Module 7. Monitoring Wildlife Populations for Management (Background, Presentation and Exercises)

Full text

Module 7
Monitoring wildlife
populations for management
Emma J Stokes, Arlyne Johnson &
Madhu Rao

Table of Contents
1. Biological monitoring in a management context...........................................................................3
2. What to monitor: Setting conservation targets for monitoring...................................................... 5
2.1. What species or groups of species to monitor?...............................................................................6
2.2. What measure to use?...............................................................................................................7
3. How to monitor: Designing monitoring programs ........................................................................9
3.1. Detectability or detection error .................................................................................................10
3.2. How to incorporate detectability into a monitoring design?.........................................................11
3.3. Spatial variation or sampling error...........................................................................................15
3.4. How much sampling effort is enough? ......................................................................................19
3.5. Determining statistical power to detect change over time............................................................. 22
3.6. Improving the efficiency of sampling designs.............................................................................. 25
3.7. Decision making: matching objectives with available resources....................................................25
4. Practical considerations in designing sustainable monitoring programs..................................... 26
4.1 Personnel and capacity building...............................................................................................26
4.2 Making a workplan and schedule.............................................................................................27
4.3 Budget planning.....................................................................................................................28
5. Data management and documentation........................................................................................29
6. Communication and dissemination of results.............................................................................. 29

3
1. BIOLOGICAL MONITORING IN A MANAGEMENT CONTEXT
Conservation managers devote a considerable amount of time and resources to preserving
wildlife populations. In recent years there has been a growing recognition amongst
conservation practitioners, donor agencies, international conservation organizations and the
scientific community of the need to measure the success of our efforts in meeting
conservation or policy objectives, and evaluating if conservation resources are well spent
(Pullin & Knight, 2001; Sutherland et al., 2004; Ferraro & Pattanayak, 2006; Kapos et al.,
2008).
In this context, wildlife monitoring programs should form a core component of any
conservation management project and, if integrated fully into the project management cycle
and decision-making process, monitoring can play three important roles:
i) firstly, it can provide managers with information on the status of wildlife populations
before deciding on the appropriate course of conservation action to take;
ii) secondly, monitoring programs can evaluate the effectiveness of management actions
relative to stated objectives; and
iii) thirdly, in an adaptive management setting, monitoring programs can provide the
important feedback loop for learning about which actions lead to the success or failure
of a particular conservation approach, in order to specifically inform and improve upon
management practice in the future (Nichols & Williams, 2006; Lyons et al., 2008).
Understanding the role of monitoring in this context can help inform the design of
monitoring programs. Monitoring data should not be collected haphazardly in the hope that
one day this might be useful for conservation. With limited budgets and staff, managers
instead need efficient monitoring programs that are focused on providing precisely the
information needed to make the right conservation decisions [see Box 1]. To this end, the
formulation of clear and explicit monitoring objectives is a key first step in the planning of any
wildlife monitoring program (Yoccoz et al., 2001; Legg & Nagy, 2006; Nichols & Williams,
2006; MacKenzie, 2009).

4
In this module we provide general guidelines for the design and implementation of
management-orientated wildlife monitoring programs. We do not aim to provide an
exhaustive list of all possible survey methodologies. Rather, we highlight common pitfalls
and potential sources of error in the design and interpretation of wildlife monitoring data,
and provide recommendations for addressing these issues directly, using examples from
different survey techniques in different contexts.
In many parts of South-East Asia the technical challenges of designing effective wildlife
monitoring programs are confounded further by over-hunting of wildlife, resulting in
severely depleted populations which themselves are often the subject of management and
recovery programs. At very low densities, the design of wildlife monitoring programs needs
to strike an important balance between technical rigor on one hand and cost-
effectiveness on the other. We provide some recommendations for redressing the balance
in such situations. Finally, the technical challenges of designing wildlife monitoring
Box 1: Targeted monitoring vs. Surveillance monitoring (from (Yoccoz et al., 2001; Nichols
& Williams, 2006)
Targeted monitoring is defined as monitoring that is integrated into conservation practice. The
ideal example of this is provided by an adaptive management framework, which is an iterative
process that directly addresses the uncertainty in biological systems by incorporating a set of
competing models about how the system responds to management interventions. Adaptive
management typically involves 5 components: 1) Clear management objectives, 2) potential
management actions to meet the objectives, 3) models of system response to different
management actions, 4) measures of confidence in the models, and 5) a monitoring program to a)
provide estimates of system state and other relevant variables to make periodic management
decisions, and b) discriminate between competing models about how the system works and adjust
our confidence in different models accordingly.
Unlike targeted monitoring, surveillance monitoring is not guided by a priori hypotheses about how
the system responds. Surveillance monitoring in conservation typically involves a two-step process.
First, population declines are identified by monitoring data by means of a statistical test of a null
hypothesis of no decline versus a decline. Following the statistical detection of a decline, either of
two actions is recommended as a second step. One is to initiate active conservation immediately,
and the other is to initiate studies to understand the ‘cause’ of the decline, followed by active
conservation. Key to both is the detection of a population decline as a trigger for initiating
management actions. This approach to monitoring is considered by some as inefficient and
frequently ineffective and has been criticized as resulting in a ‘too little, too late’ scenario.

5
programs are confounded in the tropics by the logistical challenges of accessing vast and
remote forests with low technical capacity and thinly stretched budgets. We therefore also
consider the criteria for ensuring the long-term sustainability of monitoring programs.
2. WHAT TO MONITOR: SETTING CONSERVATION TARGETS FOR
MONITORING
Deciding on what we want to monitor depends largely on the management objective or the
particular questions you want to ask. There are two aspects to the question of what to
monitor: what variable (or variables) need to be monitored, and what measure should be
used.
The management and monitoring of biological systems encompasses a variety of different
biological variables of interest, ranging in scale from species to ecosystems and including a
variety of quantitative and qualitative measures of biodiversity and populations. In this
module we focus on quantitative measures of wildlife populations, and specifically large
mammals, but the guiding principles outlined here can be equally applied to other taxa.

