Jill Landsberg, Craig D. James, Stephen R. Morton, Trevor J. Hobbs, Jacqui Stol, Alex Drew and Helen Tongway
CSIRO Division of Wildlife and Ecology
Biodiversity Convention and Strategy Section of the Biodiversity Group, Environment Australia, January 1997
ISBN 0 6422 7010 4
All the sites along each gradient were chosen to lie in similar landscape positions and included a similar array of repeating landscape elements (Section 1.2). The dominant elements formed the basis for stratification of the sites for survey. At the acacia gradients, for example, the dominant landscape elements were tree groves and the surrounding matrix of more open country. The core survey units for most plant groups and animal taxa (birds were the exception) were five replicate vegetation patches representing each landscape stratum (e.g. tree grove or open country) at each of the six sites along the gradient (Fig. 188.8.131.52). Sites varied in size depending on vegetation patchiness and the spacing between patches, but were generally around 5 ha.
Ground cover was assessed by the step-point method, at 1 m intervals along 10 m transects. Twenty transects were measured at each site, ten in each landscape stratum and at least five in each of the core replicate patches representing that stratum. Transects were located centrally, along the long axis of each patch. If a core patch was large enough, an additional 10 m of transect was located in it, end-to-end with the first. If a core patch was not large enough to accommodate 20 m of transect, the additional transect was located in the nearest similar patch. Percentage ground cover was calculated for the mutually exclusive categories of:
- bare ground (subcategories of earth, rock, lichen crust)
- loose litter (subcategories of leaf, wood, dung)
- grass (subcategories of green and dry)
- forb (subcategories of green and dry)
- shrub (subcategories of green and dry leaf and wood)
For the NT mulga gradient, which was the first gradient surveyed, we did not differentiate between grasses and forbs. The categories we recognised for the NT mulga were "standing green leaf" and "standing dry leaf". For comparison with other gradients this category is treated as comparable to the category "grass", because the ground cover at this gradient was predominantly grassy.
Cover of trees and shrubs was assessed using a Bitterlich gauge, following the method described by Friedel and Chewings (1988). Counts (by species) were recorded for the extent of cover (living canopy only) for all plants of height > 50 cm. This was done from the midpoints of each the transects assessed for ground-stratum cover.
For most gradients, a comprehensive collection was made of most plant species recorded along each gradient. If good quality reproductive specimens were available, duplicates were lodged with the relevant State and National herbaria. At least one specimen of each collected species was mounted in a field herbarium, even if reproductive material were not available, and all field and subsequent identifications were referenced to this collection.
Only partial collections were made for the SA gradients; hence it has not been possible to confirm all plants identified there. Expert field identifications for these gradients were made by Des Nelson, CSIRO Senior Technical Officer (retired), Alice Springs. Expert field identifications were also provided for species collected at the WA chenopod/ acacias gradient, by John Stretch, Western Australia Department of Agriculture, Carnarvon.
Where specimens were collected, all field identifications were confirmed at the National Herbarium, Centre for Plant Biodiversity Research, Canberra. Most herbarium identifications were undertaken by a parataxonomist (J. Stol) in consultation with herbarium staff. Expert advice and identifications were provided by:
- Ted Moore, Jo Palmer, Mike Lazarides, Lyn Craven and Lawrie Adams, National Herbarium, Centre for Plant Biodiversity Research, Canberra
- Terena Lally, Western Australian Herbarium, CALM, Perth (West Australian specimens)
- Paul Wilson, Western Australian Herbarium, CALM Perth (Chenopodiaceae family)
- Ray Cranfield, Western Australian Herbarium, CALM, Perth (Asteraceae family)
- Phillip Short, Parks and Wildlife Commission of the Northern Territory (Asteraceae family)
- Robyn Barker, Botanic Gardens of Adelaide and State Herbarium, South Australia (Malvaceae family).
Species nomenclature, geographic range and indigenous status (exotic/native) follows the advice of appropriate experts (above) or the most recent reference to the species in the following recommended sources:
Harden, G. J. 1993. Flora of New South Wales Volumes 1-4. New South Wales University Press, Kensington.
