Key words: Australia, ecological indicators, macroinvertebrates, river health ... Despite their dominance in terms of Australia's river types, dryland rivers have not ...
Hydrobiologia (2005) 552:45–56 D. Ryder, A. Boulton & P. De Deckker (eds), Conservation and Management of Australia’s Water Resources: 20/20 Vision or Blind Faith – A Tribute to the late Bill Williams DOI 10.1007/s10750-005-1504-7
Ó Springer 2005
Incorporating natural variability into the assessment of ecological health in Australian dryland rivers Fran Sheldon Centre for Riverine Landscapes, Griffith University, 4111 Nathan, Queensland, Australia E-mail: F.Sheldon@griffith.edu.au
Key words: Australia, ecological indicators, macroinvertebrates, river health
Abstract Dryland rivers occur over much of Australia’s inland and have some of the most variable patterns of flow in the world. Despite their dominance in terms of Australia’s river types, dryland rivers have not been the focus of the recent increase in research on indicators of river health, mostly owing to their spatial remoteness and relatively low levels of water resource development and human disturbance. Most rivers in arid and semi-arid regions are ephemeral, and only carry significant flows during the wetter months or following infrequent but intense rainfall events. It is not known which, if any, of the existing approaches to river health assessment can be used to accurately assess the health of these large ephemeral rivers. This paper considers why the standard methods for interpreting the currently-used indicators for river health may need to be adapted for variable systems and suggests the use of trends that recognise natural variation in indicator values for undertaking this.
Introduction Dryland rivers occur over much of Australia’s inland and have some of the most variable patterns of flow in the world (Puckridge et al., 1998; Thoms & Sheldon, 2000a). Large floods, which breach the banks and cover vast tracts of land and extensive droughts where the water in the channel dries back to a few permanent waterholes are features of these rivers. The animal and plant inhabitants of the rivers and their floodplains are well adapted to the nature of this flood-drought variability (Walker et al., 1995; Jenkins & Boulton, 2003; Arthington et al., 2005). In fact the ecological integrity of dryland rivers, particularly in lowland areas, depends upon the periodic lateral movements of water onto the floodplain (period of flooding) and the converse drying out of the channel environment (period of drought) (Walker et al., 1995).
Despite their dominance in terms of Australia’s river types (Thoms & Sheldon, 2000b), dryland rivers have not been the focus of the recent increase in research on indicators of river health, mostly owing to their spatial remoteness and relatively low levels of water resource development and human disturbance. Much of the river health assessment in Australia to date has been conducted for rivers that have year-round flows or systems for which there is a significant amount of existing data (e.g. Chessman, 1995; Bunn et al., 1999; Ladson et al., 1999; Smith et al., 1999; Smith & Storey, 2001). These tend to be in the catchments of eastern Australia and other catchments across Australia with intensive land use. Most of the rivers in the arid and semi-arid regions are ephemeral, and only carry significant flows during the wetter months, which may be winter in the southern region of the continent and summer (due to monsoonal incursions) in the
46 north, or following infrequent but intense rainfall events. It is not known which, if any, of the existing approaches to river health assessment can be used to accurately assess the health of these large ephemeral rivers. There are many definitions of ecological ‘health’ and a review of the concepts in relation to rivers can be found in Norris and Thoms (1999). ‘River health’ is usually defined in terms of ecological integrity and is used to give a measure of the overall condition of a river ecosystem. The working definition of ‘river health’ used by the Australian National River Health Program (NRHP) is: ‘The ability of the aquatic ecosystem to support and maintain key ecological processes and a community of organisms with a species composition, diversity, and functional organisation as comparable as possible to that of undisturbed habitats within the region’ (Schofield & Davies, 1996). The use of biological patterns or organisation (e.g., structure of fish and invertebrate communities) as indicators of ecological health has been widely applied within Australia (Hart et al., 1999; Smith et al., 1999; Coysh et al., 2000). Compared with water quality parameters, biological communities integrate a range of disturbances over time and give a summary of environmental conditions for the preceding period (Williams, 1980). However, focusing health assessment purely on patterns may illustrate the effect of a disturbance but will not show how a system works (Bunn et al., 1999). Measuring ecosystem processes in conjunction with patterns provides an integrated response to a broad range of catchment disturbances and also contributes to understanding how a system works (Bunn et al., 1999), which may in turn elucidate relationships between the cause and effect of a disturbance. Whether it is pattern or process that is the focus, most river health programs in Australia to date are centred on spatial comparisons without mechanisms for assessing temporal changes that are not seasonal. In dryland rivers, where parts of the same river channel may be subject to vastly different flow and flooding regimes due to the incredible variability in flow patterns, spatial comparisons among sites may be worthless. Macroinvertebrate assemblages from both the Georgina-Diamantina River system and the