6
2.1. What species or groups of species to monitor?
For management programs aimed at a particular wildlife species, such as recovery of tigers
to a particular level in a protected area, the conservation target for monitoring must clearly
include the species under management. However, for some management programs it might
also be important to include additional variables that bear some functional relationship to
the conservation target. For the example of recovering wild tiger populations, a
conservation manager also needs to ensure that there is an adequate prey base, and so the
abundance of key prey species might be a key variable to monitor in addition to the size of
the tiger population.
For more general management objectives that relate to the integrity of protected areas or
preservation of key habitats, the choice of what species or group of species to monitor
requires careful consideration and the selection of those species that will provide the most
useful and indicative information about how the system is responding to a particular
management intervention or strategy. These decisions can be made in a number of different
Box 2:
Examples of rationale
for selection of
target species
to monitor in Nam Kading
National Protected Area, Lao PDR (from Strindberg et al., 2006).
1. Tiger covers all habitat types defined for the Nam Kading landscape and is impacted largely
by the threats of ‘hunting for trade as medicine/trophies’ and ‘prey depletion’. It is
assumed that if these two threats are reduced that tigers will increase in the NPA.
2. Wild Pig is dependent on the ‘seasonal streams and pools’ habitat in the NPA and is heavily
impacted by the three types of hunting, namely ‘hunting for trade as food’, ‘hunting for
subsistence consumption’, and ‘hunting as a result of conflict from crop raiding’. It is
assumed that if seasonal streams and pools are maintained and if hunting is reduced that
wild pig will increase in the NPA.
3. Great Hornbill is threatened by ‘logging’ and ‘shifting cultivation’ in the NPA due to loss of
big nesting trees that are important for their survival. It is assumed that if habitat loss is
reduced, as well as hunting, that great hornbill populations will increase in the NPA.
4. In addition to the threats mentioned for the previous three species, White-Cheeked
Crested Gibbon is also extremely vulnerable to the threat of ‘habitat fragmentation’. It is
assumed that if habitat fragmentation is reduced, as well as other threats listed, that
gibbons will increase in the NPA.

7
ways and based on a number of key biological and conservation criteria (e.g. (Redford et al.,
2003)), but at a minimum should include a series of a priori hypotheses or assumptions
about how the species will respond to a particular management intervention (see Box 2 on
selection of what species to monitor and assumptions about their response to management)
2.2. What measure to use?
Both the specific management objective and the selection of appropriate species or taxa to
monitor have direct implications for the attribute to be measured. For example, wildlife
managers are frequently interested in measures of abundance, and specifically in density
(number of individuals/unit area) or population size (total number of individuals in a
defined area). However, population size or density is typically one of the most costly
measures to obtain, and for rare or elusive species in particular, is often precluded by the
effort required to obtain rigorous estimates that are meaningful as a monitoring tool.
In such instances, alternative measures of abundance can be used, including relative
abundance (typically an index or proxy measure that has some constant relationship to
abundance) or occupancy (proportion of area occupied by a particular population
(Mackenzie et al., 2002) (See Box 3). Whilst the decision of which measure to use is
ultimately determined by the management objective it must also be considered in terms of
cost and available budget. The choice of different measures will in turn have implications
for the design of monitoring programs (Williams et al., 2002), but these different measures
should still subscribe to a minimum standard of statistical rigor, as we discuss in Section 3.

8
The types of measures we have been describing so far are known as state variables. A state
variable is a metric that summarizes the status of a population of interest at a particular time.
Species richness, abundance, even simple presence of a species, are examples of commonly
used state variables. These types of variables are typically of most immediate interest to
management programs. However, there is now a growing interest by managers in the
dynamic processes that influence the response of state variables, and to include specific
measures of rate parameters such as reproduction, immigration or emigration in their
monitoring programs (Yoccoz et al., 2001). To continue the example of a recovering tiger
population: our primary objective is to determine if tiger numbers are increasing over time
in response to management interventions, and so we monitor tiger population size over time
as our state variable. However, at the same time we might also be interested to see if
increasing tiger numbers are due to increased breeding amongst the resident population, or
increased immigration into the protected area from outside. This in turn can have important
implications for the spatial scale at which recovery programs are targeted.
Box 3: Examples of Relative Abundance and Occupancy from protected areas in Lao PDR.
An example from Nam Et-Phou Louey NPA of relative abundance is number of camera trap photos
of tiger per camera trap day (Johnson et al., 2006). In this case, a camera trap day is defined as each
24-hour period that a camera trap is operating to capture photos of a tiger in the NPA. This
measurement does not tell us how many tigers live in the NPA but provides a relative measurement
that can be compared with other areas where camera traps are used to monitor relative abundance
of tigers.
An example of occupancy from the Nakai-Nam Theun NPA is that monitoring along line transects in
200 km2. of the Nam Chae catchment in 2007 found that Douc Langur occupied 87% of the area
(Johnson and Johnston 2007). Note that this measure of abundance does not estimate how many
Douc Langur live in the Nam Chae catchment in terms of density (individuals per km2) but provides
and estimate of the proportion of the area that is occupied by Douc Langur.

9
3. HOW TO MONITOR: DESIGNING MONITORING PROGRAMS
Managers need to have reliable information about the status of wildlife populations and
their response to interventions in order to make informed decisions. As we have seen,
monitoring programs can play a key role in providing this information, by evaluating our
assumptions about the status of populations, or how they are responding, relative to a stated
objective or target. Developing clear monitoring objectives and targets is just the first step in
the implementation of effective monitoring programs.
All too frequently however, the potential of monitoring programs to inform management
decisions is wasted during the design phase. Results from poorly designed monitoring
programs are misleading, due to poor quality information, and in some cases can do more
harm than good if conservation effort is invested poorly as a consequence (Legg & Nagy,
2006). It is therefore critical that careful consideration is given to the statistical design and
analysis of monitoring programs before substantial investment is made on their
implementation and data collection. To this end, the bridge between science and
management is an important one, and managers should be encouraged to seek appropriate
scientific advice on designing monitoring programs at the outset.
An underlying premise of successful monitoring programs is that the design is simple and
the measures are straightforward, unambiguous and replicable (Legg & Nagy, 2006).
Overly ambitious monitoring programs suffer from being unsustainable both financially and
in terms of technical staff capacity (Danielsen et al., 2005);(Poulsen & Luanglath, 2005).
Most often however, monitoring programs suffer from ‘cutting corners’ on the measures
they use, largely in a bid to save on limited conservation or management funds. Whilst cost
is one of many practical considerations to be taken into account in designing monitoring
programs, we outline here some minimum standards for the statistical design of monitoring,
which, if met, will not only ensure a minimum level of rigor and thus usefulness of the
results, but also improve cost-effectiveness in the long-run.
There are two common sources of error in population estimates that need to be taken into
account in the design phase: detection error and spatial variation or sampling error (Yoccoz et
al., 2001; Williams et al., 2002). Furthermore, we consider the importance of sample size
and sampling efficiency in determining the capacity of monitoring programs to detect true
changes in the target population with adequate statistical power [see Box 4].