Black, J. M. 1965. Flora of South Australia Parts 1-4 plus supplement. South Australian Government.
Jessop, J. 1981. Flora of Central Australia. A.H. & A.W. Reed Pty Ltd, Sydney.
Mitchell, A.A. and Wilcox, D.G. 1994. Arid Shrubland Plants of Western Australia. 2nd Edition. University of Western Australia Press, Perth.
Western Australian Department of Agriculture, 1986. Boolathana Annual Species List.
Tyler, J.P. 1987. Vegetation surveys near Lake MacLeod. Kingia 1, 49-74.
Boyland, D.E. 1984. Vegetation Survey of Queensland: South Western Queensland. Queensland Botany Bulletin No.4, Queensland Department of Primary Industries, Brisbane.
Milson, J. 1995. Plant Identification in the Arid Zone. Queensland Department of Primary Industries, Brisbane.
Neldner, V.J. 1984. Vegetation Survey of Queensland South Central Queensland. Queensland Botany Bulletin No.3, Queensland Department of Primary Industries, Brisbane.
Queensland Herbarium, 1994. Queensland Vascular Plants: Names and Distribution. Queensland Department of Environment and Heritage, Queensland Government.
Auld, B.A. and Medd, R.W. 1987. Weeds: an Illustrated Botanical Guide to the Weeds of Australia. Inkata press Pty Ltd, Melbourne.
George, A.S. (ed) 1982. Flora of Australia. Volume 4. Phytolaccaceae to Chenopodiaceae. AGPS, Canberra.
Simon, B.K. 1990. A Key to Australian Grasses. Queensland Department of Primary Industries, Brisbane.
Recent revisions, name changes and clarifications
Chapman, A.D. 1991. Australian Plant Name Index. Australian Biological Resources Study Australian Flora and Fauna Series Number 12, AGPS, Canberra.
Jacobs, S.W.L. and Everett, J. 1996. Austrostipa, a new genus, and new names for Australasian species formerly included in Stipa (Gramineae). Telopea 6, 579-595
Linder, P.H. and Verboom, G.A. 1996. Generic limits in the Rytidosperma (Danthonieae, Poaceae) complex. Telopea 6, 597-619
Morrison, S.M. and Scott, J.K. 1996. Variation in populations of Tribulus terrestris (Zygophyllaceae). 3. *Isozyme analysis. Aust. J. Bot. 44, 201-212.
Understorey plants were defined as those of height < 50cm. For most species, abundance was measured as frequency of occurrence: all species occurring in 1m² quadrats were identified and recorded for 80 quadrats at each site (40 in each landscape stratum). The quadrats were located in groups, in each of the five vegetation patches that formed the core of the survey design (Figure 184.108.40.206). In each vegetation patch eight quadrats were systematically located at 1.25 m intervals along the long axis of the patch.
In addition to plants identified in quadrats, the presence of locally rare plants (those not found in quadrats) was recorded during a 30 minute search of each landscape stratum by the two botanists who had already assessed plant frequency in the quadrats at the site. Species recorded in the rare plants search only were assigned a nominal frequency of 0.25% for the site in which they were located.
Quadrat measurements of plants in the field (both understorey and overstorey) took two people approximately six days per gradient. Herbarium identification of specimens from each gradient took one person an additional month or so, depending on the number and condition of specimens. Data entry and verification took an additional two person-weeks.
For plants growing in the overstorey (defined as plants taller than 50 cm) frequency was assessed in 100 m² (10 m x 10 m) quadrats. Twenty quadrats were measured at each site, with ten in each landscape stratum; five of these were located in the core vegetation patches (Figure 220.127.116.11) and a further five near each of these, in the nearest similar vegetation patch. Locally rare overstorey species were recorded in the same way as those in the understorey.
