2004 Sheep Report, Chapter 3

Wild sheep have been shown to have varying short term or instantaneous behavioral reactions to aircraft activity (MacArthur et al. 1982; Krausman and Hervert 1983; Stockwell et al. 1991; Krausman et al. 1993, 1998; Weisenberger et al. 1996) leading to concerns of negative impacts on Dall’s sheep (Ovis dalli) from low level military aircraft in Interior Alaska Military Operations Areas. Wildlife behavior has been used as a means for evaluating anthropogenic disturbance with the premise that changes in individual behavior may have long–term population level consequences (Klein 1973, Harrington and Veitch 1991, Frid and Dill 2002). Various metrics of animal behavior and activity have been used to evaluate the potential effects of overflights on wildlife species. Three such measures are the proportion of time animals spend active, feeding efficiency, and proportion of time (animals) are engaged in specified behavior categories (i.e., activity budgets).

One of the most basic measures of animal activity is the proportion of time animals spend active. Harrington and Veitch (1991) found an increase in daily activity in caribou (Rangifer tarandus) related to military overflights in 1 field season out of 3. Murphy et al.(1993) and Maier et al. (1998) found increased activity in caribou overflown by military jets in comparison to caribou that were not subjected to overflight activity in 2 out of 3 sampling periods during 1991. Heart rate of wild bighorn sheep (Ovis canadensis; MacArthur et al.1979) and energy expenditure in wild ungulates in general increases as animals go from a position of inactivity (bedding) to active (standing, feeding walking and running) (Fancy and White 1985). Therefore, in the instances cited above, military overflights caused an increase in energy expenditure in caribou but cumulative seasonal or annual changes in energy expenditure have not been estimated.

A reduction in feeding efficiency has also been used as an indication of disturbance (Berger et al.1983, King and Workman 1986, Stockwell et al.1991). Foraging efficiency can be defined as an index of time spent feeding relative to time spent scanning. Stockwell et al.(1991) found that in winter with intense helicopter activity, bighorn sheep foraged less efficiently in the presence of helicopters than in the absence of helicopters.

Wildlife activity budgets are useful for evaluating potential disturbance because trade-offs in behavior can be evaluated and a relative degree of disturbance can be assigned based on a specific change in behavior. MacArthur et al.(1982) found that bighorn sheep exposed to helicopter overflights at 90-250 m above ground level, responded by running. Murphy et al. (1993) found that caribou spent significantly less time lying during post-calving and insect season if they had been overflown in the last 15 minutes in comparison to a group that had not been overflown. Frid (1998) reported that 75% of Dall’s sheep groups that he observed responded to low-level helicopter flights by having at least some members of the group walk or run in response to the overflight. Krausman et al. (1998) reported limited changes in proportion of time standing, walking and foraging in bighorn sheep following military overflights and concluded that there were no measurable differences in behavior between overflown and non-overflown sheep. Similarly, Krausman et al. (2001) did not, in general, find significant differences in behavior patterns of adult pronghorn (Antilocapra americana) with and without military overflight acitivity.

We examined behavior of Dall’s sheep in relation to military overflights by evaluating the proportion of Dall’s sheep that were active, Dall’s sheep feeding efficiency, and Dall’s sheep activity budgets. Military overflight activity was quantified as the number of military jets observed during an observation bout, the time relative to the overflight event (before, during or after), and flight and sound characteristics of the overflight event. We tested the null hypothesis that the proportion of Dall’s sheep active, feeding efficiency and activity budgets would not be associated with military overflights. Because group size (Frid 2003), presence of lambs, distance to steep rocky terrain (Frid 2003), and time of year (Stockwell et al. 1991, Bleich et al. 1994) could all potentially affect behavioral responses of mountain sheep, we included these variables in our analysis. To guard against differences in study site location and differences in seasonal progression across year, we also included these variables in our analysis.

Observations of Dall’s sheep were made at 2 study sites in eastern interior Alaska in the Yukon-Tanana Uplands (Fig. 3.1). One site was mitigated for potential impacts of low level military jets (Cirque Lake) and the second site was not (West Point). The 2 study sites were approximately 35 km apart and have been previously described (Chapter 1).

Mitigation measures at Cirque Lake are in effect 10 May – 15 June. During this time frame, military aircraft are to avoid the Cirque Lake lambing area by maintaining a 5,000 foot (1,500 m) floor in a 7-nautical mile (13 km) radius circle centered at 64º48’00”N, 143º45’00”W. At other times of the year, and in the surrounding area, military aircraft may fly as low as 100 feet (30 m) above ground level (AGL; Department of the Air Force 1995). Sheep at both study sites may be exposed to military aircraft during both routine flying days as well as during Major Flying Exercises. The most prominent Major Flying Exercises are termed Cope Thunders (Department of the Air Force 1995).

The Final Environmental Impact Statement, Alaska Military Operations Areas (1995) indicates that in interior Alaska, routine military aircraft training exercises typically occur between 0800h and 2200h, Monday through Friday. Cope Thunders typically occur Monday through Friday, 0800h through 1800h but are most concentrated from 1000h through 1200h, and 1500h through 1700h. A typical Cope Thunder is 10 days long. Between 55 and 110 aircraft a day can be involved in a Cope Thunder, with each aircraft flying up to 2 sorties (a sortie is a single aircraft take-off, flight, and landing) per day for a total of up to 200 sorties per day (Department of the Air Force 1995). Cope Thunders can be distinguished from routine flying by a 50% or greater increase in the daily flying activity rate relative to routine flying days or an increase of 50 aircraft operations per day (Department of the Air Force, 1995).

Two initial assumptions of this study were that: 1) Dall’s sheep at the West Point study site were exposed to greater levels of military jet overflights than were sheep at the Cirque Lake study site; and 2) military overflight activity at both study sites was greater during major flying excercises (Cope Thunders) than on routine flying days. Overflight activity at Cirque Lake was also expected to vary in relation to the mitigation measure. These assumptions were investigated by examining the number of overflights observed by field crews while observing sheep behavior and by quantifying sound levels during these overflights.

Field crews observed Dall’s sheep behavior, observed overflight activity and collected information on sound levels during military overflights during 4 field sessions per year in 2000 and 2001. All behavioral observations of Dall’s sheep occurred in Yukon MOA 1 with the exception of 1 observation period (21 February – 5 March at the West Point study site). We attempted to distribute our field efforts between those periods with the most intense overflight activity (Cope Thunders) and those periods with reduced overflight intensity (routine flying). Data gathered included the number of aircraft observed while in the field and the number of overflight events that occurred while observing sheep. We collected data on overflights at all altitudes and proximities that we could detect. Military aircraft rarely travel alone. Because of logistical constrains in regards to sound meters and data recording, data collection focused on the nearest aircraft at any given time. We considered a pass by an aircraft or a group of aircraft to be an overflight event when either the sound associated with the overflight reached a peak or the proximity of aircraft to the observed sheep reached a minimum followed by either a decline in sound level or a decrease in proximity. Therefore, multiple passes by a single aircraft resulted in multiple overflight events. Observed overflights were categorized by type as military jet, military prop., small non-military (single engine prop.), helicopter, or other (civilian multi-engine, non-military jet or unknown). Proximity of overflights were categorized as close, moderate, or far using known elevations of mountains in the study area. An overflight was considered to be close if the observer considered the aircraft overflight to be below 1500 m AGL (5,000 feet) and to be within 1.6 km (1 mile) horizontal distance of the Dall’s sheep under observation. Overflights were considered moderate if below 1500 m AGL and within horizontal distances of 1.6-3.2 km (1-2 miles). Overflights judged to be greater than 1500 m AGL and greater than 3.2 km horizontal distant were considered far. We therefore, investigated Dall’s sheep reactions to aircraft activity that we assumed to be representative of that typically occurring in the Yukon MOA 1. Analyses of Dall’s sheep behavior presented in this report focus on reactions relative to military jets. Sample sizes of overflights by military prop aircraft and helicopters were too small to allow analysis. Statistical analysis of overflights by small single engine civilian aircraft is presented in Appendix E.

