Risk Assessment and Decision Making

1
Risk Assessment and Decision
Making Randall M. Peterman School of
Resource and Environmental Management
(REM) Simon Fraser University Burnaby, B.C.,
CANADA V5A 1S6 Annual Applied Biology
Conference Victoria, B.C. 3 May 2007
2
Many problems require decisions to be made
Fraser Valley smog
Mt. pine beetle
Loss of biodiversity
Landslides
3
Large variability
Imperfect information
Complex structure
Features shared by these decision problems
Lengthy delay in response
Confounding of potential causal mechanisms
Thus, best decisions are not obvious
4
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
5
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
6
Environmental risk assessment What is risk? -
A measure of adverse effects of human actions
or natural processes
7
Risk has two components 1. Magnitudes of
consequences from uncertain events 2. Chance
(probability) of those consequences occurring
8
Risk has two components 1. Magnitudes of
consequences from uncertain events
Uncertain events
Consequences
Summer weather conditions
Average offspring per pair of Great Blue Herons)
poor
0.5
moderate
2
good
4
9
Risk has two components 1. Magnitudes of
consequences from uncertain events 2. Chance
(probability) of those consequences occurring
Uncertain events
Consequences
Probability of occurrence
Summer weather conditions
Average offspring per pair of Great Blue Herons)
poor
0.5
0.2
moderate
0.5
2
good
0.3
4
10
Uncertain events
Consequences
Probability of occurrence
Summer weather conditions
Average offspring per pair of Great Blue Herons)
poor
0.5
0.2
moderate
2
0.5
good
4
0.3
Expected offspring per parent
(0.20.5)(0.52)(0.34)
2.3 offspring per pair of adults
where "expected" means the statistical
expectation, or
weighted average forecast
11
Uncertain events
Consequences
Probability of occurrence
Summer weather conditions
Average offspring per pair of Great Blue Herons)
Another index of risk
poor
0.5
0.2
One index of risk
moderate
2
0.5
good
4
0.3
Expected offspring per parent
(0.20.5)(0.52)(0.34)
2.3 offspring per pair of adults
12
What is environmental risk assessment?
Systematic, flexible process for -
Estimating magnitude and probability
of adverse effects - Communicating about risks
to the public and decision makers
Environmental risk assessment methods -
Evolved in 1990s still evolving
13
Why do we need risk assessment?
Uncertainties
14
Natural variability
Complexity
Imperfect forecasts of natural system's dynamics
Uncertainties
Unclear management objectives
Observation error
Implementation uncertainty (human responses)
15
Uncertainties create risks
Biological risks (e.g., plants and animals)
Social risks (e.g., resource- dependent communiti
es)
Economic risks (e.g., businesses)
16
How does environmental risk assessment work?
Model system processes - Biological -
Physical - Interactions among system's
components Consider uncertainties
17
How does environmental risk assessment work? ...
Calculate indicators of risks relevant to
management objectives, e.g. - Lowest water
quality over next 2 years - "Expected"
(weighted-average forecast) abundance in 5
years
18
How does environmental risk assessment work? ...
Early and continuous interaction among -
Risk assessors - Risk managers/decision
makers (to state goals and objectives) -
Users of resources / environment Iterative
process
19
Example guidelines U.S. EPA (1998)
"Guidelines for Ecological Risk Assessment"
at http//www.epa.gov/ncea/ecorsk.htm B.C.
MOE's Ecological Risk Assessment Guidelines
Power and McCarty (2002)
20
How does risk assessment fit into risk management?
Risk assessment (Risk analysis) System
processes Uncertainties Indicators of risks
21
How does risk assessment fit into risk management?
Risk management (Make decision) Consider
other factors Make tradeoffs
Risk assessment (Risk analysis) System
processes Uncertainties Indicators of risks
22
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
23
Quantitative decision analysis as a means
to conduct risk assessment Example of
"structured decision making" (Ian Hatter's
talk at 2006 AGM) Explicitly accounts for
uncertainties Ranks actions
24
Eight Components of Decision Analysis 1.
Management objective 2. List of alternative
actions 3. Uncertain states of nature or
uncertainties yet to be resolved 4. Probability
of each state of nature occurring 5. Model to
determine consequences for each combination
of action and state of nature
25
6. Decision tree or decision table Results 7.
Rank each management action 8. Analysis of
sensitivity of rank order of management options
to various - Assumptions - Parameter
values - Model structures - Management
objectives
Example -- Decision tree regarding a waste dump
26
1. Management objective minimize expected costs
Management Actions
2. Options/ actions
Clean up dump
Do not clean up
27
1. Management objective minimize expected costs
Human health effect?
Management Actions
Yes
3. Uncertain states of nature
Clean up dump
No
Yes
Do not clean up
No
(Parkhurst 1984)
28
1. Management objective minimize expected costs
Probabilities of states of nature
Human health effect?
Management Actions
Yes
P10.3
Clean up dump
No
(1-P1) 0.7
4. Probability of each state of nature with
sum 1.0
Yes
8.2
P10.3
Do not clean up
No
(1-P1) 0.7
(Parkhurst 1984)
29
1. Management objective minimize expected costs
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Management Actions
Yes
5. Model - Mental (experts' views)
- Quantitative -- Static -- Dynamic
P10.3
1.6
Clean up dump
No
(1-P1) 0.7
1.5
5. Model to estimate outcomes
Yes
8.2
P10.3
Do not clean up
No
(1-P1) 0.7
0.2
(Parkhurst 1984)
30
1. Management objective minimize expected costs
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Management Actions
Yes
6. Decision tree
P10.3
1.6
Clean up dump
No
(1-P1) 0.7
1.5
5. Model to estimate outcomes
Yes
8.2
P10.3
Do not clean up
No
(1-P1) 0.7
0.2
(Parkhurst 1984)
31
7. To rank actions, compare indicators with the
objective using appropriate criteria

