Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

A Journal of Conservation Biogeography Diversity and Distributions, (Diversity Distrib.

) (2016) 22, 249–262

BIODIVERSITY Use of expert knowledge to elicit


RESEARCH
population trends for the koala
(Phascolarctos cinereus)
Christine Adams-Hosking1*, Marissa F. McBride2,3, Greg Baxter4,
Mark Burgman5, Deidre de Villiers6, Rodney Kavanagh7, Ivan Lawler8,
Daniel Lunney9,10, Alistair Melzer11, Peter Menkhorst12, Robyn Molsher13,
Ben D. Moore14, David Phalen15, Jonathan R. Rhodes16,18, Charles Todd12,
Desley Whisson17 and Clive A. McAlpine18

1
School of Geography, Planning and ABSTRACT
Environmental Management, Global Change
Aim The koala is a widely distributed Australian marsupial with regional popu-
Institute, The University of Queensland,
Brisbane, Qld 4072, Australia, 2School of lations that are in rapid decline, are stable or have increased in size. This study
Botany, The University of Melbourne, examined whether it is possible to use expert elicitation to estimate abundance
Melbourne, Vic. 3010, Australia, 3Department and trends of populations of this species. Diverse opinions exist about estimates
of Biosciences, University of Helsinki, Helsinki of abundance and, consequently, the status of populations.
00014, Finland, 4School of Geography Planning
Location Eastern and south-eastern Australia
and Environmental Management, The
University of Queensland, Landscape Ecology Methods Using a structured, four-step question format, a panel of 15 experts
and Conservation Group, Brisbane, Qld 4072, estimated population sizes of koalas and changes in those sizes for bioregions
Australia, 5School of Botany, Environmental within four states. They provided their lowest plausible estimate, highest plausi-
Science, The University of Melbourne, ble estimate, best estimate and their degree of confidence that the true values
Melbourne, Vic. 3010, Australia, 6Endeavour were contained within these upper and lower estimates. We derived estimates
Veterinary Ecology, 1695 Pumicestone Rd,
of the mean population size of koalas and associated uncertainties for each
Toorbul, Qld 4510, Australia, 7The Australian
Diversity and Distributions

bioregion and state.


National University, Research School of
Biology, Canberra 0200, Australia, Niche Results On the basis of estimates of mean population sizes for each bioregion
Environment and Heritage, PO Box 2443, and state, we estimated that the total number of koalas for Australia is 329,000
North Parramatta, NSW 2150 Australia, (range 144,000–605,000) with an estimated average decline of 24% over the
8
Wildlife Heritage and Marine Division, past three generations and the next three generations. Estimated percentage of
Department of the Environment, Marine and loss in Queensland, New South Wales, Victoria and South Australia was 53%,
Freshwater Species Conservation Section, 26%, 14% and 3%, respectively.
Canberra ACT 2700, Australia, 9Office of
Environment and Heritage NSW, Hurstville, Main conclusions It was not necessary to achieve high levels of certainty or
NSW 2220, Australia, 10School of Biological consensus among experts before making informed estimates. A quantitative,
Sciences, University of Sydney, Sydney, NSW scientific method for deriving estimates of koala populations and trends was
2006, Australia, 11Koala Research Centre of possible, in the absence of empirical data on abundances.
Central Queensland, School of Medical and
Applied Sciences, CQ University, Keywords
Rockhampton, Qld 4702, Australia, bioregions, Delphi process, expert elicitation, iconic, knowledge, koala popula-
12
Department of Environment, Land, Water & tions, Phascolarctos cinereus, threatened species, uncertainty.
Planning, Arthur Rylah Institute for
Environmental Research, Heidelberg, Vic. 3084,
Australia, 13Department of Environment,
Water and Natural Resources, PO Box 39,
Kingscote, SA 5223, Australia, 14Hawkesbury
Institute for the Environment, Western Sydney
University, Locked Bag 1797, Penrith, 2751
NSW, Australia, 15Faculty of Veterinary
Science, University of Sydney, Sydney, NSW
2006, Australia, 16ARC Centre of Excellence for

DOI: 10.1111/ddi.12400
ª 2016 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/ddi 249
C. Adams-Hosking et al.

Environmental Decisions, The University of


Queensland, Brisbane, Qld 4072, Australia,
17
School of Life and Environmental Sciences,
Faculty of Science Engineering & Built
Environment, Deakin University, Burwood,
Vic. 3125, Australia, 18Landscape Ecology and
Conservation Group, School of Geography,
Planning, and Environmental Management,
The University of Queensland, Brisbane, Qld
4072, Australia
*Correspondence: Christine Adams-Hosking,
School of Geography, Planning and
Environmental Management and Global
Change Institute, The University of
Queensland, Brisbane, Qld 4072, Australia.
E-mail: c.hosking@uq.edu.au