Once crews were in place to observe sheep, a sound level meter (Larson Davis Model 812 sound level meter, Model 2560 microphone, and Model PRM826B preamp) was positioned at least 200 m from each crew. Sound meters were programmed to record A-weighted 2 minute average sound levels (L_eq) and peak sound levels in 2 minute blocks. A-weighting is a filter that adjusts sound level frequencies similarly to the human ear when exposed to low levels of sound and is most often used to evaluate environmental sounds (Larson·Davis 1997). For statistical analysis, we used the mean L_eq in 2 minute time blocks as L_eq has been suggested as an appropriate metric of sound measurements for wildlife studies (Krausman et al. 2001). Sound data presented in this report were the sound levels at the observation sites during the overflights, but do not represent exact sound levels or peaks that Dall’s sheep experienced during the overflights because the monitoring equipment was not directly associated with individual sheep.

The typical flying pattern during Cope Thunders was for pilots to familiarize themselves with the area for 1 – 2 days (range familiarization), followed by a 2 week exercise. Flight activity was anticipated to be most intense during the 2 week exercise period. Little flying activity was anticipated on weekends (Department of the Air Force 1995). During 2000, cancellation of Cope Thunders and closure of the Eielson Air Force air-strip reduced the number of Cope Thunder Exercises. If range familiarization flying and week ends are excluded from Cope Thunder dates, data were collected during 10 Cope Thunder days in 2000 and all these days were in the winter (March and April)(Table 3.1). In 2001, data were collected during 5 Cope Thunder days in the winter (March), 8 in the spring (May; and 5 of these 8 days were collected during the mitigation period), and 5 in the summer (July).

All behavior observations focused on Dall’s sheep ewe groups. Each day, field crews (typically 2 individuals) would hike from a base camp and attempt to locate ewe groups by using binoculars and scanning known sheep terrain. Once sheep were located, the field crew would move into an observation position. Observation positions were chosen based on a balance of minimizing disturbance to sheep and providing a clear view of sheep behavior. During sheep observations, one crew member would observe sheep using a spotting scope and the second would record data. Observations of sheep were made as early as 0700h and as late as 2000h but the majority of observations were made between 1000h and 1700h as military jet overflights were reported to be most concentrated at this time (Jim Hostman, USAF, personal communication). Data collected at the beginning of an observation session included temperature, wind speed and direction, precipitation, and distance from the sheep group to exposed rock. The “distance to exposed rock” was a subjective determination of potential escape terrain. All sheep within a group were classified by sex and age. For analysis, the numbers of “ewes” (ewes = ewes, young rams, and older lambs) in each group were enumerated and the numbers of young lambs were enumerated. Lambs were only categorized for analysis if they were less than 4 months old. Therefore, all statistical analyses in which lambs were included as a variable were restricted to May – August.

Dall’s sheep behavior was recorded using scan sampling (Altman 1974) of the entire group. Behavior of each ewe and lamb in the group was categorized as: running, walking, feeding, standing, bedding, and other (active behavior including playing, socializing, grooming, etc.). The “other” category was not used in statistical analysis. Scan samples were taken every 10 minutes and continued until animals moved out of sight or until 1700h. In the event of an overflight, the observer would estimate when the aircraft was closest to the group and would scan to get an immediate response of the group to the overflight event. This event would reset the scan sampling schedule and the next scan would occur 10 minutes later or when another overflight event occurred, whichever came first. We did not choose to include a behavioral category of “vigilant or alert” in our activity budgets as have some researchers (Stockwell et al. 1991, Frid 1997) because our experience indicated that this was a difficult behavior to identify. We observed sheep that were apparently vigilant (head up with ears forward) while bedded and standing, when approached by other sheep, during spontaneous rock fall, while observing potential predators (wolverines, bears and humans), for no discernible reason, and when overflown by aircraft. We do not know if the energy expenditure during these “vigilant” periods is equivalent. In addition, there are no indications of how energetically costly it is to assume a vigilant posture as there are for the behavior categories we defined for this study (Fancy and White 1985). Possibly the behavior category of vigilance could be defined by linking it directly to a physiological response. Factors such as an elevated heart rate (MacArthur et al.1979, MacArthur et al. 1982, Weisenberger et al. 1996, Krausman et al.1998,) would be optimal for defining vigilance. Stockwell et al. (1991) defined vigilance as a standing animal with head up. Based on this definition, our category of standing is synonymous with vigilant.

The ability to distinguish behavior categories of Dall’s sheep varied with proximity to the sheep, terrain characteristics, and weather. Distance between observers and sheep varied but all observations were made within 1.5 km of the sheep. Rocky terrain was particularly problematic as sheep behind rocks were hidden from view and this may have positively biased our estimates of distance to rocky terrain for sheep groups.

The proportion of observed sheep engaged in specific activities were used for analysis and sheep that were out of view were not included in these proportions. Some activities are likely under-represented in scan samples. For example, bedded sheep were more likely to be out of sight than were standing animals. This bias however, was not restricted to a specific study site, year or timing relative to an overflight event.

Another issue to consider is the number of animals sampled during observation periods. In most instances, some sheep observed during one 2-week field effort were sampled repeatedly. In our analysis, we considered each day and each overflight event to be independent.

We used paired t-tests to compare military and non-military overflights (total number and number per hour) observed during field sessions at Cirque Lake to overflights observed at West Point/Puzzle Gulch. T-tests were used to examine differences in the frequency of military and non-military overflights in the 2000 field season in comparison to the 2001 field season. All statistical tests were considered significant if P < 0.05.

Paired t-tests were used to compare the daily mean number of military overflights observed during Cope Thunder exercises to the mean number of military overflight events observed in either the 5 routine flying days before or the 5 routine flying days after these events (depending on which was available).

The daily number of military overflights observed at Cirque Lake during the mitigated time period was compared to the daily number observed during the same time period at West Point using paired t-tests.

Dall’s sheep behavior in relation to overflights was analyzed by following the steps outlined by Murphy et al. (1993). The number of animals engaged in each category of behavior was transformed to a percentage for each activity scan. Each scan was classified in relation to overflight history as: 1) within the 10 minute time period prior to an overflight event; 2) at the peak of the overflight event; 3) 10 minutes after the overflight, 4) undisturbed (no overflight event within 1 hour) or other (scan sample <1 hour of overflight but >10 minutes of overflight).

