Reduction in area of the southeastern temperate grasslands of Australia since European settlement has been accompanied by degradation of remaining remnants by various factors, including the replacement of native plant species by introduced ones. There are suggestions that these replacements have had deleterious effects on the invertebrate grassland community, but there is little evidence to support these suggestions. In the eastern Adelaide Hills of South Australia, four grassland invertebrate sampling areas, in close proximity, were chosen to be as similar as possible except for the visible amount of native grass they contained. Sample areas were surveyed in four periods (summer, winter, spring, and a repeat summer) using pitfall traps and sweep-netting. A vegetation cover survey was conducted in spring. Morphospecies richness and Fisher’s alpha were compared and showed significant differences between sample areas, mainly in the summer periods. Regression analyses between morphospecies richness and various features of the groundcover/surface showed a strong positive and logical association between native grass cover and morphospecies richness. Two other associations with richness were less strong and lacked a logical explanation. If the suggested direct effect of native grass cover on invertebrate diversity is true, it has serious implications for the conservation of invertebrate biodiversity.
The southeastern temperate grasslands of Australia have been greatly affected by European settlement and agriculture. There has been a large reduction in total area of these grasslands and resulting remnants are usually small and isolated from one another [
Yen et al. [
Our knowledge of the invertebrate communities of southeastern temperate grasslands is improving; however there has been little study of the effects of the exotic plant invasion, described above, on these invertebrate communities. Generally, knowledge of the effects of exotic plants on insect communities is limited [
For specific insects/insect groups, the effects of exotic plant invasions are variable. Grasshoppers in pampas grassland were not adversely affected by such intrusions [
Given the possibility that invasive exotic plants are adversely affecting invertebrate conservation and the lack of definitive evidence, this study considered the following questions. Do sampling areas that have been deliberately chosen to have differences in relative abundance of introduced and native grasses display differences in invertebrate diversity? If there are differences, are they consistent across seasons? If there are differences between sampling areas, can they be attributed to different degrees of replacement of native grasses?
The survey was conducted in the eastern Adelaide Hills in the Bremer River valley (Hundred of Kanmantoo; S35°00′19.4′′, E139°01′55.2′′). The survey site slopes downward towards the river in an ESE direction. Soils, predominantly clay-loams, are shallow and rock outcrops occur. The site’s elevation is 190–200 m. The area has a Mediterranean-type climate with warm to hot and dry summers and cooler winters when most of the rain falls. The long-term annual average rainfall is 468 mm at the nearest recording station (7.5 km away) and average daily maximum and minimum temperatures are 23.3°C and 9.1°C, respectively [
The site lies within the distribution of Community Type S4 (midnorth tussock grassland) [
The survey occurred between February 2007 and late February 2008, with four sampling periods: February, July, and September/October 2007 provided summer, winter, and spring sampling, respectively; February 2008 provided a repeat summer sampling.
Coincidentally, this was a period of low rainfall; the three years, 2006–8, were 20–30% below long-term average. During the sampling periods, only 12.6 mm of rain was recorded (with 7 mm being the largest single fall). The average daily maximum temperature during summer samplings was 28°C and average daily minimum 14°C. Similar recordings for winter were 14°C and 6°C and for spring 20°C and 10°C.
The work of Yen et al. [
A groundcover vegetation survey was conducted on 1-2 November 2007 in each sampling area. A point quadrat frame, 0.5 m long and with ten points (needles), was centred on 2 m intervals along transects. There were three transects per sampling area parallel to the longer side of the rectangle and placed to either side and between the two pitfall transects (see below). For each transect there were ten frames and therefore 30 per site, equivalent to 300 data points. All plants touched by a point were identified, if possible, and recorded. In addition, the nature of the ground surface reached by each point was recorded. Differences between sites, in the number of hits per frame on plants (total plant cover), were tested by ANOVA. Correlation coefficients between some of the different types of cover/surface were also determined.
Surface-active invertebrates were sampled with micropitfall (mpf) traps of diameter 43 mm. At least five days before the first sampling began, PVC sleeves with lids were inserted into the soil. The sleeves enabled the actual traps to be placed and removed with minimal soil disturbance for all sampling periods and thus minimised digging-in effects [
Because of the lack of previous surveys in South Australian grasslands, simultaneous sampling of a wide range of invertebrates was considered appropriate. It was assumed that different diameters of mpfs would create positive and negative biases towards particular invertebrate groups with no specific diameter optimal. The diameter (43 mm) chosen tends to be in the midrange of possible sizes and has been found to be efficient for sampling ants [
Invertebrates located within the grass and herb foliage were sampled by sweep-netting through the foliage (rather than above it). At each sampling area and sampling period, sweeping was done three times during the week when the micropitfalls were in place. These samplings were made as comparable and repeatable as possible by following a similar path, at a consistent pace, through each sample area. A pooter was used to remove invertebrates from the net on-site, and they were immediately placed into vials containing 70% ethyl alcohol. The consistent application of sweep-netting across sampling periods and areas enabled valid comparisons of presence/absence data. However, valid comparisons of relative abundances cannot be made because grass seeds stuck in the net made it impossible to collect all individuals of some small species. Consequently, sweep-netting abundances are only indicative.
