2007;Cederlund, 1987;Stringham, 1974). Furthermore, evidence from studies of domestic ruminants indicates that early activity and associated energy levels of neonates, olfactory and visual cues, vocalizations, and licking of fluids from neonates by dams all contribute to the maternal-neonate bond and to the establishment and maintenance of maternal investment behavior (Hersher et al., 1963;Nowak et al., 2000). Moreover, neonate behavioral effects on dams can be as profound as nipple localization and suckling stimulation increasing her prolactin concentrations, a hormone important to lactation (Nowak et al., 2000).
Proximity of cervid dams to their neonates is important to establishing and maintaining their bond and to neonate survival, but periodic separation is a natural part of this relationship and is essential to the mother's survival (Bubenik, 2007;Carstensen Powell et al., 2005;DelGiudice et al., 2018;Ralls et al., 1986;. For moose, elk (Cervus elaphus), caribou (Rangifer tarandus), and northern deer (Odocoileus spp.), parturition immediately follows the nutritional bottleneck of winter. Consequently, maternal investment must include acquiring sufficient nutrition to support energy-costly lactation and to begin recovery of the dam's body condition, which is important to defending themselves and their newborns (Poole et al., 2007;Robbins, 1993;Smith, 1987). Because the birthing season typically begins just prior to spring green-up, and condition of individual dams and habitat quality naturally vary, separation distance and time apart can also be quite variable . Furthermore, disturbances, whether natural (e.g., predation) or from anthropogenic sources (e.g., neonate capture/release), may also influence aspects of these separations, and presumably, negatively impact the interactive reciprocal stimulation between dams and neonates, ultimately compromising the bond Johnsen, 2013). Enhancing our understanding of variability of the dam-neonate bond and maternal investment of large, mobile species, such as moose, that generally inhabit rugged forested terrain and are reclusive during calving can be particularly challenging.
Recent advancements in global positioning system (GPS) collar technology and novel analytical approaches allowed us to GPScollar free-ranging, moose neonates of GPS-collared dams and monitor synchronous hourly movements for a parent study focused on calf survival, cause-specific mortality, and impacts on projections of a population that had recently and rapidly declined (DelGiudice, 2012;Severud et al., 2015;. This first-timeever opportunity to intensely monitor free-ranging dams and their neonates had facilitated an unprecedented ability to track capture-induced abandonment as it occurred in near real time, and consider new approaches to neonate capture and handling that would minimize this human-induced disturbance of the maternal instinct Severud et al., 2016). Up to this point, most of the reported accounts of this highly variable postdisturbance abandonment behavior in wild ungulates, free-ranging and in captivity, were anecdotal, and perhaps often misclassified (Livezey, 1990;Patterson et al., 2013).
Herein, we expanded our examination of the movement behavior of 50 free-ranging GPS-collared moose dams and their 74 GPScollared neonates during 48-96 hr following disturbance by our neonate capture operations to better understand the potential role proximity between them plays in maintenance and variability of the maternal-neonate bond and maternal investment.
We predicted that: 1. The 96-hr postcapture HR of the abandoned neonates would be smaller than those of neonates not abandoned, whereas HRs of the dams continuing investment would be smaller than those of abandoning dams.

| Adult capture and monitoring
On 1st May 2013 and 2014, we began computer-monitoring hourly location-fixes of 73 and 70 adult (>1.5 years) female moose, respectively. The animals had been captured by aerial darting with carfentanil, thiafentanil, and xylazine during late January-early February 2013 (73) or early February 2014 (18), as part of a companion survival and cause-specific mortality study; immobilizations were reversed with naltrexone and tolazoline (Butler et al., 2013;Carstensen et al., 2014Carstensen et al., , 2017. Handling included fitting the moose with Iridium GPS collars (Vectronic Aerospace GmbH, Berlin, Germany). Fifty of the 73 and 14 of the additional 18 adult females captured in 2013 and 2014, respectively, were determined to be pregnant by serum progesterone concentrations ≥2.0 ng/ml (Butler et al., 2013;Murray et al., 2006;Testa & Adams, 1998;M. Carstensen, MNDNR, unpublished data). A last incisor (I4) was extracted from most adults for aging by counting cementum annuli in the laboratory (Sergeant & Pimlott, 1959).
Each year, adult hourly location-fixes were transmitted 4 times daily to our base station, which permitted monitoring their movements in near real time . Our primary monitoring objective was to identify when and where pregnant females made a "calving movement," a variable, atypical long distance movement (e.g., 0.4-22.7 km over a mean 14.4 hr; Severud et al., 2015) ending with localization of spatially clustered location-fixes for 1-15 days DeMars et al., 2013;McGraw et al., 2014;Poole et al., 2007;Severud et al., 2015).
