This study has contributed to the development of the reference condition approach in disturbed landscapes. The reference condition approach has been an important development for the bioassessment of aquatic ecosystems by providing a practical tool for the accurate assessment of river condition. The selection of appropriate reference sites is critical to the success of the predictive model in terms of being able to distinguish between natural variation in biota and the effects of human disturbance. Capturing natural variability and explaining it is a key difference between the reference condition approach and other study designs (e.g. before/after/control/impact). Natural disturbances such as drought or bushfire can significantly alter the ecological condition of streams,and although the ecological condition of streams affected by natural drought or bushfire is part of the natural cycle,this natural variation of the ecological condition is rarely incorporated into many study designs because of a mismatch in time scales. Human disturbance has also significantly altered the condition of landscapes through the development of agriculture and urbanisation. In urban or agricultural landscapes it can be impossible to locate streams that have not been modified by human activity for use as a reference condition. This study looked at the effects of natural disturbance on the reference condition,in terms of the way natural disturbance affects the prediction of stream condition and also the incorporation of the condition of streams experiencing natural disturbance into a predictive model. Additionally this study identified an alternative benchmark for modified landscapes based on the presence of good management practices for river protection,and tested this benchmark for the assessment of streams impacted by urbanisation. Drought and bushfire regularly disturb aquatic ecosystems in Australia,and affected reference sites in the ACT and South Coast region of New South Wales in 2002 and 2003. Drought and bushfire conditions affected macroinvertebrates and environmental variables across these streams,and the majority of sites were assessed as significantly impaired using regional AUSRIVAS (AUstralian RIVers Assessment System) models. This indicated the existing reference conditions for these regions had not incorporated the ecological conditions of reference sites suffering these natural disturbances. Many of the environmental variables used to predict the condition of streams were also affected by drought or bushfire. The changes to environmental variables affected how sites were assessed in models,but the overall assessment was not significantly changed from the initial assessment that drought or bushfire had significantly impaired the ecological condition. To reduce potential assessment errors associated with changes to predictor variables an attempt was made to construct new models with changeable variables excluded. However,it was not possible to completely exclude these types of variables,and subsequent models were no better than the original models in terms of changes to predictor variables affecting the generation of expected taxa lists. The changes to environmental variables did not affect the actual assessment of site condition because although group membership probabilities were changed the probabilities of taxon occurrence did not change significantly. The different reference site groups all contained some common taxa that occurred at most sites and even when group probabilities changed this did not change the probability of these taxa occurring at a test site. For regional models,such as the ACT or NSW South Coast,changes to predictor variables may not significantly affect the assessment of site condition. Incorporating reference sites under drought conditions into a predictive model was an effective way of discriminating the effects of drought from human disturbance. The model only provided two different ecological conditions,a single drought measurement and a single non-drought measurement,so the model did not fully encompass the potential natural variability. The model has value as a starting point and was effective in distinguishing sites affected by human disturbance from sites affected by drought. Good Management Practice (GMP) for river protection is any intervention that minimises human impact on stream condition. Urban sites protected by GMP were used as an alternative benchmark to a minimally impacted reference condition. The criteria used to select reference sites were not sufficiently robust to detect a significant benefit of GMP on physical or chemical characteristics of protected sites,compared to sites without GMP. In general however,the physical and chemical condition of GMP sites was better than sites without GMP and there were significant differences in macroinvertebrate assemblages of GMP and non-GMP sites. A refinement to the site selection process is proposed to include a specific assessment of GMP effectiveness for the protection it is designed to provide. This will substantially improve the robustness of a GMP benchmark and provide a clearer picture of the factors controlling biota in urban streams protected by GMP. The GMP benchmark was developed into a predictive model for the assessment of urban stream health,and in terms of the assessment of test site condition,it did not differ significantly from a model using minimally impacted sites. The purpose of the GMP benchmark was to provide an alternative reference condition for the assessment of stream health in modified landscapes when minimally impacted sites are unavailable or provide an unattainable benchmark. The GMP reference condition as an alternative can provide an attainable and realistic benchmark. The development and application of the suggested site selection protocol will improve the robustness of the GMP benchmark and better account for natural variation in the biota and physical characteristics of the sites used to determine the reference condition.
|Date of Award||1 Jan 2007|