In the laboratory, all invertebrates were removed from pitfalls and identified to major taxonomic groups (usually to order). Within these groups, individuals were judged to be similar/different on the basis of external morphology and thus separated into morphospecies. The use of morphospecies has been justified in studies of this type [
Although most morphospecies were represented by less than ten individuals [
Species richness and Fisher’s alpha (
Differences between sampling areas in morphospecies richness (totals for both capture methods) were tested by 2-way ANOVA. Each cell in the data table had a single value for total richness with cells classified according to both sample area and sampling period. Three combinations of sampling periods were tested, namely, (1) all four sampling periods, (2) the three periods of 2007, and (3) the two February (summer) periods.
Fisher’s
Possible associations were explored by principal component analyses (PCA). Strong associations identified by PCA were then examined by pairwise regression analyses with richness the dependent variable. (The small number of sampling areas made the use of multiple regression invalid.) Separate analyses were done for each sampling period and two combinations of periods, namely, the three periods in 2007 and the two summer periods. In each case pitfall and sweep-netting captures were analysed separately and in combination. ANOVA were used to determine significant regressions.
Over the four sampling periods, 13,729 individuals were captured by both methods, representing 544 different morphospecies from at least 28 orders and 143 families. These data do not include known introduced species or Collembola (>20,000 springtails were captured). More detail is given in Clay and Allen [
Table
Morphospecies richness (both capture methods) for each sampling area and period and two combinations of periodsa.
Sampling period(s) | Sampling area | Range | ||||
---|---|---|---|---|---|---|
EP | EA | WP | WA | Number | % | |
February 2007 | 88 | 61 | 102 | 76 | 41 | 67 |
July 2007 | 45 | 36 | 49 | 36 | 13 | 36 |
October 2007 | 155 | 136 | 147 | 162 | 26 | 19 |
February 2008 | 84 | 61 | 88 | 74 | 27 | 44 |
2007 (3 periods) | 230 | 185 | 235 | 219 | 50 | 27 |
February 2007 and 2008 | 131 | 97 | 150 | 123 | 53 | 55 |
Morphospecies richness (both capture methods combined) for each sampling area in each sampling period or combination of periods; EP, EA, WP, and WA are the designations of the four sampling areas.
Table
Average morphospecies richnessa for each sampling area for three combinations of sampling periods,
Combined captures of both methodsb
Sampling periods | EP | EA | WP | WA |
|
---|---|---|---|---|---|
All four periods | 93.0 | 73.5 | 96.5 | 87.0 | 5.9 < 0.02 |
|
|||||
Three periods in 2007 | 96.0 | 77.7 | 99.3 | 91.3 | 2.7 > 0.10 |
Differencesd EP & WP > EA (0.01); WA > EA (0.05) | |||||
|
|||||
Both summer periods | 86.0 | 61.0 | 95.0 | 75.0 | 22.2 < 0.015 |
Differencesd WP > EA (0.01); WP > WA (0.05); EP > EA (0.05) |
Captures from pitfalls only
Sampling periods | EP | EA | WP | WA |
|
---|---|---|---|---|---|
All four periods | 67.5 | 51.5 | 63.5 | 60.0 | 2.5 > 0.10 |
|
|||||
Three periods in 2007 | 69.7 | 53.0 | 63.3 | 61.7 | 1.3 > 0.35 |
|
|||||
Both summer periods | 66.5 | 45.5 | 69.5 | 55.5 | 9.5 < 0.05 |
Differencesd EP & WP > EA (0.05) |
Captures from sweep-netting only
Sampling periods | EP | EA | WP | WA |
|
---|---|---|---|---|---|
All four periods | 30.5 | 25.3 | 37.0 | 29.3 | 15.9 < 0.001 |
Differencesd WP > EA & WA (0.01); EP & WA > EA (0.05) | |||||
|
|||||
Three periods in 2007 | 32.3 | 28.3 | 40.3 | 32.7 | 13.2 < 0.01 |
Differencesd WP > EP, EA & WA (0.01) | |||||
|
|||||
Both summer periods | 21.0 | 17.0 | 27.0 | 21.0 | 2.5 > 0.20 |
bIf a species was captured by both methods in the same sampling area and period, it was only counted once.
dDetermined by LSD (least significant difference) with the probabilities shown.