2010) and to develop management strategies for endangered


INTRODUCTION
whooping cranes (Grus americana) (Runge et al., 2011).
For many species, the available evidence in the form of field Use of expert elicitation to consolidate the current state of
surveys and modelling capabilities to predict population knowledge is valuable when the status of a species is uncer-
numbers leaves gaps in the capacity to make timely conserva- tain, yet there is general recognition that the species is sub-
tion decisions. In such contexts, managers have long relied ject to multiple, often synergistic, threats to its persistence in
on experts to translate data and available evidence into the wild. Consolidated knowledge can assist in the develop-
advice (Sutherland, 2006). Expert knowledge can be defined ment of adaptive management strategies for a species, setting
as substantive information on a particular topic that is not management goals (Lindenmayer & Burgman, 2005) and
widely known by others (Martin et al., 2012). Experts are decision-making, particularly when it complements other
distinguished from non-experts by having training or experi- information (Krueger et al., 2012). For high-profile species,
ence with respect to a topic of interest (Fazey et al., 2006). structured elicitation can be important in establishing the
Unsystematic methods for eliciting and applying expert credibility of expert knowledge, particularly where outputs
knowledge may limit the validity of this knowledge; however, inform management or policy decisions (Cooke & Goosens,
recent advances in structured, evidence-based methods for 2000; Burgman, 2004; Burgman et al., 2005) or where stake-
expert elicitation in ecology that incorporate expert knowl- holders are concerned about potential bias and manipulation
edge that is not necessarily based on empirical data (e.g. of expert judgments (e.g. U.S. EPA, 2011).
Perera et al., 2011) have provided information for decision- The koala is an arboreal and folivorous marsupial (Melzer
makers (Kuhnert et al., 2010; Martin et al., 2012). The use et al., 2000). It is widely distributed in eastern mainland
of expert knowledge is growing as a tool for synthesizing Australia with range boundaries closely linked to climatic
diverse sources of information into a reliable representation conditions (Adams-Hosking et al., 2011, 2012). At extents of
of the current state of scientific knowledge, including its 1000s of ha, its occurrence is associated with the distribution
uncertainties when elicited with the same level of rigour pro- of the genera Eucalyptus and Corymbia (Myrtaceae) (DeGab-
vided in the collection and use of empirical data (Kuhnert riel et al., 2010) which occur on relatively fertile soils (Moore
et al., 2010; Martin et al., 2012). Structured expert elicitation et al., 2004; Crowther et al., 2009). Since European settle-
has been used to address a broad range of ecological topics ment in Australia in 1788, the clearing and fragmentation of
(e.g. Grech et al., 2012; Marcot et al., 2012; McDaniels et al., forests and woodlands (McAlpine et al., 2006; Santika et al.,
2012), including potentially dealing with uncertainty in an 2014) and hunting for its fur have affected the species. In
adaptive management context (Keith et al., 2011). This has more recent times, indirect threats from urbanization (Dique
led to its increasing recognition as a valid method for cap- et al., 2003), disease (Wan et al., 2011; Jobbins et al., 2012;
turing and summarizing knowledge and characterizing Polkinghorne et al., 2013), dogs, motor vehicle collisions
uncertainty when empirical data on species distributions and (Rhodes et al., 2011, 2014) and fire have become major
abundances are limited or dispersed (Cooke, 2013). Further- threats to the koala’s survival (Dique et al., 2003; Lunney
more, expert elicitation is increasingly being accepted as a et al., 2007, Commonwealth of Australia, 2009, 2011). Now,
tool for decision support, for example, to elicit estimates of climate change (Gordon et al., 1988; Adams-Hosking et al.,
the status of Australian birds (McBride et al., 2012a), to rank 2011, 2012; Seabrook et al., 2011; Lunney et al., 2012, 2014)
threats to sea turtle populations world-wide (Donlan et al., is an additional threat. Knowledge of the status of koala

250 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd
Expert knowledge estimates koala numbers

populations across its geographical range is needed to plan S1 in Supporting Information). Some participants were not
for conservation and management actions (Commonwealth able to complete their first-round questionnaires. To be
of Australia, 2011). listed as vulnerable, the species must have a maximum
Current data on koala abundance are patchy, and there is no observed, estimated, inferred, projected or suspected reduc-
consistent monitoring method; hence, they have a high level of tion in population size of ≥30% over any 10 years or three
uncertainty when extrapolated to a bioregional scale. There are generations, whichever is longer (up to a maximum of
significant geographical gaps in abundances and trends that 100 years in the future) (sometimes known as the sliding
need addressing so that koala conservation management or moving window, e.g. RAMAS (2014)), and where the
actions can be confidently employed to secure koala popula- reduction or its causes may not have ceased, may not be
tions across their broad geographical range. There also are understood or may not be reversible.
diverse expert opinions on koala numbers in various regions. Experts were asked to estimate koala population
The aim of this study was to evaluate a structured, expert abundances and trends for each bioregion. Koalas occur in
elicitation procedure for a species for which empirical data are the states of Queensland, New South Wales [including the
limited, and where conflict may exist about estimates of abun- Australian Capital Territory (ACT)], Victoria and South Aus-
dances and population trends, and consequently its status. tralia. There are multiple bioregions within each state and
some overlap two states. Bioregions were considered the
most appropriate classification of the diverse environments
METHODS
over which the koala is distributed (Fig. 1). Bioregions are
We conducted an elicitation workshop over 4 days in July large, geographically distinct areas of land with common
2012 to assess koala population sizes and trends throughout characteristics, such as geology, landform patterns, climate,
the species’ range. The elicitation process took place in the ecological features and plant and animal communities
context of a risk analysis workshop, facilitated by an elicita- (Thackway & Cresswell, 1995; IBRA 2012) and thus present
tion specialist (MB) and assistant (MFM), neither of whom a consistent geographical unit for eliciting estimates.
are involved in koala research. They used a modified version The experts were asked to estimate population size for each
of the Delphi process for structured group decision-making bioregion, and declines or increases in koala numbers that
(Dalkey & Helmer, 1963; Burgman et al., 2011). The 15 capture the largest change in any three-generation period of
expert participants, who also co-authored this paper, are three past and three future koala generations (15–21 years).
active koala researchers in the Australian states and territory These estimates were based on a combination of past observa-
where koalas occur and have knowledge of, and experience tions or knowledgeable estimations and future knowledgeable
with, koala ecology, conservation practice and modelling. projections. The past time frame included 8 years of severe
This allowed for a holistic examination and synthesis of data drought. Along with their best estimate for each bioregion,
and analyses as well as the identification of knowledge gaps. experts also provided upper and lower bounds and an associ-
The ability of groups of diverse but knowledgeable individu- ated confidence level (Table 2). Experts were directed to spec-
als to improve on the accuracy of one or a few specialist ify confidence levels such that for a 90% confidence level, the
experts has been well documented (e.g. Arlinghaus & Krause, true value would be expected to fall between their stipulated
2013). The duration of the elicitation process was 6 months. lower and upper limits in 9 of 10 cases.
We used a structured, four-step process (Speirs-Bridge Although each expert had expertise for specific states and
et al., 2010; McBride et al., 2012b) to question the experts bioregions, they were asked to complete questions for all
(Table 1). This procedure helps improve the accuracy of states and bioregions. Responses were compiled, and linear
estimates (Burgman et al., 2011), promotes the pooling of extrapolation (Bedford & Cooke, 2001; Hora, 2007) was used
individual knowledge and reduces overconfidence (Speirs- to convert the experts’ elicited lower (l) and upper (u)
Bridge et al., 2010; McBride et al., 2012b). bounds for each parameter to standardized 80% credible
bounds. The adjusted lower (ladj) and upper (uadj) bounds
were calculated as:
Stage 1: Pre-workshop estimation
The first round of the elicitation process was conducted ladj ¼ b  ðb  lÞðcadj =cÞ
remotely. We sent the experts spreadsheets containing
questions regarding the status of koala populations. The uadj ¼ b þ ðu  bÞðcadj =cÞ
questions were drawn from the IUCN Red List criteria
(IUCN, 2012). While the round 1 questionnaire asked where b is the expert’s best estimate, c is their assigned
experts to assess koala subpopulations against the full set of confidence for their estimates, and cadj = 0.8 is the desired
IUCN Red List criteria, the decision was made prior to the confidence level (i.e. 80%). Standardizing the expert credible
workshop to proceed with further elicitation for only the intervals enabled easy viewing and comparison at the feed-
two key criteria questions deemed most relevant for the back stage of the elicitation. The choice to convert to 80%
koala (population size and percentage change in population credible intervals (as opposed to 90% or 100%, for example)
size) due to time constraints (see Queensland example Data was arbitrary, but allowed for more informative intervals to

Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd 251
C. Adams-Hosking et al.

