Montreal Indicator R&D: Indicator 4.1F Testing and refinement of AUSRIVAS for detection, assessment and interpretation of changes in stream biodiversity associated with forestry operations.

An intensive stream sampling and predictive model evaluation program was conducted in the Tasmanian southern forests to evaluate the suitability of the AUSRIVAS stream bioassessment system for sustainable forest monitoring and assessment under the Montreal Indicators program. AUSRIVAS (the Australian River Assessment Scheme) was recommended as Montreal Indicator 4.1f to assess the proportion of streams in forest areas with significant variance of biological diversity from the historic range of variability. AUSRIVAS does this by allowing a formal comparison of benthic macroinvertebrate community composition at stream sites with the composition predicted from relationships between environmental variables and composition in a set of least impacted, reference streams from the same region. AUSRIVAS produces a bioassessment score, O/E, the proportion of macroinvertebrate taxa expected at a site which actually occur there. AUSRIVAS macroinvertebrate bioassessment is established as a major stream assessment tool in a range of Australian national and state environmental regulatory and reporting frameworks. It's suitability for forest stream bioassessment in Australia required evaluation due to concerns over the level of taxonomic resolution, the spatial scale and low local reference site density of existing state-wide models, possible differences in sensitivity of models based on lab vs live sorted data, and the potential for confounding in existing state AUSRIVAS models due to the inclusion of reference sites already exposed to land clearing impacts. These issues were the focus of this project.
\n
\nForest streams in the southern forests area, southern Tasmania were sampled in the catchments of the Weld, Picton, Huon, Arve, Kermandie and Esperance Rivers. Over 60 sites were classed as reference sites and sampled between one and five occasions between spring 1999 and spring 2001. Macroinvertebrate and environmental data from these sites were used to develop a number of AUSRIVAS models with combinations of family vs genus/species level identification, live vs lab sorting, presence/absence vs rank abundance data. Models were developed at three nested spatial scales - the Warra LTER (16 km2), the Southern Forests region (325 km2) and Tasmania (ca 40000 km2). The Tasmania-wide models were re-developed from existing Tasmanian macroinvertebrate data sets collected under the National River Health Program. All macroinvertebrate sampling was conducted with the same AUSRIVAS kick sampling protocol.
\n
\nA series of test sites were also sampled for assessments with these models, including streams downstream of logging coupes and road crossings, and associated unimpacted controls. Two 'impact gradients' were developed from these data and used to comparatively evaluate the performance of all 12 AUSRIVAS models developed for this project. All models performed equally well at detecting changes associated with a taxon loss impact gradient. There were significant differences in performance between models in detecting changes associated with a 'logging impact gradient'.
\n
\nModel performance was assessed on the basis of sensitivity (ability to detect a site falling outside the range for reference sites) and reliability (consistency and lack of noise in the response to the gradient). Model performance fell in the order genus/species > family; regional (LTER and Southern Forests) > state-wide (Tasmania); live-pick > lab-sort; rank abundance > presence/absence. The greatest improvement in model performance was gained by using regional vs state models and using live pick instead of lab sort processing. This latter difference was due to the inherent bias of live pick sorting toward taxa that are more sensitive to impacts from changes in habitat and water quality. The most reliable and sensitive models were the regional southern forest models based on the live-pick protocol, and either presence/absence or rank abundance data. These models were able to detect a change in community composition resulting from logging, with O/E values falling below the model A band boundaries, when 17% of the expected taxa were lost.
\n
\nThe ability of the AUSRIVAS kick-sampling method of channel riffle habitat to detect patterns in macroinvertebrate community composition that reflect 'true' patterns in assemblage composition across all habitats in a stream was also tested. Data from kick samples was compared with data collected from quantitative sampling of all mesohabitats in 10 streams of differing sizes. The 'true' pattern of family and species richness between streams was reproduced well by kick sampling. Multivariate patterns of community composition were also effectively reproduced, surprisingly at both family and species level. While kick sampling does not produce the same data and hence exactly the same inter-sample relationships, it does appear to reproduce true pattern of stream macroinvertebrate diversity and community composition across a range of forest stream types. AUSRIVAS sampling appears to be suitable for assessing broad patterns of species and family association in steep, predominantly rocky forest streams and is a suitable surrogate for sampling the entire stream assemblage.
\n
\nOverall, we conclude that AUSRIVAS macroinvertebrate bioassessment is suitable for stream assessment in sustainable forest management (SFM) under the new Australian Forestry Standard, with only minor modification and initial investment. AUSRIVAS models for SFM applications will need to be developed regionally, with an initial sampling of reference sites, preferably with combined season (autumn plus spring) data, and at genus/species level (for at least the Plecoptera, Ephemeroptera, Trichoptera, Coleoptera, and preferably the Crustaceae and Mollusca as well). The choice of live-pick vs lab-sort models will depend on the purpose of the sampling program, and be influenced by the predominant protocol in each state. However, 'early warning' and greater sensitivity appear to be provided by live-picking. This objective may also be achieved using lab-sort data if models are based on species from sensitive families only. Use of AUSRIVAS should comply with normal monitoring program design considerations.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.