10
3.1. Detectability or detection error
Many wildlife monitoring programs assume that, if animals or signs of animals are present,
then they will always be detected. This assumption is implicitly made with simple indices of
count data as a measure of relative abundance. For example, in presence/absence counts a
survey team visits a site and records if the species was present or absent. Similarly with
count data on transects survey teams count the total number of animal signs or sightings per
distance of transect walked. The resulting index of relative abundance assumes a constant
relationship with actual abundance N. None of these measures account for the eventuality
that signs or individuals were present but undetected. In reality, few survey methods permit
100% detection of all signs of a species, or all individuals in a population (see Box 5).
In such circumstances, the estimated abundance of a population can be represented as:
Box 4: Accuracy and Precision
A major concern with the design of monitoring programs and the estimation of population
parameters (e.g. abundance) is the accuracy and precision of the survey results.
Accuracy refers to the magnitude of systematic errors or degree of bias associated with an
estimation procedure. This affects how well the estimated value represents the true value.
Systematic errors may or may not be measurable and can cause estimates to consistently under or
over-estimate the true value. Detection error and sampling error are examples of two sources of
error that can result in biased estimates.
Precision refers to the variability in estimates. High precision means that random variation
associated with the collection procedure is minimized. Generally larger sample sizes provide greater
precision than small sample sizes. If comparing estimates over time, high variation (or low precision)
makes it difficult to determine if there are statistically significant trends in the population.
Therefore, it is important to carefully choose sampling techniques and develop sampling schemes
that both meet the necessary assumptions and minimize the variation between samples
In designing a monitoring program we are generally looking to minimize bias and increase precision
of our estimates. Acceptable levels of accuracy and precision should therefore be determined prior
to conducting a survey.

11
Where is the abundance estimate, is the count statistic and is the estimated detection
probability (e.g. (Thompson, 1992; Lancia et al., 1994; Williams et al., 2002)
The probability of detecting a sign or individual animal can vary over space and time, for
example with habitat type, time of day or different observers. This, in turn suggests that
sampling designs that fail to account for probability of detection, or detection error, will
result in biased population estimates and are therefore unreliable as a tool for monitoring
true changes in populations over time.
3.2. How to incorporate detectability into a monitoring design?
There are a number of accepted and standardized methods for incorporating imperfect
detection into survey designs for monitoring programs [see Box 6]. The gold standard for
these methods is distance sampling (Buckland et al., 2001) and mark-recapture techniques
(Otis et al., 1978; White et al., 1982; Pollock et al., 1990), which incorporate detection error
into estimates of population density and true abundance.
These methods are expensive to implement and require well-trained personnel combined
with adequately large sample sizes, which often precludes their use over very large areas or
at very low population densities such as found in many areas of Lao PDR. In these
Box 5: Sampling and detectability
x
xxx
xo
xo
xxxx
xxx
xox
oxxoo
xoxooox
x
xxx
xo
xo
xxxx
xxx
xox
oxxoo
xoxooox
This box shows a forest divided into 100
units.
X = occupied cell where species is
detected.
O = occupied cell where species is not
detected.
Blank = cell where species does not occur.
Thus the observed occupancy is 0.2 or
20% of the forest. But the true occupancy
is 0.3 or 30% of the forest. The difference
is due to not detecting species when it is
present.

12
situations, indices of relative abundance are frequently used. There exist approaches for
dealing with detectability for indices of relative abundance (Conroy & Nichols, 1996); these
approaches typically rely on identifying sources of variation in detection probability (such as
time of day and other environmental conditions) and reducing them in the survey design.
A useful alternative in low-density situations or at large geographical scales is occupancy-
based methods, which have been successfully used for monitoring wildlife populations over
time in the Nam Et-Phou Louey NPA, Nam Kading NPA and Nakai-Nam Theun NPA
(Strindberg et al., 2007, Johnson and Johnston, 2007, Johnson et al., 2008). Occupancy
surveys incorporate imperfect detection into presence/absence data, and permit estimates of
the probability of detection and the proportion of area occupied (Mackenzie et al., 2002;
MacKenzie et al., 2006). Proportion of area occupied is often used as a surrogate for
abundance but is also useful as an alternative state variable for the population of interest,
and a metric with which to monitor changes in the status of a population over time.

Box 6: Examples of methods that incorporate imperfect detection into sampling design
Estimating absolute densities of tiger prey species using line transect sampling (from Karanth & Nichols,
2002)
Line transects sampling is an example of an abundance estimation approach known as distance
sampling (Buckland et al., 2001). During a line transect survey, the observer walks a series of lines
and counts any animal or a given species that he/she detects. For every animal detected the
observer measures the perpendicular distance from the animal to the survey line. In line transects we
do not assume that all animals can be detected. However, a fundamental assumption is that all
animals on the survey line are detected with certainty. Intuitively we would expect that the further
away animals are from the survey line then the harder they are to detect. The key to distance
sampling is to fit a detection function to the observed perpendicular distances and use this to estimate
the proportion detected ( ).
Figure 2. Conceptual basis for
line transect sampling. The red
curve represents the detection
function that has been fitted to
the real data. The area under
the curve represent the animals
seen and the area above it
represents the animals missed.
The proportion seen, p is
estimated from the area under
the curve divided by the total
area.
Figure 1 Diagrammatic
representation of a transect
survey. A number of transect
lines within the study area are
‘walked’ and perpendicular
distances recorded to all
animal clusters (groups of
animals) detected. Note that
not all clusters are detected.

14
.
Box 6 (cont.)
Estimating absolute densities of tigers using capture-recapture sampling (from Karanth & Nichols, 2002)
Capture-recapture is a survey method in which the total number of animals caught is counted and
the associated detection probability is the probability of being captured. Capture-recapture
methods also require that an individual animal can be reliably identified. A ‘capture’ can mean an
animal is physically caught and marked with a tag to identify it or it can mean that an animal is
captured in a photograph for example and identified by unique markings such as stripes on a tiger.
The detection probability is estimated by the pattern of captures/re-captures for each animal on
each sampling occasion over the entire survey period. To ensure that all individuals have a chance
of being captured, the survey design has to ensure that no ‘holes’ exist in the sampled area. For
example, when applying camera traps to estimate tiger abundance, sampling locations have to be
sufficiently close together to ensure a tiger could not pass between them and avoid being captured.
Figure 3. Example of a capture
-
recapture camera
trapping design for estimating tiger abundance in
Huai Kha Khaeng Wildlife Sanctuary, Thailand:
Camera-traps = 180 locations (3-4km spaced)
Camera-trap area = 981 km²
Effective area = 1745.9 km²