Many morphospecies were only captured in one sampling area over all survey periods. If species are randomly distributed across the survey area, there should be an approximately equal number of these “single occurrence” species in each sampling area. Over the four survey periods there were 289 “single occurrences” as follows: EP (84), EA (50), WP (83), and WA (72). Chi-square (testing for equal numbers) was significant, 10.36 (
Alpha indices (Table
(a) Fisher’s alpha indices of diversity for each sampling area in each sampling period or combination of periods. (b) Results of Student’s
Sampling period(s) | Sampling area | |||
---|---|---|---|---|
EP | EA | WP | WA | |
February 2007 | 20.9 | 12.1 | 28.6 | 16.7 |
July 2007 | 12.7 | 9.7 | 11.6 | 7.7 |
October 2007 | 41.0 | 33.6 | 36.2 | 45.4 |
February 2008 | 20.4 | 16.0 | 22.7 | 17.9 |
2007 (3 periods) | 52.6 | 38.3 | 44.4 | 48.0 |
February 2007 and 2008 | 28.2 | 18.5 | 32.6 | 26.4 |
Sampling period(s) |
Paired comparisons of sampling areas |
|||||
---|---|---|---|---|---|---|
EP versus EA | EP versus WP | EP versus WA | EA versus WP | EA versus WA | WP versus WA | |
February 2007 | EP |
WP |
WP |
WA |
WP | |
July 2007 | ||||||
October 2007 | WA |
|||||
February 2008 | EP |
WP |
||||
2007 (3 periods) | EP |
EP |
WA |
|||
February 2007 and 2008 | EP |
WP |
WA |
WP |
Table
Pitfall abundances and totalsa for each sampling area and period.
Sampling period | EP | EA | WP | WA | Total |
---|---|---|---|---|---|
February 2007 | 604 | 532 | 430 | 496 | 2062 |
July 2007 | 79 | 67 | 75 | 54 | 275 |
October 2007 | 644 | 629 | 471 | 475 | 2219 |
February 2008 | 444 | 338 | 374 | 395 | 1551 |
Total excluding outliers | 1771 | 1566 | 1350 | 1420 | 6107 |
Total including outliers | 6058 | 1701 | 1995 | 1706 | 11460 |
The survey was only done in spring and found that total vegetation cover differed significantly between sampling areas (
Floristic composition also differed (Figure
Vegetative cover: average number of hits per point quadrat frame on the four categories of groundcover vegetation for each sampling area. The error bars indicate the standard error of the average total number of hits per frame for all sample areas.
The soil surface was usually covered by plant litter (73 to 88% cover) but in WP it only provided 50% cover with microphytic crust contributing 37% cover. At other sites microphytic crust contributed ≤10% cover. Bare soil or rock constituted 13% of the surface in EP and WA, 6% in WP, but <2% in EA. Live plant material covering the surface was uncommon: ranging from 5% in WP to 0% in EA.
Preliminary and exploratory principal components analyses (PCA) revealed that invertebrate richness and covers (average number of hits per frame) of native grass and clover had strong influences, in the same direction, on the first principal component (PC), whereas litter cover had a similarly strong but opposite effect. The cover of introduced grasses also had an opposite, but lesser, effect. Examples of these analyses are shown in Figure
Examples of plots from exploratory principal components analyses of pitfall and sweep-netting morphospecies richness in (a) February 2007 and (b) the three surveys in 2007. Richness (% richness) is the number of morphospecies, for a particular method and sampling area, as a percentage of total morphospecies for that method over all sampling areas. For each sampling area (EP, EA, etc.) percentage richness for pitfalls (P) and sweep-netting (S) is shown. Other variables are the covers of native grass (Ngrass), clovers, total plant cover (Tot cover), introduced grass (Igrass), and the cover of plant litter on the surface.
These potential associations of cover with richness were further investigated by regression analyses for pitfall, sweep-netting, and combined data. The pairwise regressions of richness (dependent variable) on total plant cover and covers of native grasses, introduced annual grasses, clover species, and surface litter were tested for significance for each sampling period and two combinations of sampling periods (two examples are given in Figure
Results of pairwise regressions of richness (dependent variable) on groundcover/surface variables. Only significant regressions (
Sampling period(s) | Groundcover/surface variable | Significant regression positive/negative | ||
---|---|---|---|---|
Pitfalls | Sweep-netting | Both methods | ||
February 2007 | Native grass cover |
+ |
+ | |
Clover cover | + | |||
|
||||
July 2007 | Native grass cover | + | ||
Clover cover | + |
|||
Introduced grass cover |
|
|||
Litter cover |
|
|||
|
||||
October 2007 | Native grass cover | + |
||
Clover cover | + |
|||
|
||||
February 2008 | Native grass cover | + |
+ |
+ |
|
||||
2007 |
Clover cover | + |
||
Litter cover |
|
|||
|
||||
February 2007 and 2008 | Native grass cover | + | ||
Clover cover | + | |||
Litter cover |
|
Examples of pairwise regressions of richness, from both capture methods, on cover of native grass. (a) February 2007 (regression ANOVA;
Correlations between all pairwise combinations of the four variables that had significant associations with invertebrate richness were used to distinguish between possible direct effects of variables on richness, as opposed to indirect effects due to secondary associations. Clover and surface litter covers had a significant negative correlation (
Comparisons of morphospecies richness between sampling areas indicate major differences between them. A numerical ranking based on richness gave the order WP > EP > WA ≥ EA, which was observed for three of the sampling periods and the two combinations of periods. Only in the October period was there a different order: WA > EP > WP > EA.
