1. Despite significant concern about drought impacts in Australia, there have been no broad-scale studies of drought effects on river health. A severe and prolonged drought has been acting on many streams in south eastern Australia over the past decade. EPA Victoria has undertaken rapid bioassessment (RBA) of over 250 stream reference sites since 1990, providing an opportunity for a before-after-control-impact investigation of drought related changes to macroinvertebrate indices and water quality. This study uses data from 1990 to 2004 to critically evaluate the effectiveness of using RBA methods and indices, which were designed for assessment of human impacts, for monitoring streams during drought. 2. Reference stream sites across Victoria (those with minimal anthropogenic disturbances and repeatedly sampled) were classified as being 'in drought' or 'not in drought' using the Bureau of Meteorology's rainfall deficiency definition. Four biological indices (SIGNAL, EPT, Family Richness and AUSRIVAS) were calculated for combined autumn and spring samples for edge and riffle habitats for the selected sites. 3. General linear models and paired t-tests were used to detect drought related changes to index and water quality values at state-wide and bioregional scales. Changes in taxa constancy were examined to determine which taxa were sensitive to or benefited from drought conditions. Frequency of site failure against biological objectives specified in the State Environment Protection Policy (Waters of Victoria) (herein termed 'SEPP WoV') before and during drought was also examined to detect changes in a management context. 4. Few significant changes in index values were detected for riffle habitat samples. Rates of failure against biological objectives were similar before and during drought for riffle samples. In contrast, edge habitat AUSRIVAS and SIGNAL scores were significantly reduced at the state-wide scale and most indices showed significant declines in the lower altitude forests, and foothills and coastal plains bioregions. 5. Generally, more pollution tolerant, lentic taxa replaced sensitive and flow-requiring taxa in edge samples during drought. In contrast, there were few reductions in the taxa of riffle samples during drought. However, many pool preferring, but pollution sensitive taxa occurred more frequently in riffle areas. Hence, the riffle community began to resemble that of pools and edges. This was attributed to decreased flow and increased 'lentic' habitat opportunities in riffles.

Species distribution models are widely used for stream bioassessment, estimating changes in habitat suitability and identifying conservation priorities. We tested the accuracy of three modelling strategies (single species ensemble, multi-species response and community classification models) to predict fish assemblages at reference stream segments in coastal subtropical Australia. We aimed to evaluate each modelling strategy for consistency of predictor variable selection; determine which strategy is most suitable for stream bioassessment using fish indicators; and appraise which strategies best match other stream management applications. Five models, one single species ensemble, two multi-species response and two community classification models, were calibrated using fish species presence-absence data from 103 reference sites. Models were evaluated for generality and transferability through space and time using four external reference site datasets. Elevation and catchment slope were consistently identified as key correlates of fish assemblage composition among models. The community classification models had high omission error rates and contributed fewer taxa to the ‘expected’ component of the taxonomic completeness (O/E50) index than the other strategies. This potentially decreases the model sensitivity for site impact assessment. The ensemble model accurately and precisely modelled O/E50 for the training data, but produced biased predictions for the external datasets. The multi-species response models afforded relatively high accuracy and precision coupled with low bias across external datasets and had lower taxa omission rates than the community classification models. They inherently included rare, but predictable species while excluding species that were poorly modelled among all strategies. We suggest that the multi-species response modelling strategy is most suited to bioassessment using freshwater fish assemblages in our study area. At the species level, the ensemble model exhibited high sensitivity without reductions in specificity, relative to the other models. We suggest that this strategy is well suited to other non-bioassessment stream management applications, e.g., identifying priority areas for species conservation.

Key issues with defining reference condition for stream bioassessment are (1) equivocal definitions of 'minimally disturbed' pressure criteria and wide-ranging approaches to site selection, (2) highly modified regions where near-pristine areas do not exist, leading to management decisions based on inconsistent and unquantified benchmarks and (3) costly field campaigns associated with 'extensive spatial survey' approaches. We used generalised dissimilarity modelling (GDM) to classify stream segments into ecotypes, and transparently and efficiently define candidate reference conditions for the Ecosystem Health Monitoring Program (EHMP) assessment area in south-eastern Queensland, a highly modified region with a complex biogeographic history. We modelled fish presence-absence data from 396 sites with GIS-based natural and anthropogenic predictors. Stream segments were classified into ecotypes using the GDMtransformed natural variables so that (1) reference-site selection adequately covered the b-diversity of the study area and (2) we could evaluate the validity of incorporating sites from neighbouring catchments outside of the EHMP assessment area. Relationships between selected anthropogenic variables (the river disturbance index and %stream connectivity) and fish assemblages were used to define pressure criteria and map candidate reference conditions. We conclude by describing a new framework that can be used to select indicator-specific reference sites by GDM and a stratified, probabilistic sampling design. Journal compilation Ó CSIRO 2015 www.publish.csiro.au/journals/mfr B Marine and Freshwater Research P. M. Rose et al.

1. The hydrological variability of intermittent streams means that the spatial distribution of dry-season aquatic refuges within river networks and the temporal dynamics of hydrological connectivity between them are critical for the persistence of aquatic biodiversity. Here, a new approach is demonstrated to identify surface water bodies as priority refuges for efficient conservation management of freshwater biodiversity in intermittent stream networks.2. Recently developed models of surface water extent and daily streamflow were used to represent spatio-temporal variations in hydrological connectivity and surface water persistence within river networks of eastern Australia over a 107-yr period. Using this information, systematic conservation planning was applied to prioritize aquatic areas for conservation of 25 fish species under two scenarios.One scenario identified priority refuges to complement those already occurring in protected areas, whereas the other did not consider protected area status.3. The priority networks identified concentrated on the main stems of river catchments where surface water was more likely to be persistent and aquatic refuges were more likely to be connected, but also included headwaters for rare fish species. All three set conservation targets for the 25 fish species can be met in the best solution of priority networks. Although the second scenario achieved the targets with a smaller size of priority network overall, it required more new aquatic refuges and was thus less efficient than the first scenario.4. The newly developed datasets are useful for freshwater conservation prioritization because they account for hydrological variability of intermittent streams. The systematic prioritization approach applied is transferable to other regions and freshwater taxa to identify aquatic refuges for biodiversity conservation within intermittent stream systems.

No abstract