2.7.4.4 Receptor impact models


2.7.4.4.1 Percent foliage cover

Elicitation scenarios

Table 16 summarises the elicitation design matrix for the percent foliage cover in the streams landscape groups. The first three design points (design point identifiers 1, 2 and 3) address the current variability in the number of low-flow days that these two landscape groups experience. During the assessment years the model assumes that there is no groundwater drawdown due to coal resource development, and the mean annual number of days with peak daily streamflow exceeding the EventsR2.0 reference value is by definition fixed at one in 2 years.

Design points identifiers 26, 28, 48, 15, 36 and 52 capture the extremes and intermediate values of the hydrological response variables in the short-future period (2042), and identifiers 86, 61, 73, 57, 92 and 108 perform the same role for the long-future period (2102). Note that the design point identifiers are simply index variables that identify the row of the elicitation design matrix. They are included here to maintain an auditable path between analysis and reporting.

In the future-assessment periods (short, 2013 to 2042, and long, 2073 to 2102), the maximum decrease in groundwater levels has been set at 10 m following advice received from the groundwater modelling team. In these assessment years the design matrix allows for the smallest possible (zero) and a very large (356) number of annual low-flow days. The number of flood events that exceed the peak daily flow in flood events with a return period of 2 years as defined from modelled flow in the reference period (1983 to 2012) – EventsR2.0 – is varied from a very low value of one flood event every 20 years to a very high value of approximately three flood events of this magnitude every 2 years.

The reference value (Yref) in the future-assessment years spans the receptor impact model estimates of the 10th and 90th percentiles of percent foliage cover (see also the top row of Figure 22), based on the experts’ responses to the design points in the reference year. These design point values were selected by the receptor impact model software during the receptor impact modelling workshop; as soon as the experts had responded to design points 1, 2 and 3 the points were automatically included within the design for the elicitations at the subsequent design points.

Table 16 Elicitation design matrix for percent foliage cover in the streams landscape groups in the Galilee subregion zone of potential hydrological change


Design point identifier

dmaxRef

LQD

EventsR2.0

Yref

Year

1

0

177

0.50

na

2012

2

0

0

0.50

na

2012

3

0

354

0.50

na

2012

26

2.8

356

1.65

0.19

2042

28

0.2

0

0.05

0.32

2042

48

10

0

1.65

0.32

2042

15

10

178

0.85

0.19

2042

36

10

356

0.05

0.32

2042

52

0.2

356

1.65

0.32

2042

86

2.8

178

0.05

0.32

2102

61

0.2

356

0.05

0.19

2102

73

0.2

0

1.65

0.19

2102

57

10

0

0.05

0.19

2102

92

2.8

0

0.85

0.32

2102

108

10

356

1.65

0.32

2102

Design points for Yref in the future (short- and long-assessment periods) were calculated during the receptor impact modelling elicitation workshop using elicited values for the receptor impact variable in the reference period. All other design points (with identifiers) are either default values or values determined by groundwater and surface water modelling. na = not applicable, hydrological response variables are as defined in Table 14.

Data: Bioregional Assessment Programme (Dataset 4)

Receptor impact model

The receptor impact variable is the percent foliage cover of Eucalyptus camaldulensis and Melaleuca spp. in the streams landscape groups. The sample unit is a 100 m length transect along the stream reach extending from the stream channel to the top of the bank. The width is at least 10 m increasing to 15 m in areas where more than a single row of river red gum is present during the reference period. This sample frame is invariant for predictions in future periods.

The receptor impact modelling methodology allows for a very flexible class of statistical models to be fitted to the values of the receptor impact variable elicited from the experts at each of the 1010 design points (for details see companion submethodology M08 (as listed in Table 1) for receptor impact modelling (Hosack et al., 2018)).