Three measures of behavioral activity were used to examine Dall’s sheep responses to overflights: 1) percent active (percent of animals during the scan that were standing, feeding, walking or running); 2) feeding efficiency (proportion of animals feeding divided by the sum of the proportion of animals feeding and standing: sensu Berger et al. 1983, Stockwell et al. 1991), and; 3) the proportion of animals engaged in bedding, standing, feeding, walking and running (behavior or activity budget). Dall’s sheep reactions to overflights were examined using three levels of detail of overflight activity: 1) number of military overflights observed during the day (“day flight models”); 2) military overflight events (within 10 minutes before, during or 10 minutes after individual overflight events)(“event models”), and 3) characteristics of individual military overflight events (proximity of overflight event and peak dBA level of overflight event; “sound models”). A subset of the event model data set was used to test sound models. Statistical tests in the event and sound models were considered significant if P < 0.05.

Because sheep activity may vary by season, and to match other analyses presented in this report, we divided observation data into date sequences as follows: February and March, 1 – 14 April, 15 – 30 April, 3 – 18 May, 19 May – June 1, 7 – 22 June, 12 – 27 July, 28 July – 11 August. Analyses that included the effects of lambs on Dall’s sheep behavior were limited to May – August. We labeled models that included data from all date sequences and excluded lambs from the statistical analysis as “ewe models”. Models restricted to May – August and that included the number of lambs in the groups as a covariate were labeled “lamb models”.

Analysis of covariance (ANCOVA; GLM [SPSS Inc., Chicago Illinois]) was used to examine the effect of class variables (study site, year nested within study site, date sequence nested within year nested within study site, and the proximity of an overflight event) and covariates (number of ewes in the group, the number of lambs in the group, the distance of the group to rocky terrain, the number of military overflights observed per day, and the peak dBA of the overflight event) on the percent of sheep active and feeding efficiency of Dall’s sheep.

Multiple analysis of covariance (MANCOVA; GLM [SPSS Inc., Chicago Illinois]) was used to examine the effects of class variables (study site, year nested within study site, date sequence nested within year nested within study site, and the proximity of an overflight event) and covariates (number of ewes in the group, the number of lambs in the group, the distance of the group to rocky terrain, the number of military overflights observed per day, and the peak dBA of the overflight event) on the proportions of sheep engaged in different activities (standing, feeding, bedded, walking, running). Significant MANCOVA results were examined using univariate tests of between subject effects to evaluate differences in specific behavior categories.

All data were examined for linearity, normality and homoscedasticity using scatter plots, residual plots, histograms and boxplots. It is common practice to transform percentage data using an arcsine square root transformation, and biological data (often approximating a Poisson distribution; Zar 1996) using a square root transformation (Zar 1996). The intent of data transformation is to more appropriately fit data to meet assumptions of statistical models. For this data set, transformations did not appear to improve data distribution. In addition, data transformation is not always desirable unless the largest variances are in the largest samples and the largest sample is more than 5 times the size of the smallest (Budescu and Applebaum 1981, Zar 1996), and this was not the case with this set of data. For these reasons, and to simplify the interpretation of results, none of the variables included in the ANCOVA and MANCOVA models were transformed.

There was no significant difference in the number of overflight events of all aircraft types observed at Cirque Lake in comparison to West Point (t = 0.687, d.f. = 7, P = 0.514). The number of total observed overflights varied between 12 at West Point from 21 February - 5 March of 2000, and 136 at Cirque Lake from 6 - 16 May 2001 (Fig 3.3). The total number of overflights observed in 2000 was 294, and in 2001 we observed 441 overflights.

An examination of strictly military overflights found no statistical difference (t = 0.315, d.f. = 7, P = 0.76) in the number of overflights observed when comparing Cirque Lake to West Point. The number of military overflights observed during a particular field session varied considerably from a low of 2 overflights to a high of 109. Both extremes occurred at Cirque Lake (Fig. 3.2). The number of military overflights we observed during 2000 was substantially less than observed in 2001 (n = 92 in 2000 and n = 281 in 2001). The military aircraft most commonly observed during overflight events was F-16s (15% of all military overflight events) followed by F-15s (8% of all military overflight events) and A-10s (3% of all military overflight events). In most instances, military jets were distant enough that no accurate identification of aircraft type could be made. There was no difference in the number of military overflights observed when comparing the number of military overflights observed before or after Cope Thunder periods, to the number of military overflights observed during the Cope Thunder periods (t = 0.756, d.f. = 5, P = 0.48).

Sheep were not located every day while in the field. Because of this, no data were collected on the number of military overflight events observed over Dall’s sheep in the 5 days prior to the start of the mitigation period at Cirque Lake during the May 2000 field work. In addition, no field work was done in June of that year. During 2001, a daily mean of 12 (n=4) military overflights were observed in the 5 days prior to the start of the mitigation period (May 10) at Cirque Lake and a daily mean of 20 (n=3) were observed in the 5 days after the initiation of the mitigation measures. In the 5 days prior to the cessation of the mitigation period at Cirque Lake (June 15), the mean daily number of military overflights observed was 5 (n=3) and in the 5 days following the cessation of the mitigation measures, the mean daily number of military overflights events observed while watching sheep was 2 (n=4). Therefore, in the 2 instances for which data were available, the number of overflight events observed at Cirque Lake was greater during the mitigation period then before or after. There were no statistical differences (t = 0.982, d.f. = 19, P = 0.34) in the mean (+SE) number of military overflights observed during the mitigation period over Cirque Lake (9 +2.3) in comparison to West Point (11 +2.0).

Military aircraft constituted 52% of the overflight activity observed above the Cirque Lake and West Point study sites (Table 3.2). The majority of the air traffic in these areas was above 1500 m and in the “far” category and this was true of military aircraft (84%) as well as non-military aircraft (75%). Low level overflights close to observed sheep were a relatively rare event (8% of all overflight events) and a rare event when only military overflights are considered (7% of all military flights; Table 3.2). During 2000 and 2001, few overflight events occurred during the mitigated time period and fewer still occurred during a Cope Thunder event that also occurred during a mitigated time period (Table 3.2). On average sheep were exposed to 42.9 military overflights per week and 3.1 of these were low and close.

Because of differences in the amount of time spent observing sheep between study sessions and between study sites, we examined the number of overflight events (all aircraft types) as a rate process (Table 3.3). In both study years, field crews observed fewer overflights per hour at Cirque Lake (1.12 and 1.17 overflight events per hour in 2000 and 2001, respectively) in comparison to West Point (1.22 and 1.24 overflight events per hour in 2000 and 2001, respectively). The difference between the study sites in the number of overflights observed per hour paired by field session was not significant (t = 0.634, d.f. = 7, P = 0.55). No difference was detected in the number of overflights observed per hour when comparing 2000 to 2001 (t = 0.288, d.f. = 14, P = 0.78).

Sound level data were gathered with Dall’s sheep observations for over 360 hours. Success in collecting sound level data varied considerably from one outing to the next (Table 3.3). Environmental factors and equipment malfunction resulted in the loss of a substantial amount of data. In the July 2000 and the June 2001 field sessions at Cirque Lake no sound data were collected. At West Point during April 2000, just 2 hours of sound data were retrieved.

The loudest overflight events experienced by observers while watching sheep were due to military jet aircraft and only military aircraft produced sound levels > 90dBA (Table 3.4). Peak sound events > 90dBA occurred at Cirque Lake on 9 May 2001 (112 dBA; 2 unidentified aircraft with a sonic boom) and 14 May 2001 (95 dBA; 2 F-15s). At West Point, peak sound events > 90dBA occurred on 14 May 2001 (91 and 98 dBA; 2 events both by unidentified aircraft), on 11 June 2001 (132 dBA produced by four F-16 aircraft and 92 dBA produced by one F-16 aircraft), on 18 June 2001 (96 and 98 dBA; 2 events both by four F-16 aircraft), and on 23 July 2001 (92 dBA; 1 event by unknown military aircraft).

