2.7.7.2 Limitations of the receptor impact modelling approach

The strengths and limitations of the expert elicitation process used in the BAs for building qualitative ecosystem models and quantitative receptor impact models are described in some detail in Section 2.7.1 and in companion submethodology M08 (as listed in Table 1) for receptor impact modelling (Hosack et al., 2018). There is no need to revisit these here, except to acknowledge that the qualitative models and receptor impact models that were developed to represent the landscape classes (aggregated to landscape groups) in the zone of potential hydrological change for the Galilee subregion reflect the subjectivity and bias inherent in the knowledge base of the experts (e.g. in defining the scope of the model; its components and connections; the ecologically important hydrological variables; representative receptor impact variables; and magnitude and uncertainty of responses to change). Thus, each model represents ‘a view’ of a landscape group or ecosystem; a view that might brook argument about some of the specific aspects of these models, but one that would generally be accepted as an adequate high-level conceptualisation of the important components of the ecosystem(s) it represents. Importantly, given the regional-scale nature of the BA, the inherent complexity that naturally exists within each landscape group is simplified in the receptor impact modelling, although experts are encouraged to consider as fully as possible the system-level heterogeneity of each landscape group during the elicitation process. Further, Table 4 (Section 2.7.1) specifically acknowledges that the overall receptor impact modelling approach may create the potential to underestimate complex ecosystem function, and is something that needs to be considered and evaluated against observations (monitoring) in the future.

Important knowledge gaps and limitations specific to the receptor impact modelling undertaken for the Galilee subregion were identified at the expert elicitation workshops. These limit the assessment of potential impacts from hydrological changes due to additional coal resource development for some landscape groups, or components of landscape groups, within the zone of potential hydrological change. In other words, these gaps and limitations place some constraints on the comprehensiveness of the ecosystem analysis undertaken for this BA, and it is important to flag these so that further investigation can be better targeted to address these questions in the future.

Although some qualitative mathematical models include salinity and/or nutrient components, the expert elicitations to define the results for the receptor impact models are premised on changes in other types of hydrological parameters, for example, water availability or volume. Changes in water quality parameters that could potentially occur (for example, with a shift in the relative contributions of surface water runoff and groundwater contribution to streamflow, or due to enhanced connectivity between aquifers of differing water quality) are not represented. Thus, the potential ecological impacts due to additional coal resource development reported in companion product 3-4 for the Galilee subregion (Lewis et al., 2018) reflect the risk from hydrological changes that do not include water quality parameters. Consequently, the modelled impacts could differ if changes in key water quality parameters had been included in the model formulation.

The annual mean percent foliage cover of the woody vegetation, annual mean percent foliage cover of the woody riparian vegetation, and the annual mean density of the mayfly species (Offadens sp.) receptor impact variables were selected as indicators of terrestrial groundwater-dependent ecosystems (GDEs) and stream ecosystems. They have been identified as sensitive to changes in hydrology and can represent the response of other components of the ecosystem to changes in hydrology. The criteria for selecting the receptor impact variables are discussed in detail in companion submethodology M08 (as listed in Table 1) for receptor impact modelling (Hosack et al., 2018). The receptor impact variables identified here reflect those criteria and the knowledge base of the ecological experts available at the workshops, but may benefit from testing and further consideration of their optimality over time. The extent to which the receptor impact variables are suitable indicators of ecosystem response for all terrestrial GDEs and stream ecosystems across the Galilee subregion has not been established. The interpretation of results of the receptor impact models presented in companion product 3-4 for the Galilee subregion (Lewis et al., 2018) is couched in terms of risk to habitat, rather than risks to the receptor impact variables themselves.

A shortcoming of the ecological modelling process used for this BA is that it is assumed that there is sufficient hydrological modelling across the full distribution of a landscape class (or group of landscape classes) when the receptor impact models are elicited. In reality, surface water modelling particularly, is limited by the extent to which model simulation and calibration nodes exist and can be used to interpolate model output across the stream network. In the surface water zone of potential hydrological change of the Galilee subregion, there are large areas where the hydrological change to the stream network cannot be quantified. As an example, there is an absence of surface water model nodes along the upper parts of the Carmichael River and its tributary network (including Dyllingo Creek and Bimbah Creek). Similarly, there is no surface water modelling of extensive areas of Lagoon Creek and its tributary network in the southern mining cluster. The non-modelled stream network includes both the ‘Streams, GDE’ and ‘Streams, non-GDE’ landscape groups. This situation results in limited coverage of the stream network adjacent to some of the main coal mine developments both in the northern and southern parts of the zone. This limitation clearly hampers the assessment of potential risks across those parts of the landscape most affected from additional coal resource development.

A number of stream reaches in the modelled stream network of the Galilee assessment extent consist of multiple anabranches between surface water model nodes. This situation makes it difficult to reliably interpolate the surface water changes from model nodes to any particular anabranch. This is particularly the case for the Belyando River between model nodes 34 and 44 which consists of multiple anastomosing channels.

A clear knowledge gap for the Galilee subregion existed in relation to connectivity between groundwater and surface water systems. This relates both to connectivity within landscape classes and groups of classes and between ecosystems. Expert knowledge of potential surface water – groundwater interaction in the zone of potential hydrological change was limited. Similarly, there was limited understanding of the nature of groundwater interactions between riverine and terrestrial ecosystems. At the various workshops, the experts struggled with these aspects of the elicitation scenarios, in particular in relation to the ‘Floodplain, terrestrial GDE’ landscape group.

Climate change is not included directly in the receptor impact modelling, but a mid-range climate projection is used and potential changes to precipitation factored into the process through the surface water modelling of hydrological response variables.

A final shortcoming of the elicitation process was the absence of experts that had a strong knowledge of the local hydrological systems within the zone of potential hydrological change of the Galilee subregion. Most experts had not carried out field-based research within the main catchments of the zone, and had limited experience of the local hydrology and ecology. This situation resulted in multiple consequences for the receptor impact modelling process. Firstly, there were a limited number of suitable experts available for the elicitations. As an example, the elicitation process for the receptor impact variable for the ‘Non-floodplain, terrestrial GDE’ landscape group involved a single expert. Second, experts’ decisions on the choice of receptor impact variables and potential hydrological response variables were generally extrapolated from their knowledge from other regions. In the ‘best case’ scenario this information came from geographic areas adjacent to the zone. For instance the ‘high-flow macroinvertebrate’ receptor impact model was developed based on expert knowledge of a mayfly species (Offadens) obtained for the Cape River and its tributary network. This area is adjacent to the northern boundary of the zone. Lastly, for the ‘Springs’ landscape group, there was a lack of expert knowledge available at the workshops for the main spring systems within the zone of potential hydrological change. This prevented the selection of a suitable receptor impact variable for springs and consequently, it was not possible to build a receptor impact model for the ’Springs’ landscape group.