15
3.3. Spatial variation or sampling error
It is often logistically difficult or costly to survey entire protected areas or large landscapes.
As a result, sampling locations are frequently selected, and then used to make inferences
over a larger area, encompassing the population of interest (see Box 8 below). There is often
considerable spatial heterogeneity in natural systems, which if not accounted for in sampling
schemes, can introduce substantial bias or sampling error in measures of your target
population (Dixon, 1998; Yoccoz et al., 2001; Pollock & Farnsworth, 2002). Ensuring
adequate spatial coverage or spatial representativeness is therefore an important
consideration in the design of monitoring programs. Furthermore, spatial coverage should
be considered at an appropriate ecological scale for the species under study.
Let us continue our example of a tiger recovery conservation program. We are interested in
monitoring a tiger population to assess progress towards a defined increase of, say, 50%. Let
us assume our target population encompasses a landscape of 4,000km2 including a national
park and a large multiple-use buffer zone. We suppose that tiger densities are lower in the
multiple-use zone than in the protected area because of substantial hunting of tigers and
their prey – we therefore have already made the assumption that there will be spatial
variation in the distribution of tiger abundance at the landscape scale. If we are interested in
monitoring a tiger population increase at the landscape scale then we must account for this
spatial variation by adequately sampling across the landscape. For example, if we sampled
only in the protected area and then tried to extrapolate our population estimates across the
Box 6 (cont.)
Estimating occupancy rates of a species using repeat presence/absence surveys (from Mackenzie et al.,
2002)
Occupancy surveys or ‘presence-absence surveys’ involve a sampling method that requires
multiple visits to sites during an appropriate time-period when a species may be detected.
However, a species may go undetected at these sites even when present, resulting in ‘false
absences’. The patterns of detection and non-detection (presence/absence) over repeat visits
permits estimation of detection probability and the parameter of interest, proportion of sites
occupied. At each sampling occasion (visit) at each sampling site, observers can apply a number
of different sampling methods to detect the species of interest. Unlike capture-recapture
techniques, detection histories are compiled for a particular site rather than an individual. These
can be compiled through repeat visits by a single observer to a site over time, single visits by
multiple observers to a site, or visits to multiple locations in a site during a single time period.

16
entire landscape, we would have ignored spatial variation or sampling error and our
population estimates at the landscape-scale would be biased, and in this example,
overestimated. In summary, care should be taken to define precisely the target population of
interest (defined here as the scale or area at which inferences about the population are to be
made), and that this, in turn, is then used to design sampling schemes at spatially
appropriate scales. The question of scale is also of ecological relevance as management
information needs will likely vary over different geographic scales for different species. For
example, wide-ranging or migratory species may need to be monitored over a much broader
geographic area.

Box 7: Example of estimating occupancy rates of species using repeat presence / absence surveys
in the Nakai-Nam Theun NPA, Lao PDR (from Johnson and Johnston 2007)
Forest transects for arboreal mammals and hornbills in Nakai-Nam Theun NPA are focused on six indicator
species groups: Brown Hornbill – a small hornbill; large hornbills including Rufous-necked, Great and
Wreathed Hornbills; Black Giant Squirrel, Douc Langur, White-cheeked Gibbon, and macaques. In the field,
teams conduct surveys from 0600 to 1100 with observers moving slowly and silently along the transects
scanning the treetops and openings for signs and sounds of indicator species. Teams monitor along each
transect for four consecutive mornings before moving camp to the next transects. While moving along the
transects, the team leader records all observations on the field data form.
Estimates of true occupancy and standard error (SE(boot)) of the occupancy
estimate (precision) for wildlife indicators on forest transects
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Small hornbills Large hornbills
(>1sp) Black Giant
Squirrel Macaques (>1sp) Douc Langur Gibbon sp.
Occupancy estimate
Transects in 2007 regularly
detected all six wildlife indicator
species (forest hornbills (large and
small), Black Giant Squirrel, Douc
Langur, gibbon and macaque). The
analyses provided estimates of the
true occupancy for all of the
indicators ranging from a low of
60% (SE 0.07) for large hornbills to
a high of 91% (SE 0.06) for
macaques. Detection probabilities
for all six indicator species ranged
from a low of 0.38 for macaque to a
high of 0.60 for small hornbills. The
precision of the estimates for all
species was high (SE <0.07).

The potential for spatial bias in the design of monitoring programs can manifest itself in a
number of different ways. The above example refers to the error in extrapolating inferences
about populations outside the sampled area. In addition, systematic spatial bias can be
introduced into a survey design if care is not taken in the selection of sampling locations.
For example, if a particular species of interest was monitored only from roads, as these
were logistically easier for field teams to access, then all we could infer from our population
Box 8: Selecting sampling locations in the Nam Et-Phou Louey NPA (from Johnson,
Vongkhamheng, et al., 2006)
In the NEPL NPA, tiger and prey surveys were conducted using 50 camera traps set in five 100 km2
sampling blocks in the interior and proposed extension areas of the NPA as far away from enclave villages
as possible (see map below). In this case, sampling locations were selected to be representative of the
least disturbed areas of the NPA. Sampling blocks were spaced from 15-30 km apart. Each block was
divided in 25 subunits of 2 km2 and a random coordinate was chosen within each subunit. A pair of
cameras, to photograph both sides of individual tigers, was placed in an optimal location near active
animal trails within 500 m of the random coordinate.

19
estimates would be the abundance (or other state variable) of the population on roads. If our
population of interest was at the landscape-scale, then this measure would neither be
spatially representative nor particularly useful given that roads are a specific habitat feature,
which are frequently associated with hunting and other threats to wildlife. Rather, a
sampling design should identify and explicitly incorporate any potential gradients in wildlife
abundance, associated with vegetation or human factors. Given that most spatial variation
in both natural and human systems is often un-systematic, then selection of sampling sites
on a simple random or systematic (with a random starting point) basis is often sufficient to
incorporate spatial variation and address sampling error.
3.4. How much sampling effort is enough?
Recommendations on the amount of sampling effort required often have to balance the need
to collect sufficient data to make valid statistical inferences with the need to minimize cost
and time expenditures. The actual number of points, transects, sites etc. that should be
sampled and the number of times each should be revisited during a particular field season
will vary depending on the rarity of the species, variability of habitat and the objectives of
the monitoring program.
Ideally, the monitoring objectives, or the particular question you are interested in answering,
should dictate the scale, intensity, and accuracy and precision of the monitoring estimates.
Once these are identified the resources required to accomplish the surveys can be estimated.
However, because resources are often scarce, methods and specific objectives may have to
be adjusted to what is affordable.
In general, the cost of collecting data increases as the scale broadens, the focus intensifies
and/or the demand for accuracy and precision increases. The cost of implementing surveys,
as well as the need for skilled and highly trained staff, will also typically increase from
measures of occupancy and relative abundance being the least expensive, to estimates of
absolute abundance or density being the most expensive. In reality there are often trade-offs
to be made between all these factors. Mathematical equations are available to estimate the
sample sizes required to produce a reliable estimate of a population parameter. The
technique used to estimate sample size will vary according to the particular method used,
for example line transects (Buckland et al., 2001), mark-recapture (White et al., 1982;
Pollock et al., 1990) or occupancy surveys (Mackenzie & Royle, 2005) and even for a