Few military overflight events produced sound levels greater that 70 dBA. Sound events greater than 70 dBA occurred more frequently at Cirque Lake in comparison to West Point in a given year and sound events greater than 50 dBA occurred more commonly in 2001 in comparison to 2000 (Table 3.5).

Dall’s sheep were observed for over 600 hours (Table 3.3). During 2000, a total of 121 h were spent observing sheep at Cirque Lake and 130 h were spent observing sheep at West Point. In 2001, more total time was spent observing sheep at Cirque Lake (233 h) in comparison to West Point (135 h) because of differences in success locating sheep. Examining time spent during each field session (4 per year), comparable amounts of time were spent observing sheep at Cirque Lake and West Point in 2000; t = 0.861, d.f. = 3, P = 0.45). In 2001, differences spent observing sheep during field sessions between Cirque Lake and West Point were not statistically significant (t = 1.22, d.f. = 3, P = 0.31). Of particular note, however, is the large amount of time spent observing sheep in the June field session at Cirque Lake (Table 3.3).

The number of overflights observed while watching sheep did not significantly affect the percent of time Dall’s sheep ewes spent active (Table 3.6). Factors that did significantly affect the amount of time Dall’s sheep ewes were active, in models that included the number of overflights observed, were the distance ewes were from rocky terrain, year, and the date sequence (P<0.025; Table 3.6). Sheep became more active the farther they were from rocks, and sheep were more active in 2001 in comparison to 2000. Although date sequence was significant in explaining the proportion of ewes active, no trend is apparent (Fig. 3.3). As the number of ewes in the group increased, the proportion of sheep active in the group also increased significantly in ewe models (Table 3.6). Date sequence accounted for the greatest amount of variation (4% and 2% for ewe and lamb models, respectively) in the amount of time Dall’s sheep spent active when compared to the other explanatory variables, followed by year (1% for ewe and lamb models). Little of the variation in the proportion of time Dall’s sheep spent active was explained by the independent variables in the ANCOVA models (Table 3.6). Study site and year interacted with the covariates of rock and ewe (Fig. 3.4).

The number of overflights observed while watching sheep did not significantly affect the feeding efficiency of Dall’s sheep ewes (Table 3.7). Dall’s sheep feeding efficiency in the ewe/day flight model was positively associated with the distance ewes were from rocks and the number of ewes in the group (Table 3.7). Sheep at West point had a significantly higher feeding efficiency than those at Cirque Lake (0.76 and 0.73, respectively) in ewe/day flight models. Date sequence was significant in the ewe/day flight model in explaining feeding efficiency and was highest before April 1 (0.79) and lowest from 1-14 April (0.58) (Fig. 3.5). In the lamb/day flight model, distance from rocks and number of ewes in the group were positively associated with feeding efficiency and number of lambs in the group were negatively associated with feeding efficiency. In the lamb/day flight model, feeding efficiency was highest from 3-18 May (0.76) and lowest from 12-27 July (0.72) and although feeding efficiency was significantly related to date sequence, no consistent trend was apparent (Fig. 3.5). Mean feeding efficiency was higher at West Point (0.75) in comparison to Cirque Lake (0.73) and was higher in 2000 (0.75) in comparison to 2001 (0.73) but the magnitude of these differences was very small. Trends in feeding efficiency were not obvious comparing feeding efficiency across study sites, years, and date sequences (Fig. 3.5). Date sequence accounted for the greatest amount of variation (5% and 2% for ewe/day flight and lamb/day flight models, respectively) in the feeding efficiency of Dall’s sheep when compared to other independent variables. Little of the variation in feeding efficiency of Dall’s sheep was explained by the independent variables in day flight ANCOVA models (Table 3.7).

The number of overflights observed while watching sheep did not significantly affect the behavior of Dall’s sheep ewes (Table 3.7). Dall’s sheep behavior in day flight MANCOVA models was significantly affected by year and date sequence (Table 3.8). The number of ewes in the group significantly affected behavior in the ewe/day flight MANCOVA model (Table 3.8). The distance to rock had a significant effect on behavior in the lamb/day flight MANCOVA model. Study site was found to have a significant effect on behavior in ewe/day flight MANCOVA models (Table 3.8). Date sequence accounted for the greatest amount of variation (2-3%) in behavior of Dall’s sheep when compared to the other independent variables in the models (Table 3.8). The majority of the variation in behavior of Dall’s sheep was not explained by the independent variables in the MANCOVA models (Table 3.8).

Differences in behavior between years in the ewe/day flight model (Table 3.8) was due to differences in the proportion of ewes bedded (P=0.001) and walking (P<0.001). A greater proportion of sheep were observed bedded in 2000 in comparison to 2001 in ewe models and a smaller proportion of sheep were observed walking in 2000 in comparison to 2001 in the ewe model.

Significant differences between date sequences existed in the ewe model (Table 3.8) for the proportion of sheep bedded (P<0.001), feeding (P<0.001) and walking (P<0.001). Although significant differences occurred in behavior categories across date sequences, no chronological pattern was obvious (Fig.3.6; Table 3.8).

The differences in behavior associated with the number of ewes in the group (Table 3.8) was due to differences in the proportion of sheep feeding and bedded. As the number of ewes in the group increased so did the proportion of animals feeding (P<0.001). In ewe models, a smaller proportion of sheep were observed bedded with increasing numbers of ewes in the group (P=0.001).

In ewe/day flight models, behavior by sheep at the 2 study sites was significantly different but univariate tests did not indicate significant differences in specific behavioral categories (Table 3.8).

For lamb/day flight models, a smaller proportion of Dall’s sheep were observed bedded (P=0.006) and a smaller proportion were standing (P=0.007) the greater distance sheep were from rocks. As the number of lambs in groups increased a greater proportion of sheep were observed walking (P= 0.031) in the lamb/day flight model. Differences in behavior between years in the lamb/day flight model (Table 3.8) were due to differences in the proportion of sheep bedded (P=0.003), standing (P=0.025) and walking (P=0.002). A greater proportion of sheep were observed bedded in 2000 in comparison to 2001 in the lamb/day flight model and a smaller proportion of sheep were observed standing and walking in 2000 in comparison to 2001 in the lamb/day flight model (Table 3.8). Significant differences between date sequences existed in the lamb/day flight model (Table 3.8) for the proportion of sheep bedded (P<0.001), feeding (P=0.005) and walking (P<0.001). No chronological patterns in behavior categories across date sequences were obvious (Fig.3.6; Table 3.9).

Behavior in Relation to Military Overflight Events

Although there was a tendency for activity to increase during, and 10 minutes after an overflight event (Fig. 3.7), this result was not significant (Table 3.9). Factors that significantly affected the proportion of Dall’s sheep active during scan samples were the year and the date sequence during the year (P<0.05; Table 3.9) in the ewe/overflight model and in the lamb/overflight model. Sheep were more active in 2001 in comparison to 2000 in overflight models. Sheep were most active 7-22 June and least active 12-27 July and no sequential trend was apparent for date sequence and activity. Date sequence accounted for the greatest amount of variation (13% and 12%, for ewe/overflight and lamb/overflight models, respectively) in the proportion of Dall’s sheep active when compared to the other independent variables. Little of the variation in the amount of time Dall’s sheep spent active in overflight models was explained by the independent variables in the ANCOVA models (Table 3.9). There were no interactions of class variables with covariates (rock, ewe, lamb) (Table 3.9).