We employed a 3-pronged monitoring approach, which included a base station computer receiving daily-transmitted location-fixes, a web-mapping service, and automated reports (see details . This approach provided 24/7 access to raw and processed (distances between locations) location-fix data, views of location data overlaid on Google Earth (Google, Mountain View, California, USA) imagery, and automated reports, updated every 12 hr, and including plotted mean hourly distances moved for up to 10 days.

| Neonate capture and handling
During both years, we assumed parturition occurred within 12 hr of a dam beginning localization and then allowed a minimum of 24 additional hours for bonding . Mean allowable bonding times were 40.6 hr (±3.1 [SE]) and 50.4 hr (±3.7) during 2013Severud et al., 2015).
In 2013, ground captures were helicopter-assisted (Quicksilver Air, Inc., Fairbanks, Alaska), locating the calving site from overhead with dam GPS coordinates we provided, landing some distance away to allow 1-2 handlers to disembark, guiding them in from overhead via 2-way communication, and then landing again out of sight. The calfhandling protocol included fixing ear tags; collecting blood samples from the jugular vein; recording body mass, other morphological characteristics, and rectal temperature; fitting a 420-g GPS collar (GPS PLUS VERTEX Survey-1 GLOBALSTAR with expandable belt; Vectronic Aerospace GmbH, Berlin, Germany); and examination for injuries or abnormalities. These collars were programmed to collect hourly fixes in sync with those of their dams Severud et al., 2015). Calf location-fixes were transmitted to our based station 8 times per day (every third successful fix).
All fixes were also stored on-board.
We had planned to limit time for capturing calves when dealing with aggressive dams (e.g., piloerection, foot stomping, roaring, or charging) to 10 min and designed the handling protocol to be 5-6 min to limit separation from the dam (DelGiudice et al., 2015; Keech et al., 2011). Capture-related data are summarized in Table 1.
There were no differences in birth-or captured-related characteristics between newborns that were abandoned within 48 hr of release versus those not abandoned . Twins were captured, handled, and released simultaneously when encountered to otherwise limit abandonment of the first twin handled (Keech et al., 2011).
We divided our 2014 neonate capture season (no helicopter assistance) into 2 phases as summarized in Table 1. During phase 1, our team of 3-4 biologists approached the calving site (cluster of dam's location-fixes) on foot at a modest pace, until we were within 50 m, at which point we ran, captured, and handled all neonates according to the same protocol used in 2013, including the handling of twins . However, after the first dam we approached abandoned her twins subsequent to their release, we excluded blood sampling in an attempt to be less invasive . However, abandonment continued intermittently, and as in 2013, there was no informative temporal pattern of the behavior to use as guidance for modifications Severud et al., 2016).
We ceased operations to reconsider our capture approach and handling protocol. Other than highly variable, anecdotal accounts, the literature offered little understanding of captureinduced abandonment behavior in ungulates. However, studies of behavioral responses of ungulate prey to predators and human hunters (Abrams, 1993;Frid & Dill, 2002;Hugie & Dill, 1994) and one comparison of moose neonate capture techniques in Alaska (Ballard et al., 1979), proved to be of notable value. Consequently, we resumed captures on 21 May (start of phase 2) with a reduced capture team of 2 biologists, a more rapid approach time, and a protocol that limited handling to fitting the GPS collar, sexing each neonate, and visually scanning for injuries and abnormali-

| Data analysis
We recorded all locations from capture to 96 hr postcapture and employed fixed kernel density estimation (KDE) with an ad hoc optimal smoothing parameter (K href ) using AdehabitatHR in program R to delineate 95% and 50% (core) HRs (R Core Team, 2017; Worton, 1989Worton, , 1995. Two of 13,066 (0.01%) total location-fixes thought to be er-  . c Overall twinning rates were 58.1% and 32.0% in 2013 and 2014. V. St-Louis, MNDNR, unpublished data  where di d and di are displacement and direction components, respectively, and measure cohesiveness in displacement and direction separately for corresponding segments in time t for individuals a (e.g., dam) and b (e.g., neonate). The term di d ranges from 0 to 1, from no interaction (0) to maximum interaction (1), respectively. However, the di index varies between −1 for an opposing pair of directions to 1 for an aligning pair of movement vectors in the same direction (Long & Nelson, 2013). The global index (di) represents the average of localized indices throughout an entire trajectory of n points and n−1 movement vectors (Equation 4). The di statistic ranges from −1 (avoidance) to 1 (identical movement behavior).
