Long-term harvesting and male migration in a New Zealand population of Himalayan tahr Hemitragus jemlahicus

Abstract

1. Sexual differences in behaviour and ecology have important implications for the management of harvested ungulate populations.

2. This study investigated the long-term effects of harvesting on the abundance and distribution of sexes within an introduced population of Himalayan tahr in the Two Thumb and Sibbald Ranges, New Zealand. Annual counts of tahr were made from 1984 to 1996 in two catchments subject to different harvest regimes.

3. Although populations in the two catchments increased at similar rates over the study period, their adult sex ratios diverged significantly in the late 1980s. In Carneys Creek, which was subject to unrestricted harvest, the adult sex ratio was initially equivalent to the estimated population sex ratio, but from 1991 it was significantly male-biased. In North Branch, 10 km from Carneys Creek but subject only to adult (trophy) male harvest, the population was significantly female-biased in all years.

4. There was a large and male-biased harvest in Carneys Creek in 1993, suggesting that immigration of males may have been responsible for the long-term changes in sex ratio observed there. In contrast, the harvest of males in North Branch was insufficient to explain its female-biased population.

5. Monthly counts of male and female tahr in Carneys Creek between December 1993 and February 1996 revealed a consistent seasonal change in the abundance of sub-adult males, and perhaps adult males, but not females. Sub-adult males immigrated into Carneys Creek during spring, increasing to a summer maximum before emigrating during autumn; few sub-adult males were present during winter.

6. Counts at six additional sites during summer and autumn 1996 indicated that the probable source of these males was an adjacent hunting reserve. This hunting reserve, which included North Branch, contained a female-biased population with moderate female density.

7. The observed changes appear to have been a consequence of sexual differences in habitat selection and mobility. In particular, outside the rut the larger-bodied males avoided habitats containing high densities of females.

8. Harvesting has the potential to affect the spatial distribution of age–sex classes for sexually dimorphic ungulates. Managers need to consider these effects when designing and interpreting harvest and monitoring programmes.

Introduction

Many populations of sexually dimorphic ungulates are subject to intensive harvesting. Sexual differences in ecology, physiology and behaviour can have important implications for the management of such populations (Beddington 1974; Clutton-Brock, Guinness & Albon 1982; Ginsberg & Milner-Gulland 1994). Depending upon the management aim, densities of either sex can be manipulated upward or downward through the selective removal of individuals from particular age–sex classes. For example, if a population of red deer Cervus elaphus is managed to maximize the number of adult males harvested, then the female population should be high relative to carrying capacity (Clutton-Brock & Lonergan 1994; Buckland et al. 1996). However, if quality of harvested males is to be maximized then the female population should be small relative to carrying capacity (Clutton-Brock & Lonergan 1994; Buckland et al. 1996). These outcomes appear to be the result of sexual asymmetries in the effect of population density on growth and survival (Jorgenson, Festa-Bianchet & Wishart 1993, 1998).

Sexual differences in body mass are often associated with marked seasonal differences in habitat and diet selection (e.g. Putman, Culpin & Thirgood 1993; Bleich, Bowyer & Wehausen 1997), sociality (Shank 1985), and mobility (Owen-Smith 1993). Segregation of sexes outside the mating season can sometimes lead to extreme differences in habitat use at a variety of spatial and temporal scales (Main, Weckerly & Bleich 1996). For example, male and female bighorn sheep Ovis canadensis nelsoni largely use different mountain ranges outside the rut (Bleich, Bowyer & Wehausen 1997).

Although many studies have illustrated the probable effects of different harvesting strategies on population size and age–sex composition in ungulates (e.g. Beddington 1974; Fairall 1985; Ginsberg & Milner-Gulland 1994; Buckland et al. 1996), there is little information on how management actions can affect the spatial distribution of sexes. Clutton-Brock, Iason & Guinness (1987) observed that male red deer avoided high densities of females, and Main & Coblentz (1996) recorded a similar pattern for mule deer Odocoileus hemionus. Since harvesting has the potential to alter density at a variety of spatial scales, it is important to understand how such exploitation can affect the distribution, and hence availability, of age–sex classes.