The model fitted to the experts’ elicited values is summarised in Table 17 and Figure 22. The fitted model takes the form:

eta equals h open parenthesis y close parenthesis equals beta subscript 0 end subscript x subscript 0 end subscript plus beta subscript f end subscript x subscript f end subscript plus beta subscript l end subscript x subscript l end subscript plus beta subscript r end subscript x subscript r end subscript plus sum from j equals 1 to 3 of beta subscript h subscript j end subscript end subscript x subscript h subscript j end subscript end subscript

(1)

where x subscript 0 end subscript is an intercept term (a vector of ones),  x subscript f end subscript is a binary indicator variable scored 1 for the case of an assessment in the short- or long-assessment year, x subscript l end subscript is a binary indicator variable scored 1 for the case of an assessment in the long-assessment year, x subscript r end subscript is a continuous variable that represents the value of the receptor impact variable in the reference year (Yref, set to zero for the case of an assessment in the reference year), and x subscript h subscript j end subscript end subscript j equals 1 dot dot 3 are the (continuous or integer) values of the three hydrological response variables (dmaxRef, LQD and EventsR2.0).

Table 17 Mean, 10th and 90th percentile of the coefficients of the receptor impact model percent foliage cover in the streams landscape groups in the Galilee subregion zone of potential hydrological change


Mean

q10

q90

(Intercept)

–0.682

–1.07

–0.294

future1

–0.17

–0.578

0.237

long1

–0.0243

–0.372

0.323

dmaxRef

–0.0276

–0.0784

0.0233

EventsR2.0

0.084

–0.207

0.375

LQD

–0.00376

–0.00498

–0.00254

Future is a binary variable scored 1 if the analysis case is in a short- or long-assessment period. Long is a binary variable scored 1 if the analysis case is in the assessment year. dmaxRef, LQD and EventsR2.0 are as described in Table 14.

Data: Bioregional Assessment Programme (Dataset 4)

Note that the modelling framework provides for more complex models, including quadratic value of, and interactions between, the hydrological response variables but in this instance the simple linear model (Equation 1) was identified as the most parsimonious representation of the experts’ responses.

The model estimation procedure adopts a Bayesian approach. The model coefficients beta subscript 0 end subscript comma beta subscript f end subscript comma beta subscript l end subscript comma beta subscript r end subscript comma beta subscript h subscript j end subscript end subscript are assumed to follow a multivariate normal distribution. The Bayesian estimation procedure quantifies how compatible different values of the parameters of this distribution are with the data (the elicited expert opinion) under the model. The (marginal) mean and 80% central credible intervals of the three hydrological response variable coefficients (holding all other variables at their median values) are summarised in partial regression plots in Figure 22, whereas Table 17 shows the (jointly estimated) mean and 80% central credible interval (defined by the 10th and 90th percentiles) for all seven model coefficients.

Interestingly, the coefficient for Yref is not referenced in the best selected model (Table 17), which indicates that the current value of percent foliage cover is not a major predictor of future values. This is at odds with the predictions for equivalent receptor impact variables in other bioregions where the coefficient for percent foliage cover in the assessment year typically is important. It is critical to recognise, however, that percent foliage cover in the streams landscape groups starts from a very low base – the 90th percentile is approximately 30% with a mean value of about 22%. It is also important to recognise that the relatively modest changes highlighted in the summary above may still be ecologically important given the relatively low baseline condition. The model indicates that the experts’ opinion provides strong evidence that of the three hydrological response variables only LQD has a major (Bayesian credible intervals do not span zero) influence on the mean percent foliage cover in the streams landscape groups (Table 17). In the short- and long-assessment years, the model predicts that if all other model variables are held at their median values, and the groundwater depth decreases by 5 m, then the mean percent foliage cover will decrease by approximately 5% with some large uncertainty that is even larger in the long-assessment year (middle and bottom row, left column, Figure 22).