Pairs of duplicate soil cores (5 cm diameter x 5 cm depth) were taken to determine stores of readily germinable seed, from each of the 80 x 1m² quadrats in which understorey plants were identified at each site. Each group of eight pairs from a vegetation patch was bulked, resulting in ten pairs of samples representing each site, one from each of the core survey patches (Figure 18.104.22.168). These were later used in germination trials conducted in a glasshouse in Canberra. One of each pair was germinated at warm season temperatures, and subsequently the other pair was germinated at cool season temperatures. In each trial the soil samples were spread in a 0.5 cm layer over a base of steam-sterilised potting mix spread about 3 cm deep in germination trays. The trays were then kept moist for three months at the specified temperatures; and were rotated at approximately fortnightly intervals to minimise the influence of localised variations in glasshouse conditions. Warm season temperatures were calculated from the mean of the summer temperatures at the gradients and cool season temperatures from the mean winter temperatures at the gradients. Samples of all plant species that germinated were grown till they produced reproductive material, after which they were harvested and identified. Numbers of germinants were recorded weekly. The total number of germinants per species per sample was calculated by adding the counts obtained from the summer and winter trials.
Blank trays (10-20 per trial) containing only the steam-sterilised potting mix, were used to identify any species that may have originated from viable seed in the potting mix or sources local to the glasshouse. The following species (all exotics) germinated in blank trays as well as some trays spread with gradient soils, and were therefore excluded from analyses:
- Crepis capillaris Smooth Hawksbeard
- Cyperus eragrostis a sedge
- Cyperus rotundus Nutgrass
- Isolepis spA a sedge
- Malva neglecta Dwarf Mallow
- Malva spA a mallow
- Medicago spA a medic
- Modiola caroliniana Red-flowered Mallow
- Trifolium glomeratum a clover
Rangeland soils are usually air-dry to the depth we sampled, so disruption of in situ germination of seed during sampling, transport and storage was not usually a problem. Samples were packed in sealed plastic bags for transport to the laboratory, where the bags were opened, spread out to maximise air circulation, and stored in a dark cupboard.
Seedbank assessments were undertaken on soil samples collected from three of the gradients surveyed during 1994, when seasons were dry. Soil storage problems precluded assessment of samples collected from the fourth. (Samples from this gradient had been inadvertently stored in sealed plastic bags in a hot shed for several weeks, with unknown consequences for seed viability.) Seedbank assessments were also undertaken for two of the gradients surveyed in 1995, where seasonal conditions had been more mesic, but soils were still sufficiently dry for sampling. However, soils at the two gradients in Queensland were still moist from recent rains at the time of the 1995 surveys, and it was likely that some in situ germination had already commenced. Since sampling, transport and storage would have disrupted this to an unknown, but variable, extent, we did not sample the soils at these two gradients.
Field collection of soils for seedbank assessment took about four person-days per gradient. Summer and winter glasshouse trials for three gradients (which filled the glasshouse to capacity) took one person a minimum of seven months in the glasshouse plus approximately two months in the herbarium to identify specimens.
None of the plant groups was mutually exclusive. The greatest degree of overlap occurred between plants growing in the understorey and plants in the seedbank. In addition, some plant species that occurred predominantly in the overstorey as mature plants were occasionally recorded as juveniles in the understorey and/ or seedbank. The degree of overlap was explicitly measured (Section 2.2.2) and total numbers of species in the amalgamated data set were calculated for individual gradients and sites (Sections 2.1 and 3.2). Because the abundance of species was measured differently for each group (different numbers and sizes of quadrats, and different ways of expressing abundance) it was not possible to calculate meaningful abundances for the amalgamated data set. Instead the three groups were treated separately in statistical analyses of differences in abundance (Sections 3.2-3) and the potential for overlap or differentiation in their responses was discussed where appropriate (e.g. Sections 3.3-4).
The survey of bird species was not stratified according to landscape stratum, because birds moved freely between the vegetation patches at each site. Instead, counts were recorded for the whole site. Counts were conducted in the early morning and late afternoon, within a 200 m strip along timed transect walks of 1 km at each site. Transects were conducted four times per site over six days: twice in the morning and twice in the afternoon. All bird assessments at a gradient were undertaken by the same person to avoid observer bias. S. Morton did the assessments at the NT, NSW and Qld gradients and C. James did the assessments in SA and WA. The transects took one person about four hours per day for six days.