Table 1 The structured elicitation procedure (after McBride et al., 2012b)

Elicitation

Pre-elicitation Pre-workshop Workshop Post elicitation

Background information Estimate Feedback Estimate Mean and median of all experts’
compiled second round responses calculated.
Experts contacted and All experts All experts shown All experts Experts review individual and
briefed on elicitation individually answer anonymous answers individually make group status outcomes via email,
process emailed questions and discuss in second and final provide feedback and make individual
face-to-face workshop anonymous estimates corrections if needed, and sign-off on
final results

Figure 1 The bioregions outlined in


black depict where the koala occurs in
the States of Queensland, New South
Wales (including the Australian Capital
Territory), Victoria and South Australia
(Source: IBRA, 2012).

be displayed to participants than the much wider bounds assessments. These revised assessments were used in the final
typically required to encapsulate 90% or 100% confidence estimation of bioregional koala population sizes and trends
would allow. (e.g. Fig. 2).

Stage 2: Within-workshop feedback and Stage 3: Post-elicitation


re-estimation
Following the workshop, the second round expert estimates
The second round of the elicitation process was a 4-day, were compiled and used to generate aggregate estimates of
face-to-face workshop, led by the facilitator (MB) with the population size and the percentage change for each biore-
assistance of MFM. Each expert’s first-round answers from gion. Group aggregate estimates were calculated by taking
the email questionnaire were displayed anonymously to the the average of expert lower, best and upper estimates, respec-
group by the facilitator. On the first day of the workshop, tively (e.g. Lichtendahl et al., 2013). These estimates were
the first round of elicited answers was discussed in detail then summed across bioregions to generate overall estimates
within the group, with experts sharing knowledge from their of population size and percentage change for each state.
respective regions of expertise. All experts were given the Bioregional population size estimates were combined
opportunity to participate in the discussion of each region directly, but estimates of population change were combined
and, in particular, to query experts with local knowledge. by first weighting individual bioregional estimates of
The experts agreed that this information was influential in population change by the number of koalas in that
guiding their revisions of their estimates in cases where they bioregion. Aggregation is not always appropriate for a group
did not possess knowledge. On the second day of the work- of judgments (Morgan & Henrion, 1990; Kriegler et al.,
shop, experts were asked to reconsider their previous assess- 2009). For example, opinions may differ on the basis of dif-
ments in the light of the group discussions. They were given ferent interpretations of data (Morgan, 2014). However, here
the opportunity to anonymously revise their first-round the patterns of elicited responses for each bioregion were suf-

252 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd
Expert knowledge estimates koala numbers

Table 2 The parameters elicited for the koala. Experts provided


RESULTS
assessments for population size and greatest population change
by bioregion. Bioregional population estimates (group means) ranged from
55 to almost 68,000 koalas. The state mean population esti-
What is your estimate of the current population size (number of
mature individuals)? mates of the group of experts (i.e. the group mean) for
1. Realistically, what is the lowest value it could be? Queensland, New South Wales [including the (ACT)], Vic-
2. Realistically, what is the highest value it could be? toria and South Australia were 79,624, 36,350, 182,505 and
3. What is your best estimate (the most likely value)? 33,319, respectively (Table 5). The total national aggregated
4. How confident are you that the interval you provided contains mean population change was a decline of 24% (Table 5).
the truth (provide an answer in the range of 50–100%). Proportional levels of uncertainty varied between 55% and
What do you think is the biggest change in population size in any 100%, with coefficient of variation values calculated for the
three generation period over the immediate past three generations experts’ best estimates only (Table 3). The calculated average
(15–21 years ago) and immediate future three generations (15-21 80% lower bounds and upper bounds (Table 3) correspond
years hence)? (Answer in terms of % change) to an average of the group’s assessment of the uncertainty of
the responses, rather than a measure of group agreement/
1. Realistically, what is the lowest value it could be?
disagreement. These values appear high when represented in
2. Realistically, what is the highest value it could be?
3. What is your best estimate (the most likely value)? percentage form, but they reflect a level of uncertainty that
4. How confident are you that the interval you provided contains is common for expert elicitation exercises. Standard devia-
the truth (Provide an answer in the range of 50–100%) tion values were for the experts’ best estimates only
(Table 4).
For Queensland’s bioregions, the mean population esti-
ficiently consistent that, in most cases, the calculation of a mates ranged between 1,943 and 15,821 (Fig. 3). Levels of
mean and median response for the group was deemed uncertainty were between 63% and 92% (Table 3). The pre-
appropriate. dicted koala population changes were from 20% to 73%
The experts’ final assessments and the aggregate bioregion (Table 4), and the population change for the whole of
and state values were circulated to the group for comment Queensland, aggregated from bioregional estimates, was a
and final review of their individual inputs. For population decline of 53% (Table 5).
size, comparative levels of uncertainty for each aggregate For the New South Wales bioregions (including the ACT),
state and bioregion estimate were assessed as: the mean population estimates ranged between 55 and
11,133 (Fig. 4). Levels of uncertainty were between 78% and
Uncertainty r ¼ ðUr  Lr Þ=Ur  100;
100% (Table 3). The predicted koala population changes
were from 4% to 50% (Table 4), and the population
for Ur and Lr the aggregate upper and lower bounds for
change for the whole of New South Wales and the ACT,
region or state of interest r, respectively. For percentage
aggregated from bioregional estimates, was a decline of 26%
change, as estimates were already on a common, inter-
(Table 5).
pretable scale without the need for standardization, the
In the Victorian bioregions, the mean population estimates
comparative uncertainty levels were just assessed as the
ranged between 117 and 67,917 (Fig. 5). Levels of uncertainty
difference between aggregate upper and lower bounds (i.e.
were between 63% and 95% (Table 3). The predicted koala
Uncertaintyr = [Ur  Lr]). To provide a measure of the
population changes were from 2% to 28% (Table 4), and
interexpert variation in population size estimates, the coeffi-
the population change for the whole of Victoria, aggregated
cient of variation (CoV) was calculated (von der Gracht,
from bioregional estimates, was a decline of 14% (Table 5).
2012) on the basis of the standard deviation of expert
In the South Australian bioregions, the mean population
lower, best and upper estimates and the means for each
estimates ranged between 76 and 19,800 (Fig. 6). Levels of
bioregion:
uncertainty were between 55% and 89% (Table 3). The pre-
r½br  dicted koala population changes were from 29% to +16%
CoV½br  ¼
l½br  (Table 4), and the population change for the whole of South
Australia, aggregated from bioregional estimates, was a
where br is the estimate type for a given region or state r decline of 3% (Table 5).
(best estimate, upper or lower bound), r is the standard Coefficients of variation were higher for estimates of
deviation, and l is the mean of these estimates across the upper and lower bounds than for best estimates, and
expert group. For percentage change, we used the standard higher for the first- round estimates than for second-
deviation as a measure of inter-expert variation. This choice round estimates, indicating a shift towards agreement
was made because the percentage change estimates were among experts (von der Gracht, 2012). Bioregions for
already on a common scale, and because the calculation of which the level of variation among experts was relatively
the coefficient of variation is inappropriate for non-ratio low included the Lofty Block bioregions (Kangaroo Island
scale data. and Mainland) and the majority of the Victorian biore-

Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd 253
C. Adams-Hosking et al.

