Biodiversity Assessment of Mammal and Bird Species from Camera Trap Data

in Yanchiwan National Nature Reserve, Gansu Province, China

Author(s): http://orcid.org/0000-0002-1040-2578Zhang Chengcheng, http://
orcid.org/0000-0002-3140-5954Wang Jun, http://orcid.org/0000-0002-7474-9247Justine
Shanti Alexander, http://orcid.org/0000-0002-8066-4688Dou Zhigang, http://
orcid.org/0000-0002-6398-6116Wu Liji, http://orcid.org/0000-0002-3678-7760Dong Wantao, http://
orcid.org/0000-0002-0499-9881Dabuxilite, http://orcid.org/0000-0002-5636-7114Yang Jucai and
http://orcid.org/0000-0002-3731-8421Shi Kun
Source: Journal of Resources and Ecology, 9(5):566-574.
Published By: Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of
Sciences
https://doi.org/10.5814/j.issn.1674-764x.2018.05.014
URL: http://www.bioone.org/doi/full/10.5814/j.issn.1674-764x.2018.05.014

BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological,
and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books
published by nonprofit societies, associations, museums, institutions, and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of
BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use.
Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial
inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions,
research libraries, and research funders in the common goal of maximizing access to critical research.
September, 2018 Journal of Resources and Ecology Vol. 9 No.5

J. Resour. Ecol. 2018 9(5) 566-574

DOI: 10.5814/j.issn.1674-764x.2018.05.014
www.jorae.cn

Biodiversity Assessment of Mammal and Bird Species from

Camera Trap Data in Yanchiwan National Nature Reserve,
Gansu Province, China

ZHANG Chengcheng1, WANG Jun1,2, Justine Shanti ALEXANDER3, DOU Zhigang4, WU Liji4, DONG Wantao4,
Dabuxilite4, YANG Jucai4, SHI Kun1,5,*

1. The Wildlife Institute, Beijing Forestry University, Beijing 100083, China;

2. Faculty of Science and Engineering, Manchester Metropolitan University, Manchester M1 5GD, UK;
3. Snow Leopard Trust, Sunnyside Avenue, Seattle 98103, USA;
4. Administration of Yanchiwan National Nature Reserve, Jiuquan, Gansu 736399, China;
5. Eco-Bridge Continental, Beijing 100083, China

Abstract: Camera traps serve as an important tool for monitoring species diversity. We used data from camera
traps set for capturing snow leopards (Panthera uncia) in the Yanchiwan National Nature Reserve, Gansu Province,
China, to assess species richness with respect to mammal and birds species. We also assessed survey efficiency
for species detection, and conducted an initial analysis of species interactions. The survey effort of 10, 171 camera
workdays yielded 2, 868 suitable animal image events involving 17 mammal and 20 bird species. Among these, the
dhole (Cuon alpinus) is considered to be Endangered, the snow leopard and white-lipped deer (Cervus albirostris)
Vulnerable, and the Pallas’s cat (Feli smanul), mountain weasel (Mustela altaica), Himalayan griffon (Gyps hima-
layensis) and cinereous vulture (Aegypius monachus) Near Threatened under the IUCN red list. Fourteen species
were also listed as key protected wild animals according to China national standards. Both the rarefaction curves
and richness estimators suggested our sampling for mammal and pheasant species is sufficient, while more survey
efforts are still needed to detect other bird species. With a focus on the dominant species Eurasian lynx (Lynx lynx),
occupancy models were used to estimate site use and detection probability for selected species, and to investigate
predator-prey relationships between lynx on the one hand and woolly hare (Lepus oiostolus), pika (Ochotona spp.)
and Tibetan partridge (Perdix hodgsoniae) on the other. We give recommendations on how to increase efficiency in
camera-based species inventory and biodiversity monitoring.

Key words: camera trap; species richness; species inventory; species rarefaction curves; occupancy modeling

1 Introduction Biodiversity is often equated with species diversity,

Global biodiversity has been declining at an alarming rate
over the last decades, and this trend continues despite inten-
sification of conservation efforts (Butchart et al., 2010; Car-
dinale et al., 2012; Tittensor et al., 2014). Effective remedial
actions need to be supported by up-to-date information on
regional and global biodiversity status and changes (Scholes
et al., 2012).
which is commonly measured by species richness (Lande,
1996). Richness investigations are therefore important for
monitoring diversity in a specific site. In recent years, cam-
era traps are being used extensively for inventorying terres-
trial animals (Tobler et al., 2010; Akbaba and Ayaş, 2012;
He et al., 2016; Lan and Jin, 2016). They can be used as a
non-invasive method to record rare and elusive species that