We quantified the aforementioned similarity indices for each dam-neonate pair and calculated summary statistics for nonabandoning (continued investment) and rejection/abandoning groups.
The three approaches that we present are complementary, because the DI approach does not take into account the spatial proximity between a dam and its neonate; examining HR overlap and reunion statistics provided a more complete picture of the movement patterns and relative space use of dam-neonate pairs.
We compared mean 50% and 95% KDE HR areas, HR spatial overlap, proportion of reunions, and the 3 movement similarity matrices of the nonabandoning and abandoning groups using 95% CIs calculated around the means (mean ±1.96*SE) to assess significant differences. Because of the tight connection between CIs and p-values, this is equivalent to inferring statistical significance at the .05 level (Murtaugh, 2014). Because previous work showed that the general movement patterns did not differ across years,
The mean proportion of hourly location-fixes indicative of damneonate reunions (≤26 m apart) during the 96 hr postdisturbance was significantly (p ≤ .05) higher for those dams accepting and continuing to invest in their neonates (0.37, range = 0.02-0.59) than for dams in the process of abandoning their offspring (0.01, range = 0.0-0.05; Figure 3). When the maternal-neonate bond remained intact, the mean proportion of hourly reunions was highest (0.42) during the first 24 hr postdisturbance, but did not change significantly (p > .05) throughout the 96-hr monitoring interval ( Figure 3). The mean interval between these reunions was 2.7 hr (95% CL = 2.5-3.0, range = 1.8-6.9 hr). For the dams that abandoned their neonates, there was only one reunion interval for a dam-neonate pair that allowed measurement (7.5 hr), which occurred within the first 24 hr.
We measured overall di, and its two components, di θ and di d , separately for each dam-neonate pair. In analyzing the movement data of our dams and neonates, there were no significant differences in di θ during the first or second 24-hr intervals postdisturbance, or during the second 48 hr of monitoring, and mean values were relatively close to 0, indicating no similarities or strongly opposing hourly movements relative to direction (Figure 4). Because di is derived from the product of di θ and di d , the low and nonsignificant di θ findings were enough to

| D ISCUSS I ON
Linear movements of free-ranging moose dams and their neonates can be quite variable, but on average, the movements of dams that remained in close proximity to their neonates and continued parental investment after human disturbance have been F I G U R E 2 Mean (95% confidence interval) percent overlap of 50% and 95% kernel density estimator home ranges of moose (Alces alces) dams and their neonates during 96 hr postcapture/ release of neonates when maternal investment continued (circle, n = 56 pairs) versus when rejection and abandonment occurred ( These divergences in their movement dynamics, and consequently in the dam's parental investment, were mirrored by marked differences between neonates benefiting from that investment versus those abandoned. Findings from additional multidimensional metrics presented herein broaden and deepen our understanding of dam and neonate movement behavior, and how they maintain the spatial proximity that apparently is important to reinforcing their bond and supporting parental investment. This is highlighted particularly by the greater mean HR size of the seemingly more vital neonates that were not rejected, twice that of abandoned neonates, and average greater overlap of both their core and 95% KDE HRs with those of their dams. Nevertheless, interestingly, how the 95% KDE HRs of dams and their neonates overlap on the landscape, whether investment continues or not following disturbance, can vary markedly within either group. Figure 5 provides examples of 2 pairs from each group, to offer insight as to how dams and neonates with the stronger bond may move and use their environment compared with dams and neonates where the bond has functionally broken down. We previously reported that despite the disturbance of our capture operations, most GPS-collared dams remained on average <500 m from the capture sites and stayed even closer to their newborns, a mean 256 m (2013) and 102 m (2014), during the 2-4 days postcapture . Furthermore, the proximity maintained between these dams and their neonates was in part due to the steadily increasing distance their neonates put between themselves and their respective capture sites, moving in the direction of their dams . This is consistent with their more interactive and similar movement behavior and apparent reciprocal stimulation that appears to nurture and strengthen the maternal-neonate bond and encourage investment by dams (Bogomolova et al., 1992;Bubenik, 2007;Cederlund, 1987;Stringham, 1974). Our present findings suggest that dams that continued to invest in their neonates during this critical interval following parturition, and more importantly, the capture disturbance, were able to at least meet their immediately heightened nutritional needs associated with lactation (Robbins, 1993), within variable, but relatively small HR areas around their neonates ( Figure 5). This would appear critical to the "parent-offspring conflict" discussed by Trivers (1985), where the dam is selected to balance the benefits of continued investment in the neonate's rapid increase in nutritional requirements and survival, with the costs of investment in her own condition and future reproductive success.