Figure 2 Comparison of experts’ first and second round estimates for the Mulga Lands (NSW) and Lofty Block (Mainland) bioregions,
with round 1 responses shown on the left, for each bioregion pair. Uncertainty bars have been adjusted to reflect 80% credible intervals
for each expert’s response. The mean (solid black line) and median (dashed black line) were calculated using the experts’ estimates.

gions. Bioregions with high levels of variation between


DISCUSSION
expert responses included the Murray Darling Depression
and Mulga Lands in New South Wales. Disagreement in
The expert elicitation process
uncertainty intervals among experts, measured in terms of
the proportion of non-overlap of expert intervals (for each The main achievement of this expert elicitation process was
variable calculated as the proportion of all possible pair- a koala population estimate, at 2012, for each bioregion
wise comparisons between experts for which intervals did through the application of expert elicitation that did not
not overlap), decreased from an average of 35% for initial require formal consensus.
population estimates to 4% for final estimates, indicating Lack of agreement among scientists is often seen as
that a reasonable agreement in the expert responses was problematic by policymakers and the public. However,
achieved when taking into account their uncertainty. A experts are often hesitant to offer quantitative estimates
similar result was achieved for estimates of percentage due to the frequently high uncertainties involved. Past
change in population size, with a shift from 44% for ini- work on koalas, including the 2005 IUCN expert work-
tial responses down to 16% non-overlap in the final round shops on the status of Australian mammals, did not seek
estimates. estimates of population size, information about uncertainty

254 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd
Expert knowledge estimates koala numbers

Table 3 Group mean population estimates, average of 80% lower and upper bounds, percentage uncertainty and coefficient of variation
of the best estimates of each bioregion’s koala population size (mature individuals only). The estimates of mean population size were
used to calculate an aggregate mean density estimate for each state.

Mean Average of Average of Uncertainty Coefficient


Queensland bioregions population lower bounds upper bounds (%)* of variation†

Brigalow Belt North 15,179 6541 28,810 77 0.3


Central Mackay Coast 8857 3196 18,777 83 0.4
Desert Uplands 6357 1138 14,490 92 0.6
Einasleigh Uplands & Wet Tropics 4750 942 12,426 77 0.2
Mitchell Grass Downs 1943 364 4366 92 0.5
Mulga Lands 15,286 8234 22,476 63 0.5
South Brigalow 11,071 3719 27,608 87 0.3
Southeast Queensland 15,821 9194 224,871 63 0.2
Queensland 79,264

Mean Average of Average of Uncertainty Coefficient


New South Wales bioregions population lower bounds upper bounds (%) of variation

Murray-Darling Depression 55 0 298 100 2.5


South East Corner 655 182 1454 88 0.8
Cobar Peneplain & Riverina 2354 864 5326 84 0.5
Darling-Riverine Plains 964 238 2699 91 0.9
Mulga Lands 711 94 2033 95 0.9
New England Tableland 2771 468 5831 92 0.6
NSW North Coast 8367 4048 14,618 72 0.4
NSW Southwestern Slopes 2310 870 4895 82 0.6
South Brigalow & Nandewar 11,133 4262 21,604 80 0.4
South-Eastern Highlands 1363 363 2950 88 0.5
Sydney Basin 5667 2436 11,307 78 0.4
New South Wales 36,350

Mean Average of Average of Uncertainty Coefficient


Victorian bioregions population lower bounds upper bounds (%) of variation

Murray-Darling Depression 117 16 319 95 0.6


Naracoorte Coastal Plain 4233 1812 7941 77 0.2
Furneaux 330 89 620 86 0.4
Riverina 2833 864 5265 84 0.3
South East Coastal Plain (East) 11,357 6264 18,464 66 0.3
South East Coastal Plain (West) 6000 2872 10,511 73 0.3
South East Corner 2768 463 6897 93 0.4
Otways 67,917 29,762 103,746 71 0.3
Strezleckis 4667 1869 10,739 83 0.2
South-Eastern Highlands/ Northern Ranges 46,667 19,988 85,710 77 0.3
Victorian Midlands 24,231 6747 57,286 88 0.3
Victorian Volcanic Plain 11,385 7017 19,017 63 0.2
Victoria 182,505

Mean Average of Average of Uncertainty Coefficient


South Australian bioregions population lower bounds upper bounds (%) of variation

Murray Darling Depression 475 218 905 76 0.6


Eyre and Yorke Blocks (West) 76 33 168 81 0.7
Lofty Block (Kangaroo Island) 12,600 8158 18,119 55 0.2
Lofty Block (Mainland) 19,800 10,512 30,969 66 0.2
Naracoorte Coastal Plain 368 116 1059 89 0.8
South Australia 33,319

*Uncertainty %: (calculated as [(Upper bound – Lower bound)/Upper bound*100].


†Coefficient of variation values reported here are for the experts’ best estimates only.

Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd 255
C. Adams-Hosking et al.

Table 4 Estimates of percentage change in population size in each bioregion either in the past or future, 80% lower and upper bounds,
percentage uncertainty and inter-expert standard deviation.