• Examples
• Maximize expected value of
• Minimize expected value of

32
1. Management objective minimize expected costs
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Management Actions
"Expected" weighted average
Yes
P10.3
1.6
Clean up dump
No
(1-P1) 0.7
1.5
Expected costs
Yes
8.2
P10.3
Do not clean up
2.6
(0.3 x 8.2) (0.7 x 0.2)
No
(1-P1) 0.7
0.2
(Parkhurst 1984)
33
1. Management objective minimize expected costs
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Management Actions
Yes
P10.3
1.6
Clean up dump
1.53
No
(1-P1) 0.7
1.5
Expected costs
Yes
8.2
P10.3
Do not clean up
2.6
No
(1-P1) 0.7
0.2
(Parkhurst 1984)
34
1. Management objective minimize expected costs
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Choose for "minimizing expected cost"
Management Actions
Yes
P10.3
1.6
Clean up dump
1.53
No
(1-P1) 0.7
1.5
Expected costs
Yes
8.2
P10.3
Do not clean up
2.6
No
(1-P1) 0.7
0.2
(Parkhurst 1984)
35
7. To rank actions, compare indicators with the
objective using appropriate criteria

• Examples
• Maximize expected value of
• Minimize expected value of
• Choose the action with the most favorable
  worst-case outcome ("mini-max" or "minimum
  regret")

Dont necessarily need full decision analysis
36
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Management Actions
Yes
P10.3
1.6
Worst-case outcome for "Clean up"
Clean up dump
(1-P1) 0.7
Choose for "minimum regret"
1.5
No
Yes
8.2
P10.3
Do not clean up
Worst-case outcome for "Do not clean up"
(1-P1) 0.7
0.2
No
37
Probabilities of states of nature
Human health effect?
Costs of outcomes (millions of )
Management Actions
Yes
8. Sensitivity analysis on
P10.3
1.6
Clean up dump
No
(1-P1) 0.7
1.5
Yes
8.2
P10.3
Do not clean up
No
(1-P1) 0.7
0.2
38
Sensitivity analysis
Cost if you "don't clean" up, but there actually
is a human health effect
Probability that there is a human health effect
Baseline case
39
How does decision analysis relate to risk
assessment and risk management?
Risk assessment (Risk analysis) System
processes Uncertainties Indicators of risks
40
How does decision analysis relate to risk
assessment and risk management?
Risk assessment (Risk analysis) System
processes Uncertainties Indicators of risks
Decision analysis Management objectives
Rank management options
41
How does decision analysis relate to risk
assessment and risk management?
Risk assessment (Risk analysis) System
processes Uncertainties Indicators of risks
Decision analysis Management objectives
Rank management options
Risk management (Make decision) Consider
other factors Make tradeoffs
42
Useful books on decision analysis Clemen,
R.T. and T. Reilly. 2001. Making Hard Decisions
with DecisionTools. Duxbury/Thomson Learning
Press. Hammond. J.S., R.L. Keeney, H. Raiffa.
1999. Smart Choices A Practical Guide to Making
Better Decisions. Harvard Business School Press.

