Professor Hugh Possingham is currently the Director of the Ecology Centre at The University of Queensland. Hugh has over 290 publications, 5300 Web of Science citations and a lab of 32 students and staff. Work from his lab helped stop land clearing ("the Brigalow Declaration") in Queensland and NSW securing at least 1 billion tonnes of CO2.
"We generally assume that all monitoring is good. However there are numerous examples of people monitoring things to extinction and monitoring with no clear objective. Hugh Possingham will present a completely different way of looking at environmental monitoring - using decision science thinking. This approach enables us to work out how much of our precious budget should be spent monitoring, if any! The problem with existing monitoring, aside from doing too little, is that ecologists have been trained within a classical null hypothesis testing framework - great for pure science, rubbish for solving environmental problems."
Russian Escort Service in Delhi 11k Hotel Foreigner Russian Call Girls in Delhi
Hugh Possingham- Why Monitor the Environment
1. The Environment Institute
Where ideas grow
Hugh Possingham
„Why Monitor the Environment? - A Decision Science
Approach‟
2. How much and why should we monitor?
Monitoring is an optimisation problem
first and a statistical problem second
Hugh Possingham, lab and friends
The Ecology Centre and Centre for Applied
Environmental Decision Analysis – a CERF
Read www.aeda.edu.au/news
The University of Queensland
Australia
the ecology centre
university of queensland
australia
www.uq.edu.au/spatialecology
h.possingham@uq.edu.au
3.
4. Who pays for all the work?
• Australian Research Council grants
(19), UQ, UofA, Australian Federal
Government Environment Department
(CERF), TNC, PEW, CI, state govts
(several), local governments, mining
companies, TWS, WWF, BA, CRCs, +
innumerable minor grants
5. Some “straw men” of applied
monitoring/data collection
• We need to monitor all conservation
interventions with sufficient power to detect
significant effects
• I have just monitored frog species Y to extinction
• We need to learn about how the system works =
science
• Count first, ask questions later
• Getting more data on biodiversity is always a
good investment
Balmford A. & Gaston K.J. (1999). Why biodiversity surveys are good value.
Nature, 398, 204-205
6. Heretical views
• Most monitoring programs have no clearly
stated objectives and hence can‟t be optimised
(Joseph et al. 2010, Optimal monitoring for conservation)
• Surveillance monitoring is a waste of time
• (Nichols, J. D., and B. K. Williams. 2006. Monitoring for conservation. Trends in
Ecology & Evolution 21:668-673.)
• All monitoring for conservation should be based
in a decision-making framework
Possingham, H. P., Andelman, S. J., Noon, B. R., Trombulak, S. and Pulliam, H. R.
2001. Making smart conservation decisions. In: Research priorities for
conservation biology. Eds. Orians, G. and Soule, M. Island Press
7. Monitoring costs money that could be used for
solving the problem = managing
Optimal allocation
to monitoring 100%
Expected Efficiency
outcome of
from manage-
managing Net gain ment
if efficient
0% 0%
100% 0%
Percentage of budget spent on management
0% 100%
Percentage of budget spent on monitoring
9. Monitoring marine reserves
Control
How many times do we have to reject the Impact
null
hypothesis that fishing does not kill fish? Or dead
fish grow?
Before After Big fish?
More fish?
What marine reserve monitoring could we do
that would influence future decisions?
10. “Classical” approach to optimal
monitoring – alpha = 5%
Blue
monitoring
strategy
Statistical
power Predetermined
level of power
Purple
we want
monitoring
strategy
Fixed budget
Investment in monitoring strategy
11. “Classical” approach to optimal
monitoring – alpha =Why?
5%
Blue
monitoring
strategy
Statistical
power Predetermined
level of power
Purple
we want
monitoring
strategy
Why?
Fixed budget
Investment in monitoring strategy
12. How much monitoring should we do for
management/policy? The answer requires an
objective. 7 reasons to monitor (Joseph et al.)
1. Audit the to see if actions taken or legislative
requirements met or make donors happy
2. State-dependent management – (e.g. setting
fisheries quotas, acting to save a threatened species)
3. To learn for learning‟s sake
4. Active adaptive management – optimal management
accounting for the benefits of learning
5. Inform the public and/or politicians of an issue so
policy and allocations may change
6. Serendipity, so many breakthroughs have come from
just looking
7. People like it and do it for free
13. How much monitoring should we do for
management/policy? The answer requires an
objective. 7 reasons to monitor (Joseph et al.)
1. Audit the to see if actions taken or legislative
Boring
requirements met or make donors happy
2. State-dependent management – (e.g. setting
fisheries quotas, acting to save a threatened species)
3. To learn for learning‟s sake
Irrelevant
4. Active adaptive management – optimal management
accounting for the benefits of learning
5. Inform the public and/or politicians of an issue so
How much is enough?
policy and allocations may change
6. Serendipity, so many breakthroughs have come from
?
just looking
7. People like it and do it for free
Great
14. 2 State Dependent Management –
how much monitoring?
