This document discusses using expert judgement and uncertainty analysis for biodiversity management. It outlines several types of uncertainty including variability, incertitude, ambiguity, vagueness, and underspecificity. It provides examples of quantitative and linguistic uncertainty and challenges in expert judgement, including overconfidence, insensitivity to sample size, testing risk classifications, and appeals to authority. Effective uncertainty analysis is important for making robust decisions for biodiversity management under conditions of uncertainty.
1. EIANZ October 2008
Using
U i expert j d
t judgement and
t d
uncertainty analysis for
y y
biodiversity management
Mark Burgman
2. Outline
A taxonomy of uncertainty
Making decisions under uncertainty: practical
applications of uncertainty analysis
• ‘T ki ’ Sindh Ibex
‘Taking’ Si dh Ib
• Habitat models
• How many samples?
• IUCN conservation assessments
3. ‘Classical’ taxonomy of
uncertainty
Variability is naturally occurring, unpredictable
change, differences in parameters attributable to
‘true’ or ‘inherent’ variation. i.e. natural variation
Incertitude is lack of knowledge about parameters or
models. i.e, measurement error, systematic error,
model uncertainty, subjective judgement
4. Quantitative methods
Natural variation
• depends on unjustifiable assumptions
• can’t deal with non-statistical uncertainty
• deals poorly with Lack of knowledge
dependencies Risk
• creates an illusion of false precision
t t+1
Time
Probability arithmetic, ‘classical’ decision theory, Monte Carlo
5. Linguistic uncertainty
g y
• Ambiguity – words have two or more
meanings, and it is not clear which is
meant.
• V
Vagueness – b d li cases.
borderline
(Regan et al 2002)
• Underspecificity – unwanted generality
generality.
• Context dependence – a failure to specify
context.
6. Underspecificity
There is a 70% chance
of rain.
Possible interpretations
• rain during 70% of the day (time)
• rain over 70% of the area (area)
• on 70% of days such as this one, there will be at
least some rain at a particular point (the weather
station) (days such as this one)
7. Underspecificity
There’s a 70% chance of rain
Gigerenzer, H t i van d B k F
Gi Hertwig, den Broek, Fasolo, & K t ik
l Katsikopoulos, Ri k
l Risk
Analysis (in press)
12. How do you
y
decide that a
species is absent
from an area?
13. Overconfidence
Settlement depth Height to failure
of a clay layer of an earth embankment
1
0.8
08
udgments
0.6
Ju
0.4
0.2
0
1 2 3 4 5 6 7 1 2 3 4 5 6 7
Experts
(Hynes and Vanmarche 1977)
14. Insensitivity to sample size
Tversky and Kahneman (1974)
‘the law of small numbers’: most people (including
experienced scientists) expect small samples to
represent the population from which they were
pop lation hich the ere
drawn to a degree that can only be assumed with
much larger samples.
samples
15. Challenge p
g problem
• H many 0 1 h quadrats d you need to b
How 0.1 ha d do d be
90% certain of encountering 95% of the
woody plants in a 40 ha area?
d l i h ?
17. Testing Risk Classifications
g
Mountain pygmy possum Desert tortoise
(Burramys parvus)
(B ) (Gopherus agassizii)
(G h i ii)
IUCN: Endangered Heritage: Apparently Florida scrub jay
Secure (G4) (Aphelocoma
coerulescens coerulescen)
(Regan et al 2005) Millsap: Threatened
18. IUCN Critically endangered
IF Decline of >= 80% in 10 years or 3 generations
OR Range <100 km2 or occupied habitat <10 km2
AND
at least 2 of the following 3 conditions are met:
1) severely fragmented or in 1 subpopulation
2) continuing to decline
3) fluctuations > 1 order of magnitude
OR number of mature individuals < 250
b f t i di id l
AND
at least 1 of the following 2 conditions are met:
1) >=25% decline in 3 years / 1 generation
25%
2) continuing decline and 1 subpopulation or <=50 per subpopulation
OR < 50 mature individuals
OR >= 50% risk of extinction in 10 years / 3
generations.
19.
20. IUCN
1
2
3
4
5
ecies
6
7
Spe
8
9
10
11
12
13
14
LR VU EN CR
IUCN 2000
(Regan et al 2005)
21. IUCN
1
2
3
4
5
ecies
6
7
Spe
8
9
10
11
12
13
14
LR VU EN CR
IUCN 2000
(Regan et al 2005)
22. IUCN
Why?
1
2
•3 Unwillingness to make inferences
4
•5 Parameter estimates spanning thresholds
ecies
6
7
Spe
•89
Variation i
V i i in qualitative judgements (language)
li i j d (l )
10
•
11 Inconsistent logic and data entry mistakes
12
13 (relatively rare)
14
LR VU EN CR
IUCN 2000
(Regan et al 2005)
23. Expert judgement?
A specialist knows... ‘a lot of things, … details … not covered in the
general texts...’
‘I am not thinking of general gut feeling but knowledge and
t thi ki f l t f li b t k l d d
experiences … in the light of which [the expert] can judge’
24. Appeals to authority
Irrefutable, unspecified wisdom and
unassailable expert status generates a
culture of technical control
‘There is a naked assertion that the identity
of the expert warrants acceptance of the
proposal’ (Walton).
Appeal to authority is legitimate only if
it can be challenged.