The objective of this study was to evaluate the long-term effects of two harvesting regimes on the distribution and abundance of male and female Himalayan tahr Hemitragus jemlahicus Smith in the Two Thumb & Sibbald Ranges, Southern Alps, New Zealand. From 1984 to 1996 the northern catchments of the Two Thumb Range were subject to unlimited harvesting of all age–sex classes, but only a small harvest of trophy males occurred in the remainder of the study area. In this paper (i) 13 years of population monitoring in two catchments are examined, one within each harvesting regime; (ii) annual harvesting patterns within these catchments are described; and (iii) the seasonal movements of male and female tahr within one catchment are analysed. Unexpected changes in the abundance and distribution of male tahr within the study area appear to have been a consequence of interactions between harvesting-induced changes in female density and sexual differences in behaviour.

Himalayan tahr in new zealand

Himalayan tahr were liberated into New Zealand's Southern Alps to provide a hunting resource during 1904–19 (Donne 1924; Caughley 1970a) and currently occupy ≈4300 km² of the central Southern Alps (Fraser, Cone & Whitford 1996). Adult male tahr are popular hunting trophies (Davys, Forsyth & Hickling 1999), and recreational hunting is now the primary means of maintaining populations at low density (Department of Conservation 1993).

Male and female tahr are strongly dimorphic; adult males average 73 kg whereas adult females weigh 36 kg (Tustin 1990). Mating peaks in May and the median birth date of tahr in the eastern Southern Alps is 30 November; twins are rare (Caughley 1971). Habitats utilized by tahr in New Zealand are similar to those used in central Nepal (Caughley 1970b). Females in New Zealand are sedentary on rock bluffs (Tustin & Parkes 1988) with home ranges of ≈ 2 km² (Tustin 1990). The limited female dispersal that occurs appears to be density-dependent (Parkes & Tustin 1985). In contrast, adult males are frequently recorded many kilometres from the nearest female group (Anderson & Henderson 1961; Caughley 1970a).

The sexes aggregate on snow-free bluffs during winter, but are segregated outside this period (Tustin & Challies 1978; Tustin 1990; Forsyth 1997). Adult and sub-adult males form loose social groups in late winter, and move into summer habitat which may be well separated from, or interspersed between, habitat used by female groups (Caughley 1967).

Study areas

The Two Thumb and Sibbald Ranges extend eastward from the central Southern Alps and are drained by the Godley and Macaulay Rivers in the south, and by the Havelock and Rangitata Rivers in the north (Fig. 1). Elevations range from 500 m a.s.l. to > 2500 m peaks near the Main Divide. Terrain is typical of tahr habitat in the eastern Southern Alps (Tustin & Challies 1978), with extensive areas of scree and tussock intergrading with rock bluffs and, at lower altitudes, shrubland and small patches of forest (Forsyth 1997). The region receives 4000–5000 mm of precipitation annually, with rain or snow recorded on 2 days in 3 (Canterbury Regional Council, unpublished data).

Figure 1. Location of study areas (&U25CF;) where tahr were counted annually in the Two Thumb and Sibbald Ranges, South Island, New Zealand, during 1984–96.

Following the establishment of an overseas market for tahr meat in 1970, government and commercial helicopter-based hunting coupled with recreational hunting dramatically reduced tahr densities throughout the Southern Alps (Parkes & Tustin 1985; Parkes, Nugent & Warburton 1996; Forsyth & Hickling 1998; this study). Helicopter-based hunting continued in the Two Thumb and Sibbald Ranges until 1983 when a government moratorium prohibited the practice in this area (Hughey & Parkes 1995). Densities of tahr have subsequently increased over much of the eastern Southern Alps range (Forsyth & Hickling 1998).

Annual monitoring (see Methods) of tahr began in Carneys Creek and North Branch in 1984. These two catchments occupied different land tenures subject to contrasting harvest regimes and were considered to be representative of tahr habitat in the Two Thumb and Sibbald Ranges. North Branch was in an area of pastoral lease (termed the Godley Hunting Reserve) that leaseholders managed as a trophy-hunting area until 1996. Only adult male tahr (≥5 years old; see Parkes & Tustin 1988) were harvested in North Branch during 1984–92 (G. Joll, unpublished data) but, beginning in 1993, a new leaseholder conducted some helicopter-based control of females and juveniles.