Counting moose or kangaroos (Hauser et al. 2006, Mansson et al.)
Survey roughly
Survey well
Happiness
Quota
Count
Reality
Number of moose or kangaroos
Hauser CE, Pople AR, Possingham HP. 2006. Should managed populations be
monitored every year? Ecological Applications 16:807-819.
15. 4 Active adaptive management
The holy grail of applied ecology – where we
try to gain knowledge only in so far that
the benefit of that knowledge gain is
expected to outweigh the costs of fiddling
with the system and learning about how it
works
19. Enter Reverend Thomas Bayes
and the
incredible
beta distribution
.
Thomas Bayes (pronounced: beɪz), (c. 1702 –
17 April 1761) was a British mathematician and
Presbyterian minister, known for having
formulated a specific case of the theorem that
bears his name: Bayes' theorem, which was
published posthumously.
20. Treatment A: 2/1
Treatment B: 1/1
The chance of survival Treatment A
Treatment B
Likelihood of probability
0.1
0.08
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Probability
21. Do what is best for the poor little
Treatment A: 24/18
wallabies
Treatment B: 1/1
The chance of survival Treatment A
Treatment B
Likelihood of probability
0.1
0.08
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Probability
22. No, I am a scientist, randomised
sequential clinical trial
Treatment A: 80/70
Treatment B: 90/50
The chance of survival Treatment A
Treatment B
Likelihood of probability
0.1
0.08
0.06
0.04
0.02
0
0 0.2 0.4 0.6 0.8 1
Probability
23. No, I am a scientist, randomised
sequential clinical trial
Treatment A: 80/70
Treatment B: 90/50
The chance of survival Treatment A
Treatment B
Don‟t worry, I just
Likelihood of probability
0.1
0.08
discovered treatment C
0.06 which is a lot better
0.04
0.02 than A or B, stop the trial
0
0 0.2 0.4 0.6 0.8 1
Probability
24. Answer
• There is an optimal state dependent
allocation of wallabies to treatments that is
a compromise between doing what is best
now and reducing uncertainty so we make
better decisions in the future = perfectly
optimal active adaptive management
Rout T.M., Hauser C.E. & Possingham H.P. (2009). Optimal adaptive
management for the translocation of a threatened species. Ecol. Appl.,
19, 515-526
McCarthy M.A. & Possingham H.P. (2007). Active adaptive management
for conservation. Conserv. Biol., 21, 956-963
25. 5 A tricky objective
Keep the public and/or politicians
happy, or provide them with
enough information to drive actions
26. Another new problem: How much monitoring do
we need to keep the masses/politicians happy?
Many
people
cranky
More rigorous
Public’s approach
level of
discontent
Publicise
with the casual
monitoring observations
investment
People who are
never happy
Few Level of funding
people that is legislated for
Amount of investment in
cranky
monitoring strategy
28. Thoughts
• Many things should not be monitored because the costs
outweigh the benefits
• Monitoring is first and foremost an optimisation problem.
Statistics is part of the mechanics but should not proceed
without being nested in a decision theory problem
• Ecological stats is taught in the context of pure science not
applied science which is why we are in a mess
• Is monitoring a political displacement activity intended to
keep scientists busy?
• How much data do we need to convince the masses that
everything is bad/ok? Is some data more compelling than
other data?
• Is there an optimal amount of surveillance?
• What should I tell TNC to do?
29. Some more of our papers on optimal
monitoring and information gain
• How long should I monitor a fix stock before fixing the
reserve size?
– Gerber, L. R., M. Beger, M. A. McCarthy, and H. P.
Possingham. 2005. A theory for optimal monitoring of
marine reserves. Ecology Letters 8:829-837
• Monitor or manage? – uses POMDPs
– Chades I., McDonald-Madden E., McCarthy M.A.,
Wintle B., Linkie M. & Possingham H.P. (2008). When
to stop managing or surveying cryptic threatened
species. PNAS, 105, 13936-13940
• More recent papers by McDonald-Madden et al.
30. Before you monitor
• Stop, Think
• Maybe monitor less, better and longer
• Work out what you might do with the information
that could alter future actions (even public opinion)
and increase the chance of delivering a net
conservation outcome relative to other forms of
expenditure
• Place it in a decision theory or forecasting context
and work out how long it will take and how much it
will cost – can you afford it? Maybe you should act
with what you know now?
Read Decision Point (monthly): www.aeda.edu.au/news
32. 2 Trading type I and type II errors
Mapstone (1995), Field et al. (2004)
Truth
Species OK Species declining
Species Type II
OK
Great
Data
Species Type I Great
declining
Field, S. A., A. J. Tyre, N. Jonzén, J. R. Rhodes, and H. P. Possingham. 2004.