Table 3.9. Summary of ANCOVA results examining factors affecting the proportion of time Dall’s sheep were active including the effect of military overflight events during field sessions in 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs, and a second model (restricted to May, June and July) included lambs in the analysis. Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska.

Dependent variable	Independent Variable*	d.f.	F	P	Partial eta squared
Proportion Active (ewe/overflight model)
	Rock	1	3.308	0.069	0.006
	Ewe	1	3.152	0.076	0.005
	Overflight	2	1.408	0.245	0.005
	Site	1	3.204	0.074	0.006
	Year(site)	2	3.338	0.036	0.012
	Sequence(year(site))	15	5.918	0.000	0.134

Proportion Active (lamb/overflight model)
	Rock	1	3.295	0.070	0.009
	Ewe	1	2.844	0.093	0.008
	Lamb	1	0.088	0.767	0.000
	Overflight	2	1.128	0.325	0.006
	Site	1	3.325	0.069	0.009
	Year(site)	2	5.875	0.003	0.031
	Sequence(year(site))	7	6.766	0.000	0.115

* Independent variables used in model were: 1) distance from steep rocks; 2) number of ewes in group; 3) number of lambs in group; 4) before, during, or after an overflight event; 5) study site; 6) year nested within study site; and 7) date sequence nested within year nested within study site.

^aA covariate effect was inferred if the significance of a class effect was different when comparing Type I and Type III Sums of Squares.

Figure 3.7. Mean (+SE) proportion of Dall’s sheep active during scan sampling in: 1) the 10 minutes before a military overflight event (n=85 for ewe model and n=55 for lamb model); 2) during a military overflight event (n=375 for ewe model and n=245 for lamb model); and 3) 10 minutes after the overflight event (n=135 for ewe model and n=82 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Dall’s sheep feeding efficiency was significantly affected by overflight events (Table 3.10; Fig. 3.8). The significant differences in feeding efficiency occurred when comparing feeding efficiency before the overflight to the feeding efficiency after the overflight (P= 0.001) for the ewe/overflight model. For the lamb/overflight model, there were significant differences in feeding efficiency comparing before to during overflight events (P= 0.028) and when comparing before to after overflight events (P= 0.028). In the ewe/overflight model, date sequence was significant. As the number of lambs in groups increased in the lamb/overflight models, feeding efficiency significantly declined. Date sequence accounted for the greatest amount of variation (5% and 3%, for ewe/overflight and lamb/overflight models, respectively) in the feeding efficiency of Dall’s sheep when compared to the other independent variables. Overflight events accounted for 3% of the variation in feeding efficiency models. The majority of the variation in feeding efficiency of Dall’s sheep was not explained by the independent variables in the ANCOVA overflight models (Table 3.10). The class variables of year and sequence interacted with the covariates (rock, ewe, lamb) in the ewe/overflight model. In the lamb/overflight model no indication of interactions of class variables with covariates were indicated.

Table 3.10. Summary of ANCOVA results examining factors affecting the feeding efficiency of Dall’s sheep including the effect of military overflight events during field sessions in 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs, and a second model (restricted to May, June and July) included lambs in the analysis. Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska.

Dependent variable	Independent variable*	d.f.	F	P	Partial eta squared
Feeding Efficiency (ewe/overflight model)
	Rock	1	0.455	0.500	0.001
	Ewe	1	0.133	0.716	0.000
	Overflight	2	5.660	0.004	0.025
	Site	1	0.166	0.684	0.000
	Year(site)	2	0.427	0.652^a	0.002
	Sequence(year(site))	15	1.731	0.042^a	0.055

Feeding Efficiency (lamb/overflight model)
	Rock	1	0.150	0.698	0.001
	Ewe	1	2.054	0.153	0.007
	Lamb	1	5.134	0.024	0.017
	Overflight	2	3.738	0.025	0.025
	Site	1	1.103	0.294	0.004
	Year(site)	2	0.466	0.628	0.003
	Sequence(year(site))	7	1.310	0.245	0.031

^a A covariate effect was inferred if the significance of a class effect was different when comparing Type I and Type III Sums of Squares.

Figure 3.8. Mean (+SE) feeding efficiency of Dall’s sheep during scan sampling in: 1) the 10 minutes before a military overflight event (n=63 for ewe model and n=41 for lamb model); 2) during a military overflight event (n=305 for ewe model and n=201 for lamb model); and 3) 10 minutes after the overflight event (n=105 for ewe model and n=65 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Although there was a suggestion of differences in behavior before, during and after overflight events, these results were not statistically significant (Fig. 3.9). Observed Dall’s sheep tended to be bedded more with in the 10 minutes before an overflight event than they were during or 10 minutes after an overflight event. Observed Dall’s sheep tended to walk more during and 10 minutes after an overflight event in comparison to before the overflight event (Fig. 3.9).

There were significant differences in behavior of Dall’s sheep associated with the number of ewes in groups (Table 3.11). In the ewe/overflight model, univariate tests did not indicate significant differences in specific behavioral categories. In the lamb/overflight model, the proportion of ewes feeding increased as the number of ewes increased (P=0.046).

There was a significant difference in the behavior of Dall’s sheep associated with the number of lambs in the group (Table 3.11). A greater proportion of animals were standing (P= 0.025) as the number of lambs in groups increased. In addition, a smaller proportion of sheep were observed feeding as the number of lambs in the group increased (P= 0.008) in the lamb/overflight MANCOVA models.

Significant differences between date sequences existed in ewe/overflight and lamb overflight MANCOVA models (Table 3.11). In the ewe/overflight model, significant differences were detected in the proportion of sheep bedded (P<0.001), feeding (P<0.001), and walking (P<0.001). In the lamb/overflight model, significant differences were detected in standing (P<0.001), feeding (P<0.001) and running (P<0.001). Although significant differences occurred in behavior categories across date sequences, no chronological pattern was obvious (Fig. 3.6).

The class variable year interacted with covariates (rock, ewe, lamb) in the ewe/overflight and the lamb/overflight (Table 3.11).

The number of ewes in the group explained the most variation in behavior of Dall’s sheep in the ewe/overflight model (6%) and the number of lambs in the group explained the most variation in the lamb/overflight model (5%). Less than 25% of the variation in Dall’s sheep behavior observed was accounted for by the independent variables we used to construct overflight event models (Table 3.11).

Table 3.11. Summary of MANCOVA results examining factors affecting the behavior of Dall’s sheep (percent bedding, standing, feeding, walking, and running) including the effect of military overflight events during field sessions in 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs, and a second model (restricted to May, June and July) included lambs in the analysis. Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska.

Dependent Variable	Independent Variable*	d.f.	F	P	Partial eta squared
Behavior
ewe/overflight model	Rock	5	2.058	0.069	0.018
	Ewe	5	7.401	0.000	0.061
	Overflight	10	1.692	0.078	0.015
	Site	5	0.715	0.613	0.006
	Year(site)	10	1.408	0.171^a	0.012
	Sequence(year(site))	75	2.305	0.000	0.059

Behavior (lamb/overflight model)
	Rock	5	1.507	0.187	0.020
	Ewe	5	2.287	0.046	0.031
	Lamb	5	4.117	0.001	0.054
	Overflight	10	1.369	0.190	0.019
	Site	5	1.173	0.322	0.016
	Year(site)	10	1.773	0.062^a	0.024
	Sequence(year(site))	35	2.275	0.000	0.042

^a A covariate effect was inferred if the significance of a class effect was different when comparing Type I and Type III Sums of Squares.