Received: 2018-03-27 Accepted: 2018-06-02

Foundation: National Natural Science Foundation of China (31470567)
*Corresponding author: SHI Kun, E-mail: kunshi@bjfu.edu.cn
Citation: ZHANG Chengcheng, WANG Jun, Justine Shanti ALEXANDER, et al. 2018. Biodiversity Assessment of Mammal and Bird Species from
Camera Trap Data in Yanchiwan National Nature Reserve, Gansu Province, China. Journal of Resources and Ecology, 9(5): 566–574.
ZHANG Chengcheng, et al.: Biodiversity Assessment on Mammal and Bird Species from Camera Trap Data in Yanchiwan National…
sign surveys often fail to detect (Voss and Emmons, 1996)
and can support longer-term biodiversity monitoring (O’Brien
et al., 2010).
In China, although many nature reserves have been es-
tablished for years, systematic wildlife inventories and reg-
ular monitoring are yet to be launched, and information gaps
on species population status persist (Xiao, 2016). Lan and
Jin (2016), Xue et al. (2016), Chen et al. (2016) and many
other research groups have conducted camera trap surveys
to investigate terrestrial animals in several nature reserves in
China. However, those surveys have mostly focused on
assessing biodiversity in forest habitat and information on
biodiversity in China’s high altitude areas remains scarce.
Recent research efforts have used camera traps to
investigate species residing at high elevations, such as the
snow leopard (Panthera uncia) (Alexander et al., 2015a),
providing a good entry point and opportunity to look for the
presence of other species in snow leopard landscapes.
Since the Yanchiwan Nature Reserve was established in
1982, reports of its species inventory (Liu et al., 2010) have
been based mostly on sign surveys and recorded sightings.
Systematic survey has not yet been conducted to reliably
update the species inventory and provide information on
species richness and composition in any part of the reserve.
Our study took advantage of a survey designed for estimat-
ing snow leopard density and occupancy to obtain records
of the presence of other species. The aim of our study was to
1) assess overall species richness of mammals and birds;
2) evaluate the survey effort needed to obtain a certain per-
centage of the total species assemblage at this area; 3) dis-
Fig.1 Study area, systematic grids (white-lined squares) and camera trap stations (black triangle spots) in Yanchiwan National
Nature Reserve, Gansu Province, China
567
cuss species interactions, with a focus on the dominant spe-
cies Eurasian lynx (Lynx lynx).
2 Material and methods
2.1 Study area
Yanchiwan National Nature Reserve (YCWNNR, Fig. 1)
lies to the west of the Qilian Mountain range (Liu et al.,
2010). It is located on the north-eastern margin of the
Qinghai-Tibet Plateau and forms the border between Gansu
Province and Qinghai Province. The YCWNNR was ex-
panded to cover a total area of 13600 km2 (38°33N–
39°10N and 95°19E–97°13E) when it was upgraded to
national nature reserve in 2006 by China’s State Council.
The mean annual temperature is –0.8C with the minimum
average in January of –14.4C and the maximum average in
July of 11.7C. The mean annual rainfall is 202.5 mm with a
marked dry winter season from October to March.
This study focused on a 400 km2 zone with elevation
ranges from 3184 m to 4207 m on the eastern edge of
YCWNNR. The habitat primarily consists of grassland
(Halogeton arachnoideus, Kochia melanoptera, Thylaco-
spermum caespitosum) and shrub (Ephedra przewalskii,
Nitraria sphaerocarpa, Kalidium foliatum) vegetation in a
rugged rocky terrain. There are no human settlements within
the area but some livestock herding is permitted.
2.2 Camera trap survey
Camera-trapping was carried out from 5th November
2015 to 8th June 2016. The survey was designed to meet the
requirements for estimating snow leopard abundance. The
568
entire study area was divided into 4 km × 4 km grid cells to
ensure even camera coverage (Fig. 1). In each grid, two to
three stations (with one camera trap per station) were placed
along trails with animal signs. A total of 62 camera stations
(LTL ACORN-5210 & 6210) were set up across the 400
km2 area. Cameras were placed with a minimum spacing of
one km in order to increase independence of detection, with
exceptions of 11 cameras which were separated by 194, 262,
277, 390, 470, 732, 764, 818, 869, 900 and 948 m.
Each camera trap was left to operate for 24 hours per day
in order to detect both nocturnal and diurnal animals, and
was programmed to take 3 continuous photos when trig-
gered. The infrared sensor was set to 50 cm above ground.
2.3 Data analyses
2.3.1 Rarefaction curves and richness estimators
All images were entered into a database. For each image,
the camera station, date and time were recorded. Images of
mammal species were identified using Smith (2008), while
images of bird species were identified using MacKinnon
and Phillipps (2000) and Zheng (2017). As jerboa, pika and
rosefinch were hard to identify at the species level through