Percentage Average of Average of Uncertainty Inter-expert


Queensland bioregions change (%) lower bounds upper bounds (%)* standard deviation†

Brigalow Belt North 63 87 29 58 9


Central Mackay Coast 35 64 5 59 26
Desert Uplands 20 46 1 45 19
Einasleigh Uplands & Wet Tropics 41 75 4 79 29
Mitchell Grass Downs 39 74 2 76 43
Mulga Lands 73 88 41 47 50
South Brigalow 56 78 22 56 47
Southeast Queensland 51 74 20 54 41

Percentage Average of Average of Uncertainty Interexpert


New South Wales bioregions change (%) lower bounds upper bounds (%) standard deviation

MurrayDarling Depression 12 29 4 33 24


South East Corner 46 71 16 55 40
Cobar Peneplain & Riverina 9 34 15 49 16
DarlingRiverine Plains 34 69 57 126 21
Mulga Lands 31 69 1 70 22
New England Tableland 6 19 28 47 10
NSW North Coast 50 72 32 40 18
NSW Southwestern Slopes 23 59 14 73 21
South Brigalow & Nandewar 35 64 7 71 30
SouthEastern Highlands 19 40 9 49 20
Sydney Basin 4 33 15 48 15

Percentage Average of Average of Uncertainty Interexpert


Victorian bioregions change (%) lower bounds upper bounds (%) Standard deviation

MurrayDarling Depression 22 52 1 18 17


Naracoorte Coastal Plain 9 40 19 18 16
Furneaux 9 37 6 10 8
Riverina 9 36 22 20 17
South East Coastal Plain (East) 2 20 17 25 25
South East Coastal Plain (West) 28 58 8 18 22
South East Corner 5 27 15 13 14
Otways 6 51 34 25 26
Strezleckis 20 57 4 0 13
SouthEastern Highlands/ Northern Ranges 21 61 4 10 14
Victorian Midlands 20 53 6 16 13
Victorian Volcanic Plain 15 47 5 19 17

Percentage Average of Average of Uncertainty Interexpert


South Australian bioregions change (%) lower bounds upper bounds (%) standard deviation

Murray Darling Depression 26 56 23 79 51


Eyre and Yorke Blocks (West) 10 32 7 39 37
Lofty Block (Kangaroo Island) 29 51 13 64 24
Lofty Block (Mainland) 16 15 53 68 50
Naracoorte Coastal Plain 5 33 21 54 27

*Uncertainty %: calculated as [(Upper bound-Lower bound)]. Note that this differs from the method used for calculating the uncertainty for
population size estimates, as reported in Table 3.
†Standard deviation values reported here are for the experts’ best estimates only.

in the form of a range of possible values or information the koala as “Least Concern” in view of its wide distribu-
at the regional level from the expert group, two of whom tion, presumed large population (Gordon et al., 2008). It
are co-authors on this current paper. That workshop listed became apparent that much more detail was required for

256 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd
Expert knowledge estimates koala numbers

Table 5 State and national koala population sizes and trends for 2012, aggregated from bioregional estimates. The group mean koala
population sizes and uncertainty range are rounded to the nearest thousand. The mean population change and estimated total national
population for 2012 are based on the Red List criterion that captures the largest change in any three generation period, that is in the
past three koala generations (from 15 to 21 years ago) to the future three koala generations (15–21 years into the future).

Queensland New South Wales Victoria South Australia National total

Aggregated mean population 79,000 36,000 183,000 33,000 331,000


Aggregated range ~33,000–153,000 ~14,000–73,000 ~77,000–327,000 ~19,000–51,000 ~144,000–605,000
Aggregated mean population change (%) 53 26 14 3 24

Mean populaon Populaon change (%) Uncertainty for mean pop. size
(past and future) (%)

Figure 3 Estimated mean koala population size, population change and uncertainty for each bioregion in Queensland.

this widespread species to enable appropriate and effective


Contribution to koala conservation and management
conservation priority setting (Commonwealth of Australia,
2011) and that is now available as a result of this expert Expert knowledge plays a role in conservation decision-mak-
elicitation process. ing and is regularly utilized to inform listing decisions under
In this study, we observed broad agreement among all the IUCN Red List criteria (McBride et al., 2012a). This
experts, which allowed for the calculation of aggregate group study is the first to apply this type of structured expert elici-
estimates. However, in situations where experts have made tation procedure to quantifying the status of widely dis-
very different judgments about the relevant underlying tributed koala populations. It can inform the mapping of the
science, or if the variable being assessed will be used as an koala’s conservation status at a regional, state and national
input to a nonlinear model, aggregation may not be suitable scale, supply estimates of uncertainty and contribute to
and could lead to a misleading picture of the true nature of setting priorities for data acquisition. The koala population
expert opinion (Keith, 1996; O’Hagan et al., 2006; Morgan, elicitation process incorporated published accounts of local
2014). In these cases, running separate analyses using each of koala population sizes and trends (e.g. Kavanagh & Barrott,
the key schools of thought could be used to gain insight into 2001; Seabrook et al., 2011; Kavanagh & Barrott-Brown,
the impact that different opinions have on the resulting deci- 2014; Lunney et al., 2014; Santika et al., 2014; McAlpine
sion or model outputs (Morgan & Henrion, 1990). In Crome et al., 2015).
et al. (1996), for example, rather than generate an aggregate Our results also highlight the differing levels of uncertainty
prior, the authors compared the effects of using priors eli- associated with populations within each of the bioregions. In
cited from proponents in the logging industry, and priors South Australia, for example, the management goal for the
elicited from a conservationist in their model predicting the Kangaroo Island koala population (55% uncertainty) is a
impacts of logging on birds. Given that the two groups held decrease in abundance to reduce browsing levels, and the
such vastly different opinions about the impact of logging, abundance is regularly monitored. Even so, estimates for this
investigating the sensitivity of the model outputs to their dif- bioregion had upper and lower bounds of 30,000 and 5000,
ferent opinions was considerably more informative than respectively. Although these bounds appear to represent high
attempting to combine opinions from individuals with levels of uncertainty when expressed as percentages, they are
conflicting worldviews. typical of such elicitation exercises (e.g. O’Neill et al., 2008;

Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd 257
C. Adams-Hosking et al.

Mean populaƟon PopulaƟon change (%) Uncertainty for mean pop. size
(past and future) (%)

Figure 4 Estimated mean koala population size, population change and uncertainty for each bioregion in New South Wales.

Mean populaƟon PopulaƟon change (%) Uncertainty for mean pop. size
(past and future) (%)

Figure 5 Estimated mean koala population size, population change and uncertainty for each bioregion in Victoria.

Mean populaƟon PopulaƟon change (%) Uncertainty for mean pop. size
(past and future) (%)

Figure 6 Estimated mean koala population size, population change and uncertainty for each bioregion in South Australia.