43
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
44
Example of decision analysis and risk assessment
Sockeye salmon from Cultus Lake, B.C.
Thousands of spawners
45
Problem many stocks mixed in fishing areas
Other stocks
Cultus
46
Management objectives 1. Recovery of Cultus
Lake sockeye abundance 2. Economic value for
commercial harvesters - Conflicting
objectives ecological vs. economic
47
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8
48
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8 General features of ideal
objectives
49
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8 General features of ideal
objectives- Indicator variable
50
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8 General features of ideal
objectives- Indicator variable- Target
condition
51
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8 General features of ideal
objectives- Indicator variable- Target
condition- Time frame ...
52
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8 General features of ideal
objectives- Indicator variable- Target
condition- Time frame ... - Recognition of
uncertainty
(Pestes 2006)
53
1. Recovery objective
Pr(Spawners gt 20,000 within 20
years) 0.8 General features of ideal
objectives- Indicator variable- Target
condition- Time frame ... - Recognition of
uncertainty
2. Maximize economic value of commercial harvest
as long as first objective has been met
(Pestes 2006)
54
Management action (state-dependent and
time-independent harvest rule)
Target proportional harvest rate
1
0
0
Abundance of adult Cultus Lake sockeye
(Pestes 2006)
55
Actions
Harvest rules
1
2
193
(Pestes 2006)
56
Shape parameters of this relation
?1 a1, d1, K1 ?2 a2, d2, K2 ?3360 a3360,
d3360, K3360
3000
?
Juvenile (smolt) abundance (1000s)
2000
?
?
1000
0
0
10
20
30
40
50
60
Spawner abundance (1000s)
(Pestes 2006)
57
Uncertainties (simplified)
Actions
Shapeof spawner-to-smoltrelation
Harvest rules
?1
...
?2
1
?3
2
?4
?5
193
...
?n
(Pestes 2006)
58
Uncertainties (simplified)
Actions
Bayesian posterior proba- bilities
Shapeof spawner-to-smoltrelation
Harvest rules
?1
...
?2
Pr1
Pr2
1
?3
Pr3
2
Pr4
?4
Pr5
?5
193
Prn
...
?n
(Pestes 2006)
59
Uncertainties (simplified)
Actions
Bayesian posterior proba- bilities
Marine surv. rates N(0,? 2)
Shapeof spawner-to-smoltrelation
Harvest rules
?1
...
...
...
poor
?2
PrOC,1
Pr1
Pr2
1
?3
moderate
PrOC,20
Pr3
2
PrOC,m
good
Pr4
?4
...
...
Pr5
?5
...
...
193
Prn
...
?n
...
...
(Pestes 2006)
60
Uncertainties (simplified)
Actions
Actual harvest rate compared to the
target beta(?,?)
Bayesian posterior proba- bilities
Shapeof spawner-to-smoltrelation
Marine surv. rates N(0,? 2)
Harvest rules
?1
...
...
...
...
poor
low
?2
PrOC,1
PrIE,1
Pr1
Pr2
1
?3
moderate
on target
PrOC,20
PrIE,20
Pr3
2
PrOC,m
PrIE,q
good
high
Pr4
?4
...
...
...
Pr5
?5
...
...
...
193
Prn
...
?n
...
...
...
(Pestes 2006)
61
Outcomes
Uncertainties (simplified)
Actions
Actual harvest rate compared to the
target beta(?,?)
Shapeof spawner-to-smoltrelation
Bayesian posterior proba- bilities
Marine surv. rates N(0,? 2)
Indicators of recovery and economic value
Harvest rules
?1
...
...
...
...
Pr(Spawn. gt 20,000 by year 20)
poor
low
?2
PrOC,1
PrIE,1
Pr1
Pr2
1
?3
moderate
on target
PrOC,20
PrIE,20
Pr3
2
Annual average economic value of catch
PrOC,m
PrIE,q
good
high
Pr4
?4
...
...
...
Pr5
?5
...
...
...
193
Prn
...
?n
...
...
...
(Pestes 2006)
62
Harvest rules that met the recovery objective
Proportion harvested