camera data, they were grouped at the genus level. The
global and regional conservation status of each species was
determined based on the IUCN Red List of Globally
Threatened Species (IUCN, 2017), as well as Chinese Na-
tional key protected wild animal list (The State Council,
1988). To avoid pseudo replication, we considered consecu-
tive images as independent images when the same species
was photographed more than once during a period of less
than 30 minutes (O’Brien et al., 2013).
Two non-parametric species richness estimators, the in-
cidence-base estimator (ICE) and Chao 2, were calculated
using the program EstimateS (Colwell, 2013). For all ran-
domizations we used 1000 runs. Species rarefaction curves
were generated, to evaluate the completeness of our survey
for inventorying animals.
2.3.2 Species site use and species interaction
Detection/non-detection matrix data of Eurasian lynx,
woolly hare (Lepus oiostolus), pika (Ochotona spp.) and
Tibetan partridge (Perdix hodgsoniae) were inputted into
PRESENCE (version 11.5) to explore species site use,
probability of detection (MacKenzie et al., 2002), and spa-
tial interaction between species (MacKenzie et al., 2004).
As our analysis was based on the scale of the camera trap
station, occupancy is equated to site use, and probability of
detection is understood as the probability that the species is
present and detected (MacKenzie et al., 2004). The site uses
of snow leopard have been widely searched in other studies
(Alexander et al., 2015b; Alexander et al., 2016; Rovero et
al., 2018), while Eurasian lynx has been few studied (Laze
and Gordon, 2016), thus we give a particular focus on Eura-
sian lynx. In our interaction analysis, we looked at the in-
teraction between Eurasian lynx and wooly hare, pika and
Journal of Resources and Ecology Vol. 9 No. 5, 2018
Tibetan partridge. The Eurasian lynx was identified as the
dominant predator and its potential prey species on the basis
of more than 150 independent images. To increase the
probability of detection and to reduce zero inflation of the
detection/non- detection matrix, we combined 5 camera trap
days into 1 survey occasion, resulting in 43 sampling occa-
sions. We ran simple single-season occupancy models for
each single species with elevation as a covariate for site use
and probability of detection. As our study area was located
in a depopulated zone with no settlements and few livestock,
we didn’t look at other covariates such as human distur-
bance in our single species modeling. Based on Akaike’s
information criterion (AIC), we selected the best detection
model for Eurasian lynx. We then ran single-season
two-species models between lynx - woolly hare, lynx - pika
and lynx - Tibetan partridge to investigate interactions
among these species. The magnitude of the interaction be-
tween species could be estimated from the species interac-
tion factor φ = ΨAB/ (ΨA×ΨB) (MacKenzie et al., 2004).
3 Results
3.1 Presents of mammal and avian species
From 5th November 2015 to 8th June 2016, 57 camera traps
were collected with a total sampling effort of 10171 camera
workdays, 5 camera traps were lost or destroyed under se-
vere weather conditions. We obtained 3161 suitable inde-
pendent events from 44 278 images, of where 2868 (90.7%)
were wildlife independent events and 293 (9.3%) were live-
stock or human activity-related independent events.
We photographed 17 wild mammalian species represent-
ing 5 orders and 11 families, 20 wild avian species repre-
senting 4 orders and 8 families (Table 1). The mammalian
species included 10 carnivores, three ungulates, two rodents
and at least two lagomorphs with one jerboa and 186 pika
independent images which we were unable to identify at the
species level. The avian species included four pheasants,
one pigeon, three accipiter and 12 passerines, with one in-
dependent image of rosefinch that we were unable to iden-
tify. The most commonly detected mammalian species were
woolly hare, followed by snow leopard and blue sheep
(Pseudois nayaur), while Tibetan partridge was the most
frequently detected avian species, followed by Himalayan
snowcock (Tetraogallus himalayensis) and black redstart
(Phoenicurus ochruros).
Under the IUCN conservation status (Table 1), there are
several recorded species of high conservation value. Dhole
(Cuon alpines) is listed as Endangered, and the snow leop-
ard and white-lipped deer are listed as Vulnerable by the
IUCN. There were 4 species listed as Near Threatened:
Pallas’s cat (Felis manul), mountain weasel (Mustela al-
taica), Himalayan griffon (Gyps himalayensis) and ciner-
eous vulture (Aegypius monachus). We also identified four
species belonging to the Class I protected species and 10
species belonging to the Class II designated by the Chinese
central government (Table 1).
ZHANG Chengcheng, et al.: Biodiversity Assessment on Mammal and Bird Species from Camera Trap Data in Yanchiwan National… 569