In marked contrast, the smaller HRs of the rejected neonates, which had remained close to the disturbance sites, typically showed little to no overlap with the much more expansive 50% and 95% KDE HRs (Figures 1 and 5) of their frequently more distant and far-ranging dams . This appeared to reflect less vigor in offspring in the process of being rejected and the markedly diminished willingness of their dams to accept or invest in them, both negatively impacting their bond (Langenau & Lerg, 1976). Previous reports showed maternal deer reduced their daily movement rates immediately postparturition and remained close (means of 130-140 m) to their neonates (Carstensen Powell et al., 2005;Hawkins & Klimstra, 1970;Ozoga et al., 1982;Peterson et al., 2018); however, to the best of our knowledge, there are no other reported data documenting HR size of moose or other cervid newborns, or the degree to which they share the HR of their dams (Figures 1, 2, and 5). This is where the ability to GPS-collar newborns and their dams , and record synchronous hourly location-fixes, provided an invaluable advantage for our unique, more in-depth examination.
The value of close proximity to strengthening and sustaining the maternal-neonate bond was indicated further by the high frequency F I G U R E 4 Mean (95% confidence interval) interval and global dynamic interaction (di) of movement data of moose (Alces alces) dams and their neonates, including movement direction (di θ ) and displacement (di d ), during the first and second 24 hr, second 48 hr, and first 96 hr postcapture/release of neonates when maternal investment continued (circle, n = 52−56 pairs) versus when rejection and abandonment occurred (triangle, n = 8−18 pairs), May−June 2013 and 2014 (pooled), northeastern Minnesota. Small solid symbols represent individual data points of reunions to within 26 m of nonabandoning dams and their neonates. Importantly, our synchronous hourly location-fixes likely provided a minimum number of reunions; the actual frequency of reunions during the 96-hr monitoring interval may be higher and more variable. Bogomolova et al. (1992) noted that captive moose dams remained within 50 m of their neonates for 7-8 days and contended that a strong maternal-neonate bond required constant contact. Our estimates of HR overlap of dams and neonates and hourly distances between them, provide useful context for contact and close proximity, relative to bonding for free-ranging moose. Still, it is likely that functionally, these metrics may be quite variable because of the mixed influence that olfactory, visual, and auditory cues, and numerous other factors have on this bond and parental investment (Nowak et al., 2000;Stringham, 1974). Indeed, the quality of these interactions, which may in part be related to the distance separating dams and their neonates, and time apart notably influence neonate survival (Estes & Estes, 1979). Based on location-fixes of these dam-neonate pairs, we estimated a mean reunion interval of about every 2.7 hr, although as indicated by our hourly data, this interval can be quite variable. Although we cannot confirm that calves were nursing during these reunions, we are confident it was highly likely. Periodic nursing bouts not only provide nutrition vital to the calf's survival, but the associated suckling activity, sustained energy level, supported vitality, and associated aforementioned cues are all intricate contributors to reinforcing and sustaining the maternal-neonate bond and assuring parental investment (Langenau & Lerg, 1976;Lickliter, 1985;Nowak et al., 2000;Stringham, 1974).
In captive and semicaptive moose, frequency and duration of nursing bouts for moose neonates have been variable and change as the neonates age and begin foraging on solid foods, which occurs within 2-3 weeks (Bubenik, 2007;Schwartz, 1992). Stringham (1974) emphasized the absence of a strict nursing schedule, but observed about 1 nursing bout per hour during the first week of life and 1 bout every 3 hr by week 9, but these were semicaptive moose. Similarly, in captivity, deer neonates were observed to nurse at least hourly, and frequently up to 3 bouts per hour, whereas free-ranging does may seek out and attend their neonates from 2 to 10 times per day (Robbins, 1993;White et al., 1972). Noteworthy, the mean high F I G U R E 5 Examples of overlapping 95% kernel density estimator home ranges of moose (Alces alces) dams (light gray) and their neonates (lined) during 96 hr postcapture/release of neonates when maternal investment continued (top left, ID 26698565, 92.7% overlap;top right, ID 26698665, 74 (Henry et al., 2009). The moose calf's metabolic rate is greatest during the first 2 weeks of life, and its associated energy requirement can be increased by not only cold weather and precipitation, but by the stress of the dam's absence (Henry et al., 2009;Robbins, 1993). These neonates require a sizable volume of highly digestible milk during their first week (estimated 3-4 L/24 hr) to meet their high energy requirements (Gasaway & Coady, 1974;Reese & Robbins, 1994;Robbins, 1993). The progressive nutritional deprivation of neonates in the process of being rejected steadily reduces their energy levels and vitality, as does the absence of nurturing (e.g., licking) from the dam (Hersher et al., 1963), which in turn, importantly may reinforce the dam's instinctive response to reject (Langenau & Lerg, 1976;Lent, 1974;Nowak et al., 2000;Stringham, 1974;Verme, 1962).
