Correspondence J. M. Goodrich, 50 Years October Street, 5-1, Terney, Primorsky Krai, Russia 692150. Email: firstname.lastname@example.org
We examined causes of mortality and survival rates for Amur tigers on and near the Sikhote-Alin Biosphere Zapovednik. Our objectives were to estimate and compare survival rates among sex and age classes, estimate cause-specific mortality, identify conservation issues related to tiger mortality and provide recommendations for reducing human-caused mortality. We used two separate datasets; one based on radio-tracking tigers from 1992 to 2005 and one based on reports of dead tigers from 1976 to 2000. We examined causes of mortality for both datasets and used a Cox proportional hazards models to estimate survival rates using data from 42 radio-collared tigers. Mortality was predominantly human-caused for both datasets (83% for the telemetry dataset and 78% for the other, n=24 and 53 mortalities, respectively), and 75% of collared animals were poached. All collared subadult tigers that dispersed were poached (n=6). Annual survival of adult females (0.81±0.10) was greater than that of adult males (0.63±0.20) (z=1.52, P=0.13) and subadult males (0.41±0.46) (z=2.07, P=0.04). Survival rates were precariously low on our study area, which included the largest protected area within Amur tiger range. Efforts to reduce human-caused mortality should focus on poaching and reducing deaths from tiger–human conflicts.
Knowledge of rates and causes of mortality is important to understanding of carnivore population dynamics. Mortality and survivorship data provide real input for population models, an opportunity to assess whether observed mortality rates are sustainable, and insight into ways to reduce mortality. Humans cause most mortality in most large carnivore populations and are one of the greatest threats to their survival worldwide (Noss et al., 1996; Woodroffe & Ginsberg, 1998; Woodroffe, 2001). Most large carnivores evolved under conditions of high survival of breeding adults (Weaver, Paquet & Ruggiero, 1996), but human-caused mortality often takes a heavy toll on this cohort (e.g. Fuller, 1989; Mclellan et al., 1999). Small changes in adult survivorship may have serious consequences for persistence, especially of isolated carnivore populations and long-term survival of adults, especially females, is critical to population well-being (Knight & Eberhardt, 1985; Weaver et al., 1996). For example, Kenney et al. (1995) estimated that extinction risk would increase from 5 to 95% with an increase in poaching mortality from 4 to 8% for small, isolated tiger populations.
As a species at the northern edge of its range, Amur tigers may be particularly extinction prone. They occur at very low densities – a result of low prey biomass – and have large area requirements (Smirnov & Miquelle, 1999). Viable populations cannot be contained within the boundaries of protected areas and rely on multiple use lands for persistence (Miquelle et al., 1999a). On these unprotected lands, conflicts between tigers and people and associated mortality, are common; hence, Amur tigers may be more extinction prone than populations that exist within the boundaries of protected areas (Woodroffe & Ginsberg, 1998; Woodroffe, 2001; Miquelle et al., 2005).
While considerable data exist on Amur tiger mortality, samples are biased because they rely on reports, and tigers killed in tiger–human conflicts are often reported whereas most poached animals are not (e.g. Nikolaev, 1985; Nikolaev & Yudin, 1993; Pikunov, 1994; Matyushkin et al., 1996; Miquelle et al., 2005). Further, natural deaths of this elusive animal are rarely detected by hunters and others who provided information for the above sources. Radio-tracking provides a less biased sample because mortalities are usually detected regardless of cause of death. We examined patterns and causes of mortality, and survival rates for radio-collared tigers on and near the Sikhote-Alin Biosphere Zapovednik (SABZ), 1992–2004. Our objectives were to estimate and compare survival rates among sex and age classes, estimate cause-specific mortality, identify conservation issues related to tiger mortality and provide recommendations for reducing human-caused mortality.