Maximizes gross revenue from fishing
Abundance of adult Cultus Lake sockeye salmon
(1000s)
(Pestes 2006)
63
Tradeoff
50
Expected mean annual gross revenue ( millions)
48
46
44
0.5
0.6
0.7
0.8
0.9
1
Desired probability of recovery for Cultus
sockeye stock
(Pestes 2006)
64
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
65
Example 2 A decision analysis to help resolve
conflicts (Maguire Boiney 1994) -
Northern white rhinoceros in Zaire -
Endangered - Dispute about next steps --
Wildlife officials in Zaire -- Western zoo
officials -- Hypothetical example based on
real situation
- Iterative revision of decision tree -
Negotiations based on decision analysis
66
P2 (anti-poaching is effective wild
animals survive)
Maguire Boiney 1994
67
Zaire wildlife officials
Anti-poaching best
P2
Captivity best
Western zoo officials
Maguire Boiney 1994
P2 (anti-poaching is effective wild animals
survive)

68
Maguire Boiney 1994

69
Maguire Boiney 1994
70
Example 3 A decision analysis to help resolve
conflicts (Peters and Marmorek 2001) - Snake
River chinook salmon
71
Seven populations of Snake River chinook salmon
- Listed under U.S. Endangered Species Act
Pacific Ocean
72
Potential causes of decline in salmon
abundance and inhibitors of recovery -
Hydroelectric dams and operations - Hatcheries
(spreading diseases) - Harvest - Habitat use
(cattle ranching) - Ocean conditions
73
How to achieve recovery?
Conflicting views among 8 groups
(Hydroelectric agency, states and federal
government, native tribes, conservation
groups) Led to different models and
interpretations of data - Deadlocked But
... decision analysis can consider many
hypotheses - Impasse broken!
74
Main purpose of this decision analysis
- Identify best management action to obtain
recovery, given many hypotheses and
uncertainties Other purposes - Single focus
for collaboration - "Neutral" ...
75
Management objective
For 6 of 7 populations, for i1-7
Pr(Spawnersi gt Goali over 48 years)0.5
76
Actions
Manage-ment option
A1 (status quo)
A2 (max. barging)
A3 (remove dams)
... other options
77
Uncertainties (simplified)
Actions
Survival inside of hydropower system
Manage-ment option
poor
8 hypotheses
A1 (status quo)
good
poor
A2 (max. barging)
8 hypotheses
good
poor
A3 (remove dams)
8 hypotheses
good
(Peters and Marmorek 2001)
78
Uncertainties (simplified)
Actions
Survival inside of hydropower system
Proba- bility or degree of belief
Manage-ment option
poor
0.7
A1 (status quo)
good
0.3
poor
0.7
A2 (max. barging)
good
0.3
poor
A3 (remove dams)
0.7
good
0.3
(Peters and Marmorek 2001)
79
Uncertainties (simplified)
Actions
Survival outside of hydropower system
Survival inside of hydropower system
Proba- bility or degree of belief
Manage-ment option
poor
...
...
0.7
poor
A1 (status quo)
good
0.3
0.3
9 hypotheses
poor
moderate
0.7
0.6
A2 (max. barging)
0.1
good
good
...
...
0.3
poor
...
...
A3 (remove dams)
0.7
...
good
...
0.3
(Peters and Marmorek 2001)
80
Uncertainties (simplified)
Actions
Survival outside of hydropower system
Survival inside of hydropower system
Proba- bility or degree of belief
Timing and effects of removing dams
Manage-ment option
poor
...
...
0.7
poor
A1 (status quo)
good
0.3
0.3
poor
moderate
0.7
0.6
A2 (max. barging)
0.1
good
good
...
...
0.3
poor
...
...
...
A3 (remove dams)
0.7
6 hypotheses
...
...
good
...
0.3
(Peters and Marmorek 2001)
81
Outcomes
Uncertainties (simplified)
Actions
Survival outside of hydropower system
Survival inside of hydropower system
Proba- bility or degree of belief
Timing and effects of removing dams
Number of the 7 stocks recovered in 48 years