20
particular method, e.g. occupancy surveys, the estimation of sample size will depend on the
underlying assumptions of the distribution of abundance (Royle & Nichols, 2003; Joseph et
al., 2009), for example whether a species is typically randomly distributed or in a clumped
distribution, which in turn is likely to vary between species and between habitats. In
summary, determining the optimum number of samples needed should be an initial step of
every wildlife population survey or monitoring program, regardless of the state variable (e.g.
occupancy, abundance etc.) that is being measured (see Box 9 for an example of
determining sample effort for the wildlife monitoring program in the Nakai-Nam Theun
NPA in Lao PDR).

21
Box 9: Determining sample effort for wildlife monitoring along forest transects in the Nakai-Nam
Theun NPA (from Johnson, O’Brien, et al., 2005)
To determine the sampling effort for wildlife monitoring along forest transects in the Nakai-Nam Theun (NNT)
NPA, patrol reports by Village Conservation Monitoring Unit (VCMU) teams were compiled to estimate how
frequently wildlife were encountered along trails in various sectors of the NPA (Table 1). The VCMU data
indicated that wildlife populations were severely depressed and infrequently encountered. For example,
arboreal mammals and birds were encountered at rates ranging from 0 to 10.63 for every 100 km of transect.
By comparison, the encounter rates for small and large hornbills in the Bukit Barisan National Park in southern
Sumatra are 2 times and 10 times greater. The NNT VCMU data sets indicated that extensive sampling effort
would be required to accurately estimate the occupancy of arboreal mammals and hornbills along the dry
season forest transects.
Table 1. Encounter rates (sign or observation/100 km of transect) for arboreal mammals and
hornbills during VCMU patrols between 2000 and 2002.
Species Thameuang
Navang Xonglek Makfeuang Teung Nameuy
Km. Transects 1,537 1,543 923 2,631 1,832 837
Giant squirrel 5.46 4.67 10.30 6.00 5.29 10.63
Douc Langur 2.80 1.23 3.79 4.03 1.04 7.05
Gibbon 6.90 4.80 4.88 2.58 1.09 2.39
Large Hornbills 4.50 2.66 8.34 3.76 0.87 4.78
Small hornbills 4.10 2.85 4.12 4.67 5.18 7.77
To estimate the sampling effort, the VCMU encounter data was used to simulate the detection of an indicator
species on a forest transect. For instance, if small hornbills have an encounter rate of 0.05 per km on transect
walks, we specified sampling designs of different lengths of transect, different number of transects and
different intensities of visits to the transect. Hornbill detections were assigned randomly to transects and
visits such that detections summed to 0.05/km over the sampling design. If we assume 100 transects of 5-km
transects were visited 5 times, we would expect to encounter hornbills 0.25 times per transect or 125 times
over the entire survey. The simulated data was then used to estimate detectability for an encounter rate under
different levels of occupancy and sampling intensity. This allowed us to estimate the minimum and maximum
detectability we could expect and then apply the range of detection probabilities to estimate precision and
accuracy of occupancy statistics that result from different levels of sampling. The result of the simulations was
a set of tables (see Appendices 2-4 in Johnson et al., 2005) that track the accuracy and precision of sampling
under different levels of occupancy, transects and visits, modeling the performance of sampling x transects on
y visits under the assumption of true occupancy and a detection probability. In this way we answered the
question of how much effort was needed to estimate occupancy accurately and precisely and how much
sensitivity
there would be to
detect change in the system over time.

22
3.5. Determining statistical power to detect change over time
For long-term monitoring programs, we are interested in not only estimating a population
parameter in a single field season, but also in detecting changes in the population parameter
towards a desired target over time. Statistical estimates of sampling effort required to detect
changes or trends over time tend to focus on the concept of statistical power (Field, 2005;
Legg & Nagy, 2006). In this instance, statistical power refers to the probability of detecting a
true change if present (1-β). In general, the power of a test is influenced by the probability of
Type 1 error (α), which is the probability of falsely detecting a change when one isn’t
present (a ‘crying wolf’ scenario), the probability of Type 2 error (β), the probability of
failing to detect a true change when one is present, sample size (n), variability or precision
of the population estimate, and the strength of the trend or magnitude of the desired
change to be detected, also called the effect size (often denoted as r, or rate of change). The
relationship between these parameters depends on the ecological process producing the
trend and the techniques used to detect it. For this reason, the selection of an appropriate
model to evaluate power is critical (Gerrodette, 1987), see Box 10).
Conducting an a priori power analysis during the planning of a monitoring program can
provide guidance on designing sampling schemes that ensure adequate sample size. This
exercise will help to prevent the implementation of a monitoring program which is too weak
and unable to discriminate a meaningful difference over time. This is often due to a sample
size that is too small and/or has high variability in the study population. It is possible, and
indeed a recommended approach, to account for and reduce some of this variability with
improved efficiency of sampling design (see below) or by increasing the number of replicates
or sampling occasions (see Box 10).
Whilst a power analysis is essentially a statistical process, it requires meaningful input from
managers or scientists (Lenth, 2001). In this context, a good starting point is to ask the
question: “What change (or effect size) do I expect – or hope - to see?” Meaningful effect sizes in
turn need to be both biologically feasible for the species under study as well as for the given
time-frame of the study. They also need to be of direct relevance to the management
objective. In our tiger recovery example, we have explicitly stated that our overall objective
is a 50% increase in our tiger population over 10 year. We can them examine various
scenarios of sampling schemes with different sampling effort to evaluate if our proposed
monitoring design is able to detect this defined change with adequate statistical power. Bear

23
in mind that, in general, reducing the effect size will increase the amount of sampling effort
required, and that reducing the effect size for a given level of sampling effort will reduce the
statistical power to detect change (see Box 10).
Statistical power becomes particularly important when the information resulting from the
monitoring program will go on to influence management recommendations. For example, a
particular monitoring objective might be to detect a decline in an endangered species in an
area under a particular logging practice. Failure to detect any true decline due to low
statistical power provides ‘evidence’ that forest-cutting is having no effect on this species and
thus the recommendation is for this management practice to continue.
Recently many free user-friendly software packages have become available for power
analysis, such as TRENDS
(http://swfsc.noaa.gov/textblock.aspx?Division=PRD&ParentMenuId=228&id=4740;
(Gerrodette, 1987) and MONITOR
(http://www.mbr-pwrc.usgs.gov/software/monitor.html) (see also (Thomas, 1997) for a
review).