Materials and methods
We studied tigers on and near the 390 184 ha SABZ, Primorski Krai (Province), Russia (44°46′N, 135°48′E). SABZ had minimal human disturbance, with access restricted to scientists and forest guards, but the eastern edge was bisected by a paved public road which provided access for poachers (Kerley et al., 2002). The Sea of Japan borders SABZ to the east and its central feature is the Sikhote-Alin Mountains, which parallels the coastline and has elevations reaching 1600 m. The land surrounding SABZ is sparsely populated (about 13 000 people in five villages) and contains a 70 350 ha buffer zone (1–8 km wide) where human activities include fishing, hunting and some agricultural practices (e.g. livestock grazing). Potential predators of tigers and their cubs include brown bears Ursus arctos, Asiatic black bears Ursus thibetanus, wolves Canis lupus, lynx Lynx lynx and a variety of smaller carnivores. Descriptions of the region and environmental variables influencing tiger mortality can be found elsewhere (Knystautas & Flint, 1987; Newell & Wilson, 1996; Miquelle et al., 1999b; Kerley et al., 2002).
Causes of mortality
We examined causes of mortality for two datasets, one based on radio-tracking and one based on reports. We used seven categories to define cause of death: (1) tiger–human conflict (legally killed because it was a threat to human life or livelihood); (2) poached (illegally killed); (3) suspected poached (for radio-collared animals only, see below); (4) road-kill (animal hit by a vehicle); (5) trapped (animals incidentally captured in legal traps set for furbearers); (6) natural death (cause of death not directly related to human actions); (7) unknown.
The first dataset (hereafter the ‘telemetry dataset’) included mortalities of radio-collared tigers and their offspring detected through our radio-tracking activities, February 1992 to January 2005 (Goodrich et al., 2001; Kerley et al., 2002, 2003). Although radio-collared animals are not a true random sample because capture probabilities vary, their deaths are the best available approximation thereof. We primarily detected mortality when radio signals indicated inactivity for several hours to days and we investigated the site to determine cause of death. We necropsied dead tigers when possible, but when animals were poached, we usually found only a collar cut from the tiger. However, often when tigers were poached the signal disappeared because the poachers destroyed the collar. In some cases, informants later provided data (e.g. a tattoo number) confirming a poaching.
We categorized tigers as ‘poached’ when we found a dead tiger or cut radio-collar or if the animal disappeared and we received verifiable evidence of poaching. Resident adults were categorized as ‘suspected poached’ if the signal disappeared despite extensive ground and aerial searches, ground searches in snow failed to locate tracks, and a new tiger took over the missing animal's territory. Because tigers are territorial and habits and track patterns of study animals were well known, it was possible to detect the presence of an individual by its tracks, even when collars were non-functional (Yudakov & Nikolaev, 1987; Kerley et al., 2002). It is possible that in some cases the collar was destroyed, for example, in a fight with another tiger that resulted in the collared animal's death. However, we never detected the death of a subadult or adult tiger from predation or fighting. We detected fighting between tigers only once, and of 53 tigers handled, we only once found scaring suggestive of fighting. For dispersing subadults, we categorized an animal as ‘suspected poached’ if extensive aerial searching did not produce a signal. In six cases when we classified tigers as suspected poached, we later received information confirming our suspicions, whereas we never detected that a ‘suspected poached’ animal was alive, suggesting that we were usually correct in our classification. We did not classify all missing tigers as ‘suspected poached’; in two cases we suspected premature collar failure and in two cases signals were lost after batteries were due to fail.
We detected mortality of unmarked cubs (<1 year old) of collared tigresses in three ways. First, remains of cubs were found when we searched den sites of collared tigresses. Second, we assumed cubs <6 months old died when their mother was poached or would have died in two cases when we intervened to save the cubs (Kerley et al., 2003). Third, we estimated litter size of collared tigresses by tracks and/or visual observation of cubs associated with their mother. On average, we determined litter size when cubs were 4.1 months old (Kerley et al., 2003). We monitored cubs in this way until 12 months old and assumed cubs had died if repeated observations failed to detect them with their mothers. After 12 months of age, cubs begin moving independently of their mothers, so their absence no longer indicated mortality. We treated data on cubs separately, but pooled data for adults and subadults because of small sample sizes.