Table 1 Numbers of photographed camera stations, number of independent images and conservation status for all species
from our camera trap survey in Yanchiwan National Nature Reserve, Gansu Province, China
Species No. of camera No. of independ- IUCN conserva- China conserva-
Species Common name
classification stations (%) ent images tion status tion status
Mammalia
Rodentia
Sciuridae Marmota himalayana Himalayan marmot 10 (17.54) 61 LC –
Dipodidae – Jerboa 1 (1.75) 1 – –
Lagomorpha
Ochotonidae – Pika 23 (40.35) 186 – –
Leporidae Lepus oiostolus Woolly hare 44 (77.19) 1188 LC –
Carnivora
Felidae Felis manul Pallas’s cat 3 (5.26) 3 NT Ⅱ
Lynx lynx Eurasian lynx 34 (59.65) 186 LC Ⅱ
Panthera uncia Snow leopard 52 (91.23) 340 VU Ⅰ
Canidae Canis lupus Gray wolf 15 (26.32) 28 LC –
Cuon alpinus Dhole 20 (35.09) 64 EN Ⅱ
Vulpes ferrilata Tibetan fox 2 (3.51) 2 LC –
Vulpes vulpes Red fox 17 (29.82) 64 LC –
Ursidae Ursus arctos Brown bear 9 (15.79) 13 LC Ⅱ
Mustelidae Martes foina Beech marten 3 (5.26) 4 LC Ⅱ
Mustela altaica Mountain weasel 13 (22.81) 65 NT –
Perissodactyla
Equidae Equus kiang Kiang 8 (14.04) 53 LC Ⅰ
Artiodactyla
Cervidae Cervus albirostris White-lipped deer 2 (3.51) 2 VU Ⅰ
Bovidae Pseudois nayaur Blue sheep 45 (78.95) 254 LC Ⅱ
Aves
Galliformes
Phasianidae Tetraogallus tibetanus Tibetan snowcock 1 (1.75) 1 LC Ⅱ
Tetraogallus himalayensis Himalayan snowcock 13 (22.81) 79 LC Ⅱ
Alectoris chukar Chukar 15 (26.32) 39 LC –
Perdix hodgsoniae Tibetan partridge 27 (47.37) 188 LC –
Columbiformes
Columbidae Columba rupestris Hill pigeon 1 (1.75) 6 LC –
Falconiformes
Accipitridae Gyps himalayensis Himalayan griffon 1 (1.75) 6 NT Ⅱ
Aegypius monachus Cinereous vulture 1 (1.75) 1 NT Ⅱ
Aquila chrysaetos Golden eagle 1 (1.75) 1 LC Ⅰ
Passeriformes
Corvidae Pseudopodoces humilis Tibetan ground tit 6 (10.53) 13 LC –
Pyrrhocorax pyrrhocorax Red-billed chough 11 (19.30) 39 LC –
Pyrrhocorax graculus Yellow-billed chough 1 (1.75) 1 LC –
Muscicapidae Phoenicurus ochruros Black redstart 16 (28.07) 66 LC –
Phoenicurus erythrogaster White-winged redstart 6 (10.53) 11 LC –
Sylviidae Leptopoecile sophiae White-browed tit warbler 1 (1.75) 1 LC –
Phylloscopus fuscatus Dusky warbler 1 (1.75) 1 LC –
Passeridae Montifringilla adamsi Tibentansnowfinch 9 (15.79) 16 LC –
Motacilla alba White wagtail 1 (1.75) 1 LC –
Prunella fulvescens Brown accentor 11 (19.30) 23 LC –
Fringillidae Linaria flavirostris Twite 2 (3.51) 2 LC –
Carpodacus spp. Rosefinch 1 (1.75) 1 - –
Note: IUCN (2017) status categories: EN = Endangered, VU = vulnerable, NT = Near Threatened, LC = Least Concern.
570
3.2 Species richness estimators
The species rarefaction curve is shown in Fig. 2 and Fig. 3.
The curves of mammal and pheasant species plateau, while
the curves of avian (include pheasants) and total terrestrial
species are not asymptotic. The camera survey for mammal
and pheasants were clearly more complete than that for
birds or overall species.
Fig. 4 shows the behavior of richness estimators of the
different species catalogue with increasing survey efforts.
The estimated overall terrestrial species richness is 49.37
from ICE estimator and 44.2 from Chao 2 estimator, with an
observation of 37 species. The estimated number of species
for mammal reaches a maximum at around 5000 camera
workdays, then slightly declines to near observed number of
Fig.2 Mammal species rarefaction curve (comparison with
total species) with 1000 randomization runs from our camera
trap data in Yanchiwan National Nature Reserve
Fig.3 Avian and pheasant species rarefaction curves (com-
parison with total species) with 1000 randomization runs from
our camera trap data in Yanchiwan National Nature Reserve
Journal of Resources and Ecology Vol. 9 No. 5, 2018
species as the survey effort increases (Obs = 17, ICE =
17.45, Chao 2 = 17). For avian species, the estimated spe-
cies numbers from both ICE and Chao 2 are increasing and
end at a number higher than the observed number (Obs = 20,
ICE = 44.61, Chao 2 = 36), while the numbers of estimated
pheasant species is identical to the observed number (Obs =
4, ICE = 4, Chao 2 = 4).
3.3 Species occupancy modeling
We used 186 Eurasian lynx, 1188 woolly hare, 186 pika and
188 Tibetan partridge independent images to produce detec-
tion/non-detection matrixes for each species. Elevation of
camera traps, ranging from 3184 m to 4002 m (3641 m ±
215 m, mean ± SD), was used as a covariate affecting site
use and probability of detection in our models.
Based on AIC values, the simple single-season occu-
pancy models for the four species were ranked (Table 2).
For Eurasian lynx and woolly hare, elevation did not have a
major influence on site use (summed AIC weight of 0.08
and 0.06 accordingly) and the probability of presence and
detection (summed AIC weight of 0.11 and 0.00 accord-
ingly). For pika and Tibetan partridge, elevation provided
better explanations for the probability of presence and det-
ection (summed AIC weight = 0.88 and 0.66 accordingly)
but not for site use (summed AIC weight = 0.35 and 0.54
accordingly). Table 3 gives the estimates of parameters from
our top-ranked models. The Ψ estimates for Eurasian lynx,
woolly hare and pika are slightly higher than naïve occu-
pancy, and were overestimated by 3.3%, 2.6% and 2.5%
respectively. For Tibetan partridge, although the Ψ estimate
overestimated occupancy by 15.0%, the predictive power of
this estimate should be interpreted with caution given its
large standard error.
As the single species models demonstrate, elevation did
not influence site use and detection probability of lynx. El-
evation was therefore not used as a covariate for the two-
species models (Table 4). The magnitude of the interacttion
between species could be estimated from the species inter-
action factor φ = ΨAB/(ΨA × ΨB), where ΨA is the prob-
ability occupied by species A, ΨB is the probability occu-
pied by species B, ΨAB is the probability occupied by both
species (MacKenzie et al. 2004). Eurasian lynx are slightly
avoidant of woolly hare (φ = 0.94 ± 0.04) and Tibetan par-
tridge (φ = 0.83 ± 0.12), as they co-occur less frequently
than if they were distributed independently. Or rather, woo-
lly hare and Tibetan partridge are slightly avoidant of Eura-
sian lynx. Pika seem to use the sites independently of lynx
with φ value close to 1 (φ = 1.03 ± 0.11). With regard to
probability of detection, Eurasian lynx are more likely to be
detected when they use the sites together with woolly hare
(rA = 0.13 ± 0.01) and Tibetan partridge (rA = 0.17 ± 0.01)
than when they use the sites alone (pA = 0.01 ± 0.01, pA =
0.04 ± 0.01, respectively). While there seems no difference
on detection probability if lynx use the site only (pA = 0.18
± 0.02) or together with pika (rA = 0.19 ± 0.01).
ZHANG Chengcheng, et al.: Biodiversity Assessment on Mammal and Bird Species from Camera Trap Data in Yanchiwan National… 571