258 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd
Expert knowledge estimates koala numbers

Czembor & Vesk, 2009; Morgan, 2011; Cooke, 2013). There Adams-Hosking, C., McAlpine, C., Rhodes, J.R., Grantham,
was a reasonably high level of best -estimate agreement H.S. & Moss, P.T. (2012) Modelling changes in the distri-
among experts, with 7,000–15,000 for Kangaroo Island and bution of the critical food resources of a specialist folivore
7,000–20,000 for south-east Queensland. in response to climate change. Diversity and Distributions,
Structured elicitation workshops are used for assessing the 18, 847–860.
current state of expert knowledge and providing a forum for Arlinghaus, R. & Krause, J. (2013) Wisdom of the crowd and
discussion and a baseline for further research and assessment natural resource management. Trends in Ecology and
(e.g. Bamber & Aspinall, 2013; Durant et al., 2014). In addi- Evolution, 28, 8–11.
tion to highlighting areas of agreement, our process also Australian Government (2012) Koala populations in Queens-
highlighted knowledge gaps. The knowledge from such exer- land, New South Wales and the Australian Capital Territory
cises always needs to be treated as a snapshot of current sci- and national environment law. Department of Sustainability,
entific opinion and used with the intent of periodic revision Environment, Water, Population and Communities. Available
as new information comes to light. This means that expert at: https://www.environment.gov.au/system/files/resources/1b
knowledge, including the uncertainties, and how and when 863ab8-39ae-4f21-bd27-7e18952daad6/files/bio220-0213-koa
they were reported, must be properly documented. la-listing-factsheet-general.pdf (accessed May 2014).
The workshop presented here suggested regional variations Bamber, J.L. & Aspinall, W. (2013) An expert judgement
in koala population numbers and trends, including the assessment of future sea level rise from the ice sheets.
disparity between the northern and southern states and among Nature Climate Change, 3, 424–427.
bioregions within states. The mostly abundant southern koala Bedford, T. & Cooke, R.M. (2001) Mathematical tools for
populations are not necessarily secure because of excessively probabilistic risk analysis. Cambridge University Press,
high levels of herbivory (Duka & Masters, 2005), low genetic Cambridge.
diversity (Cristescu et al., 2009) and, in more recent times, cli- Burgman, M. (2004) Expert frailties in conservation risk
mate change, including extreme weather events and more fre- assessment and listing decisions. Threatened species legisla-
quent fires (CSIRO, 2014). Estimates of the rate of loss, what tion: is it just an Act? (ed. by P. Hutchings, D. Lunney and
the rate is likely to be in the future and where changes are tak- C. Dickman), pp. 20–29. Royal Zoological Society of NSW,
ing place may inform the drafting and implementation of a Mosman, NSW, Australia.
National Recovery Plan for this species now that the National Burgman, M.A., Lindenmayer, D.B. & Elith, J. (2005)
Koala Conservation and Management Strategy 2009–2014 Managing landscapes for conservation under uncertainty.
(Commonwealth of Australia, 2009) has expired. Ecology, 86, 2007–2017.
By providing population abundances and trends, the Burgman, M.A., McBride, M., Ashton, R., Speirs-Bridge, A.,
expert elicitation process we used has informed the debate Flander, L., Wintle, B., Fidler, F., Rumpff, L. & Twardy, C.
over how to manage koalas. This is demonstrated by post- (2011) Expert status and performance. PLoS ONE, 6, e22998.
workshop consultations with decision-makers in the state Commonwealth of Australia (2009) National Koala Conserva-
and Commonwealth governments that informed the Com- tion and Management Strategy 2009–2014. Department of
monwealth listing on the koalas as vulnerable in Queensland, the Environment, Water, Heritage and the Arts, Canberra.
New South Wales and the ACT (Australian Government, Commonwealth of Australia (2011) The koala – saving our
2012). While the coastal populations have previously been national icon. The Senate Environment and Communica-
the public and policy focus, because that is where most peo- tions References Committee, Canberra.
ple live, the status of koalas at the more arid, inland edge of Cooke, R. (2013) Expert judgement assessment: quantifying
their range has become clearer. uncertainty on thin ice. Nature Climate Change, 3, 311–312.
Cooke, R.M. & Goosens, L.H.J. (2000) Procedures guide for
structured expert judgement in accident consequence mod-
ACKNOWLEDGEMENTS
elling. Radiation Protection Dosimetry, 90, 303–309.
We are grateful to Alison Specht and the Australian Centre Cristescu, R., Cahill, V., Sherwin, W.B., Handasyde, K., Car-
for Ecological Analysis and Synthesis (ACEAS) a Facility lyon, K., Whisson, D., Herbert, C.A., Carlsson, B.L.J., Wil-
within the Terrestrial Ecosystem Research Network (TERN) ton, A.N. & Cooper, D.W. (2009) Inbreeding and testicular
for providing funding and logistic support for this abnormalities in a bottlenecked population of koalas
workshop. (Phascolarctos cinereus). Wildlife Research, 36, 299–308.
Crome, F.H.J., Thomas, M.R. & Moore, L.A. (1996) A novel
Bayesian approach to assessing impacts of rain forest log-
REFERENCES ging. Ecological Applications, 6, 1104–1123.
Adams-Hosking, C., Grantham, H.S., Rhodes, J.R., McAl- Crowther, M.S., McAlpine, C.A., Lunney, D., Shannon, I. &
pine, C. & Moss, P.T. (2011) Modelling climate-change- Bryant, J.V. (2009) Using broad-scale, community survey
induced shifts in the distribution of the koala. Wildlife data to compare species conservation strategies across
Research, 38, 122–130. regions: a case study of the Koala in a set of adjacent

Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd 259
C. Adams-Hosking et al.