Figure 22

Figure 22 (Top row) Predicted mean (black dot or line) and 80% central credible interval (grey line or polygon) of percent foliage cover under reference hydrological conditions. (Middle and bottom rows) Predicted future effects (mean = black line, 80% central credible interval = grey polygon) of each hydrological response variable on percent foliage cover

In the middle and bottom rows, the uncertainty in the percent foliage cover in the reference year has been integrated out, and all other hydrological response variables are held constant at the midpoint of their elicitation range (during the risk estimation process all hydrological response variables vary simultaneously). The dashed vertical lines show range of hydrological response variables used in the elicitation workshop. Reference = period from 1983 to 2012, Short = assessment period 2013 to 2042, Long = assessment period 2073 to 2102. dmaxRef, LQD and EventsR2.0 are as described in Table 14.The range of the y axis is from just below 10% to just above 50%. The numbers on the y-axis range from 0 to 1 as the receptor impact model was constructed using the proportion for the statistical modelling. They should be interpreted as a percent foliage cover ranging from 0 to 100%.

Data: Bioregional Assessment Programme (Dataset 4)

If all other model variables are held at their median values, the model predicts a somewhat more dramatic decrease of approximately 10% in mean percent foliage cover if the number of low-flow days increases by 100 days/year from a mid-value of 177 days/year to 277 days/year. Again, the uncertainty associated with these predictions appears to increase slightly (larger credible interval) in the long-assessment year (middle and bottom row, middle column, Figure 22).

Finally, the model suggests that if all other model variables are held at their median values, the mean percent foliage cover should increase as the number of flood events with peak daily flow exceeding the 1983 to 2012 2-year return period increases. The magnitude of the mean increase in foliage cover will be a few percent. There is, however, a large amount of uncertainty associated with this prediction as indicated by the relatively broad credible intervals in Figure 22 and the order of magnitude difference between the 10th and 90th percentiles of the Events2.0 coefficient in Table 17.

2.7.4.4.2 Density of baetids

Elicitation scenarios

Table 18 summarises the elicitation design matrix for the mean density of baetids (Offadens) in the streams landscape groups of the Galilee subregion. Here, the first six design points (design point identifiers 1 to 6) address the current variability in the number of low-flow days, and mean maximum spell duration (in days) per year with low streamflow, that the streams landscape groups experience.

Design points identifiers 7, 21, 41, 9 and 11 capture the extremes and intermediate values of the hydrological response variables in the short-future period (2042), and identifiers 22, 17 and 18 perform the same role for the long-future period (2102). Note that the design point identifiers are simply index variables that identify the row of the elicitation design matrix. They are included here to maintain an auditable path between analysis and reporting. Note also that the design points are the minimum required to estimate the coefficients of a linear model that allows for a quadratic relationship between hydrological response variables and receptor impact variables, and for interactions between hydrological response variables, in the reference and future assessment periods (Section 5.2, companion submethodology M08 (as listed in Table 1) for receptor impact modelling (Hosack et al., 2018)). The absence of any particular combination of hydrological response variables (for example, LQD = 0 and LME = 0) in any particular year (for example 2102) does not imply that this scenario cannot occur or is not modelled. The purpose of the receptor impact modelling is to enable the analysis to extrapolate across what is an infinite set of possible combinations.

Again, the reference value (Yref) in the future assessment years spans the receptor impact model estimates of the 10th and 90th percentiles of density of baetids (see also the top row of Figure 23), based on the experts’ responses to the design points in the reference year.

Table 18 Elicitation design matrix for density of baetids (Offadens) in the streams landscape groups in the Galilee subregion zone of potential hydrological change


Design point identifier

LQD

LME

Yref

Year

3

343

15.3

na

2012

1

0

0

na

2012

6

354

201

na

2012

4

177

101

na

2012

5

354

101

na

2012

2

177

0.90

na

2012

7

0

0

240

2042

21

178

0.95

30

2042

41

178

115

30

2042

9

343

16

240

2042

11

356

115

240

2042

22

178

115

240

2102

17

356

115

30

2102

18

356

230

30

2102

Design points for Yref in the future (short- and long-assessment periods) were calculated during the receptor impact modelling elicitation workshop using elicited values for the receptor impact variable in the reference period. All other design points (with identifiers) are either default values or values determined by groundwater and surface water modelling. na = not applicable, LQD and LME are as described in Table 14.