We tested for seasonal variation in mortality to determine if mortality and factors influencing it (e.g. poaching or legal hunting seasons for ungulates) varied seasonally. We tested (χ2) four separate seasonal classifications: (1) year split into winter (November–April) versus summer (May–October); (2) year split into four seasons (January–March, etc.); (3) year split into three seasons (January–April, etc.); (4) hunting season (November–February, etc.) versus non-hunting season (March–October). We tested for seasonal variation in poaching alone in the same way. We also compared causes of mortality between sexes. Post hoc, we compared mean poaching rates (animals killed/year/total animals in sample at beginning of year) across three intervals of our study period (1992–1996, 1997–2000 and 2000–2004) because data indicated that poaching was higher during 1997–2000.
The second dataset included reports of dead tigers from SABZ and a 5000 km2 surrounding, 1976–2000 (hereafter the ‘long-term dataset’). Two tigers from the telemetry dataset were included because their deaths were detected regardless of our radio-tracking activities. We did not make the above comparisons for this dataset because data on sex and age were often missing.
Estimation of survival rates
Cox proportional hazards models (Venables & Ripley, 1999) were applied to data collected from 11 February 1992 to 12 November 2004 from 42 radio-collared tigers, to estimate and compare survival of five sex-age classes of tigers: cubs (<18 months old), female subadults (18–36 months old), male subadults, female adults (>36 months old) and male adults. For all animals included, age at capture (Goodrich et al., 2001) and date of death or censoring was recorded. Animals were censored from a given age class if contact with the animal was lost (e.g. the radio failed or fell off) or if the animal was alive on 12 November 2004.
Animals captured before adulthood and that lived long enough to enter the next age class were included in all appropriate age classes. Survival rates in different age classes of one animal were assumed independent. For example, F01 was captured at 1 year of age in 1992 and was still alive and collared on 12 November 2004. She was initially entered in the ‘cub’ category and subsequently in the subadult and adult categories. The survival rates in the three categories were assumed to be independent and the sample size was increased by three based on data from one animal. Thus, while 42 tigers were included in the analysis, overall sample size was 55.
Preliminary analyses indicated interactions between age and sex (e.g. survival rates for a single age class varied between sexes), but there were insufficient data to fit a complicated model using sex, age and their interactions. Therefore, we assumed interactions existed and fit a separate Cox proportional hazards model to each of the five sex-age classes, with the assumption that survival rates for male and female cubs were the same. We then combined classes that were judged not significantly different (using a liberal P-value of 0.15 because of small sample sizes), and compared the rest using an approximate normal z-statistic and two-sided test.
Causes of mortality
Mortality detected through radio-tracking (n=24) was primarily human-caused (83.3%), with most (75%) associated with poaching (i.e. sum of ‘poached’ and ‘suspected poached’). Natural deaths accounted for <17% of mortalities (Fig. 1). Even when ‘suspected poachings’ (the only unconfirmed deaths) were removed from the data, human-induced mortality predominated (poaching=62.5% and vehicle collisions=12.5%), with natural deaths representing only 25% of mortalities (n=16). We detected no differences in causes of mortality between males and females (χ2=0.69, d.f.=2, P=0.7) and no seasonal trends in poaching or overall mortality (P>0.50 for all tests). However, mean proportion of radio-collared tigers poached per year was greater from 1997 to 2000 than during other years (Fig. 2) (ANOVA F=4.8, d.f.=2, P=0.04).
Of four natural mortalities, one tiger died of an unknown disease, one when a tree fell on it, one when it fell through the ice on a stream and froze to death and one from injuries from an unknown cause. When the latter (M50) was captured, the right side of his head and his left rear knee were swollen and he had a minor puncture wound on his head. He died a month later and necropsy revealed that his left rear leg and left front foot had both been broken several months before. The diseased tiger (M34) was a resident adult male found dead in an emaciated condition with no other apparent pathology; a condition consistent with canine distemper which has been confirmed in the population (Goodrich et al., 2005a).
Forty-three per cent of cub mortality was human-caused, although much indirectly, that is, cubs died or were removed from the wild when their mothers were poached (Fig. 3). One cub was shot by a SABZ guard in an aggressive encounter. This cub was thin and necropsy revealed a large abdominal hernia that likely caused his poor condition.