Box
10
:
Power analysis for detecting trend
s in sea otter populations under different sampling scenarios
using TRENDS (from Gerrodette, 1987)
(Gerrodette, 1987) examined the feasibility of monitoring trends in sea otter populations in California, USA
using aerial strip transects conducted by plane. They first conducted a pilot study of 7 aerial transects to
determine the precision of aerial counts. They estimated the co-efficient of variation (CV) as 0.13. They
assumed that CV was proportional to the inverse of the square root of abundance and that sea otter
growth was likely to be exponential. They also assumed that α = 0.05.
Various sampling scenarios were investigated. Firstly they investigated the power of detecting various
annual rates of increase (r) at different sampling intensities (number of flights/year) for a monitoring
program of 5 year duration (Fig 1). Then they supposed that the total monitoring time was not fixed, and
wished to know how many annual surveys would be required to detect a given trend at different survey
intensities (flights/yr), with 95% power (Fig 2).
Finally, they asked whether annual surveys
were the optimal sampling frequency and,
if the population is growing slowly at 5%/year,
would it be more efficient to survey every 2 or
3 years at a level of survey intensity of 2 flights/
survey year (Table 1).The number of surveys
(and therefore costs) could be reduced by half
if done every 3 years rather than every year.
However the total number of years to detect a
change will increase from 7 to 9 years.
The additional ‘cost’ of waiting longer to detect a change (and the potential conservation risks of delaying
potential management action) needs to be weighed against the financial benefits of reducing survey
frequency. This will depend on your specific objectives and species under study.
Fig 2
. Minimum number of annual surveys requir
ed
to detect various rates of annual increase in
population size of California sea otters. More flights
permit more precise esti- mates of population size
during each survey, hence fewer annual surveys
required.
Table 1. Effect of different sampling frequencies on
the number of surveys required to detect a mean
5% annual increase in sea otter populations
Fig 1.
Power curves for detecting various rates of
annual increase in population size of sea otters in
central California using five annual aerial surveys.
More flights per year permit more precise
estimates of population size, hence greater
power to detect a given rate of increase.

3.6. Improving the efficiency of sampling designs
For very low density or rare species as in our protected areas in Lao PDR, the amount of
effort required to obtain adequate sample sizes with sufficient power to detect change over
time can be daunting. The choice of sampling scheme therefore also needs to be evaluated
in terms of its efficiency. Efficiency of sampling design refers to the precision of the resulting
population estimate for a given level of survey effort (Yoccoz et al., 2001). Precision, as we
have seen, can influence the power of the monitoring program to detect true changes in the
status of a population over time, or to detect a true response of a population to a specific
management intervention.
The efficiency of a sampling design depends largely on the characteristics of the target
population. If the target population can be divided into different spatial units that are
relatively homogenous in nature, for example large blocks of different forest types, then
stratification (Thompson, 1992) of sampling by forest type would result in a more efficient
sampling design and more precise population estimates by forest type. However, in such
cases adequate sample sizes need to be maintained for each stratum (rather than the
population as a whole), and for low density populations or rare species, this is frequently not
a feasible option. In these situations sampling designs can be tailored to maximize the
number of observations (or sample size), for example by standardizing the timing of surveys
at a particular time of day or during a particular season when individuals are more visible,
thus increasing detection probability, or by employing adaptive sampling techniques
(Thompson, 2004) where the intensity of sampling is dependent on initial sampling results.
3.7. Decision making: matching objectives with available resources
Even after sampling efficiency has been taken into consideration, the level of survey effort
required may still be prohibitively expensive for available budgets. In these situations,
managers need to re-consider their monitoring objectives and reflect on whether their
proposed monitoring design is the most cost-effective approach to take. This is often the
most difficult part of the planning process but an all-too-common situation faced by
managers seeking to manage low density or depleted populations over large areas. In these
situations the following guidelines can provide some assistance:
- Reflect upon whether the parameter to be measured needs to be density or true abundance
and if minimum sample sizes can be achieved in order to satisfy the assumptions of these

26
methods. If not, then consider if the objective can be still be met with alternative and
perhaps cheaper measures, such as occupancy or presence/absence.
- Consider if the monitoring objective itself can be realistically achieved with the available
budget, and trained staff to hand, over the desired timeframe. If not, then objectives may
need to be modified either in terms of geographic scale (e.g. reducing the total survey
area), or expected outcomes (e.g. adjusting the state variable or effect size you are hoping
to measure)
Try to avoid the temptation of cutting corners by ignoring some of the fundamental
principles of monitoring design that we have outlined in this module. Whilst certain data
collected using opportunistic or ad-hoc methods can often prove useful to managers, for
example opportunistic observations of tigers by field patrol teams can be extremely useful in
confirming presence of very rare species such as tigers in a particular area, these data
should be valued for what they are and not seen as a replacement for carefully designed
monitoring programs intended to inform managers about the status of wildlife populations
and/or how these are changing over time relative to management objectives.
4. PRACTICAL CONSIDERATIONS IN DESIGNING SUSTAINABLE
MONITORING PROGRAMS
It is important to ensure that the design of any wildlife monitoring program is sustainable –
or, in other words, to ensure that it can be implemented and replicated reliably over the
long-term. In order to achieve this, a monitoring program needs to ensure the availability of
sufficient and adequately trained staff, a feasible timeframe and workplan for
implementation, and an adequate budget that supports all associated costs.
4.1 Personnel and capacity building
This addresses the important question of who will be responsible for implementing the
monitoring program. For some protected areas and landscapes there are multiple actors or
institutions responsible for management and monitoring implementation, thus it is
important to identify which institution is responsible for which part of the monitoring
program: the design, the field implementation and the analysis and communication of results.