These animals were surveyed by pit-fall trapping in pit-trap-arrays located in each of the core vegetation patches representing each landscape stratum at a site (Figure 22.214.171.124). Each array consisted of two 20 litre buckets (40 cm tall x 30 cm diameter) buried 8 m apart, with their openings at ground level, and connected by 10 m of partly-buried drift fence. This design is recommended as particularly effective for sampling reptiles in arid Australia (Hobbs et al. 1994). Buckets were usually dug into place 1-3 months before surveys were undertaken to allow disturbed soil to settle, but they were sealed with lids that were only removed at the time of the survey. During the survey, the buckets were left open for seven nights and checked at least every morning for the presence of animals. All captured animals were marked so that re-captures could be detected, and released close to the point of capture after positive identification.
Checking pit-traps, identifying and returning captures took one to two people about five hours per day for seven days. Setting up the trap arrays took two to five people about five days in the field, plus the time taken to travel to the gradients.
Invertebrates active on the ground surface were sampled in small pit-traps containing preservative fluid. Traps were 7 cm in diameter and 12 cm deep. Two small pit-traps were located near each vertebrate pit-trap array (Figure 126.96.36.199), 5 m out from the ends of the drift fence. Invertebrate pit-traps were left open for 5 nights, after which the catch was transferred to glass jars containing 100% ethanol, for transport and storage.
Sweep netting was used to sample invertebrates active in the understorey vegetation. A sweep net consisted of a tapered calico bag, 50 cm diameter at the mouth and 90 cm long, with a 70 cm long handle. The sweeping technique was a double-handed back-and-forth sweep which was repeated 20-30 times while the sweeper walked steadily through a landscape stratum. The total number of sweeps per sample varied between gradients. The number was set at the start of each gradient's survey following trials to determine how many sweeps were needed to catch at least 20 individuals. If this was more than 30 sweeps the sample was collected in a series of walks; the maximum number was 4 walks with 20 sweeps each, resulting in 120 sweeps per sample. After each walk the catch was transferred to an extracting box (a modified plastic rubbish bin with a glass jar fitted into a hole near the base). After 15 to 30 minutes the animals in the jar were transferred to a storage vial containing 100% ethanol, and a careful search was made of the debris in the extractor box for any additional animals. Sweeps were made between 8 am and 11 am on sunny days when winds were light. Two replicate sweep samples were collected from each landscape stratum (four per site), on different days.
In the laboratory all invertebrates were sorted and counted to Class (Arachnida), Order (most) or Family (Formicidae). Identifications to lower taxonomic levels were undertaken only for pit-trap samples. Sweep net sampling had been considered desirable for sampling those invertebrates, particularly orthopterans, which are active predominantly above the ground and therefore not sampled effectively by pit traps. However, sweep captures of all target taxa, including orthopterans, were too low for statistical analyses by species to be performed (see Table 2.1.4). Sweep samples could not be combined with pit trap samples for analysis because the techniques were so different. Hence, none of the samples collected by sweep netting have been analysed to species level.
For the pit-trap samples from the first four gradients surveyed, four taxa, ants (Formicidae), springtails (Collembola), beetles (Coleoptera) and grasshoppers and crickets (Orthoptera) were sorted to species level. Ants and springtails were chosen because they have been shown to dominate the ground layer invertebrates in other arid environments in Australia and to be particularly responsive to disturbance; beetles because of their relatively high species diversity and potential as indicators; and grasshoppers because they have been shown to be influenced by grazing in other rangeland studies (Appendix 1; Greenslade and Greenslade 1989; Majer and Beeston 1996; New 1996). Few species of springtails were found, however, and grasshoppers proved difficult to identify, so these groups were not identified to species level for the second four gradients surveyed.
Initial identifications of morphospecies were undertaken by practiced technicians (H. Tongway, A. Drew), and subsequently confirmed or revised by expert entomologists (Dr Alan Andersen for ants, Dr Penelope Greenslade for springtails, Dr John Lawrence for beetles and Dr David Rentz for grasshoppers and crickets).