The upper Havelock River, and Carneys Creek in particular, has traditionally been a popular area for recreational tahr hunting (Challies & Thomson 1989). During 1984–96, recreational hunting was actively encouraged as a means of controlling tahr by the New Zealand Forest Service (1984–87) and the New Zealand Department of Conservation (1987–96). Free hunting permits were issued to recreational hunters with no restrictions on the number, age, or sex of tahr harvested. The area east of Carneys Creek, in the northern Two Thumb Range, was pastoral lease. In contrast to North Branch, recreational hunters had virtually unrestricted access to this lease. For simplicity, the portion of the northern Two Thumb Range that was subject to unrestricted recreational hunting is termed the Rangitata Hunting Area.

North Branch and Carneys Creek, ≈ 10 km apart, are of similar size (20·4 and 19·1 km², respectively) and habitat. Seasonal habitat selection by tahr was evaluated in Carneys Creek in an associated study (Forsyth 1997). Habitat availability was estimated in 1995–96 by digitizing aerial photographs in three seasons (spring, summer–autumn, and winter). In winter, only the steepest rock and grass bluffs were free of snow. Grasslands and shrublands became available at lower altitudes in spring. During summer and autumn only a few small areas of permanent snow and ice remained.

Methods

Annual trends in carneys creek and north branch

An annual count of tahr was made in Carneys Creek and North Branch during February, March, or April 1984–96. However, failure to gain permission from the leaseholder resulted in North Branch not being counted in 1985–87 and 1994–95. In Carneys Creek, hunting was prohibited from early January until the count was completed so that the tahr would be undisturbed and more easily counted.

Tahr were counted by experienced observers based at observation sites located midway to the ridgeline (Tustin & Challies 1978; Challies 1992; Fig. 2). The sites provided complementary and overlapping views of each catchment, and the same sites were used each year. Counts were usually made on two different days by different observers, although at least one person was the same from year to year. All counts were made during the 3–4 h post-dawn and pre-dusk when tahr were most active (Tustin & Parkes 1988).

Figure 2. Observation sites (○) used to count tahr annually during February, March, or April from 1984 to 1996, and monthly from December 1993 to February 1996 in Carneys Creek, New Zealand. Shading indicates the five areas searched monthly from the lettered observation sites across the valley.

Tahr were located with binoculars (8–10×) and then classified using spotting scopes (20–60×) into one of five age–sex classes using a variety of physical and behavioural cues (Forsyth 1997). The age–sex classes, based on Caughley (1967) and Tustin (1990), were adult males (> 4 years), sub-adult males (2–4 years), females (≥ 2 years), yearlings (1–2 years), and kids (< 1 years). Location, time observed, and the age–sex composition of each tahr group were plotted on aerial photographs and 1 : 50000 scale maps. This enabled observers to cross-reference observations after each count, and the largest and most accurate counts for all groups were summed to give the total count (Tustin & Challies 1978; Challies 1992). Although not all tahr present were likely to have been counted using this technique (Forsyth & Hickling 1997), it was assumed that the method was adequate for detecting trends in abundance and sex ratio. Yearlings and kids were excluded from subsequent analyses because it proved difficult to differentiate accurately between the two classes (see below), and because sex could not be determined even with spotting scopes.

Log-linear regression was used to calculate rate of increase for tahr in the two catchments (Eberhardt & Simmons 1992), and trends in the adult sex ratio were tested using logistic regression (McCullagh & Nelder 1989). In the logistic regression model the independent variables were location and year, and the dependent variable was the proportion of males (adult and sub-adult pooled) counted.