‘catchments’. Ecological Management and Restoration, 10 assurance. Technological Forecasting and Social Change, 79,
(Suppl s1), S88–S96. 1525–1536.
CSIRO (2014) State of the Climate 2014. Australian Govern- Grech, A., Chartrand-Miller, K., Erftemeijer, P., Fonseca, M.,
ment Bureau of Meteorology, Canberra. McKenzie, L., Rasheed, M., Taylor, H. & Coles, R. (2012)
Czembor, C.A. & Vesk, P.A. (2009) Incorporating between- A comparison of threats, vulnerabilities and management
expert uncertainty into state-and-transition simulation approaches in global seagrass bioregions. Environmental
models for forest restoration. Forest Ecology and Manage- Research Letters, 7, 024006.
ment, 259, 165–175. Hora, S. (2007) Eliciting probabilities from experts. Advances
Dalkey, N. & Helmer, O. (1963) An experimental application in decision analysis: from foundations to applications (ed. by
of the Delphi method to the use of experts. Management E. Ward, R.F. Miles Jr and D. von Winterfeldt), pp. 129–
Science, 9, 458–467. 153. Cambridge University Press, Cambridge.
DeGabriel, J.L., Moore, B.D., Marsh, K.J. & Foley, W.J. IBRA (2012) Australia’s Bioregions (IBRA). Australian
(2010) The effect of plant secondary metabolites on the Government, Department of the Environment.
interplay between the internal and external environments IUCN (2012) IUCN red list categories and criteria: version
of marsupial folivores. Chemoecology, 20, 97–108. 3.1, 2nd edn. Gland, Switzerland and Cambridge, UK:
Dique, D.S., Thompson, J., Preece, H.J., Penfold, G.C., IUCN. iv + 32 pp.
De Villiers, D.L. & Leslie, R.S. (2003) Koala mortality on Jobbins, S.E., Sanderson, C.E., Griffith, J.E., Krockenberger,
roads in south-east Queensland: the koala speed-zone trial. M.B., Belov, K. & Higgins, D.P. (2012) Diversity of mhc
Wildlife Research, 30, 419–426. class ii dab 1 in the koala (Phascolarctos cinereus).
Donlan, C.J., Wingfield, D.K., Crowder, L.B. & Wilcox, C. Australian Journal of Zoology, 60, 1–9.
(2010) Using expert opinion surveys to rank threats to Kavanagh, R., & Barrott, E., (2001) Koala populations in the
endangered species: a case study with sea turtles. Conserva- Pilliga forests. Pp. 93–103 Perfumed Pineries: Environmental
tion Biology, 24, 1586–1595. history of Australia’s Callitris forests, ed. by J. Dargavel, D.
Duka, T. & Masters, P. (2005) Confronting a tough issue: Hart and B. Libbis. CRES, Australian National University,
fertility control and translocation for over-abundant Koalas Canberra.
on Kangaroo Island, South Australia. Ecological Manage- Kavanagh, R. & Barrott-Brown, G. (2014) Pilliga Koala Sur-
ment and Restoration, 6, 172–181. veys 2013. Report to NSW Office of Environment and Her-
Durant, S.M., Wacher, T., Bashir, S., Woodroffe, R., De itage. Niche Environment and Heritage Pty Ltd,
Ornellas, P., Ransom, C., Newby, J., Abaigar, T., Abdelga- Parramatta.
dir, M., El Alqamy, H., Baillie, J., Beddiaf, M., Belbachir, Keith, D.W. (1996) When is it appropriate to combine
F., Belbachir-Bazi, A., Berbash, A.A.., Bemadjim, N.E., Beu- expert judgments? Climatic Change, 33, 139–143.
dels-Jamar, R., Boitani, L., Breitenmoser, C., Cano, M., Keith, D.A., Martin, T.G., McDonald-Madden, E. & Wal-
Chardonnet, P., Collen, B., Cornforth, W.A.., Cuzin, F., ters, C. (2011) Uncertainty and adaptive management for
Gerngross, P., Haddane, B., Hadjeloum, M., Jacobson, A., biodiversity conservation. Biological Conservation, 144,
Jebali, A., Lamarque, F., Mallon, D., Minkowski, K., Mon- 1175–1178.
fort, S., Ndoassal, B., Niagate, B., Purchase, G., Sama€ıla, S., Kriegler, E., Hall, J.W., Held, H., Dawson, R. & Schellnhu-
Samna, A.K., Sillero-Zubiri, C., Soultan, A.E., Stanley Price, ber, H.J. (2009) Imprecise probability assessment of tipping
M.R. & Pettorelli, N. (2014) Fiddling in biodiversity hot- points in the climate system. Proceedings of the National
spots while deserts burn? Collapse of the Sahara’s mega- Academy of Sciences USA, 106, 5041–5046.
fauna. Diversity and Distributions, 20, 114–122. Krueger, T., Page, T., Hubacek, K., Smith, L. & Hiscock, K.
Fazey, I., Fazey, J.A., Salisbury, J.G., Lindenmayer, D.B. & (2012) The role of expert opinion in environmental mod-
Dovers, S. (2006) The nature and role of experiential elling. Environmental Modelling and Software, 36, 4–18.
knowledge for environmental conservation. Environmental Kuhnert, P.M., Martin, T.G. & Griffiths, S.P. (2010) A guide
Conservation, 33, 1–10. to eliciting and using expert knowledge in Bayesian ecolog-
Gordon, G., Brown, A.S. & Pulsford, T. (1988) A koala ical models. Ecology Letters, 13, 900–914.
(Phascolarctos cinereus Goldfuss) population crash during Lichtendahl, K.C. Jr, Grushka-Cockayne, Yael & Winkler,
drought and heatwave conditions in south-western R.L. (2013) Is it better to average probabilities or quantiles?
Queensland. Austral Ecology, 13, 451–461. Management Science, 59, 1594–1611.
Gordon, G., Menkhorst, P., Robinson, T., Lunney, D., Mar- Lindenmayer, D. & Burgman, M. (2005) Practical conserva-
tin, R. & Ellis, M. (2008) Phascolarctos cinereus. In: IUCN tion biology. CSIRO Publishing, Vic., Australia.
2013. IUCN Red List of Threatened Species. Version Lunney, D., Gresser, S., O’Neill, L.E., Matthews, A. &
2013.1. Available at: http://www.iucnredlist.org (accessed Rhodes, J. (2007) The impact of fire and dogs on koalas
June 2014) at Port Stephens, New South Wales, using population
von der Gracht, H.A. (2012) Consensus measurement in Del- viability analysis. Pacific Conservation Biology, 13,
phi studies: review and implications for future quality 189–201.

260 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd
Expert knowledge estimates koala numbers