Cooper Creek in the Lake Eyre Basin have been shown to vary considerably between sites with the same level of anthropogenic disturbance but markedly different hydrological connection histories (Sheldon & Puckridge, 1998; Sheldon et al., 2002). One of the key features of dryland systems is inherent variability. As high flow periods shift to dry times, there are corresponding natural declines or changes in assemblage structure as well as ecosystem process. A challenge for managing and assessing the health of these dryland systems will be to use indicators, or a method of interpretation, that has the power to detect human-induced impact over and above the enormous natural changes that are related to natural hydrological fluctuations (drought, drying and flooding). This paper explores these ideas in relation to large ephemeral rivers, and uses the rivers of the Lake Eyre Basin in central eastern Australia as an example. The ideas presented here are equally applicable to other dryland rivers in Australia, including those of the Murray-Darling Basin (e.g. Macquarie, Darling and Lachlan Rivers) and the large semi-arid rivers of the southern Gulf of Carpentaria. This paper explores why the standard methods for interpreting the currently used indicators for river health may need to be adapted for variable systems and offers an approach for undertaking this.
Assessing river health in Australia There are many river and catchment health schemes being used within Australia. They range from generic ‘Australia-wide’ schemes such as AUSRIVAS to more focused regional and catchment-based schemes. AUSRIVAS (Australian River Assessment System) is an Australia-wide rapid prediction system used to assess the biological health of Australian rivers (Schofield & Davies, 1996; Coysh et al., 2000). AUSRIVAS is similar to the British bioassessment scheme RIVPACS (Wright, 1995) except instead of a species level of taxonomic resolution, AUSRIVAS uses a rapid, standardised approach at the taxonomic level of family for assessing riverine ecological health. The macroinvertebrate AUSRIVAS models have been used for a number of years to assess the health of a range of stream and river
47 types (e.g. Marchant et al., 1997, 1999; Smith et al., 1999). AUSRIVAS uses a ‘reference condition’ approach, which relies on comparing test sites (those with likely disturbance) with ‘reference’ sites (Reynoldson et al., 1997). Reference sites may or may not be in a natural (or unimpacted) condition. In many instances, and especially in upland streams, reference condition is defined as natural. However, in many lowland rivers or entirely urban catchments, reference condition is based on ‘best available’ natural habitats (Norris & Thoms, 1999; Coysh et al., 2000). Different sources of information (e.g., historical references and expert opinion) can be used to improve the description of ‘reference’ for any given site or catchment. The use of a referential approach does not equate with returning rivers to a pristine condition. The AUSRIVAS models predict an invertebrate assemblage that is expected to occur at test sites in the absence of impact – the expected (E) value. The comparison between invertebrates predicted to occur at the test site (E) with those that were observed (O) provides a measure of ecological impairment at the test site (Coysh et al., 2000). The AUSRIVAS score is believed to be a reasonable reflection of ecosystem health in those rivers and streams with permanent and seasonal flow. However, there are conceptual difficulties in applying this approach using spatial reference sites to large dryland rivers such as those in the Lake Eyre Basin where natural condition is often exceptionally variable and natural declines in condition are frequent. SIGNAL (Stream Invertebrate Grade Number – Average Level) is another macroinvertebrate scoring system used to infer the health of Australian rivers and streams (Chessman, 2003). SIGNAL was first developed for use in the Hawkesbury-Nepean River system in New South Wales with a specific emphasis on detecting the impacts of discharges from sewage treatment plants (Chessman, 1995). The SIGNAL score gives an indication of the average sensitivity of individual taxa in terms of the water quality in the river from which the sample was collected. High SIGNAL scores suggest samples from sites with low salinity, turbidity and nutrients and possibly high levels of dissolved oxygen. When combined with taxa richness at a site, SIGNAL potentially
provides an indication of the types of pollutants and other physical and chemical factors affecting the community (Chessman, 2003). The sampling protocol for SIGNAL uses standard macroinvertebrate sampling procedures (as in the AUSRIVAS scheme) with the desired aim of collecting more than 100 macroinvertebrates from any one site, and collecting as many different types as possible. Each macroinvertebrate type recorded is then given a grade number and the number of specimens of each type recorded (Chessman, 2003). Using the abundance value for each type collected, an abundance-weighted SIGNAL score is calculated for each site. The resulting SIGNAL score is then plotted on a bi-plot against the number of taxa (families or orders) in the sample. The bi-plot is then divided into quadrants with the borders of each quadrant dependent on the geographic area and habitat type sampled. Each quadrant reflects a different ‘condition’ of the site. While AUSRIVAS and SIGNAL focus on macroinvertebrate assemblages and are applied in various situations to monitor river health across Australia, they are increasingly being used only as individual measures in broader ecosystem monitoring programs