27
Different monitoring methods require different levels of training and skills. Advanced
method such as mark-recapture techniques or Distance-based sampling require a higher
level of skill and formal education than more simple presence/absence surveys for example.
Regardless of the methods used, all field staff should receive the necessary training from a
qualified trainer in the appropriate data collection protocols. Building up a strong
monitoring team of trained and experienced field staff is an important component of
ensuring the long-term sustainability of a monitoring program.
In addition to identifying the necessary field staff for implementation, it is also important to
identify the appropriate and qualified technical support staff to provide advice and oversight
on monitoring methods and analysis. Finally, it is important to identify the institutions and
personnel responsible for data management and the logistics personnel required for
coordinating the implementation of the data collection and ensuring that teams get out into
the field when they are supposed to.
4.2 Making a workplan and schedule
Once the appropriate personnel have been identified, it is important to determine the time-
frame of the monitoring program and to develop a detailed workplan for help in planning
the required resources. Remember also to allocate sufficient time for the design phase of the
monitoring program, as well as the field implementation and analysis phase.
It is important also to think about the time-frame of the monitoring program, including how
quickly you need the results, and whether this is an appropriate time-frame given the
biology and reproductive potential of the species. For example, for conservation targets such
as elephants, our monitoring time-frame would need to be much longer to see results of a
recovery program than for species such as muntjac, as elephants reproduce slower and
populations would take a longer time to increase. Related to this is the question of frequency
of surveys: would they be repeated annually, once every two years etc. These are also
important issues to address in the design phase, but they often have practical and cost-
related implications too.
For each survey year it’s important to make a detailed workplan and schedule for
implementation. Think carefully about the timing of your surveys and whether this
maximizes the potential for observations and increasing sample size. Consider also if the
timing is appropriate from a logistical perspective – for example it might not be possible to
access certain parts of the protected area in the wet season. Consider also any other

28
potential seasonal impacts on your conservation targets, such as large-scale migrations,
breeding cycles or seasonal variation in certain threats, such as hunting. Ensure these
impacts are accounted for and minimized by planning your data collection to occur within a
particular season, and across the same seasons over multiple years. Finally, consider
whether the timeframe of the proposed field implementation is sufficient to ensure that all
data can be collected appropriately and at all spatial sampling locations, given the number
of field staff you have available and the ease with which they can access and traverse
different parts of the landscape. Make sure you also include necessary rest days for field
teams as fieldwork can often be physically and mentally challenging.
4.3 Budget planning
Conservation and management budgets are often limited. It is therefore important to ensure
that adequate funds are available for the proposed monitoring program design. As we have
seen, some monitoring methods are more costly to implement than others, and the design of
the monitoring program may need to be rethought if sufficient funds are not available.
When planning your monitoring budget, the following costs should be considered:
Personnel – Sufficient staff and support need to be recruited for all stages of the monitoring
process (see above)
Training – All staff need to be adequately trained in appropriate data collection methods and
analytical procedures. Ensure the appropriate trainers are identified and that training
programs and/or workshops are planned accordingly
Implementation – This includes field costs, such as food, camping equipment, navigation
equipment (such as compasses, maps) medical supplies, and any specialized field equipment
needed for certain monitoring methods (for example GPS units or cameras for photographic
mark-recapture techniques, and the batteries needed to run this equipment). Ensure also that
the monitoring teams have sufficient materials for recording and analysing field data, for
example field notebooks for recording observations and computers with the required
software programs installed for managing and analysing data (see below under Data
Management and Documentation).
Logistics – Remember that the monitoring field teams need to move around the landscape!
Ensure that sufficient vehicles or boats are available, together with an estimate of fuel costs

29
and other logistical support that is needed to achieve the workplan and schedule you have
developed.
5. DATA MANAGEMENT AND DOCUMENTATION
All aspects of the monitoring program should be carefully documented and stored in a
clearly marked and accessible location (for example as electronic files on a central computer
or server within the protected area, rather than on a personal laptop). This applies to the
monitoring program goals and objectives, the monitoring design and associated assumptions,
the data collection protocols and methods and the analytical techniques used. Monitoring
programs can be adaptable and may change as new techniques evolve and more information
becomes available. To adapt and refine the monitoring methods it is important to have a
clear record of the development and assumptions that underly the original monitoring
design, to ensure institutional knowledge is retained as new staff are taken on into the
program.
A system of storing and managing field data is also required to ensure both integrity and
quality of data is maintained. If field data are recorded in notebooks or on hard-copy forms,
then a system should be made available that transcribes these data into an electronic format
that can be stored on a central computer. This will greatly facilitate and speed-up data
analysis as well as ensuring that data is not lost following general deterioration or wear and
tear of paper forms. The electronic format may take the form of a simple Excel-based
database with standardized column headings and pre-defined data entry codes, or,
depending on the needs and capacity of the site, it may be in the form of a more
sophisticated Access-type database or purpose-built management information system (ie
MIST for ranger-based law enforcement data). Regardless, the database should be regularly
backed up and the backup copy stored on a separate computer or location, to ensure that
data is protected against any computer breakdown or virus.
6. COMMUNICATION AND DISSEMINATION OF RESULTS
Data analysis and communication of the results are the final and important stages in the
management cycle. It is absolutely critical that all the hard work, time and effort put into
designing and implementing rigorous monitoring programs is not wasted by failing to get

30
the results to the key decision makers in a timely manner. Implicating all stakeholders at the
outset and ensuring that monitoring programs are integrated as a core component of
management planning and decision-making will greatly facilitate this process.
The presentation of monitoring results needs to assess the findings in the light of the
monitoring goals and objectives. Furthermore, accepted and peer-reviewed analytical
techniques should be employed wherever possible. It is recommended that the analysis of
monitoring data is reviewed by an independent and scientific technical advisor or group to
ensure its reliability and utility for management.
Management decision-makers and/or donors might not always be familiar with the
technical details of the monitoring methods used. Depending on who the results are being
presented to, it may be necessary to modify the format. For example, if presenting to an
external or non-technical audience it will be important to ensure that the key results are
presented as clearly as possible, using maps and charts wherever possible to facilitate
communication of key findings.
Finally, be prepared to assess and review the monitoring design in the light of the results
and to adapt and improve the design where appropriate. Monitoring programs are intended
to be dynamic in nature and should be able to respond to changes in threats or management
action.