Statistical analysis of average species richness data confirmed the greater richness of P areas over A areas. Averages of the former were often significantly greater than those of the latter, but the reverse was never true. Furthermore, Fisher’s
It should be noted that differences in diversity appear to be based more on morphospecies numbers rather than differences in abundance, since there was no significant difference in the latter between sample areas averaged over sample periods.
The spring sampling period had the highest richness and
It may be argued that the lower diversity in WA and EA in summer sampling periods is not of consequence because differences tend to disappear in the spring sampling period. However, when seasonal turnover of morphospecies is considered [
There seems little doubt that there are differences in invertebrate diversity between the four sampling areas. The four areas also have considerable differences from one another with respect to total vegetative cover, floristic composition, and, to a lesser extent, the nature of the surface. Are any of these differences associated with the differences in invertebrate diversity?
PCA and regression analyses revealed a negative association between litter cover and richness but positive associations for covers of native grass and clover with richness. Have any of these factors affected richness directly?
If the extent of litter cover was negatively affecting species richness directly, it is logical to suggest that this effect would be most evident in captures of surface-active invertebrates, that is, in pitfall captures. However, there were no significant regressions between litter cover and pitfall richness but three significant ones between litter cover and sweep-netting richness. This suggests a secondary association between these two factors via another unknown factor(s) rather than any direct effect of litter cover on richness.
If native grass cover has a direct effect on richness, such an effect could be expected to be greatest in summer and winter. The groundcover/surface survey was done only in spring, but floristic composition and plant cover are assumed to be very different in summer. In the Mediterranean-type summer of the survey site, perennial native grasses usually maintain some growth providing live plant material for invertebrates. Their tussocks and dried inflorescences may also provide shelter. In spring there are many other sources of plant material suggesting that native grass will have less influence. On the other hand, the clover species are introduced annuals with little, or no, presence in summer and likely to be represented only by seedlings or small plants in winter. Their maximum influence can be expected in spring when plants are large and flowering. They also have the notable advantage of high nitrogen content which may attract invertebrates in spring and may have some lingering influence as their litter decomposes in following seasons. However, the clovers have a spreading, prostrate habit which may limit their positive influence on sweep-netting captures, since they are below the physical level of the net.
If this reasoning is accepted and if both native grass cover and clover cover are having a direct influence on richness, regression of richness (dependent variable) on native grass cover should be greatest in summer and winter and least in spring, and the opposite should be true for similar regressions involving clover cover. Also, any direct effect in the latter case should be more obvious in pitfall captures than sweep-netting captures. The regressions for native grass cover and richness follow the suggested pattern and suggest a direct effect of native grass on richness. However, regressions for clover cover and richness do not follow the suggested, opposite pattern and also show positive associations in sweep-netting captures. This does not suggest a direct effect of clover on richness but rather a secondary association, possibly because of the positive correlation between native grass and clover covers. Of course, this does not prove a direct effect for native grass, but it does show a strong and logical association between native grass cover and invertebrate morphospecies richness.
With only four sampling areas and one survey site, any conclusions must be tentative. However, the results of summer sampling show greater morphospecies richness and/or diversity in areas with greater cover of native grasses. Statistical analyses show a strong positive association between cover of native grass and richness and there is a logical explanation for this association, which suggests a direct effect of native grass cover on invertebrate richness. If this direct effect exists, the loss of native grasses has serious implications not only for plant biodiversity but also for the invertebrates, which may be one of the largest components of biodiversity in these southern temperate grasslands.
The author declares that there is no conflict of interests regarding the publication of this paper.
The author wishes to thank Annette Scanlon for her help in the field and assistance with laboratory work. The author is grateful for review of some statistical analyses by Ross Frick. Peter Allen provided critical comment on an earlier draft of this paper. The Native Grasses Resources Group of South Australia generously provided financial assistance to support this project.