adapted for regions and catchments. One such program is the Ecosystem Health Monitoring Program (EHMP) for southeast Queensland where the health of the streams and rivers is determined using a range of biological, physical and chemical indicators (Smith & Storey, 2001). The program used the ‘disturbance gradient’ approach to select indicators, with land clearance being the primary disturbance gradient. Of the above macroinvertebrate indices, SIGNAL was originally selected with the AUSRIVAS score being further tested. The design phase of the EHMP was the first time that this approach had been used in Australia to objectively compare a range of indicators for freshwaters (Smith & Storey, 2001). The EHMP continues to use a disturbance gradient in conjunction with reference sites, scaling results against guideline values that range from minimally to heavily disturbed on an annual basis (EHMP, 2004). Another regional approach is the Index of Stream Condition (ISC), which was developed in Victoria as a tool to assist the management of Victoria’s waterways (Ladson et al., 1999). It was
48 designed to assess the ‘health’ or condition of rural streams, with results reported approximately every 5 years for stream reaches of lengths between 10 and 30 km. For each stream reach the ISC provides a summary of the extent of changes to hydrology, physical form, streamside zone, water quality and aquatic life (diversity of macroinvertebrates – the AUSRIVAS score is used). The ISC was designed to provide measures of the health of the aquatic biota and the drivers that may impact on this health. It uses a ‘synthetic reference condition’ approach where current stream condition is compared with that thought to have existed before European settlement in Australia (Ladson et al., 1999). Comprehensive river health assessment programs are still being developed for Australia’s arid rivers (Sheldon et al., 2005). However, in Queensland, the rivers of the arid Lake Eyre Basin were assessed using geomorphic and habitat data under the State of the Rivers Assessment (Moller, 1994). The aim of this assessment was to obtain data that accurately described the condition of the streams surveyed, providing a snapshot of the physical and environmental state of rivers and streams across Queensland. The assessments are related back to a generic ‘natural’ state rather than a region- or river-specific expected ‘natural’ state. When this approach was used to report the condition of channels and waterways within the Cooper Creek system, Lake Eyre Basin, the assessment suggested 95% of the channels showed low or very low channel diversity, 39% of the channels had poor or very poor riparian vegetation, and 77% of the channels had poor or very poor aquatic habitat (Moller, 1994). Overall, these assessments suggest a catchment in poor condition, despite the fact that for the indicators measured, it is probably little changed from natural condition and the low diversity is a reflection of its character as a desert river. The State of the Rivers Assessment procedure in Queensland, as currently applied and reported, is misleading for dryland rivers where lower levels of diversity (vegetation and channel, habitat) are natural phenomena as expected for these rivers. This highlights the need to recognise the high levels of variability in ‘expected’ condition in these rivers.
Assessing the health of dryland rivers in Australia Variability and temporal change: the role of resistance and resilience Lake (2003) suggests there are two kinds of drying disturbance in lotic systems. One form is the seasonal (periodic), or predictable, drought, which can be considered a ‘press’-like disturbance that elicits press responses in the biota such as high resistance and strong resilience. The other is the unpredictable aseasonal or supra-seasonal drought which has a continual increasing ‘ramp’like disturbance effect and elicits ramp-like responses, which include low to moderate resistance and variable resilience. Resilience (cf. Holling, 1973) is defined as the measure of the ability of a system to absorb change and still persist whereas resistance is a measure of the ability of a system to resist disturbance and/or persist during disturbance. While ‘press’ and ‘ramp’ may explain the response of permanent and ephemeral streams to drought, the same may not be the case for large and variable dryland rivers. These rivers are subject to both season-induced droughts as well as longer and more severe aseasonal droughts of varying frequency (see Fig. 1). Although the seasonally-induced droughts are more likely to occur during winter, the highly variable hydrology of the rivers means they can occur at any time of the year. The disturbance impact of these different drought regimes (seasonal and aseasonal) may be severe. However, the flora and fauna of dryland rivers have had a long evolutionary exposure to this disturbance regime and are adapted to rapidly recover from even quite severe droughts (see Arthington et al., 2005). Dryland rivers have variable patterns of drying; during dry times, the riverine landscape comprises dry channels and tributaries interspersed with waterholes of varying sizes and permanence. These waterholes have been termed dry-time ‘refugia’ for many organisms (Morton et al., 1995; Bunn et al., 2003; Arthington et al., 2005). Research on the fauna of these dryland rivers suggests that waterhole permanence and the landscape connection/ disconnection regime of the waterholes may influence assemblage composition (Sheldon & Puckridge, 1998; Sheldon & Waker, 1998; Sheldon
49
Figure 1. Location of the Lake Eyre Basin in Australia and the 50-year hydrograph of the Cooper Creek (monthly average discharge Gl) at the Currareva gauging station.