Testing for biased sex ratios

Because the true adult (i.e. ≥ 2-year-old) sex ratio of the study population was unknown, the ratio was estimated from 2775 adult tahr shot in the eastern Southern Alps from helicopters during winter 1972–75 (K.G. Tustin, unpublished data). The tahr were shot during May–September, when the sexes were aggregated, and were aged by counting horn rings (Caughley 1965). The limited hunting that occurred prior to 1972 (Caughley 1970c; Tustin 1990) is unlikely to have significantly biased the age–sex structure of such a large sample. Furthermore, the resulting age–sex distribution was similar to estimates obtained from other ungulates exhibiting similar sexual size dimorphism (Clutton-Brock, Albon & Guinness 1985; Owen-Smith 1993; Clutton-Brock & Lonergan 1994). Departure of the observed annual summer sex ratios from this population estimate was tested using a G-test for goodness-of-fit (Sokal & Rohlf 1981).

Carneys creek harvest

A mail and telephone survey was used to determine the legal harvest of tahr in Carneys Creek during 1993. A list of all hunters who obtained permits to hunt within the Rangitata Conservation Area (which included Carneys Creek) during the 1993 calendar year was obtained from the New Zealand Department of Conservation, Canterbury. A mail survey, followed by a telephone survey of all non-respondents 6 weeks later, was used to determine who hunted in Carneys Creek and how many tahr they harvested. Both surveys asked identical questions.

Because some hunters do not gain permits before hunting, illegal harvest was also measured. Legal hunters were asked to list other party members, and records were cross-referenced to determine whether they had held hunting permits. Illegal harvest was also measured by telephoning people who had entered their names in hut books in and around Carneys Creek and specified the purpose of their visit as ‘hunting’.

The total design method (Dillman 1978) was used to format the mail survey; this involved asking questions that were easily understood and could be answered with minimal effort. Respondents were asked to list separately their monthly harvests and those of other members of their party. Many hunters could not accurately differentiate between adults and sub-adults of either sex, so four classes were used to measure harvest; ‘males’ (> 1 years), ‘females’ (> 1 years), ‘kids’ (≤1 year), and ‘unknown’ (i.e. shot animals that were not recovered).

Seasonal index of abundance in carneys creek

Seasonal migration of adult male, sub-adult male, and female tahr may have been responsible for long-term changes in the abundance and sex ratio of tahr recorded in Carneys Creek. To investigate this hypothesis, five observation sites (Fig. 2) in the headwaters of Carneys Creek were visited during 25 months from December 1993 to February 1996. These five sites encompassed the habitat of 99·7% of adult and sub-adult male tahr and 65·6% of female tahr observed in the annual counts during 1993–96. Extreme weather conditions prevented access to the study area in two months. Hunting was prohibited within Carneys Creek from January 1994 until the study ended.

A discrete non-overlapping area was searched from each site with binoculars (10×40) for 3 h post-dawn or pre-dusk during periods with good visibility. All searching was done by the author using a spotting scope (20×50) to classify tahr into one of the five age–sex classes.

Sites where female tahr were observed (A, C and E; Fig. 2) and not observed (B and D) during winter were differentiated because males formed mixed groups with females in winter (Forsyth 1997) but tended to avoid female habitat when segregated (Caughley 1967).

Each year was divided into four, 3-month seasons: spring (September–November); summer (December–February); autumn (March–May); and winter (June–August). These seasons have biological meaning for mountain ungulates in the Southern Alps; Clarke & Frampton (1991) recorded significant seasonal changes in the abundance of marked Alpine chamois Rupicapra rupicapra in Basin Creek (70 km north-east of Carneys Creek) using these pooled months.

The numbers of different individuals sighted within each 3-h count were log-transformed and pooled into seasons, giving three replicate counts per season per year. Each of the five sites was assumed to be an independent unit of measurement. Differences in abundance between years, seasons, and site × season interactions were explored for each age–sex class using repeated measures anova in systat (Wilkinson 1990). Interaction effects are not reported unless P < 0·1.

Densities in adjacent catchments

The density and age–sex class composition of tahr in catchments between Carneys Creek and North Branch were unknown, but these areas may have been important in understanding trends within the study area. Consequently, the largest-count method (described above) was used to count tahr at six additional sites in the Two Thumb and Sibbald Ranges during January, February and March 1996. Density was assessed as the number of tahr counted divided by the area of the site. Site area was calculated from 1 : 50000 scale Department of Survey and Land Information maps using ARC/INFO GIS (Environmental Systems Research Institute, Inc. 1991).