Fig.4 Comparison of ICE and Chao 2 richness estimators for different species groups
Note: A-Total species; B-Mammal species; C-Avian species; D-Pheasant species.

Table 2 Summary of simple single-season occupancy models for Eurasian lynx, woolly hare, pika and Tibetan partridge from
our camera trap survey in Yanchiwan National Nature Reserve, Gansu Province, China

No. of No. of
AIC Model -2log AIC Model -2log
Model AIC ΔAIC parame- Model AIC ΔAIC parame-
weight likelihood likelihood weight likelihood likelihood
ters ters

Eurasian lynx Pika

Ψ(.), p(.) 1059.86 0.00 0.88 1.00 2 1055.86 Ψ(.), p(E) 757.39 0.00 0.57 1.00 2 753.39
Ψ(E), p(E) 1064.61 4.75 0.08 0.09 2 1060.61 Ψ(E), p(E) 758.58 1.19 0.31 0.55 2 754.58
Ψ(.), p(E) 1066.32 6.46 0.03 0.04 2 1062.32 Ψ(.), p(.) 761.49 4.10 0.07 0.13 2 757.49
Ψ(E), p(.) 1159.3 99.44 0.00 0.00 2 1155.3 Ψ(E), p(.) 762.52 5.13 0.04 0.08 2 758.52
Tibetan
Woolly hare
partridge

Ψ(.), p(.) 2021.11 0.00 0.94 1.00 2 2017.11 Ψ(E), p(E) 836.06 0.00 0.36 1.00 2 832.06
Ψ(E), p(.) 2026.79 5.68 0.06 0.06 2 2022.79 Ψ(.), p(E) 836.41 0.35 0.30 0.84 2 832.41
Ψ(.), p(E) 2081.35 60.24 0.00 0.00 2 2077.35 Ψ(E), p(.) 837.38 1.32 0.18 0.52 2 833.38
Ψ(E), p(E) 2086.97 65.86 0.00 0.00 2 2082.97 Ψ(.), p(.) 837.69 1.63 0.16 0.44 2 833.69

Notes: AIC is the AIC (Akaike’s Information Criterion) value for each model; ΔAIC is the relative difference in AIC values between each model and the
currently top-ranked model; AIC weight is a measure of support for each model being the best model; Model likelihood is the ratio of each model AIC
weight over the model weight for the top-ranked model; -2log likelihood is twice the negative log-likelihood evaluated at the maximum likelihood estimates
(MLEs). Ψ is the probability of site use, p is the probability of present and detected; The terms in parentheses represent the factors in the model for respec-
tive parameters, with ‘.’ indicating the parameter is constant and ‘E’ indicating elevation has been used as a factor.
572 Journal of Resources and Ecology Vol. 9 No. 5, 2018

Table 3 Summary of parameter estimates for selected species from our camera trap survey in Yanchiwan National Nature
Reserve, Gansu Province, China
Species Model Ψ± SE p± SE Naïve occupancy
Eurasian lynx Ψ(.), p(.) 0.62 ± 0.07 0.13 ± 0.01 0.60
Woolly hare Ψ(.), p(.) 0.79 ± 0.06 0.29 ± 0.01 0.77
Pika Ψ(.), p(E) 0.41 ± 0.07 0.15 ± 0.03 (average) 0.40
Tibetan partridge Ψ(E), p(E) 0.54 ± 0.27(average) 0.13± 0.03 (average) 0.47
Notes: Ψ is the probability of site use, p is the probability of present and detected; The terms in parentheses represent the factors in the model for respective
parameters, with ‘.’ indicating the parameter is constant and ‘E’ indicating elevation has been used as a factor.

Table 4 Summary of two-species single-season occupancy models for selected species, and parameter estimates from the
best model. ΨBa/rBa and φ/ indicate two custom parameterizations
Model No. of -2Log
Model AIC ΔAIC AIC weight φ ± SE pA± SE rA± SE
Likelihood parameters Likelihood
Lynx (A) × Woolly hare (B)
ΨBa/rBa 2971.47 0.00 0.51 1.00 8 2955.47 0.94 ± 0.04 0.01± 0.01 0.13 ± 0.01
φ/δ 2971.54 0.07 0.49 0.97 8 2955.54
Lynx (A) × Pika (B)
φ/δ 1816.30 0.00 1.00 1.00 8 1800.30 1.03 ± 0.11 0.18± 0.02 0.19 ± 0.01
ΨBa/rBa 1834.64 18.34 0.00 0.00 8 1818.64
Lynx (A) × Tibetan partridge (B)
ΨBa/rBa 1849.03 0.00 0.93 1.00 8 1833.03 0.83 ± 0.12 0.04 ± 0.01 0.17 ± 0.01
φ/δ 1854.34 5.31 0.07 0.07 8 1838.34
Note: AIC is the AIC (Akaike’s Information Criterion) value for each model; ΔAIC is the relative difference in AIC values between each model and the
currently top-ranked model; AIC weight is a measure of support for each model being the best model; Model likelihood is the ratio of each model AIC
weight over the model weight for the top-ranked model; -2log likelihood is twice the negative log-likelihood evaluated at the maximum likelihood estimates
(MLEs); φ is the species interaction factor; pA is the probability of detecting species A, given only A is present; rA is the probability of detecting species A,
given both are present.