31
Literature cited
Buckland, S. T., Anderson, D., Burnham, K., Laake, J., Borchers, D. & Thomas, L. (2001)
Introduction to Distance Sampling: Estimating abundance of biological populations, Oxford
University Press, Oxford.
Conroy, M. & Nichols, J. (1996) In Measuring and monitoring biological diversity: Standard methods for
mammals (eds D. Wilson, F. Cole, J. Nichols, R. Rudran & M. Foster), pp. 41-49.
Smithsonian Institution Press, Washington, DC, USA.
Danielsen, F., Jensen, A. E., Alviola, P. A., Balete, D. S., Mendoza, M., Tagtag, A., Custodio, C. &
Enghoff, M. (2005) Does Monitoring Matter? A Quantitative Assessment of Management
Decisions from Locally-based Monitoring of Protected Areas. Biodiversity and Conservation, 14,
2633-2652(2620).
Dixon, P. M., A. R. Olsen, and B. M. Kahn (1998) Measuring trends in ecological resources.
Ecological Applications, 8, 225-227.
Ferraro, P. J. & Pattanayak, S. K. (2006) Money for Nothing? A Call for Empirical Evaluation of
Biodiversity Conservation Investments. PLoS Biology, 4, 482-488.
Field, S. A., Tyre, A.J. & Possingham, H.P. (2005) Optimizing allocation of monitoring effort under
economic and observational constraints Journal of Wildlife Management, 69, 473-482.
Gerrodette, T. (1987) A power analysis for detecting trends. Ecology, 68, 1364-1372.
Johnson, A. and Johnston, J. (2007). Biodiversity Monitoring and Enforcement Project in the Nam
Theun 2 Watershed. Final Report V1.1. November 2007. Vientaine, Lao PDR: Wildlife
Conservation Society.
Johnson, A., O'Brien, T., Bezuijen, M. R., Robichaud, W. G. and Timmins, R. J. (2005).
Recommendations for Wildlife Monitoring Design and Implementation in the Nakai-Nam
Theun National Protected Area. 77. Vientiane: Wildlife Conservation Society.
Johnson, A., Vongkhamheng, C., Hedemark, M. and Saithongdam, T. (2006). Effects of human-
carnivore conflict on tiger (Panthera tigris) and prey populations in Lao PDR. Animal
Conservation 9: 421-430.
Joseph, L., Elkin, C., Martin, T. & Possingham, H. (2009) Modeling abundance using N-mixture
models: the importance of considering ecological mechanisms. Ecological Applications, 19,
631–642.
Kapos, V., Balmford, A., Aveling, R., Bubb, P., Carey, P., Entwistle, A., Hopkins, J., Mulliken, T.,
Safford, R., Stattersfield, A., Walpole, M. & Manica, A. (2008) Calibrating conservation:
new tools for measuring success. Conservation Letters, 1, 155–164.
Karanth, K. U. & Nichols, J. D. (eds.) (2002) Monitoring tigers and their prey: A manual for researchers,
managers and conservationists in tropical Asia, Centre for Wildlife Studies, Bangalore.
Lancia, R., Nichols, J. & Pollock, K. (1994) Estimation of number of animals in wildlife populations, The
Wildlife Society, Behesda, MD, USA.
Legg, C. & Nagy, L. (2006) Why most conservation monitoring is, but need not be, a waste of time.
Journal of Environmental Management, 78, 194-199.
Lenth, R. V. (2001) Some Practical Guidelines for Effective Sample-Size Determination. The
American Statistician, 55, 187-193.
Lyons, J. E., Runge, M. C., Laskowski, H. P. & Kendall, W. L. (2008) Monitoring in the Context of
Structured Decision-Making and Adaptive Management. The Journal of Wildlife Management
72, 1683-1692.
MacKenzie, D. I. (2009) Getting the biggest bang for our conservation buck. Trends Ecol. Evol., 24,
175-177.

32
Mackenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Royle, J. A. & Langtimm, C. A.
(2002) Estimating site occupancy rates when detection probabilities are less than one. Ecology,
83, 2248-2255.
MacKenzie, D. I., Nichols, J. D., Royle, J. A., Pollock, K. H., Bailey, L. L. & Hines, J. E. (2006)
Occupancy Estimation and Modeling: Inferring patterns and dynamics of species occurrence, Elsevier
Academic Press.
Mackenzie, D. I. & Royle, J. A. (2005) Designing occupancy studies: general advice and allocating
survey effort. Journal of Applied Ecology, 42, 1105-1114.
Nichols, J. D. & Williams, B. K. (2006) Monitoring for conservation. TRENDS in Ecology and
Evolution 21, 668-673.
Otis, D., Burnham, K., White, G. & Anderson, D. (1978) Statistical inference from capture data on
closed animal populations. Wildlife Monographs, 62, 1-135.
Pollock, K., Nichols, J., Brownie, C. & Hines, J. (1990) Statistical inference for capture-recapture
experiments. Wildlife Monographs, 107, 1-97.
Pollock, K. H., J. D. Nichols, T. R. Simons, G. L. & Farnsworth, L. L. B., and J. R. Sauer. (2002)
Large scale wildlife monitoring studies: statistical methods for design and analysis.
Environmetrics, 13, 105–119.
Poulsen, M. K. & Luanglath, K. (2005) Projects come, projects go: lessons from participatory
monitoring in southern Laos. Biodiversity and Conservation, 14, 2591–2610.
Pullin, A. S. & Knight, T. M. (2001) Effectiveness in Conservation Practice: Pointers from Medicine
and Public Health. Conservation Biology, 15, 50-54.
Redford, K. H., Coppolillo, P., Sanderson, E. W., Fonseca, G. A. B. D., Dinerstein, E., Groves, C.,
Mace, G., Maginnis, S., Mittermeier, R. A., Noss, R., Olson, D., Robinson, J. G., Vedder,
A. & Wright, M. (2003) Mapping the Conservation Landscape. Conservation Biology, 17,
116–131.
Royle, J. A. & Nichols, J. D. (2003) Estimating abundance from repeated presence-absence data or
point counts. Ecology, 84, 777–790.
Strindberg, S., Johnson, A., Hallam, C., Rasphone, A., Helm, F. V. D., Xiongyiadang, P. and
Sisavath, P. (2007). Recommendations for monitoring landscape species in the Nam Kading
National Protected Area. A report to the Integrated Ecosystem and Wildlife Management
Project. Vientiane: Wildlife Conservation Society (WCS) and the Integrated Ecosystem and
Wildlife Management Project (IEWMP).
Sutherland, W. J., A. S. Pullin, Dolman, P. M. & Knight, T. M. (2004) The need for evidence-based
conservation. Trends in Ecology & Evolution, 19, 305-308.
Thomas, L., and C.J. Krebs (1997) A review of power analysis software. Bulletin of the Ecological
Society of America, 78, 126-139.
Thompson, S. K. (1992) Sampling, John Wiley & Sons, New York.
Thompson, W. (2004) Sampling Rare or Elusive Species: Concepts, Designs, and Techniques for Estimating
Population Parameters, Island Press.
White, G., Nichols, J. & Boulinier, T. (1982) Capture-recapture and removal methods for sampling closed
populations, Los Alamos National Laboratory Publication, Los Alamos, NM, USA.
Williams, B. K., Nichols, J. D. & Conroy, M. J. (2002) Analysis and Management of Animal
Populations: Modeling, Estimation and Decision Making, Academic Press, San Diego, California,
USA.
Yoccoz, N. G., Nichols, J. D. & Boulinier, T. (2001) Monitoring of biological diversity in space and
time. TRENDS in Ecology & Evolution 16, 446-453.