Results

Reliability of age–sex classes

Of 33 tahr shot by the author during 1993–96 for an associated diet study, one kid was misclassified (as a yearling) when first sighted (Forsyth 1997). Observers in annual counts were experienced government or recreational hunters and their classifications were assumed to be similarly accurate.

Annual trends in carneys creek and north branch

Although both Carneys Creek (r = 0·15 ± 0·01SE, t₁₁ = 18·34, P < 0·001) and North Branch (r = 0·17 ± 0·03SE, t₆ = 4·94, P = 0·003) populations increased at statistically similar rates (Fig. 3; comparison of slopes, t₁₇ = – 0·81, P > 0·2), the trend changed when analysed by sex. In Carneys Creek, females did not increase during the 13 years of monitoring (r = 0·02 ± 0·02, t₁₃ = 1·12, P = 0·29), but males did (r = 0·26 ± 0·02, t₁₃ = 14·41, P < 0·0001). In North Branch, both females (r = 0·28 ± 0·06, t₆ = 4·47, P = 0·004) and males (r = 0·14 ± 0·04, t₆ = 3·96, P = 0·007) increased between 1984 and 1996. Logistic regression confirmed that in both catchments there was a significant trend in sex ratio, but no evidence that the trend differed between sites (Table 1). The North Branch population was female-biased in every year, whereas in Carneys Creek there was a change from an estimated population adult sex ratio of 1 adult male: 1·53 adult females to an increasingly male-biased ratio from 1991. Thus, although the proportion of males increased in both catchments, the adult sex ratios were significantly different. The large residual deviance indicates lack of fit (

• 

 = 44·32, P < 0·0001). A probable explanation for this is that tahr were observed in single-sex groups during the counts and the assumption of independence may have been violated. Under the assumption that the mean structure of the model is correct but that variance is kp(1 – p) rather than np(1 – p), where k is an arbitrary variance-scaling constant, the F-values should be approximately correct (McCullagh & Nelder 1989). From 1990 to 1994 the number of females counted in Carneys Creek declined (r = – 0·18 ± 0·06, t₃ = – 2·90, P = 0·06), but increased again in 1995 and 1996 following the local prohibition of hunting.

Figure 3. Number of male (adult and sub-adult pooled) and female tahr counted in (a) Carneys Creek, and (b) North Branch, New Zealand, during February, March, or April 1984–96. During 1984–93 recreational hunting in Carneys Creek was prohibited from January until the count was completed, but from January 1994 hunting was prohibited in all months; North Branch was subject only to a small adult male harvest. The symbol above each bar indicates that the sex ratio in that year was significantly (P < 0·05) male- or female-biased relative to the population estimate of 1 male: 1·53 females (see text). Bars without symbols indicate years in which there was no significant sex bias.

Table 1.  Logistic regression analysis of changing adult tahr sex ratios in Carneys Creek & North Branch, New Zealand 1984–96

Variable	d.f.	Deviance	F	P
Location	1	294·39	12·91	<0·0001
Time	1	118·61	45·49	<0·0001
Time * Location	1	0·21	0·08	0·78
Residual	17	44·32
Total	20	457·53

Hunter harvest in carneys creek in 1993

Of 181 legal hunters, 164 (85%) were surveyed. The remainder had either provided insufficient addresses and could not be traced, or had moved with no forwarding address. Hence, the harvest presented here is probably conservative. Out of the 164 individuals contacted, 37 had hunted in Carneys Creek during the 10 months (March–December) in which hunting was permitted in 1993. In addition, 27 hunters had hunted in Carneys Creek without a permit.

Hunters harvested more tahr than were observed in Carneys Creek during March 1993 (Table 2), when 84 tahr (22 females, 8 kids, 45 sub-adult males, and 9 adult males) were counted. There was a strong seasonal bias in the numbers reported harvested, with few animals shot during winter. The number of males harvested exceeded females in all seasons. Although these data suggest that few males were available to be harvested during winter, the small number of tahr harvested in that season makes inference difficult; when the winter and spring harvests were pooled there was no seasonal difference in the proportion of males and females harvested (G₂ = 0·64, P > 0·1).