Figure 3.9. Mean (+SE) behavior of Dall’s sheep during scan sampling in: 1) the 10 minutes before a military overflight event (n=85 for ewe model and n=55 for lamb model); 2) during a military overflight event (n=375 for ewe model and n=245 for lamb model); and 3) 10 minutes after the overflight event (n=135 for ewe model and n=82 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Behavior in Relation to Proximity and Peak Sound Level of Military Overflight Events

In the lamb/sound model, as the overflights became closer more sheep were active, (Fig. 3.10; Table 3.13). As sound levels (2 min. mean L_eq) increased, a higher proportion of sheep were active (P= 0.012 and P=0.046 for ewe/sound and lamb/sound models, respectively) (Fig. 3.11). Date sequence was also significantly (P<0.05) related to the proportion of Dall’s sheep active in both the ewe/sound and lamb/sound models but no trend was apparent. In the lamb model, a greater proportion of sheep were active the farther they were from the rocks (Table 3.12). Date sequence accounted for the greatest amount of variation (20% and 12%, for ewe/sound and lamb/sound models, respectively) in the proportion of Dall’s sheep active when compared to the other independent variables (Table 3.12). Independent variables entered into the ewe/sound and lamb/sound models explained 31% and 29% of the variation in the proportion of time sheep were active. Year interacted with the covariates (rock, ewe, L_eq [dBA], lamb) in the ewe/overflight and the lamb/overflight models (Table 3.12).

Table 3.12. Summary of ANCOVA results examining factors affecting the proportion of time Dall’s sheep were active including proximity and sound level (2 min. mean L_eq [dBA]) of military overflights during field sessions in 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs, and a second model (restricted to May, June and July) included lambs in the analysis. Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska.

Dependent variable	Independent Variable*	d.f.	F	P	Partial eta squared
Proportion Active (ewe/sound model)
	Rock	1	3.676	0.057	0.018
	Ewe	1	0.471	0.493	0.002
	L_eq (dBA)	1	6.386	0.012	0.031
	Proximity	3	1.613	0.188	0.023
	Site	1	1.267	0.262	0.006
	Year(site)	2	2.619	0.075^a	0.025
	Sequence(year(site))	11	4.705	0.000	0.204

Proportion Active (lamb/sound model)
	Rock	1	4.042	0.046	0.026
	Ewe	1	1.549	0.215	0.010
	Lamb	1	1.266	0.262	0.008
	L_eq (dBA)	1	4.031	0.046	0.026
	Proximity	3	2.693	0.048	0.050
	Site	1	2.304	0.131	0.015
	Year(site)	2	2.313	0.102^a	0.029
	Sequence(year(site))	5	4.193	0.001	0.121

* Independent variables used in model were: 1) distance from steep rocks; 2) number of ewes in group; 3) number of lambs in group; 4) proximity of overflight to sheep; 5) sound level (2 minute mean L_eq [dBA]); 6) study site; 7) year nested within study site; and 8) date sequence nested within year nested within study site.

^a A covariate effect was inferred if the significance of a class effect was different when comparing Type I and Type III Sums of Squares.

* All observed close overflight events occurred between May and August.

Figure 3.10. Mean (+SE) proportion of Dall’s sheep active during scan sampling during military jet overflights grouped by proximity: 1) close = jets lower than 1,500 m AGL and within 1.6 km horizontal distance from sheep (n=20 for lamb model); 2) moderate = jets lower than 1,500 m AGL and between 1.6 and 3.2 km horizontal distance from sheep (n=38 for ewe model and n=27 for lamb model); 3) far = jets higher than 1,500 m AGL and greater than 3.2 km horizontal distance from sheep (n=317 for ewe model and n=199 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Figure 3.11. Mean (+SE) proportion of Dall’s sheep active during scan sampling during military jet overflights grouped by 2 minute mean L_eq (dBA): 1) < 40 dBA (n=90 for ewe model and n=65 for lamb model); 2) 41 – 50 dBA (n=60 for ewe model and n=48 for lamb model); 3) 51 – 60 dBA (n=56 for ewe model and n=45 for lamb model); and 4) > 61 dBA (n=20 for ewe model and n=13 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Dall’s sheep feeding efficiency was significantly affected by date sequence in the ewe/sound model (Table 3.13). In the lamb/sound model, feeding efficiency declined with increasing numbers of lambs in the group. When 2 minute mean L_eq sound levels from overflight events were > 61 dBA, feeding efficiency tended to decline in comparison to quieter overflight events but this result was not statistically significant (Fig. 3.11). Feeding efficiency in both the ewe/sound model and the lamb/sound model tended to decline as military jet overflight got closer but again, this was not significant (Fig. 3.12). Independent variables entered into the ewe/sound and lamb/sound models explained 23% and 22% of the variation in feeding efficiency of Dall’s sheep. Date sequence accounted for the greatest amount of variation in the ewe model (15%) and the number of lambs in the group accounted for the most variation in the lamb model (8%; Table 3.14) in the feeding efficiency of Dall’s sheep when compared to the other independent variables in the models. In the ewe/sound model, year interacted with the covariates (rock, ewe, L_eq [dBA]).

Table 3.13. Summary of ANCOVA results examining factors affecting the feeding efficiency of Dall’s sheep including proximity and sound level (2 min. mean L_eq [dBA]) of military overflights during field sessions in 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs, and a second model (restricted to May, June and July) included lambs in the analysis. Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska.

Dependent variable	Independent variable*	d.f.	F	P	Partial eta squared
Feeding Efficiency (ewe/sound model)
	Rock	1	0.427	0.515	0.003
	Ewe	1	0.023	0.879	0.000
	L_eq (dBA)	1	2.847	0.094	0.018
	Proximity	3	1.784	0.153	0.033
	Site	1	3.438	0.066	0.022
	Year(site)	2	0.400	0.671^a	0.005
	Sequence(year(site))	11	2.436	0.008	0.147

Feeding Efficiency (lamb/sound model)
	Rock	1	0.064	0.801	0.001
	Ewe	1	0.344	0.559	0.003
	Lamb	1	9.693	0.002	0.075
	L_eq (dBA)	1	0.804	0.372	0.007
	Proximity	3	1.000	0.395	0.024
	Site	1	0.064	0.800	0.001
	Year(site)	2	2.919	0.058	0.046
	Sequence(year(site))	5	1.727	0.133	0.067

^a A covariate effect was inferred if the significance of a class effect was different when comparing Type I and Type III Sums of Squares.

* All observed close overflight events occurred between May and August.