4 Discussion workdays to detect 90% of its species accordingly. The spe-

4.1 Species inventory and sampling effort
Our results demonstrate the efficiency of camera traps in
producing inventories of terrestrial animal species in a
sub-frigid zone of Qinghai-Tibetan Plateau. A total of 16
other mammal species and 20 bird species were detected
during our snow leopard-focused camera survey. We re-
corded seven species that were not listed in previous inven-
tory of YCWNNR (Liu et al., 2010), including dhole, Ti-
betan fox, red fox, mountain weasel, Tibetan partridge, yel-
low-billed chough and dusky warbler. This suggests that
camera trap surveys are superior for detecting elusive spe-
cies or species with low density in certain area. Some spe-
cies that recorded in YCWNNR were not detected in our
survey, most likely because the camera stations did not in-
clude habitats used by these species (Chutiponget al., 2014).
As our survey covered from winter to spring, the lack of full
year survey may also resulted in non-detection of some spe-
cies.
We found that 6644 camera workdays of survey would
detect 90% of the animal species in our study area. For
mammals, birds and pheasants, it took 4019, 7808 and 6086
cies rarefaction curves of mammal and pheasant were al-
most leveled off after around 5000 camera workdays, sug-
gesting that the community of mammals and pheasants was
nearly completely sampled during our survey. However, five
mammals and one pheasant species were represented by less
than five independent images in the 7-month survey period,
showing that a substantial survey effort is still needed to
record some species. The rarefaction curve of birds was not
asymptotic as sampling efforts increased. As small body size
animals are more likely to pass in front of a camera without
trigging it (Tobler, et al., 2010), camera traps may not suf-
fice for the detection of bird species. Furthermore, owing to
different behaviors or habitat requirements, multiple meth-
ods are required to detect the entire bird community. The
ICE and Chao 2 estimators also indicated that some species,
especially bird species, were still missing from our camera
survey.
4.2 Occupancy modeling and species composition
Occupancy modeling provides reliable methods to assess
probability of occupancy and detection while accounting for
imperfect detection and site characteristics. It has been in-
ZHANG Chengcheng, et al.: Biodiversity Assessment on Mammal and Bird Species from Camera Trap Data in Yanchiwan National…
creasingly used in monitoring programs with the flexibility
of allowing for missing observations (MacKenzieet al.,
2002; Alexander et al., 2015b; Bohnettet al., 2015). In our
study, the site use and probability of detection of Eurasian
lynx were not associated with elevation, in contrast to other
studies (Laze and Gordon, 2016). One of the possible rea-
sons could be the lack of heterogeneity with regard to the
elevation of our camera sites (from 3184 to 4002 m). There
was a low discrepancy between naïve occupancy and the
estimated probability of occupancy for Eurasian lynx, indi-
cating sufficient detection by camera traps.
The two-species occupancy modeling between Eurasian
lynx with woolly hare, pika and Tibetan partridge provides
insights on predator-prey relationships. Woolly hare and
Tibetan partridge appear to be slightly avoiding site use with
Eurasian lynx, while pika seems use the sites independently
from lynx. The detection probability estimates (pA, rA) sup-
port this hypothesis from another aspect. Lynx are much
more likely to be detected when using the sites together with
woolly hare and Tibetan partridge, indicating strong preda-
tion, while there is no difference on detection probability
when use the sites together with pika. However, caution is
required when inferring this interpretation to other contexts,
as camera trapping is not the optimal method for surveying
woolly hare and pika. Besides, the scale of our snow leopard
targeted survey may introduce bias in the analysis of prey
species.
4.3 Study design and management implications
Our camera traps survey, although not originally designed
for wildlife diversity monitoring, could produce a reliable
inventory of mammal and bird species as a useful byproduct.
As many studies (Tobleret al., 2008; McCarthy et al., 2010;
Nainget al., 2015) demonstrated, a survey designed for one
purpose may produce data that are useful for another. Our
study was successful in identifying species around snow
leopard and provided important insights on richness estima-
