Slide 1

Slide 1 text

Predictive invasion ecology and management decisions under uncertainty Corey Chivers McGill University

Slide 2

Slide 2 text

Ecological methodology for decision support Linking theory with data to make Inference, Prediction, & Decisions.

Slide 3

Slide 3 text

No content

Slide 4

Slide 4 text

No content

Slide 5

Slide 5 text

Prediction is very hard, Prediction is very hard, especially about the future especially about the future -Niels Bohr, Danish physicist (1885-1962) -Niels Bohr, Danish physicist (1885-1962)

Slide 6

Slide 6 text

Outcome A Outcome B Outcome C Outcome E Outcome F Do A Do B Do C Do D Do Nothing t = 0 t = 1 Time In a changing world, not making a decision has consequences, intended or otherwise.

Slide 7

Slide 7 text

How do we make the best use of the data and theory that we have to make predictions which take into account (potentially large) inherent uncertainties?

Slide 8

Slide 8 text

No content

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

No content

Slide 11

Slide 11 text

No content

Slide 12

Slide 12 text

What information do I have? What can I go out and observe? Data The Process

Slide 13

Slide 13 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Data The Process

Slide 14

Slide 14 text

All models are wrong, All models are wrong, But some are useful. But some are useful. -George E.P. Box -George E.P. Box

Slide 15

Slide 15 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Data The Process

Slide 16

Slide 16 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes Data The Process

Slide 17

Slide 17 text

No content

Slide 18

Slide 18 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes Data The Process

Slide 19

Slide 19 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Data The Process pseudo-data

Slide 20

Slide 20 text

No content

Slide 21

Slide 21 text

Sampling periods

Slide 22

Slide 22 text

No content

Slide 23

Slide 23 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Data The Process pseudo-data

Slide 24

Slide 24 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Data The Process pseudo-data

Slide 25

Slide 25 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Data The Process pseudo-data

Slide 26

Slide 26 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Does it fit the real data? Data The Process pseudo-data

Slide 27

Slide 27 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Does it fit the real data? Data The Process pseudo-data

Slide 28

Slide 28 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Does it fit the real data? Test Hypotheses Make forecasts (Forward Simulation) Data The Process pseudo-data

Slide 29

Slide 29 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Does it fit the real data? Test Hypotheses Make forecasts (Forward Simulation) Optimize Decisions Scenario Analysis Data The Process pseudo-data

Slide 30

Slide 30 text

No content

Slide 31

Slide 31 text

Invasive forest insects 1. International trade has many externalities 2. Total damages of existing pests 3. Estimate the probability of new high impact pest A. Guilds: which pathways? B. Economic sectors: who pays the costs?

Slide 32

Slide 32 text

• Base line information lacking – Compile all known non-indigenous forest pests – Identify short list of intermediate damaging pests – • National economic estimates lacking – In depth analysis of the most damaging pests – 3 guilds (borers, sap suckers, foliage feeders) – 3 economic cost sectors (government, households, market) Emerald Ash Borer Hemlock Woolly Adelgid Gypsy Moth

Slide 33

Slide 33 text

( ) ( ) ) ( ) | ( ) ( | | Pr 1 ϑ ϑ ϑ ϑ ϑ P c f P P M = m m       ∝ ∝ ∏ c c If we knew the cost of each pest, we can fit our models using the simple likelihood function. 0 1 2 3 4 5 6 7 8 0 2 4 6 8 Cost ($) Frequency of pests

Slide 34

Slide 34 text

0 1 2 3 4 5 6 7 8 0 2 4 6 8 Cost ($) Frequency of pests 78 13 1 Pr (ϑ∣d)∝ [∏ i=1 I P(low∣ϑ) x∏ j=1 J P(intermediate∣ϑ) x∏ k=1 K P(high∣ϑ) ]P(ϑ) What we have are frequencies of species in different impact ranges.

Slide 35

Slide 35 text

The Framework: A) Species Frequencies in 3 categories. B) Which Model? C) Model estimation D) Probability distribution of derived variable of interest (total cost, probability of new high impact pest) Aukema JE, Leung B, Kovacs K, Chivers C, Britton KO, et al. (2011) Economic Impacts of Non-Native Forest Insects in the Continental United States. PLoS ONE 6(9): e24587

Slide 36

Slide 36 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Does it fit the real data? Test Hypotheses Make forecasts (Forward Simulation) Optimize Decisions Scenario Analysis Data The Process pseudo-data

Slide 37

Slide 37 text

Simulation Validation 1 Can we identify the Correct model?

Slide 38

Slide 38 text

Simulation Validation 2 Given a model, can we recapture the parameters, and derived quantities of interest?