Table 2.  Seasonal harvest of tahr by recreational hunters in Carneys Creek, New Zealand, during March–December 1993

Season
Age–sex class (see Methods)	Summer*	Autumn	Winter	Spring	Total
* No hunting was permitted during January and February.
Males	34	21	2	8	65
Females	10	7	0	5	22
Yearlings/kids	2	3	0	0	5
Unknown	0	3	0	0	3
Total	46	34	2	13	95
Percentage of annual harvest	(48)	(36)	(2)	(14)	(100)

Yearly and seasonal patterns of abundance in carneys creek

The abundance of sub-adult males and females in Carneys Creek did not increase over the 2 years that monthly counts were made (F_1,3 = 3·82, P = 0·15 and F_1,3 = 0·37, P = 0·59, respectively), but adult male abundance did (F_1,3 = 24·98, P = 0·02). There were no seasonal differences in the abundance of females (Fig. 4a; F_3,9 = 0·92, P = 0·47), but there was a suggestion of a seasonal trend in adult males (Fig. 4b; F_3,9 = 3·06, P = 0·08), although the numbers of tahr observed in these two age–sex classes were small.

Figure 4. Mean number (+ 95% CL) of tahr counted at three sites with females and two sites without females in Carneys Creek, New Zealand, during 1994–96. Counts were made monthly from each site and are presented back-transformed; the lower CL would be conservative. Seasonal differences were statistically significant (P < 0·05) only for sub-adult males.

Sub-adult males displayed a strong seasonal trend in abundance (Fig. 4c; F_3,9 = 16·64, P = 0·001) that varied according to whether or not the site had females (F_3,9 = 3·84, P = 0·05). There were significantly fewer sub-adult males present in winter compared to the other three seasons (F_1,3 ≥ 15·31, P ≤ 0·03), and fewer sub-adult males in autumn compared to summer (F_1,3 = 15·31, P = 0·03). These changes were significantly greater in the sites without females from summer to winter (F_1,3 = 12·53, P = 0·04) and from autumn to winter (F_1,3 = 15·67, P = 0·03).

Population structure in adjacent catchments

Murphy Stream and Camp Creek, both adjacent to Carneys Creek and within the Rangitata Hunting Area, contained male-biased populations of tahr, and low densities of females (Table 3; Fig. 5). In contrast, the four sites within the Godley Hunting Reserve were of moderate (by historical standards) female density, and composed almost exclusively of females.

Table 3.  Number of male and female tahr (≥2 years) counted at eight sites in the Two Thumb and Sibbald Ranges, New Zealand, in January, February and March 1996

Site*	Site area†	Males	Females	Female density‡	Sex bias	G1	P <
Site locations are shown in Fig. 5. † ^km^2. ‡number km^–2.
Rangitata Hunting Area
Carneys Creek	19·1	93	32	1·7	Male	62·63	0·001
Murphy Stream	10·2	32	9	0·9	Male	25·30	0·001
Camp Creek	17·0	31	28	1·7	Male	4·08	0·05
Godley Hunting Reserve
Toms Stream	11·6	10	61	5·3	Female	22·20	0·001
East Macaulay	10·1	10	67	6·6	Female	26·49	0·001
West Macaulay	11·1	11	130	11·7	Female	73·95	0·001
North Branch	20·4	4	77	3·8	Female	53·02	0·001
Weka Stream	9·3	10	52	5·6	Female	16·09	0·001

Figure 5. Densities of male and female tahr (≥2 years) during January, February and March 1996 in the Two Thumb and Sibbald Ranges, New Zealand. The hypothesized spring migration routes of sub-adult male tahr are indicated by arrows; sub-adult (2–4 years) males born in the moderate female density Godley Hunting Reserve emigrated to the low female density Rangitata Hunting Area during spring, and returned during autumn.

Long-term changes in the summer sex ratio in carneys creek

The sex ratios of tahr populations in Carneys Creek during February were significantly different in 1965 and 1995, with more males and fewer females present in the latter count (Table 4).

