Development of AusRivAS Models for New South Wales
Monitoring River Health Initiative Technical Report Number 10
E. Turak and N. Waddell - NSW Environment Protection Authority
Environment Australia, May 2002
ISBN 0 642 54835 8 ISSN 1447-1280
About the report
Supporting document to the Australia-Wide Assessment of River Health - NSW Bioassessment Report
Development of the α-1 AusRivAS models for NSW was completed in mid-2000. The development of earlier versions i.e. the b-1 and b-2 versions of AusRivAS models provided a good basis for identifying the gaps in the spatial coverage of sampling sites and for determining the attributes needed for effective predictive modelling in NSW. Following the development of the b models, sampling site selection was guided by this. Reference sites were selected to fill in spatial gaps and a sub-set of reference sites was continually sampled in all seasons from 1994-1999. The desired attributes of the new models were as follows:
- Free of poor quality data,
- Not unduly affected by temporal variation, i.e., the new models should provide consistent results at sites (where no new disturbances occurred) for different years over the period of sampling,
- Robust to spatial variation i.e. the models should work well for river systems where few or no reference sites were located,
- Usable for all major river types in all parts of NSW,
- Sensitive to the types of disturbance which are of management interest.
The extensive data set collected from 1994 to 1998, including data from the long-term sampling sites, made it possible to develop a-version models with these desired attributes. In order to proceed with the development of new models, however, data needed to be reviewed and new approaches in model development had to be adopted. It was necessary to tighten the definition of reference condition, revise the test/reference status of all sites, remove all poor quality data and add data from new reference sites. It was also necessary to refine the procedures for defining biological groups and choosing predictor variables, and to exclude particular sites from the model development process.
The steps taken in developing the new models in NSW are listed under four major headings: data preparation, definition of biological groups, choice of predictor variables and testing the performance of the models, as explained in the report.