All subadult tigers that dispersed from their natal home ranges were poached or suspected poached (n=6), whereas philopatric cubs survived (n=4) (Fisher's exact test, d.f.=1, P=0.005). All but one dispersing subadults were males, and all subadults that settled within their natal areas were females.
Humans caused 78% (n=53 deaths) of deaths in the long-term dataset (Fig. 4). Of eight natural deaths, six were cub remains found in scats of male tigers and were likely killed by these males, one adult male died when it fell through the ice into a river and drowned, and one subadult male died of disease clinically diagnosed as canine distemper.
Estimated annual survival rates varied among sex-age classes (Fig. 5). No collared tigers died while they were cubs, so cubs were dropped from further analyses. Annual survival rate of adult females (0.81±0.10) did not differ from that of subadult females (0.72±0.25) (z=0.56, P=0.58), but was greater than those of adult males (0.63±0.20) (z=1.52, P=0.13) and subadult males (0.41±0.46) (z=2.07, P=0.04). We detected no differences between survival rates for adult males and subadult males (z=0.986, P=0.32), adult males and subadult females (z=0.394, P=0.69) and subadult males and subadult females (z=1.107, P=0.27).
The estimated probability that a subadult female would reach 36 months was 0.65 given that she reached 18 months, and the estimated probability that a subadult male would reach 36 months of age was 0.32 given that he reached 18 months. Estimated probabilities of annual survival (Fig. 5) are consistent with expected lower survival rates for males versus females in each age class.
Survival rates were highest for adult females, and progressively lower for subadult females, adult males and finally subadult males. Whereas some adult females were expected to survive 7 years, given that they reach their third birthday, few individual adult males were expected to survive that long. Standard errors were large indicating that precisions of the estimates were low. However, differences between survival rates as great as those observed between adult females and subadult males (0.81 vs. 0.41), are a strong indication of real differences that cannot be attributed to random variation and small sample sizes. Precision of these statistics is low because of small samples, but new data would have to indicate no differences in survival rates or even opposite results to those in the current analysis to reverse the trends.
Deaths from natural causes were rare in either dataset. Of four radio-collared tigers that died of natural causes, two may have died from human-related causes. M34 likely died of canine distemper, a disease usually fatal to tigers that is common in and probably transmitted by domestic dogs in Russia (Goodrich et al, 2005a). While impacts of this disease on tiger populations are unknown, distemper caused 34% mortality in Serengeti lions (Roelke-Parker et al., 1996), so the issue warrants further investigation. M50's pathologies were consistent with a vehicle collision, but could also have been obtained during a predation attempt. We detected six cases of probable infanticide based on cub remains in tiger scats. Infanticide may also have caused the death of two litters born to radio-collared mothers, whose father died shortly before the cubs were born. Both litters disappeared shortly after they left their natal dens and about the same time when a new male immigrated to the area. Infanticide has been reported for Amur and Bengal tigers elsewhere (Smith & McDougal, 1991; Nikolaev & Yudin, 1993) and probably increases reproductive success of immigrating males by cycling resident females into estrus more quickly (Smith & McDougal, 1991; Smith, 1993).
Poaching was the most common cause of death in both datasets, but was higher for radio-collared animals. This dataset was less biased because we could detect deaths regardless of cause. The long-term dataset depended on reports which favored conflict situations because they were widely publicized and reported by people seeking government assistance (Miquelle et al., 2005); whereas, poachers maintained secrecy. Nonetheless, it is clear that conflict is an important cause of mortality. No collared tigers died in tiger–human conflicts, but most lived on SABZ where there was a low potential for conflict because it was closed to the public and livestock. On non-protected areas, we would expect higher mortality from both poaching and tiger–human conflict (Miquelle et al., 2005), an alarming fact, considering that human-caused mortality made up 83% of radio-collared tiger deaths. All other studies of Amur tiger mortality detected similar rates of human-caused mortality (e.g. Nikolaev, 1985; Nikolaev & Yudin, 1993; Pikunov, 1994; Matyushkin et al., 1996; Miquelle et al., 2005) and it was the leading cause of tiger death in Nepal and India (Schaller, 1967; Sunquist, 1981; Smith, 1993).