Slide 39

Slide 39 text

Results Highest impact are the Borer guild. Costs are born primarily by local governments. ~1.7 Billion USD per annum Single most damaging pests in each guild accounts for 25-50% of the total impacts. At current establishment rates ~32% chance of another high impact pest in the next ten years. Aukema JE, Leung B, Kovacs K, Chivers C, Britton KO, et al. (2011) Economic Impacts of Non-Native Forest Insects in the Continental United States. PLoS ONE 6(9): e24587

Slide 40

Slide 40 text

More imports, this time of the fishy sort. ● Given a proxy measure of propagule pressure, how well can we estimate the risk of establishment? Bradie, J., Chivers, C. & Leung, B. (2013) Importing risk: quantifying the propagule-pressure establishment relationship at the pathway level. in press Diversity and Distributions.

Slide 41

Slide 41 text

● Not all species are equally likely to establish ● In the absence of species specific lifehistory information, how well can we estimate overal pathway risk?

Slide 42

Slide 42 text

Bradie, J., Chivers, C. & Leung, B. (2013) Importing risk: quantifying the propagule-pressure establishment relationship at the pathway level. in press Diversity and Distributions. Effects of unaccounted variability

Slide 43

Slide 43 text

Results At an import level of 100,000 individuals, establishment risk of 19% Importing 1 million individuals leads to just under a 1 in 2 chance of establishment. Bradie, J., Chivers, C. & Leung, B. (2013) Importing risk: quantifying the propagule-pressure establishment relationship at the pathway level. in press Diversity and Distributions.

Slide 44

Slide 44 text

No content

Slide 45

Slide 45 text

Alternative models of human behaviour ● Gravity Model – 'Pull' of attractive lakes ● Random Utility Model – Rational utility maximizers (Schneider et al. 1998, Leung et al. 2004, 2006) (Moore et al. 2005, Timar and Phaneuf 2009)

Slide 46

Slide 46 text

Alternative models of human behaviour ● Gravity Model – 'Pull' of attractive lakes ● Random Utility Model – Rational utility maximizers PGM T nj =A n W j e D nj −d , n=1,... ,n , j=1,..., J. A n =1/∑ k =1 L W k e D nk −d . U nj =V nj +ϵ nj , n=1,... , N , j=1,... J V nj =   X nj PRUM T nj = expV nj  ∑ k=1 J expV nk  , n=1,... , N , j=1,... , J (Schneider et al. 1998, Leung et al. 2004, 2006) (Moore et al. 2005, Timar and Phaneuf 2009)

Slide 47

Slide 47 text

Observing boater choice ● Online survey ● Which lakes did you visit? ● How many times?

Slide 48

Slide 48 text

Chivers & Leung (2012)

Slide 49

Slide 49 text

Figure A1: Simulated trip outcomes in a landscape of lakes with induced spatial auto-correlation. Size of circle is proportional to the size of the simulated lake.

Slide 50

Slide 50 text

Figure A3: Generating vs maximum likelihood estimates for the four parameters (panels) of the random utility model. The 1:1 line is also plotted for comparison. Figure A2: Generating vs maximum likelihood estimates for the four parameters (panels) of the gravity model. The 1:1 line is also plotted for comparison. Re-capture the parameter values?

Slide 51

Slide 51 text

Chivers & Leung (2012)

Slide 52

Slide 52 text

∑ i=1 I w i =13 Low Entropy High Entropy Predicted Dispersal Networks ∑ i=1 I w i =13

Slide 53

Slide 53 text

Chivers & Leung (2012)

Slide 54

Slide 54 text

Managment of spreading aquatic invasives

Slide 55

Slide 55 text

No content

Slide 56

Slide 56 text

Behavioural Model The effect of costly cleaning impossed at a lake on boater choice. -Redistribute visits? -Reduce visits? -Both?

Slide 57

Slide 57 text

1b =  P 1b  P 1b ∏ b=1 B N 1b N 1b '  1b N 1b ' 1− 1b N 1b −N 1b '  .N 2b ' N 2b  2b N 2b 1− 2b N 2b ' −N 2b  2b = P 2b P 2b  L ,|D= Where: , Policy Lake Non-Policy Lake Behavioural observation model

Slide 58

Slide 58 text

Counterfactual stated preference

Slide 59

Slide 59 text

Can I recapture model parameters from the amount and type of data I have?

Slide 60

Slide 60 text

Scenario analysis

Slide 61

Slide 61 text

What information do I have? What can I go out and observe? What are the hypothesized processes which generated the data? Theory/models Simulate Hypothesized Biological Processes How well can we recapture patterns and processes? (parameter estimation, model discrimination, & derived variables) Does it fit the real data? Test Hypotheses Make forecasts (Forward Simulation) Optimize Decisions Scenario Analysis Data Methodology for decision support pseudo-data

Slide 62

Slide 62 text

Thank you Supervisors: Dr. Brian Leung Dr. Elena Bennett Dr. Claire De Mazancourt Dr. Gregor Fussman Lab Mates: Johanna Bradie Paul Edwards Kristina Marie Enciso Andrew Sellers Lidia Della Venezia Erin Gertzen