Pit-trapping of invertebrates took a team of two to four people approximately two days for set-up and transfer to alcohol. Sweeps occupied two people approximately one hour for three days. Laboratory sorting and identification of samples to morphospecies took two people approximately two months per gradient. Data were entered into a computer database as they were collected, but verification and checking usually took one person at least an additional week.
The significance of any relationships between cover and distance from water was tested in a series of linear regressions, with distance from water as the independent variable and each of the major categories of cover as dependent variables. Separate regressions were tested for each gradient, and each landscape stratum within each gradient. Subcategories of ground cover were tested separately, and also tested as composites grouped in the following combinations:
- Bare = bare rock + bare earth + bare lichen
- Litter = leaf litter + wood litter + dung litter
- Grass = standing green grass + standing dry grass
- Forb = standing green forb + standing dry forb
- Shrub = standing green and dry shrub leaf and wood
- Green = standing green grass + standing green forb + standing green shrub
- Dry = standing dry grass + standing dry forb + standing dry shrub.
In addition, regressions against distance from water were tested for the canopy cover of the upper layer of vegetation (trees and shrubs > 0.5 m tall), measured by Bitterlich gauge. All analyses were performed using Systat computer software (SYSTAT 1992).
To determine the nature of the variation in species richness with distance from water, the statistical significance of linear and second-order polynomial regressions was tested for each plant group or animal taxon identified to species level (understorey plants, overstorey plants, seedbank plants, birds, reptiles, small mammals, ants, springtails, beetles and grasshoppers). Distance from water was treated as the independent variable, and the number of different species per site as the dependent variable. The analyses were performed in StatView v.4.5 (Abacus Concepts 1992).
For species-rich and abundant plant groups and animal taxa a two-staged analysis routine was undertaken to determine patterns in abundance of individual species along the gradients. Firstly, correspondence analysis (Greenacre 1984) was used to determine associations between sites and the abundance of species, for each plant group or animal taxon (plants growing in the understorey, plants growing in the overstorey, plants in the seedbank, birds, reptiles and ants) and each gradient. The correspondence analysis compared the profile of abundance of each species across the gradient against the profile of all species combined for that gradient. It resulted in two-dimensional graphical displays in which species were plotted close to the sites at which they were prominent (Appendix 3). When groups of species were prominent at several sites, these sites and species were plotted close together.
The following objective procedure was adopted for grouping species and sites. Species were grouped with the site to which they were closest on the graph. A set of rules based on the distance between sites and the graphical quadrant in which they occurred was used to group sites and their corresponding species. (Sites closer than 0.1 graphical units were always grouped, sites further than 1.0 were never grouped, and sites between these extremes were grouped if the distance between consecutive sites was <0.5, or if they occurred in a separate graphical quadrant from all other sites.) The Genstat 5 computer program (Genstat 1988) was used to perform correspondence analyses and subsequent calculations to determine groupings of species and sites.
For each of the groups identified by correspondence analysis, the relationship between distance from water and the abundance of the species in that group was investigated by stepwise linear regression. Distance from water was treated as the independent variable in each analysis and "abundance" per species as the dependent variable. (For plants growing in the field, abundance was represented by frequency of occurrence; for seed bank plants and animal taxa abundance was represented by counts of individuals.) Abundance was transformed to the log of its value [frequency or count +1] to stabilise variances. The different species were treated as different levels of a factor "species" in the analyses. Terms were added to the model stepwise, and the significance of each added term was assessed from the change in the variance ratio of the summary analysis of variance. The highest order of regression fitted was:
Log (abundance) = constant + distance + distance x species
where "abundance" is frequency (understorey and overstorey plants) or count (seedbank plants and animals), "distance" is distance from water in km, and "species" is a factor with a different level for each species in the group.
If the only significant terms in the final model were "distance" and the factor "species", this indicated that each of the species had different intercepts but the same slope. That is, each of the species had characteristically different levels of abundance, but the same overall trend in change of abundance with distance. If, however, there was a significant interaction between distance and species, it indicated that the slope of the relationship between distance and abundance varied between species. Only linear terms were fitted, and only those terms that significantly improved the fit of the regression were retained in the final regressions presented. All analyses were performed in Genstat 5 (Genstat 1988).