Lunney, D., Crowther, M.S., Wallis, I., Foley, W.J., Lemon, Morgan, M.G. (2011) Certainty, uncertainty, and climate
J., Wheeler, R., Madani, G., Orscheg, C., Griffith, J.E., change. Climatic Change, 108, 707–721.
Krockenberger, M., Retamales, M. & Stalenberg, E. (2012) Morgan, M.G. (2014) Use (and abuse) of expert elicitation
Koalas and climate change: a case study on the Liverpool in support of decision making for public policy. Proceed-
Plains, north-west NSW. Wildlife and climate change: ings of the National Academy of Sciences USA, 111, 7176–
towards robust conservation strategies for Australian fauna 7184.
(ed. by D. Lunney and P. Hutchings), pp. 150–168. Royal Morgan, M.G. & Henrion, M. (1990) Uncertainty: a guide to
Zoological Society of NSW, Mosman NSW, Australia. dealing with uncertainty in quantitative risk and policy anal-
Lunney, D., Stalenberg, E., Santika, T. & Rhodes, J. (2014) ysis. Cambridge University Press, New York, NY.
Extinction in Eden: identifying the role of climate change O’Hagan, A., Buck, C.E., Daneshkhah, A.R., Eiser, J.R.,
in the decline of the koala in south-eastern NSW. Wildlife Garthwaite, P.H., Jenkinson, D.J., Oakley, J.E. & Rakow, T.
Research, 41, 22–34. (2006) Uncertain judgments: eliciting expert probabilities.
Marcot, B.G., Allen, C.S., Morey, S., Shively, D. & White, R. John Wiley and Sons Ltd., Chichester.
(2012) An expert panel approach to assessing potential O’Neill, S.J., Osborn, T.J., Hulme, M., Lorenzoni, I. &
effects of bull trout reintroduction on federally listed sal- Watkinson, A.R. (2008) Using expert knowledge to assess
monids in the Clackamas River, Oregon. North American uncertainties in future polar bear populations under cli-
Journal of Fisheries Management, 32, 450–465. mate change. Journal of Applied Ecology, 45, 1649–1659.
Martin, T.G., Nally, S., Burbidge, A.A., Arnall, S., Garnett, Perera, A.H., Drew, C.A. & Johnson, C.J. (eds) (2011) Expert
S.T., Hayward, M.W., Lumsden, L.F., Menkhorst, P., knowledge and its application in landscape ecology. Springer-
McDonald-Madden, E. & Possingham, H.P. (2012) Acting Verlag, New York, NY.
fast helps avoid extinction. Conservation Letters, 5, 274–280. RAMAS (2014) Red List version 3.0. Available at: http://
McAlpine, C.A., Rhodes, J.R., Callaghan, J.G., Bowen, M.E., www.ramas.com/redlist.htm (accessed June 2014).
Lunney, D., Mitchell, D.L., Pullar, D.V. & Possingham, Rhodes, J.R., Fei Ng, C., de Villiers, D.L., Preece, H.J., McAl-
H.P. (2006) The importance of forest area and configura- pine, C.A. & Possingham, H.P. (2011) Using integrated
tion relative to local habitat factors for conserving forest population modelling to quantify the implications of mul-
mammals: a case study of koalas in Queensland, Australia. tiple threatening processes for a rapidly declining popula-
Biological Conservation, 132, 153–165. tion. Biological Conservation, 144, 1081–1088.
McAlpine, C.A., Lunney, D., Melzer, A., Menkhorst, P., Ste- Rhodes, J., Lunney, D., Callaghan, J. & McAlpine, C. (2014)
phen Phillips, S., Phalen, D., Ellis, W., Foley, W., Baxter, A few large roads or many small ones? How to accommo-
G., de Villiers, D., Kavanagh, R., Adams-Hosking, C., date growth in vehicle numbers to minimise impacts on
Todd, C., Whisson, D., Molsher, R., Walter, M., Lawler, I. wildlife. PLoS ONE, 9, e91093.
& Close, R. (2015) Conserving koalas: a review of the con- Runge, M.C., Converse, S.J. & Lyons, J.E. (2011) Which
trasting regional trends, outlooks and policy challenges. uncertainty? Using expert elicitation and expected value of
Biological Conservation, 192, 226–236. information to design an adaptive program. Biological
McBride, M.F., Garnett, S.T., Szabo, J.K., Burbidge, A.H., Conservation, 144, 1214–1223.
Butchart, S.H.M., Christidis, L., Dutson, G., Ford, H.A., Santika, T., McAlpine, C.A., Lunney, D., Wilson, K.A. &
Loyn, R.H., Watson, D.M. & Burgman, M.A. (2012a) Rhodes, J.R. (2014) Modelling species distributional shifts
Structured elicitation of expert judgments for threatened across broad spatial extents by linking dynamic occupancy
species assessment: a case study on a continental scale models with public-based surveys. Diversity and Distribu-
using email. Methods in Ecology and Evolution, 3, 906–920. tions, 20, 786–796.
McBride, M.F., Fidler, F. & Burgman, M.A. (2012b) Evaluat- Seabrook, L., McAlpine, C., Baxter, G., Rhodes, J., Bradley,
ing the accuracy and calibration of expert predictions A. & Lunney, D. (2011) Drought-driven change in wildlife
under uncertainty: predicting the outcomes of ecological distribution and numbers: a case study of koalas in south
research. Diversity and Distributions, 18, 782–794. west Queensland. Wildlife Research, 38, 509–524.
McDaniels, T., Mills, T., Gregory, R. & Ohlson, D. (2012) Speirs-Bridge, A., Fidler, F., McBride, M., Flander, L., Cum-
Using expert judgments to explore robust alternatives for ming, G. & Burgman, M. (2010) Reducing overconfidence
forest management under climate change. Risk Analysis, 32, in the interval judgments of experts. Risk Analysis, 30,
2098–2112. 512–523.
Melzer, A., Carrick, F., Menkhorst, P., Lunney, D. & St. Sutherland, W.J. (2006) Predicting the ecological conse-
John, B. (2000) Overview, critical assessment, and conser- quences of environmental change: a review of the methods.
vation implications of koala distribution and abundance. Journal of Applied Ecology, 43, 599–616.
Conservation Biology, 14, 619–628. Thackway, R. & Cresswell, I.D. (1995) An Interim Biogeo-
Moore, B.D., Wallis, I.R., Wood, J.T. & Foley, W.J. (2004) graphic Regionalisation for Australia: a framework for setting
Foliar nutrition, site quality, and temperature influence priorities in the National Reserves System Cooperative Pro-
foliar chemistry of tallowwood (Eucalyptus microcorys). gram. Reserve System Unit, Australian Nature Conserva-
Ecological Monographs, 74, 553–568. tion Agency, Canberra.

Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd 261
C. Adams-Hosking et al.

U.S. EPA (2011) Expert elicitation task force white paper. BIOSKETCH
Science and Technology Policy Council, U.S. Environmen-
tal Protection Agency, Washington, DC. 20460 Christine Adams-Hosking Christine’s main research inter-
Wan, C., Loader, J., Hanger, J., Beagley, K., Timms, P. & ests have involved using a range of systematic conservation
Polkinghorne, A. (2011) Using quantitative polymerase planning tools to model the predicted distribution of terres-
chain reaction to correlate Chlamydia pecorum infectious trial species under future climate change and identify ‘cli-
load with ocular, urinary and reproductive tract disease in mate change refugia’, using koalas and their key Eucalyptus
the koala (Phascolarctos cinereus). Australian Veterinary food trees as case study species. Currently, she is investigat-
Journal, 89, 409–412. ing the potential impacts of climate change on future agri-
culture to assist Natural Resource Management bodies in
SUPPORTING INFORMATION developing climate change adaptation plans.

Additional Supporting Information may be found in the Author contributions: All co-authors participated equally in
online version of this article: the workshop and contributed to the writing which was led
by Christine Adams-Hosking.
Data S1. Queensland example of the questionnaire provided
to the experts.
Editor: David Richardson

262 Diversity and Distributions, 22, 249–262, ª 2016 John Wiley & Sons Ltd

You might also like