Study design
The ultimate aim of this thesis was to investigate the effects of post-fire succession on the avifauna of Tasmanian buttongrass moorlands in order to provide recommendations for their conservation to fire and wildlife managers. The literature is replete with studies
investigating the effects of fire on Australian fauna, and on birds in particular (for a review see Woinarski 1999a, 1999b). However, since it is clear that reported patterns and their underlying processes are both complex and highly variable in space and time (Whelan et al. 2002; Burbidge 2003; Gill and Bradstock 2003), it is critical to conduct site- and species- specific research to develop appropriate management recommendations. There was a clear management need for this study as no previous studies have explicitly focused on the fire ecology of the Tasmanian moorland avifauna, and only one (Bryant 1992) has explored the effects of fire age on the resident Ground Parrot. Accordingly, the primary aims of this Chapter were: 1) to provide a rationale for the study and describe its design and the site selection process; 2) to present background information on the study locations and sites; and 3) to collect, compile, and evaluate data on fire regimes for the study sites.
Studies with comparable aims to this one have been conducted both on the Australian mainland and abroad. The inferences that can be drawn from many of these studies are limited by a host of issues associated with study design, implementation, analyses, and reporting. A number of different study designs are often used to assess the effects of fire on birds, including designed experiments, before-after-control-impact (BACI), space-for-time (SFT), and inferential studies (Stewart-Oaten et al. 1986; Pickett 1989; Loyn 1999; Block et al. 2001), each of which has its pros and cons. Replicated manipulative field experiments are considered to be the most desirable design since they make it possible to draw strong inferences regarding responses to fire regime parameters per se and to develop a more process-based understanding of population changes (Whelan et al. 2002). BACI studies that include randomly allocated replicate burnt and control sites and adequate pre- and post-fire surveying can likewise provide a basis for strong inferences (Whelan 1995; see Chapter 6). Despite their obvious advantages, studies that employ such rigorous designs are exceedingly rare (e.g. Loyn et al. 2003; Woinarski 1990) due to real-world constraints such as limited time, personnel, and funding. Furthermore, it is often not feasible to impose the treatment (i.e. fire) for designed experiments at the appropriate time, location, and scale (Whelan 1995). Thus, many studies must use sites that were either burnt in the past (e.g. Meredith et
Chapter 2 Study design and site descriptions
al. 1984; McFarland 1988b) or opportunistically utilise previously surveyed sites subjected to contemporaneous fire events and accept the associated compromises in study design and applicability (e.g. Brooker and Rowley 1991; Recher 1997; Loyn 1997).
A SFT design was deemed the most appropriate and feasible option to utilise, considering the primary aims and specific constraints imposed on this study. SFT designs are commonly used within ecosystems that exhibit marked successional dynamics in relation to disturbance events, such as buttongrass moorlands, and are based on the assumption that spatial and temporal variations are homologous (Pickett 1989; Brown et al. 2002). In other words, many researchers are interested in identifying temporal trends over the medium to long term (i.e. ~ 10 to > 50 years) but do not have the ability to monitor sites over commensurate time scales. For that reason, sites of different ages (i.e. time since the last fire) are ‘substituted’ for time to create a chronosequence of sites from which retrospective temporal trends can be extrapolated (Pickett 1989). In this sense, SFT studies fall within the general class of comparative mensurative experiments, as defined by Hurlbert (1984). In a comprehensive review and annotated bibliography of Australian birds and fire by Woinarski (1999a, 1999b), he noted that a large proportion of studies used SFT or similar designs. Perhaps the most serious inherent limitation of SFT studies is the possibility of inter-site variability in abiotic (e.g. geology, topography, climate) and biotic (e.g. vegetation, disease, predation) factors confounding or compounding observed differences in populations from effects of fire regimes per se (Loyn 1999; Woinarski 1999b). Another limitation is that both population and fire regime differences may covary with these and other factors that may be either unmeasured or unknown (Whelan 1995). Specific deficiencies that have been identified for previous avian SFT studies in particular, and fire ecology studies in general, include failures to:
- incorporate adequate (or any) replication, control sites, survey timespans, and statistical analyses (Whelan 1995; Burbidge 2003);
- survey an adequate range of fire ages (Baker and Whelan 1994);
- clearly report and assess the quality of fire regime data and report the sizes and specific locations of fires, sites, replicates, and sampling units (Parr and Chown 2003);
- investigate the impacts of fire regime parameters other than just time since last fire on populations (Woinarski 1999b);
- quantify and investigate the impacts of other habitat variables (e.g. vegetation, productivity, food resources) (Smith 2000);
- and form explicit collaborations between researchers and land management authorities to facilitate the research process and implementation of management recommendations (Whelan 1995).
Chapter 2 Study design and site descriptions Despite some of these limitations and deficiencies, SFT and other mensurative studies are still deemed to be valuable and pragmatic approaches to assess post-fire changes in faunal communities, provided they are carefully designed, implemented, and interpreted (Twigg et al. 1989; Whelan 1995; Loyn 1999). Furthermore, they are viewed as a necessary
prerequisite for much needed experimental research into ecosystems, such as buttongrass moorlands, for which such baseline information does not currently exist (Saab and Powell 2005).
To address several of the above design deficiencies, this SFT study included two study locations stratified by productivity and vegetation type, and comprising a broad chronosequence of replicated fire ages. To the extent possible, sites were selected to minimise inter-site variability in abiotic and biotic factors which have been described and quantified in detail. This study was designed and implemented in close collaboration with the Biodiversity Conservation Branch (BCB) and the Parks and Wildlife Service (PWS) (Department of Primary Industries, Parks, Water and Environment), which are responsible for fauna, flora, and fire management within the Tasmanian Wilderness World Heritage Area (TWWHA). This multidisciplinary collaboration included a synchronous SFT study
conducted by BCB on the effects of fire on moorland invertebrates and was inclusive of all sites used in this study except Gingerbread Track at Lake St Clair (GIT) and Airstrip West (AIW) at Lake Pedder (Table 4) (M. Driessen unpublished data). To make the results from these studies directly applicable to current fire management practices in buttongrass moorlands, the design was based on two of the primary variables in the fuel and fire behaviour models currently used by PWS and described by Marsden-Smedley et al. (1999) (see Chapter 1).
The first variable is vegetation age (i.e. years since fire). Although the predictive models do not include other fire regime parameters such as frequency and season, they are used in the prescription guidelines. As such, they are considered herein to the extent that reliable data were available for inferences. The second variable is site productivity as it pertains to fuel accumulation rates, loads, and spatial continuity. Sites are classified in the models as either low or moderate productivity based on geological substrate. Investigating productivity as it relates to intrinsic edaphic differences between sites appears to be a unique approach for an avian fire ecology study as neither the literature review by Woinarski (1999a, 1999b) nor that conducted for this thesis identified any such previous research (see References). While not an explicit model variable, the blanket and eastern moor community groups were also considered since they are closely associated with site productivity within the study area (i.e. low and moderate, respectively), and the latter group presents major issues for fire management (e.g. high fire spread rates and flame heights) (Marsden-Smedley and Catchpole 1995b; Marsden-Smedley et al. 1999). Although productivity is obviously a continuum and can vary over both time and space, and time since fire is likewise only one
Chapter 2 Study design and site descriptions
factor in an extremely complex system, these variables have proven to be effective in modelling and predicting fuel loads and fire behaviour in buttongrass moorlands. Therefore, these two variables are used as part of the underlying paradigm of this study.