The type of response was determined from the slope of the best-fit model. If regression models showed that the species in a correspondence group all had negative slopes they were designated "Increaser species", because the model indicated that their abundance increased with increasing proximity to water. If the species in a correspondence group all had positive slopes they were designated "Decreaser species", because the model indicated the reverse trend. The remaining species were designated "Not Determined". This was a composite group including some species that may have had zero linear slope and others that may have had quadratic responses (e.g., a tendency toward greatest abundance at middle distances); however, quadratic regressions have yet to be tested for these groups.
Some species of birds, reptiles and ants were excluded from correspondence and regressions analyses (designated "Excluded" in listings of species; Appendix 4). Bird species that were considered to be vagrants or just flying over were automatically excluded (e.g. Western Gerygone, Fork-tailed Swift). In addition any bird species for which the total count for the gradient was less than five was excluded if it was considered to be either highly mobile (e.g. Brown Falcon, Nankeen Kestrel) or so difficult to detect that its inclusion was not considered reliable (e.g. Stubble Quail). Bird species that could be reliably detected (e.g. Spiny-cheeked Honeyeater, Grey Shrike-thrush) were included, even if their count was low. Among reptiles, elapid snakes (e.g. Orange-naped Snake, Yellow-faced Whip Snake) were excluded from the analyses, because pit-traps were not considered an efficient or reliable way of surveying their abundance. Three species of ants (Iridomyrmex galbanus, I. mayri and I. spCD) were excluded from the analyses on those gradients where they were super-abundant; ie. where their total count was 10-100 times greater than the count for any other species, because their inclusion obscured any trends in all the less-abundant species. These super-abundant species did not themselves show any sign of variation with distance from water, but were usually clumped in one or two vegetation patches at a single site.
No species of plants were excluded from the analyses, because we were confident that virtually all species growing at a site were identified as present, either in quadrats or during the subsequent search for locally rare species.
For animal taxa with very few species per gradient, or with very low counts of individuals for most species, there were too few data points to make correspondence analysis meaningful. Since there were not many species with counts of 5 or more individuals per gradient, individual regression analyses were undertaken for each of the more abundant species. When regressions were significant, species were described according to "response types" based on the shape of the most significant regression curve. The regressions tested were of the form:
- log(count) = a + b1 (distance) + b2 (distance)².
Response types were designated "Increaser" when log (count) = a - b1 (distance) and "Decreaser" when log (count) = a + b1 (distance). Two additional response types were recognised to describe the shapes of curves with significant quadratic terms: "Intermediate" when the curve was hump-shaped (log (count) = a + b1 (distance) - b2 (distance)²) and "Inverted" for U-shaped curves (log (count) = a - b1 (distance) + b2 (distance)²). Species with total counts of < 5 animals at a gradient were not analysed.
Among the more abundant and diverse plant groups and animal taxa, some species were found at only one of the six sites along a gradient, often the reference site. We undertook a statistical analysis to determine the potential significance of this pattern. To overcome spatial confounding (sites near water were closer together, and therefore more likely to share species in common) we restricted the analyses to a subset of equidistant sites.
For all the gradients except NT mulga, sites 2, 4, 5 and 6 were approximately 2.5 km apart (Table 188.8.131.52). A subset of species-presence data was compiled for the diverse and abundant groups and taxa at seven gradients (excluding NT mulga), consisting of species found at at least one of the four equidistant sites. From these data, the number of species found at one only of the four sites was calculated. Generalised linear models were then undertaken for each group or taxon to test the significance of differences among sites. Variation among gradients was treated as a blocking factor, by first adding a term for gradient (a factor with seven levels), followed by a term for site, which was treated as a factor with four levels. Species numbers were expressed as proportions of the total number of species found at all four sites and analysed assuming binomial errors and a logit link function. All analyses were performed in Genstat 5 (Genstat 1988).