Data: Bioregional Assessment Programme (Dataset 4)

Receptor impact model

The receptor impact variable is the density of mayfly nymphs (order Ephemeroptera) in the genus Offadens of the family Baetidae (Webb and Suter, 2011). The measurement unit for the receptor impact variable is the number of mayfly nymphs per m2. Density would be measured in a 2 m by 0.5 m quadrat in flowing water. The measurement period would be 3 months after the end of the wet season.

The model fitted to the experts’ elicited values is summarised in Table 19 and Figure 23 . The fitted model takes the form:

eta equals h open parenthesis y close parenthesis equals beta subscript 0 end subscript x subscript 0 end subscript plus beta subscript f end subscript x subscript f end subscript plus beta subscript l end subscript x subscript l end subscript plus beta subscript r end subscript x subscript r end subscript plus sum from j equals 1 to 2 of beta subscript h subscript j end subscript end subscript x subscript h subscript j end subscript end subscript

(2)

where x subscript 0 end subscript is an intercept term (a vector of ones), x subscript f end subscript is a binary indicator variable scored 1 for the case of an assessment in the short- or long-assessment year, x subscript l end subscript is a binary indicator variable scored 1 for the case of an assessment in the long-assessment year, x subscript r end subscript is a continuous variable that represents the value of the receptor impact variable in the reference year (Yref, set to zero for the case of an assessment in the reference year), and x subscript h subscript j end subscript end subscript j equals 1 dot dot 2 are the values of the two continuous hydrological response variables (LQD and LME).

Table 19 Mean, 10th and 90th percentile of the coefficients of the receptor impact model for density of baetids (Offadens) in the streams landscape groups in the Galilee subregion zone of potential hydrological change


Mean

q10

q90

(Intercept)

5.26

3.92

6.6

future1

0.0112

–1.13

1.15

long1

0.0241

–1.47

1.52

LQD

–0.00365

–0.00853

0.00123

LME

–0.00319

–0.0118

0.00544

Future is a binary variable scored 1 if the analysis case is in a short- or long-assessment period. Long is a binary variable scored 1 if the analysis case is in the assessment year. LQD and LME are as described in Table 14.

Data: Bioregional Assessment Programme (Dataset 4)

Again, the modelling framework provides for more complex models, including quadratic value of, and interactions between, the hydrological response variables but in this instance the simple linear model (Equation 2) was identified as the most parsimonious representation of the experts’ responses.

The model indicates that the experts’ opinion provides no strong evidence that either the number of low-flow days or the mean maximum spell of low-flow days have a significant effect on mean baetid density.

The model also predicts that baetid density under reference conditions does not influence outcomes under the different low-flow conditions in the future assessment years (Table 19). This conclusion is consistent with receptor impact model predictions for other relatively short-lived species (such as Hydropyschidae larvae) in other bioregions.

Figure 23

Figure 23 Predicted mean (black line) and 80% central credible interval (grey polygon) of density of baetids (Offadens) under reference hydrological conditions. (Middle and bottom rows) Predicted future effect (mean = black line, 80% central credible interval = grey polygon) of each hydrological response variable on density of baetids in the Galilee subregion zone of potential hydrological change

In the middle and bottom rows, all other hydrological response variables are held constant at the midpoint of their elicitation range (during the risk estimation process all hydrological response variables vary simultaneously). The dashed vertical lines show range of hydrological response variables used in the elicitation workshop. Reference = period from 1983 to 2012, Short = assessment period 2013 to 2042, Long = assessment period 2073 to 2102. LQD and LME are as described in Table 14.

Data: Bioregional Assessment Programme (Dataset 4)


Last updated:
6 December 2018
Thumbnail of the Galilee subregion

Product Finalisation date

2018
PRODUCT CONTENTS

ASSESSMENT