Figure 3.12. Mean (+SE) feeding efficiency of Dall’s sheep during scan sampling during military jet overflights grouped by proximity: 1) close = jets lower than 1,500 m AGL and within 1.6 km horizontal distance from sheep (n=17 for lamb model); 2) moderate = jets lower than 1,500 m AGL and between 1.6 and 3.2 km horizontal distance from sheep (n=32 for ewe model and n=21 for lamb model); 3) far = jets higher than 1,500 m AGL and greater than 3.2 km horizontal distance from sheep (n=256 for ewe model and n=141 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Figure 3.13. Mean (+SE) feeding efficiency of Dall’s sheep during scan sampling during military jet overflights grouped by 2 minute mean L_eq (dBA): 1) < 40 dBA (n=68 for ewe model and n=49 for lamb model); 2) 41 – 50 dBA (n=46 for ewe model and n=37 for lamb model); 3) 51 – 60 dBA (n=47 for ewe model and n=38 for lamb model); and 4) > 61 dBA (n=17 for ewe model and n=13 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

There were significant differences in behavior of Dall’s sheep associated with the sound level (2 minute mean L_eq [dBA]) and the date sequence in the ewe/sound MANCOVA model (Table 3.14). Fewer sheep were bedded (P= 0.012) and more sheep were standing (P=0.028) with increasing sound levels in the ewe/sound model. There were significant differences in the proportion of animals observed bedding (P< 0.001), standing (P= 0.002) and feeding (P= 0.027) associated with date sequence in the ewe/sound model. Although significant differences occurred in behavior categories across date sequences, no chronological pattern was obvious (Fig. 3.6).

In the lamb/sound MANCOVA model, there was a significant difference in the behavior of Dall’s sheep associated with the number of lambs in the group (Table 3.14). A greater proportion of animals were standing (P= 0.005) as the number of lambs in groups increased. In addition, a smaller proportion of sheep were observed feeding as the number of lambs in the group increased (P= 0.003) in the lamb/sound MANCOVA models. Significant differences in behavior occurred due to the proximity of military jet overflights in the lamb/sound models. Significant differences were detected in the proportion of sheep bedded (P=0.048) and walking (P=0.020; Table 3.15). As military jets got closer, sheep tended to be bedded less often and walking more often in the lamb/sound model (Fig. 3.14) and although not statistically significant, this pattern was also present in the ewe/sound model. In both the ewe/sound model and the lamb/sound model, sheep tended to walk more as military overflights got louder but these results were not statistically significant. (Fig. 3.15).

Year interacted with the covariates (rock, ewe, L_eq [dBA]) in the ewe/sound model (Table 3.14).

Date sequence explained the most variation in behavior of Dall’s sheep in the ewe/sound MANCOVA model (11%) and the number of lambs in the group explained the most variation in the lamb/sound MANCOVA model (8%). Total variation explained by the independent variables in the ewe/sound and lamb/sound MANCOVA models was 28% and 35%, respectively (Table 3.14).

Table 3.14. Summary of MANCOVA results examining factors affecting the behavior of Dall’s sheep (percent bedded, standing, feeding, walking, and running) including proximity and sound level (2 min. mean L_eq [dBA]) of military overflights during 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs and a second model (restricted to May - July) included lambs in the analysis. Observations were made at 2 sites in the Yukon-Tanana Uplands of Interior Alaska.

Dependent Variable	Independent Variable*	d.f.	F	P	Partial eta squared
Behavior (ewe/sound model)
	Rock	4	1.878	0.116	0.036
	Ewe	4	0.170	0.953	0.003
	L_eq (dBA)	4	2.423	0.050	0.046
	Proximity	12	1.750	0.054	0.034
	Site	4	1.092	0.362	0.021
	Year(site)	8	1.464	0.169^a	0.029
	Sequence(year(site))	44	2.162	0.000	0.106

Behavior (lamb/sound model)
	Rock	4	1.689	0.155	0.043
	Ewe	4	0.559	0.693	0.015
	Lamb	4	3.391	0.011	0.083
	L_eq (dBA)	4	1.550	0.191	0.040
	Proximity	12	1.845	0.040	0.047
	Site	4	0.974	0.424	0.025
	Year(site)	8	1.759	0.085^a	0.045
	Sequence(year(site))	20	1.546	0.052	0.050

^a A covariate effect was inferred if the significance of a class effect was different when comparing Type I and Type III Sums of Squares.

* All observed close overflight events occurred between May and August.

Figure 3.14. Mean (+SE) behavior of Dall’s sheep during scan sampling during military jet overflights grouped by proximity: 1) close = jets lower than 1,500 m AGL and within 1.6 km horizontal distance from sheep (n=20 for lamb model); 2) moderate = jets lower than 1,500 m AGL and between 1.6 and 3.2 km horizontal distance from sheep (n=38 for ewe model and n=27 for lamb model); 3) far = jets higher than 1,500 m AGL and greater than 3.2 km horizontal distance from sheep (n=317 for ewe model and n=199 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

Figure 3.15. Mean (+SE) behavior of Dall’s sheep during scan sampling during military jet overflights grouped by 2 minute mean L_eq (dBA): 1) < 40 dBA (n=90 for ewe model and n=65 for lamb model); 2) 41 – 50 dBA (n=60 for ewe model and n=48 for lamb model); 3) 51 – 60 dBA (n=56 for ewe model and n=45 for lamb model); and 4) > 61 dBA (n=20 for ewe model and n=13 for lamb model). Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska. Observations were made between late February and early August for the ewe/model and a subset of this data (observations between May and August) was used for the lamb/model.

DISCUSSION

Overflight actvity

Based on 2 field seasons of observation, and attempted observations in a third (1999) military overflight activity in interior Alaska was extremely variable between years, seasons and study areas. We were not able to evaluate the impacts on Dall’s sheep of a typical MOA flying schedule as presented in the Final Environmental Impact Statement, Alaska Military Operations Areas (Department of the Air Force 1995) because of the Balkan war (1999), Cope Thunder schedule changes (2000 and 2001), and construction projects (2000). Based on military flying schedules during 2002 and 2003, this uncertainty and fluctuation in the number and extent of military aircraft activity in the MOAs is normal. Therefore, this variability needs to be considered when evaluating mitigation measures designed to reduce potential impacts of military overflights on wildlife species as well as when interpreting the results of this study.

We assumed that the number of observed military overflights would be greater at the West Point study site in comparison to the Cirque Lake study site based on the presence of a mitigation measure at Cirque Lake, the closer proximity of West Point to bombing ranges and Eielson Air Force Base, and personal communication with a individuals familiar with the study areas (Skip Ambrose, National Park Service, personal communication). This assumption was not supported thereby eliminating our ability to compare overall differences in behavior between 2 sites receiving different levels of long-term military overflight activity. In addition, we did not find any statistical differences in the number of military overflights observed during Cope Thunder exercises in comparison to routine flying days. Therefore, it would be difficult to assume that any differences in Dall’s sheep behavior observed between Cope Thunder and routine flying days were a product of differences in exposure to military aircraft.

During the mitigation time period (10 May – 15 June), there were no statistical differences in the number of military overflights observed at the two sites suggesting that the reduction in airspace (1,500 m floor in a 7-nautical mile radius circle) did not reduce the air traffic in this area. The altitude restriction also did not eliminate loud overflights from the Cirque Lake Study area. One of the loudest overflight events we experienced (95 dBA) was measured at the mitigated site during the mitigated time period (14 May 2001). Four additional overflights over 80 dBA were measured at the same site on that day.

