Question 1.

The figure below shows a cognitive map capturing understanding of the drivers of algal blooms in an inland river. The system could also be described using a Bayesian Belief Network (BBN).

What are the strengths and weaknesses of cognitive maps and BBNs in this context?

Question 2.

Biological surveys for rare and threatened species are usually required prior to development of a site containing native vegetation. The chance of failing to detect a rare or threatened species depends on how hard we look.

Let's say a site looks like good habitat for the threatened Corroboree Frog. Ecologists believe that Corroboree frogs occur at 25% of sites with good habitat. The frog is difficult to detect. If it is in fact present, there is a 50% chance of detecting it in any one survey. Three surveys fail to detect the species.

(a) What is the probability that all three surveys will fail to detect the species, if in fact it is present? (three marks)

The frog is distinctive and cannot be confused with other species. It's reasonable to assume that the frog is never recorded as present when in fact it is absent (the false positive rate is zero).

Bayes' rule is given as:

(b) If the frog is recorded as absent in each of three surveys, use Bayes' rule to calculate the posterior probability that it is in fact present at the site.

Question 3.

Expert judgment is often used to inform risk assessments, however, this may introduce bias and error.

(a) When using expert judgment an important expert frailty to consider is over-confidence. What is meant by ‘over-confidence'?

(b) Estimates are usually elicited as intervals. Describe a procedure the risk analyst can use to potentially reduce over-confidence in an individual expert judgment of an interval.

(c) In a workshop setting, describe three ways in which biases in expert judgment can be minimised?

Question 4.

The Environmental Protection Authority claims that its regulatory procedures for use of a new pesticide are safe. The basis of the claim is that regulation is built around the 1 mg/L no observable effect concentration recorded in toxicological testing.

(a) What is meant by no observable effect concentration in the context of toxicology testing?

Concerned stakeholders carry out their own toxicology tests and summarise results using 95% confidence intervals as shown below.

(b) Do these results support the Authority's claim that regulatory procedures for use of a new pesticide are safe? In your answer, explain the term ‘statistical power' and its relevance here.

Question 5.

A cohort study was initiated to examine the effect of exposure to oxalic acid in chocolate on osteoporosis. There were 5368 people in the cohort exposed to oxalic acid (they regularly ate chocolate). There were 8264 people in the cohort not exposed to chocolate.

The number of cases of osteoporosis in the two cohorts was compared. There were 310 cases in the exposed cohort and 403 cases in the unexposed cohort.

(a) What is the rate of disease in these two groups? (two marks)
(b) What is the correct measure of effect between exposure and outcome for this study? (two marks)
(c) What is the numerical value of this measure? (two marks)
(d) Interpret this result in one plain language sentence. (two marks)

Question 6.

Two alternative views of risk are Risk = likelihood × consequence. Risk = hazard + outrage.

Discuss these two views from a technical and social perspective.

Question 7.

Managers of Wilson's Promontory National Park have identified three alternatives for improving the condition of Coastal Grassy Woodlands. In good condition these woodlands comprise (native) Banksia trees and a ground layer of native grasses. Some of the alternatives involve controversial actions that may spark public outrage. Objectives for the managers include:
- Maximise Banksia cover
- Maximise native grass cover
- Minimise costs of implementation
- Minimise public outrage

The performance matrix below describes the merit of each alternative against each objective. The managers of the park ask you to facilitate discussion with stakeholders.

In your role as the facilitator, how would you go about progressing decision-making?

Question 8.

A government agency is reviewing existing regulations for emissions of airborne particulates. A hazard elicitation workshop with health experts identified smoke from domestic wood-fired heaters as a relatively high risk; although, there was substantial difference of opinion among the experts. The agency is considering new regulations that would: ban the use of old wood- fired heaters that do not comply with Australian Standard AS4013, and designate "Wood Fire Ban" days when atmospheric conditions are unfavourable for the timely dispersion of emissions.

Given the disagreement among the experts, and the potential public backlash against the new regulations, the agency has produced a Monte Carlo simulation model. The model compares the current particulate (PM2.5) levels with the predicted PM2.5 levels after implementation of the proposed new regulations. The model predicts a significant decrease in the daily average PM2.5 levels during winter under the new regulations. The agency believes this modelling proves the public health benefits of the proposal.

The experts disagree about the dominant mechanism by which wood-fired heaters cause increased PM2.5 levels. One suggested mechanism is the high density of old heaters, the other mechanism is the presence of weather conditions that trap emissions.

(a) What type of uncertainty does this disagreement create for the Monte Carlo model?

The inputs for the Monte Carlo model were specified by fitting probability distributions to data, as follows:
- Climate data were drawn from records containing hourly observations over 30 years.
- The age and density of wood-fired heaters came from an online survey with 17 responses.
- Current PM2.5 levels came from a dataset containing half-hourly observations at 15 locations over 10 years.

(b) Would you expect a well-fitted probability distribution from any of the above data sources? If so, which one(s)?

(c) There are four key types of uncertainty we need to consider when using Monte Carlo simulation. Which type(s) of uncertainty relate(s) to the fitting of the distributions?

(d) The government agency states that 1,000,000 simulations were performed with their Monte Carlo model. They claim this makes the results extremely accurate and reliable. Is this true? Why or why not?

The government agency receives feedback from the independent scientific review board that their Monte Carlo analysis is unreliable and that they have neglected aspects of all four key types of uncertainty.

(e) What are two ways the government agency could improve their analysis to make it more reliable?

Question 9.

Demographic stochasticity poses a substantial risk to the viability of a small population of a critically endangered snail.

(a) Define ‘demographic stochasticity'.

(b) Why is demographic stochasticity important in small populations?

The per annum fecundity (f) of each individual in the population is estimated to be 0.4. The per annum survivorship (s) of each individual is estimated to be 0.7. The current population size (Nt) is 10 individuals.

(c) Using the formula below, calculate the mean expectation for the population size next year.

N_t+1 = (f+s)N_t.

(d) A computer simulates demographic stochasticity by drawing random numbers from a uniform distribution bounded by zero and one. A single iteration of the simulation draws the random numbers shown in the table below. Assuming again that Nt = 10, calculate the population size next year for this iteration.

Question 10.

(a) Briefly explain why you might use causal models in an adaptive management framework.

(b) What is ‘critical uncertainty' in the context of adaptive management?

(c) Even if adaptive management is warranted, what factors can influence an organisation's ability to implement adaptive management?