Minimizing the cost of environmental management decisions by optimizing
statistical thresholds. Ecology Letters 7:669-675
33. 2 Trading type I and type II errors
Mapstone (1995), Field et al. (2004)
Truth
Species OK Species declining
Species Type II
OK
Great
Data
Species Type I Great
declining
Field, S. A., A. J. Tyre, N. Jonzén, J. R. Rhodes, and H. P. Possingham. 2004.
Minimizing the cost of environmental management decisions by optimizing
statistical thresholds. Ecology Letters 7:669-675
34. History: Gum sites
Bob Howe, David
Paton, Drew Tyre,
Tim and Patrick
Three 20min 2ha
Stringybark
counts - c160 sites sites
from 1999 to now
the ecology centre
university of queensland
australia
www.uq.edu.au/spatialecology
h.possingham@uq.edu.au
35. The canary of the
Statistically significant decline in stringybark
canaries. All is not well
for Scarlet Robins in
stringybark.
This is not surprising as
there is ample local and
national evidence that
Not stat significant decline in gum woodland this species is going
downhill steadily.
36. 1. Specify project objectives
No 3. Implement research to identify
2. Do I know the threats and management options? threats and/or management options.
Yes
5. Do I know which
4. Does my choice of
Yes management option Yes 6. Use decision analysis to evaluate
management action depend on the
is best given each options for monitoring the state of the
state of the system?
state of the system? system.
No
No
8. Implement this management option.
7. Is my best management option clear? Yes
No monitoring recommended.
No 1. Use decision analysis to evaluate
management options. Implement best
9. Do I have sufficient time to make No
management option from this analysis.
changes to management? No monitoring recommended.
Yes
1. Monitor and manage within an
active adaptive management
1. Do we have the resources to Yes framework to determine the best
implement active adaptive management option over time.
management?
No 1. Monitor and manage within a
passive adaptive management
Yes framework.
1. Use decision analysis to evaluate Use decision analysis to identify initial
options for monitoring the performance management option.
of my management options.
Has an effective monitoring option
emerged? No 1. Use decision analysis to evaluate
management options.
Implement best management option
from this analysis.
Figure 1: Decision tree for deciding when to monitor No monitoring recommended.
to improve conservation management.
37. 1. Specify project objectives
No 3. Implement research to identify
2. Do I know the threats and management options? threats and/or management options.
Yes
5. Do I know which
4. Does my choice of
Yes management option Yes 6. Use decision analysis to evaluate
management action depend on the
is best given each options for monitoring the state of the
state of the system?
state of the system? system.
No
No
8. I.mplement this management option.
7. Is my best management option clear? Yes
No monitoring recommended.
No 1. Use decision analysis to evaluate
management options. Implement best
9. Do I have sufficient time to make No
management option from this analysis.
changes to management? No monitoring recommended.
Yes
1. Monitor and manage within an
active adaptive management
1. Do we have the resources to Yes framework to determine the best
implement active adaptive management option over time.
management?
No 1. Monitor and manage within a
passive adaptive management
Yes framework.
1. Use decision analysis to evaluate Use decision analysis to identify initial
options for monitoring the performance management option.
of my management options.
Has an effective monitoring option
emerged? No 1. Use decision analysis to evaluate
management options.
Implement best management option
from this analysis.
Figure 1: Decision tree for deciding when to monitor No monitoring recommended.
to improve conservation management.
38. a) 1 1
Half protection rate
0.95 0.99 Protection rate
Retention in Landscape
Representation in PAs
0.9 0.98 Double protection rate
Number of species 0.85 0.97
0.8 0.96
saved as a function of 0.75 0.95
0.7 Half protection rate
years spent collecting
0.94
Protection rate
0.65 0.93
Double protection rate
0.6
protea data 0 2 4 6
Survey Period
8 10
0.92
0 2 4
Survey Period
6 8 10
c) 1 d) 1
0.95 0.98
Retention in Landscape
Data on habitats
Representation in PAs
0.9 0.96
0.85
and proteas 0.8
0.94
0.92
0.75
Half habitat loss rate 0.9 Half habitat loss rate
0.7
Habitat loss rate Habitat loss rate
0.65 0.88
Double habitat loss rate
Build Double habitat loss rate
Get more 0.6 0.86
reserve 0 2 4 6 8 10 0 2 4 6 8 10
data? Survey Period Survey Period
system
(Grantham H.S., Wilson K.A., Moilanen A., Rebelo T. & Possingham H.P.
(2009). Delaying conservation actions for improved knowledge: how long
should we wait? Ecology Letters, 12, 293-301) – similar concept in Gerber et al.
2004.
39. The Environment Institute
Where ideas grow
Hugh Possingham
For more information about this event or other events, please
visit our website at www.adelaide.edu.au/environment