tors and species interaction in a particular snow leopard
habitat.
However, if a survey is targeted primarily to produce a
species inventory, it could use a more flexible camera de-
sign covering station placements suitable for other species
as well. Species inventories serve a foundational purpose for
wildlife management. For programs that aim to monitor the
community composition over time, it is important to keep in
mind that a large survey effort is needed to register certain
species (Tobler, et al., 2010), thus a survey span a whole
year containing multiple seasons will be necessary for fur-
ther biodiversity monitoring.
Camera traps play an important role in biodiversity
monitoring (SoberónandLlorente, 1993), and basic informa-
tion on species presence is crucial for planning conservation
strategies in protected areas (O’Brien, 2008; Tobler, et al.,
2010; Nainget al., 2015). The results of this first systematic
573
camera survey in Yanchiwan National Nature Reserve show
its efficacy in species inventorying.
The nature reserve network can play a key role in such
monitoring in China (Li et al., 2010), as the camera traps
could be applied a regular survey method in other areas of
the reserve, for long-term monitoring projects and for bio-
diversity comparison between years. With a comprehensive
biodiversity and richness trends monitoring, it would con-
tribute to wildlife managements and conservation actions.
Acknowledgements
We acknowledge with thanks the support received from the Second
National Survey of Terrestrial Wildlife in China, the State Forestry
Administration of China. We also thank the Yanchiwan National
Nature Reserve administration for their support to the fieldwork.
We are grateful to HAN Xuesong for identifying avian species,
Alice Laguardia for providing constructive suggestions on data
analysis.
References
Akbaba B, Ayas Z. 2012. Camera trap study on inventory and daily activity
patterns of large mammals in a mixed forest in north-western Turkey.
Mammalia, 76(1): 43–48.
Alexander J S, Gopalaswamy A M, Shi K, et al. 2015a. Face Value: To-
wards Robust Estimates of Snow Leopard Densities. Plos One, 10(8):
e0134815.
Alexander J S, Shi K, Tallents L, et al. 2015b. On the high trail : examining
determinants of site use by the Endangered snow leopard. Oryx, 50(2):
231–238.
Alexander J S, Gopalaswamy A M, Shi K, et al. 2016. Patterns of Snow
Leopard Site Use in an Increasingly Human-Dominated Landscape.
Plos One, 11(5): e0155309.
Bohnett E, Riordan P, Shi K. 2015. Initial Assessment on Large and Me-
dium Sized Terrestrial Mammal Assemblage Using Camera Trapping in
Nangunhe Nature Reserve in Yunnan, China. Journal of Resources and
Ecology, 6(5): 331–344.
Butchart S H M, Walpole M, Collen B, et al. 2010. Global biodiversity:
Indicators of recent declines. Science, 328: 1164–1168.
Cardinale B J, Duffy J E, Gonzalez A, et al. 2012. Biodiversity loss and its
impact on humanity. Nature, 486: 59–67.
Chutipong W, A J Lynam, R Steinmetz, et al. 2014. Sampling mammalian
carnivores in western Thailand: Issues of rarity and detectability. Raffles
Bulletin of Zoology, 62(July): 521–535.
Chen S, Yu J, Chen X, et al. 2016. Camera-trapping survey on the diversity
of mammal and pheasant species in Gutianshan National Nature Re-
serve, Zhejiang Province. Acta Theriologica Sinica, 36(3): 292–301. (in
Chinese)
Colwell R K. 2013. Estimate: Statistical estimation of species richness and
shared species from samples. Version 9. User’s Guide and application
published at: http://purl.oclc.org/estimates.
He B, Sun R, Chen P, et al. 2016. Baseline survey of mammal and bird
diversity using camera-trapping in the Changqing National Nature Re-
serve of Shaanxi Province. Theriologica Sinica, 36(3): 348–356. (in
Chinese)
IUCN 2017. IUCN Red List of Threatened Species. Version 2017. 2.
Lande R. 1996. Statistics and partitioning of species diversity, and similar-
ity among multiple communities. Oikos, 76(1): 5–13.
Lan H, Jin K. 2016. Infrared-triggered camera technology application in the
investigation of mammals in Beijing Wulingshan National Nature Re-
serve. Acta Theriologica Sinica, 36(3): 322–329. (in Chinese)
Laze K, Gordon A. 2016. Incorporating natural and human factors in habi-
tat modelling and spatial prioritisation for the Lynx lynxmartinoi. Web
574
Ecology, 16: 17–31.
Li S, Wang D, Gu X, et al. 2010. Beyond pandas, the need for a standard-
ized monitoring protocol for large mammals in Chinese nature reserves.
Biodiversity and Conservation, 19(11): 3195–3206.