Table 4.  Number and densities (number km^–2) of male (≥2 years) and other (females, yearlings, and kids) tahr counted within Carneys Creek during February 1965 and 1995. Compared to the estimated population ratio of 1 male: 3·35 females, yearlings, and kids the ratio was female-biased in 1965 and male-biased in 1995 (G₁ ≥ 57·23, P < 0·001), and was significantly different between the two years (G₁ = 209·51, P < 0·001)

Year	Males	Other	Other density
* Data from Tustin & Challies (1978).
1965*	40	670	32·9
1995	63	49	2·3

Discussion

Female tahr give birth to similar numbers of male and female offspring (Caughley 1966), so in 1991–96 the Carneys Creek female population was insufficient to produce the number of sub-adult males present during the annual summer or autumn counts (Fig. 3). Sub-adult, and perhaps adult, males immigrated into Carneys Creek in spring, stayed for summer, and then emigrated during autumn; few males were present in winter (Fig. 4). The probable source of these males was the Godley Hunting Reserve (Fig. 5). During the 1993–96 monthly counts, adult and sub-adult male tahr were frequently observed moving in and out of Carneys Creek along the axial ridge of the Two Thumb Range during spring, summer, and autumn. In January and February 1996 the Godley Hunting Reserve contained, relative to the high densities of the 1960s, moderate densities of females but few males. From 1984 to 1992, only 30–40 adult trophy males (≥ 5 years) and no females were harvested annually from this area (G. Joll, unpublished data). However, during winter 1995 c. 2000 tahr were culled from throughout Godley Hunting Reserve by the leaseholder, so the densities of females recorded there during summer 1996 were likely to have been lower than during the previous few years.

The extreme male-biased recreational hunter harvest recorded in Carneys Creek in 1993 is unlikely to represent only hunter selectivity for males (Table 2). A subsequent postal survey of legal recreational hunters (Davys, Forsyth & Hickling 1999) recorded a significantly male-biased harvest in the Havelock watershed during 1994–95 (see Fig. 1), but not in the more northern Clyde watershed. The Clyde watershed would have been too geographically separated to receive immigrant males from the Godley Hunting Reserve, suggesting that the harvest recorded in Carneys Creek in 1993 was facilitated by immigration of males. The annual counts (Fig. 3) indicate that significant immigration of male tahr born within the Godley Hunting Reserve into Carneys Creek began about 1989. In that year the number of males counted in Carneys Creek began to increase.

The seasonal migration of sub-adult male tahr recorded in this study appears to be related to long-term changes in the density of females within the Godley Hunting Reserve. There is strong circumstantial evidence that female density affects habitat use by male tahr when the sexes are segregated. In 1965, when tahr were at high density throughout the Two Thumb and Sibbald ranges (Caughley 1967; Tustin & Challies 1978), few males summered in Carneys Creek and large female-juvenile groups were observed in the two areas (sites B and D in Fig. 2) that did not contain females in this study. During 1993–96 the seasonal changes in abundance of sub-adult males in Carneys Creek were significantly greater in the non-female sites compared to the female sites. Furthermore, few sub-adult males were observed within the Godley Hunting Reserve during summer 1996, despite this class not being harvested there. These results suggest that the larger-bodied sub-adult male tahr avoid habitats utilized by females when sexually segregated. Illius & Gordon (1987) showed that a sexual size dimorphism of only 20% could lead to swards becoming unprofitable for males if grazing reduced mean sward height below some critical threshold (see also Clutton-Brock & Harvey 1983), and Clutton-Brock, Iason & Guinness (1987) argued that such a mechanism was responsible for increased sexual segregation in red deer at high population density. Permanent vegetation monitoring plots were established in North Branch in 1990–92, and in Carneys Creek in 1992, and were remeasured in 1996 and 1997, respectively (Parkes & Thomson 1998). Most indices of tussock biomass and health declined in North Branch, but not in Carneys Creek (Parkes & Thomson 1998). Male tahr thus appear to avoid areas of high female density when segregated due to the overgrazing of preferred food species (see also Caughley 1970c,d).