et al., 2002, 2003; Jenkins & Boulton, 2003; Arthington et al., 2005; Marshall et al., in press). Sheldon et al. (2003) suggested that community composition in dryland rivers reflected landscape level fluctuations in hydrology whereby falling water levels fragment the channel network with the assemblage composition of isolated waterholes diverging in a manner that reflects those species present at the time of disconnection and the biotic and abiotic selective forces at work since disconnection. Each isolated waterbody therefore may behave as a separate ‘mesocosm’ differing from other sites in ways not easily explained by measured physical and chemical variables. As time since disconnection increases, the assemblage composition of isolated waterholes may merge reflecting a community of tolerant generalists, many of which are vagile and show little habitat
specificity. In this way, assemblage composition changes dramatically in response to hydrological fluctuations (Fig. 2). In large dryland rivers, site-specific invertebrate species richness tends to be lower than in equivalent sized permanent rivers, and is often correlated with the duration of flow, water permanence, suitability of habitats and proximity to permanent refugia (Boulton et al., 2006). The invertebrate fauna is dominated by both mobile insect taxa including the Hemiptera, Coleoptera and Diptera as well as many desiccation resistant life-stages of crustaceans (Boulton et al., 2006). A similar pattern of biodiversity is seen in the fish assemblage with low levels of diversity and dominance of generalist taxa (Unmack, 2001; Arthington et al., 2005).
50
?
Connection
Disconnection
Prolonged Disconnection Figure 2. Hypothetical ordination diagram of site-based faunal assemblage composition in a dryland river. Assemblages of connected sites are similar in ordination space. After disconnection, each site diverges and behaves as a separate ‘mesocosm’. If disconnection is prolonged, the assemblage composition of isolated waterbodies may merge reflecting an assemblage of tolerant generalists. What is still unknown is whether any particular waterhole has a static disconnection assemblage (i.e., the assemblage during each disconnected phase is similar to the previous disconnection phase) or if the assemblage of an individual waterhole during each disconnected phase differs from the phase before.
The highly variable nature of dryland rivers and the dynamic changes in site-specific assemblage composition reflecting hydrological fluctuations provide a challenge in developing and interpreting measurements of ecosystem health. Most existing river health schemes measure ecosystem health spatially at one point in time (usually seasonal) and results are compared either between reference sites using seasonal models (AUSRIVAS – Coysh et al., 2000) or across a disturbance gradient (Smith & Storey, 2001). These approaches work well if assemblage composition and changes in composition vary seasonally with strong seasonal hydrological signals, which is the case for many river systems and regions. In large dryland rivers, however, assemblage composition is driven more by aseasonal hydrological changes than seasonal shifts (Puckridge, 1999; Puckridge et al., 2000; Capon, 2003; Arthington et al., 2005; Marshall et al., in press). The overriding influence of hydrological regime in dryland rivers suggests the use of seasonal models for ecosystem health assessment may be open to misinterpretation. For example, the
macroinvertebrate-based AUSRIVAS observed vs. expected (O/E) score for sites on the GeorginaDiamantina River system in the Lake Eyre Basin ranged from 0.28 (very few expected taxa – and thus assumed impacts) to 1.6 (more than expected taxa when compared to reference sites) (Fig. 3). Although pastoralism has a widespread impact over the Georgina-Diamantina catchment all sampled sites within the system are presumed to be
12 Number of Samples
Static measures of ecosystem health in dryland rivers
10 8 6 4 2 0