In general, loud or low military overflight events were rare during our 600 hours of Dall’s sheep observation. This was analytically undesirable as small samples sizes could adversely affect statistical power when examining behavior of Dall’s sheep exposed to loud or low aircraft. From a management perspective, however, this was a positive condition as exposure of an individual sheep in interior Alaska MOAs to loud and low aircraft is uncommon under the level of military overflight activity we observed in 2000 and 2001. It does, however, raise the question of the potential for sheep in the MOA structure to habituate to military aircraft. Startle and panic behavior occur in most wildlife species evaluated at noise levels greater than 95 dB (Dept. of the Air Force 1992). We observed overflight events >95 dB during this study in six situations. In these six situations, most Dall’s sheep exhibited behavior that may be considered disturbed (standing without feeding, walking and running). Dall’s sheep, therefore do apparently react to loud overflights. Alternatively, interior Alaska MOA airspace originated in 1976. Over the 28 years of the existence of this airspace, multiple generations of Dall’s sheep have been exposed to military aircraft overflights and the sheep currently living underneath the MOA airspace have been exposed to military aircraft activity their entire lives. This long exposure without direct consequences associated with the sight or sound of military jet aircraft may have allowed Dall’s sheep to habituate to overflights by military aircraft. Habituation to military aircraft activity has been suggested in other populations of wild sheep (Weisenberger 1996, Krausman et al. 1998).

To summarize military overflight activity in interior Alaska MOAs during 2000 and 2001, we observed considerable variation in military overflight activity both temporally and spatially. Differences in aircraft activity associated with Cope Thunders, routine flying days, and mitigated airspace did not necessarily result in differences in exposure to military aircraft by observers on the ground. In those instances when Dall’s sheep ewes were exposed to very loud overflight events (>95dBA) Dall’s sheep apparently reacted. However, low and loud military jet overflights were rare events. The long exposure of Dall’s sheep to military jet overflights in interior Alaska MOAs may have habituated sheep to military overflights thereby minimizing potential negative behavioral responses.

Dall’s sheep behavior in relation to military jet overflights

During the 2000 and 2001 field seasons, variables other than military overflights explained more variation in the proportion of Dall’s sheep active, the feeding efficiency of Dall’s sheep, and the activity budgets of Dall’s sheep (Table 3.15). Interactions between some of these variables make interpretation difficult in some instances but some patterns are clear. Date sequence (nested within year and study site) was the most notable of all the independent variables as it was significant in 15 out of 18 models we examined and also tended to explain the most variation in the proportion of Dall’s sheep active, the feeding efficiency of Dall’s sheep, and the activity budgets of Dall’s sheep. The year (nested within study site) was significant in explaining variation in the proportion of sheep active and the activity budget in 6 out of 12 models. The number of lambs and the number of ewes in the group were significant in explaining variation in feeding efficiency and the activity budget of Dall’s sheep in 6 out of 6, and 5 out of 12 models, respectively. In lamb models, the distance sheep were from rock was significant in explaining variation in the proportion of sheep active, feeding efficiency and the activity budget in 4 out of 9 models.

The number of overflight we observed over Dall’s sheep on a given day was not found to influence Dall’s sheep behavior. However, significant differences occurred in the feeding efficiency of Dall’s sheep when behavior was examined in the 10 minutes before, during and 10 minutes after an overflight event. Feeding efficiency declined approximately 7.5% during an overflight, then increased approximately 10% by 10 minutes after an overflight. Thus, recovery was quick. Proximity and sound level of overflight events also influenced the proportion sheep active and the activity budget of Dall’s sheep (Table 3.15). As sound levels increased, sheep transitioned from bedded to standing (Fig. 3.15) but the response to overflight proximity was complex (Fig. 3.14).

Table 3.15. Summary of significant statistical results from examining factors that may have affected behavior of Dall’s sheep during field sessions in 2000 and 2001. Two data sets were considered. One model was examined without considering the presence of lambs, and a second model (restricted to May, June and July) included lambs in the analysis. Observations were made at 2 study sites in the Yukon-Tanana Uplands of Interior Alaska.

Behavior	Independent Factor	Overflight metric
			Observed overflight		Event^a		Overflight Character^b
			Ewe^c	Lamb^c	Ewe^c	Lamb^c	Ewe^c	Lamb^c
Proportion	Rock		*	*				*
of sheep	Ewe		*
active	Lamb		na		na		na
	Site
	Year(site)		*	*	*	*
	Sequence(year(site))		*	*	*	*	*	*
	Overflight						2*	1,2

Feeding	Rock		*	*
Efficiency	Ewe		*	*
	Lamb		na	*	na	*	na	*
	Site		*
	Year(site)
	Sequence(year(site))		*	*	*		*
	Overflight				*	*

Activity	Rock			*
Budget	Ewe		*		*	*
	Lamb		na	*	na	*	na	*
	Site		*
	Year(site)		*	*
	Sequence(year(site))		*	*	*	*	*
	Overflight						2*	2*

* Significant result.

^a Event is a categorical variable: 1) within 10 minutes before an overflight event; 2) during an overflight event, and; 3) 10 minutes after the event.

^b Overflight were characterized by: #1) proximity

SUMMARY

The number of military overflights that we observed flying over Dall’s sheep during 2 field seasons did not significantly affect the proportion of sheep active, the feeding efficiency or the activity budgets of observed sheep.

Dall’s sheep feeding efficiency was significantly higher 10 minutes following the overflight event than in the 10 minutes before the event in models that considered only ewes and in models that included lambs. Feeding efficiency is a function of feeding and standing. To lower feeding efficiency, Dall’s sheep would need to increase standing relative to feeding. If time spent standing is a relative indication of vigilance, Dall’s sheep ewes were less vigilant 10 minutes following the overflight event than before, the opposite pattern of what one would expect if military overflights were disturbing sheep.

Little of the variation in feeding efficiency could be attributed to the before, during or after overflight variable (3% in both models). Although small, feeding efficiency was lower during overflight events in comparison to the before and after feeding efficiencies in both the ewe model and in the lamb model. In the lamb model, this decline in efficiency was statistically significant. This decline in feeding efficiency is an important finding because, by far, most overflights we observed over Dall’s sheep were high and far away. We found evidence that Dall’s sheep react more strongly to close and loud jets than to quiet jets far away. Therefore, an increase in the frequency of low overflights over Dall’s sheep could further reduce feeding efficiency during the overflight event. The duration of this impact, however, may be short lived based on the high, 10 minute post military overflight event feeding efficiency. Behavioral responses of short duration are consistent with those from other studies investigating the impacts of military overflights on wildlife. Magoun et al. (2004) Murphy et al. (1993) and Harrington and Veitch (1991) found that overt behavioral responses by caribou to overflights were short-term. Weisenberger et al. (1996) and Krausman et al. (1998) found few behavioral changes when comparing activity of bighorn sheep before an overflight event to 3 minutes after the overflight event. Heart rate data supports the conclusion that responses to military overflights are short lived (Weisenberger et al. 1996, Krausman et al. 1998).

The proportion of Dall’s sheep active during an overflight event increased as military jets got closer (Fig. 3.10) and louder (Fig. 3.11). The variables of sound level and jet proximity explained 5% and 8% of the variation in the proportion of animals active in the ewe model and lamb model respectively. Sheep bedded less and stood more in the ewe model as the sound of an overflight increased, and in the lamb model, sheep bedded less and walked more when jets were closer. Military aircraft overflights can therefore exert an energetic cost on Dall’s sheep. The magnitude and biological significance of this cost however would depend on the sound level produced by the overflights, the proximity of the overflights, the frequency of exposure to overflights, and the nutritional status of the animal as well as the interactions of these variables.

Chapter 2 | Chapter 4

2004 Sheep Report
http://www.nps.gov/yuch/Expanded/key_resources/sheep/sheep_2004.htm
Doug Beckstead
December 8, 2004