For cubs, the second most common cause of death was associated with poaching of their mothers. Four of eight tigresses poached had young cubs, and in three cases, the cubs were too young to survive on their own. Tigresses with cubs may be more susceptible to poaching. They move less and more slowly, making them easier for poachers to track down in snow. Also, tigresses often aggressively protect their cubs, making them easier targets for poachers.
Poaching and related cub mortality was clearly associated with roads (Kerley et al., 2002). High human-induced mortality along roads has been demonstrated for other large carnivores, for example grizzly bears, cougars and wolves; often the result of increased hunting and poaching (reviews in Noss et al., 1996; Kerley et al., 2002). If the current trend of increasing and improved roads, and increased vehicles per capita continues within Amur tiger habitat, we expect human-induced mortality to increase.
Following high levels of poaching from 1997 to 2000, SABZ improved anti-poaching activities on the public road through SABZ, including increased patrols and construction of a guard station and two observation cabins (Astafiev, 2005); poaching subsequently decreased. However, while such intensive anti-poaching activities may be sustainable within protected areas, the remainder of tiger habitat is too vast and the number of roads too numerous to patrol adequately. From 1998 to 2000, over $800 000 were invested in tiger-related law enforcement activities in Russia, yet poaching remains high (Christie, 2006). Thus, solutions must include attacks on other fronts, such as changes in Russian legislature to increase fines for poaching and eliminate loopholes allowing possession of endangered species, road restrictions, educational programs and economic incentives not to poach.
Dispersal is essential to maintaining genetic connectivity within the Amur tiger population (Miquelle et al., 1999b). However, all dispersing collared tigers were poached before they settled and reproduced. That successful dispersal does occur is clear because some tigers immigrated to our study area, suggesting they dispersed from somewhere, settled, and survived to reproduce (Goodrich et al., 2005b). However, our data indicate that successful dispersal is rare and if current levels of poaching continue, maintaining suitable habitat for dispersal corridors alone may not be enough to retain genetic exchange across the range of Amur tigers.
One model of tiger population dynamics suggests that tiger populations will decline human-induced mortality exceeds 10% and if survival of breeding adults is <80% (Chapron et al., in press). Estimated survival for our population was 0.81 and 0.63 for adult females and adult males, respectively; that is, precariously low for a protected area that should act as a source population. However, our survival estimates reflect means from a 13-year period and survival was much lower from 1997 to 2000, when 34% of our radio-collared tigers were poached per year. Kenney et al. (1995) suggested that for small (<120 individuals) isolated populations of tigers, annual poaching levels of 4% were acceptable, but twice that level would result in a 95% probability of extinction. However, Karanth et al. (2006) estimated survival of 77% in a stable population in India. While it is unclear what level of poaching may be sustainable by the Amur tiger population, which is relatively large (about 450 individuals) and unfragmented (Miquelle et al., 1999a), there is little doubt that poaching rates from 1997 to 2000 (34%) were unsustainable. Efforts to reduce human-caused mortality should focus on poaching and reducing deaths from tiger–human conflicts.
Funding was provided by the Wildlife Conservation Society, 21st Century Tiger, National Geographic Society, National Fish and Wildlife Foundation Save the Tiger Fund, National Wildlife Federation, Exxon Corporation, the Charles Engelhard Foundation, Disney Wildlife Fund, Turner Foundation, US Fish and Wildlife Service Rhino and Tiger Conservation Fund, Richard King Mellon, Avocet Charitable Lead Unitrust, Robertson Foundation, Starr Foundation and Goldman Environmental Foundation. I. Nikolaev, B. Schleyer, N. Rybin, A. Rybin, A. Kostirya, I. Seryodkin, V. Melnikov, A. Saphonov, V. Schukin and E. Gishko assisted with data collection. Director A.A. Astafiev and Assistant Directors M. Gromyko and Y. Potikha of SABZ provided logistical and administrative support, and the Russian Ministry of Natural Resources provided permits for capture and political support.