The most notable finding from our examination of dynamic interaction of movement data was the stable (mean di d = 0.45-0.48), relevant correlation of mean hourly distances moved (di d ) during the 96-hr monitoring period overall, and particularly during the first and second 24-hr intervals for those dams and neonates that remained in close proximity. This is perhaps indicative of a self-reinforcing and stronger maternal-neonate bond. Maximum di d values documented were 0.85, 0.65, and 0.56 during the first and second 24-hr and F I G U R E 6 Examples of local analysis of dynamic interaction of movements (distance, di d ; direction, di θ; di = di d × di θ ) of moose (Alces alces) dams and their neonates during 96 hr postcapture/release of neonates when maternal investment continued (nonabandoning) versus when rejection (abandonment) occurred, mid-May−mid-June 2013 and 2014, northeastern Minnesota. Dam-neonate pair IDs also correspond to those in Figure 5 second 48-hr intervals, respectively. Corresponding maximum values for rejecting dam-neonate pairs were 0.55, 0.58, and 0.56, respectively. The much higher interaction with their neonates by the group of investing dams during the first 24 hr postdisturbance compared with the second 24 hr and second 48 hr (Figure 4) may reflect a degree of recovery from capture and handling-induced stress subsequent to the first 24 hr. We did not expect to calculate a mean di d of the abandoning dams and their neonates as high as 0.41 (96 hr), given the often markedly greater distance between them compared with dams that consistently invested in their offspring. It may be that all of the mean di d values are somewhat low, attributable to the hourly synchronous location-fixes rather than more frequent sampling (S. Dodge, Department of Geography, University of California-Santa Barbara, personal communication). Additional research should explore whether more frequent location-fixes would be more sensitive to the dynamic interaction of movements of moose dams and their offspring. Figure 6 provides examples comparing the stronger movement interactions of individual dam-neonate pairs, as indicated by di, di θ , and di d , associated with persistent maternal-neonate bonds versus when neonates were rejected and ultimately abandoned by their dams. These are the same pairs used to reflect the differences in HR overlap in Figure 5 and clearly illustrate the variable, but more frequent influence that dams maintaining close proximity can have on the movements of their offspring (note the greater frequency of values between 0.5 and 1.0).

| CON CLUS IONS
The location-fix and movement data collected during our study supported all of our predictions but one. The value of close proximity of dams to their neonates for nurturing and maintaining a strong bond and parental investment, particularly following disturbance, is reflected by our calculations of core and 95% KDE HR size, percent overlap of dam and neonate HRs, frequency of dam-neonate reunions, and mean temporal intervals between reunions. Furthermore, we believe that the interactive movements and reciprocal stimulation facilitated by this proximity, as characterized, contributed to nurturing neonate vitality that likely positively reinforced or encouraged dams to continue investment, promoting a positive feedback loop.
In contrast, the collective data associated with rejection strongly indicate that the absence of close proximity and limited reciprocal stimulation (including nursing), fed a negative feedback loop, so that as neonate vitality diminished (lower movement rates), the dam's decision to reject and abandon was reinforced. We have an enhanced understanding of a functional meaning of close proximity between dams and neonates relative to successfully maintaining a strong maternal-neonate bond, which has implications for survival. We believe our approach, metrics, and findings may have applicability relative to other disturbances, depending on their specific dynamics. Clearly, the strength of the maternal-neonate bond of moose and other cervids can be further assessed and better understood (e.g., parent-offspring conflict) following any variety of disturbances by more frequent, synchronous sampling of location-fixes and using metrics reflective of dam-neonate proximity.

ACK N OWLED G M ENTS
We thank the many individuals that assisted in so many ways to facilitate this more thorough examination of maternal-neonate bond, parental investment, and rejection of moose neonates, including K. J.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available at the Data Repository for the University of Minnesota (DRUM), https://doi.org/10.13020/ rkpf-6h68.