There is considerable evidence that harvesting may have subtle, but important, effects on age- and sex-specific survival and behaviour of many ungulates (Clutton-Brock, Guinness & Albon 1982; Clutton-Brock, Albon & Guinness 1985; Clutton-Brock & Lonergan 1994). A long-term study by Jorgenson et al. (1997) suggests that harvesting of prime-age bighorn sheep rams may lead to increased mortality of younger rams, probably due to the high costs associated with earlier participation in rut. Similarly, increasing the population density of female red deer leads to greater competition for food between females and males, increasing mortality of both juvenile and adult males (Clutton-Brock, Albon & Guinness 1985; Clutton-Brock & Lonergan 1994). Modelling has also demonstrated how highly male-biased harvests could lead to lower fecundity and population declines through disruption of the mating system (Ginsberg & Milner-Gulland 1994).

In this study two harvest regimes, unrestricted harvest vs. limited adult male harvest, created low and moderate female densities, respectively. Emigration of male tahr from the moderate female density area to the low female density area during the period of sexual segregation resulted in an extremely male-biased harvest in the latter area. Age- and sex-specific differences in seasonal distribution apparently resulted from asymmetries in philopatry, mobility, and resource selection.

Harvesting activities that impact on population density can significantly affect the spatial distribution of age–sex classes in sexually dimorphic ungulates. In situations where ungulates are controlled to protect environmental values, such as tahr in the Southern Alps of New Zealand, management will often necessitate large and instantaneous population reductions within management units. These reductions may alter habitat use by segregated males and females in a non-linear fashion relative to the population reduction. The potential for such effects is particularly evident near the border of management units (i.e. areas subject to different control intensities, as occurred in this study). Managers need to understand how age- and sex-specific differences in behaviour interact with changing population density, and should consider the potential for such effects when designing and interpreting both harvest and monitoring programmes.

Recommendations

Considerable money is spent attempting to mitigate the impacts of tahr on indigenous flora in the Southern Alps of New Zealand (Tustin 1990). Current management involves maintaining tahr populations below specified ‘intervention densities’ at which grasslands are modified (Department of Conservation 1993). Densities of tahr are monitored by annual summer or autumn counts in sample catchments spread throughout the breeding range. This study demonstrates how careful interpretation of spatial and temporal trends in the abundance of age–sex classes can illuminate population changes at larger spatial scales. However, this ability relies on observers being able to differentiate accurately between age–sex classes. It is thus recommended that observers be sufficiently trained to be able to classify accurately tahr as ‘sub-adult male’, ‘adult male’, ‘female’, or ‘juvenile’.

Although the total summer population in Carneys Creek has exceeded the intervention density specified for that management unit (2·5 tahr km^–2; Department of Conservation 1993) since 1990, a control operation that did not target the source of the migrant males (i.e. Godley Hunting Reserve) would have achieved only a short-term reduction in density, with seasonal immigration of males continuing to occur. Because of the spatial segregation of males from females outside winter, intervention densities monitored through summer counts should be defined and interpreted in terms of age–sex classes. In particular, using female density rather than total density would remove uncertainty resulting from the mobility of male tahr outside winter and their apparent preference for areas containing low densities of females.

Acknowledgements

I thank the many New Zealand Forest Service, New Zealand Forest Research Institute, New Zealand Department of Conservation, and Landcare Research New Zealand employees who participated in the annual Carneys Creek and North Branch counts, in particular: D.C. Anderson, N. Bolton, C.N. Challies, M.C. Coleman, S. Harraway, K. Lange, C. Pearson, and C. Thomson. K.F.D. Hughey, J. Andrew, and D.C. Anderson (New Zealand Department of Conservation, Canterbury) were instrumental in prohibiting hunting within Carneys Creek during 1994–96. L. Prouting provided periodic aerial transport to Carneys Creek, the use of huts on Mesopotamia Station, and permission to count tahr in Camp Creek. I thank R.J. Barker and C.M. Frampton for statistical advice, and K.G. Tustin for permission to cite unpublished data. Comments by G.J. Hickling, J.P. Parkes, V.C. Bleich, R.J. Barker, J. Hone, M. Festa-Bianchet, and two anonymous referees improved the manuscript. Funding was provided by New Zealand Department of Conservation Research Grant 1894 & a Lincoln University Doctoral Scholarship.

Received 10 June 1998; revision received 30 December 1998