Liu N, Zhang H, Dou Z. 2010. Synthetical Scientific Investigation on Yan-
chiwan National Reserve in Gansu Province. Lanzhou: Lanzhou Uni-
versity Press. (in Chinese)
Mackenzie D I, Nichols J D, Lachman G B, et al. 2002. Estimating Site
Occupancy Rates When Detection Probabilities Are Less Than One.
Ecology, 83(8): 2248–2255.
MacKenzie D I, Bailey L L, Nichols J D. 2004. Investigating species
co-occurrence patterns when species are detected imperfectly. Journal
of Animal Ecology, 73(3): 546–555.
MacKinnon J, Phillipps K. 2000 (1st. ). A Field Guide to the Birds of China.
Changsha: Hunan Education Publishing House. (in Chinese)
McCarthy J L, McCarthy K P, Fuller T K, et al. 2010. Assessing variation
in wildlife biodiversity in the tien shan mountains of kyrgyzstan using
ancillary camera-trap photos. Mountain Research and Development,
30(3): 295–301.
Naing H, Fulle T K, Sievert P R, et al. 2015. Assessing large mammal and
bird richness from camera-trap records in the Hukaung valley of North-
ern Myanmar. Raffles Bulletin of Zoology, 63: 376–388.
O’Brien TG, Kinnaird M F, Wibison H T. 2003. Crouching tigers, hidden
prey: Sumatran tiger and prey populations in a tropical forest landscape.
Animal Conservation, 6(2): 131–139.
O’Brien TG. 2008. On the use of automated cameras to estimate species
richness for large- and medium-sized rainforest mammals. Animal
Conservation, 11(3): 179–181.
O’Brien TG, Baillie J E M, Krueger L, et al. 2010. The wildlife picture
index: Monitoring top trophic levels. Animal Conservation, 13(4):
335–343.
Rovero F, Augugliaro C, Havmøller R W, et al. 2018. Co-occurrence of
基于红外相机数据的甘肃盐池湾国家级自然保护区鸟兽生物多样性评估
张成成 1，王 君 1,2，Justine Shanti ALEXANDER3，窦志刚 4，乌力吉 4，董万涛 4，达布西力特 4，杨巨才 4，
时 坤 1,5
1. 北京林业大学野生动物研究所，北京 100083；
2. 曼彻斯特城市大学，科学与工程学院，曼彻斯特 M1 5GD，英国；
3. 国际雪豹基金会，西雅图 98103，美国；
4. 甘肃盐池湾国家级自然保护区管理局，甘肃酒泉 736399；
5. 北京市海淀区陆桥生态中心，北京 100083
摘 要：红外相机在物种多样性监测中有着重要的应用。我们于 2015 年 11 月至 2016 年 6 月间，在盐池湾自然保护区雪豹
（Panthera uncia）栖息地约 400km2 的范围内系统性地布置了 62 台自动红外相机，以期了解该地区的野生动物资源现状，估测该
地区的鸟兽物种总丰富度，评估红外相机在该地区鸟兽调查监测中的应用效果，并对关键物种的栖息地利用及种间相互作用进行
探究。调查共回收红外相机 57 台，累积相机工作日 10171 天，拍摄到野生动物有效独立事件 2868 次。记录到至少 17 种哺乳动
物，分属 5 目 11 科，记录到 20 种鸟类，分属 4 目 7 科，其中豺（Cuon alpinus）为 IUCN 濒危物种，雪豹和白唇鹿（Cervus albirostris）
为易危物种，兔狲（Felis manul）、香鼬（Mustela altaica）
、高山兀鹫（Gyps himalayensis）和秃鹫（Aegypius monachus）为近危
物种。使用 Chao2 和基于出现率的盖度估计（incidence-based coverage estimate, ICE）两个统计量进行物种总丰富度的非参数估计，
并做物种稀疏曲线，对取样强度进行评估。结果表明红外相机对兽类和雉类物种的取样较为充分，而对鸟类的取样还不完全。使
用占有率模型（occupancy modelling）对欧亚猞猁（Lynx lynx）及其潜在猎物物种的栖息地利用与种间相互作用进行分析。表明
调查范围内，欧亚猞猁对栖息地的利用与海拔因素无关；高原兔（ Lepus oiostolus）和高原山鹑（ Perdix hodgsoniae）在栖息地
利用上倾向于与欧亚猞猁相回避，而鼠兔（Ochotona spp.）对栖息地的利用则与欧亚猞猁相独立。
关键词：红外相机；物种编目；物种丰富度；物种累积曲线；占有率模型
Journal of Resources and Ecology Vol. 9 No. 5, 2018
snow leopard Pantherauncia, Siberian ibex Capra sibirica and livestock:
potential relationships and effects. Oryx, 1–7.
Scholes R J, Walters M, Turak E, et al. 2012. Building a global observing
system for biodiversity. Current Opinion in Environmental Sustainabi-
lity, 4(1): 139–146.
Si X, Kays R, Ding P. 2014. How long is enough to detect terrestrial ani-
mals? Estimating the minimum trapping effort on camera traps. PeerJ,
2: e374.
Smith A T, Xie Y. 2008. A guide to the mammals of China. Changsha:
Hunan Education Publishing House. (in Chinese)
Soberón M J, Llorente B J. 1993. The use of species accumulation func-
tions for the prediction of species richness. Conservation Biology, 7:
480–488.
The State Council. 1988. The wildlife under special state protection.
Tittensor D P, Walpole M, Hill S L, et al. 2014. A mid-term analysis of
progress toward international biodiversity targets. Science, 346(6206):
241–244.
Tobler M W, Carrillo-Percastegui S E, Pitman R L, et al. 2010. An evalua-
tion of camera traps for inventorying large- and medium-sized terres-
trial rainforest mammals. Animal Conservation, 11(3):169-178.
Voss R S, Emmons L H. 1996. Mammalian Diversity in Neotropical Low-
land Rainforests : a Preliminary Assessment. New York: American Mu-
seum of Natural History, 25.
Xiao Z. 2016. Wildlife resource inventory using camera-trapping in Natural
Reserves in China. Acta Theriologica Sinica, 36(3), 270–271. (in Chi-
nese)
Xue M, Jiang B, Li W, et al. 2016. Wildlife survey of Taihangshan Macaq-
ues National Nature Reserve using camera-trapping in Jiyuan City, Henan
Province, China. Acta Theriologica Sinica, 36(3), 313–321. (in Chi-
nese)
Zheng G. 2017. A Checklist on the Classification and Distribution of the
Birds of China. Beijing: Science Press. (in Chinese)