By: James Bernard Supervised By: Charles Todd (Department of Sustainability and Environment) Simon Nicol (Department of Sustainability and Environment) Charles Twardy (Monash University) David Green (Monash University) Building Bayesian Models for the Analysis of Critical Knowledge Gaps in Australian Freshwater Fish
2 Introduction Aim Growth Curves New Growth Curves New Curves using Data Clustering Future Work
3 Aim Overall Goal (Big Picture): –Predict the sustainability of the Murray Cod > Growth Curves > Survival Rate (Mortality) > Population Modelling
4 Growth Curves Considered various curves: –von Bertalanffy, Gompertz, Logistic Reviewed previous experts curves: –Anderson (1992) –Gooley (1995) –Rowland (1998) –Todd (unpublished)
5 Existing Growth Curves: Rowland
6 Existing Growth Curves: Anderson
7 Existing Growth Curves: Todd
8 Existing Growth Curves: Gooley
9 Existing Growth Curves (equations) RowlandToddAnderson k Parameters: Equation: (von Bertalanffy)
10 Difference (0-5)
11 Difference (0-5) continued… RowlandToddAnderson What happens to the differences between these curves if is set to zero?
12 New Growth Curves RowlandToddAnderson k (0) Parameters: Equation: (von Bertalanffy) = 0 Note:
13 New Growth Curves: Rowland 0
14 New Growth Curves: Anderson 00
15 New Growth Curves: Todd 00 0
16 Evaluating the New Curves New Curves vs Old Curves 0 0 0
17 New Growth Curves: Using Data Clustering New Data Set: Only lengths (no age) Data Clustering provides: Length-Classes –using Minimum Message Length (MML) approach Expert Knowledge: Assign approximate ages to the classes Results: Three New Growth Curves modelling different amounts of uncertainty
18 New Growth Curves: Achieved by Data Clustering Class 1: Length: mm -> Age: 0-1 Class 2: Length: mm -> Age: 1-2 Class 3: Length: mm -> Age: 2-5 Class 4: Length: mm -> Age: 3-9 Class 5: Length: mm -> Age 9+ Length (mm) Number (fish in each class) Data Clustering Murray Cod lengths
19 New Growth Curves: Using Data Clustering D/Clus 1D/Clus 2D/Clus k Parameters: Equation: (von Bertalanffy)
20 New Growth Curves: Using Data Clustering D/Clus 1D/Clus 2D/Clus k (0) Parameters: Equation: (von Bertalanffy) =0 Note:
21 New Growth Curves: Using Data Clustering 000
22 Comparing Existing Curves to New Curves 00
23 Summary Improved Existing Curves –Using old data sets Created New Curves –Using new data sets and data clustering –The curve modelling the most uncertainty provided the best fit to otolith data –In all cases setting = 0 provided the best fit to recapture data
24 Future Work We do plan on modelling the entire population –Our next step is developing a Bayesian model for determining survival rates! Stay tuned: –