Manage-ment option
poor
...
...
...
0.7
poor
A1 (status quo)
1
good
0.3
0.3
poor
moderate
3
0.7
0.6
A2 (max. barging)
0.1
good
4
good
...
...
...
0.3
poor
...
...
...
A3 (remove dams)
0.7
6 hypotheses
...
...
good
...
0.3
(Peters and Marmorek 2001)
82
Baseline results of decision analysis
Probability of successful recovery
Best action
A1 (status quo)
A3 (remove dams)
A2 (max. barging)
(Peters and Marmorek 2001)
83
Sensitivity analyses Option A3 (removing dams)
was still best option over wide range of
assumptions and management objectives.
(Peters and Marmorek 2001)
84
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
85
Special challenges to communicating about
uncertainties and risk 1.Public's perceptions
about risks Affected by - Control over
risk - Catastrophic potential - Immediacy of
effect - Trust of "experts" What can we
do? - Improve education about issues
(Slovic 1987)
86
2. Confusion about the term "probability" Six
interpretations of "probability" (Teigen
1994) 1. Chance Outcome of a random process 2
... 6 (e.g., confidence, control, ...) What
can we do? - Use frequency format
87
"Chance" of an outcome for a given set of
management regulations Probability
format "There is a probability of 0.2 (or a 20
chance) that a local mountain caribou population
will drop below a critical abundance."
88
"Chance" of an outcome for a given set of
management regulations Probability
format "There is a probability of 0.2 (or a 20
chance) that a local mountain caribou population
will drop below a critical abundance." Frequency
format "Two out of every 10 situations like this
will lead to a local mountain caribou population
dropping below a critical abundance."
89
"Chance" of an outcome for a given set of
management regulations Probability
format "There is a probability of 0.2 (or a 20
chance) that a local mountain caribou population
will drop below a critical abundance." Frequency
format "Two out of every 10 situations like this
will lead to a local mountain caribou population
dropping below a critical abundance." Anderson
(1998a,b), Gigerenzer and Hoffrage (1995)
90
Outline
1. Environmental risk assessment - What is it
and why do we need it? 2. Quantitative decision
analysis - What is it and how does it help? 3.
Examples - Quantify tradeoffs - Resolve
conflicts 4. Communication about uncertainties
and risks 5. Benefits and limitations of
risk assessment and decision analysis
91
Benefits of risk assessment and decision
analysis 1. Provide systematic structure for
analyses of complex problems 2. Allow
comparison of alternative management actions,
taking uncertainties into account 3. Quantify
tradeoffs 4. Improve confidence in, and quality
of, decisions
92
Benefits ... 5. Provide documentation of inputs
to decisions 6. Can be a focal point for -
Stakeholder involvement - Communication -
Conflict resolution 7. Can help identify
priorities for future research through
sensitivity analyses
93
Limitations of risk assessment and decision
analysis 1. Cannot guarantee the desired
outcome 2. Poor data 3. Bounded scope of
analyses 4. Method may create overconfidence 5.
Limited technical expertise
94
Limitations ... 5. Time constraints 6.
Intangibles and other hard-to-quantify
variables are not considered explicitly 7.
Difficult to communicate technical information
95
References Anderson, J.L. 1998a. Conservation
Ecology 2(1)2-30, (www.consecol.org./vol2/iss1
/art2). Anderson, J.L. 1998b. B.C. Min. Forests
Research Program, Extension Note 221-6. Clemen,
R.T. and T. Reilly. 2001. Making Hard Decisions
with DecisionTools. Duxbury/Thomson Learning
Press. Gigerenzer, G. and U. Hoffrage. 1995.
Psychol. Rev. 102(4)684-704. Hammond. J.S., R.L.
Keeney, H. Raiffa. 1999. Smart Choices A
Practical Guide to Making Better Decisions.
Harvard U. Press. Maguire, L.A. and L.G. Boiney.
1994. Journal of Environmental Management
4231-38. Pestes, L. 2006. Master's thesis,
School of Resource and Environmental Management,
Simon Fraser University. Peters, C.N. and D.R.
Marmorek. 2001. Canadian J. of Fisheries and
Aquatic Sciences 58(12)2431-2446. Power, M.
and McCarty, L.S. 2002. Human and Ecological Risk
Assessment 87-18. Slovic, P. 1987. Science
236280-285.