Half a century of change in floristics and structure in an urban grassy woodland: implications for conservation management

Study area

The Domain is situated adjacent to the Hobart CBD, the inner suburbs of Hobart and the Derwent Estuary (Fig. 1). Climatic means are somewhat meaningless, except for global comparison, as temperatures in Hobart have increased exponentially since the mid-20th century and rainfall has declined markedly in a context of high inter- and intra-annual variability (Fig. 2). However, the average annual rainfall was 559 mm between 1974 and 2022, mean monthly rainfall in the same period varied from a mean of 34 mm in the driest month (February) to 61 mm in the wettest month (August), the mean daily maximum temperature was 17.5°C, with the mean for the warmest month (January) averaging 22.4°C and the coldest month (July) averaging 12.4°C. The mean daily minimum temperature was 8.9°C, with the coldest month (July) averaging 4.9°C and the warmest (January) averaging 12.8°C.

Fig. 1.

Distribution of the 69 plots on the Queens Domain.

Fig. 2.

Variation in (a) rainfall (mm) and (b) temperature (°C) from 1964 to 2022 on the Queens Domain, with the survey years highlighted. Source: Bureau of Meteorology (bom.gov.au/tas/climate/data station 094030, accessed December 2022).

A large part of the Domain is covered by vegetation dominated by native species (Fig. 1). The major dominants are eucalypts (Eucalyptus), sheoaks (Allocasuarina) and tussock grasses in the genera Themeda and Poa (see vegetation map in Kirkpatrick (2004); vascular plant nomenclature follows de Salas and Baker (2023)) The substrate is almost entirely Jurassic dolerite, on which there have formed black prairie soils (dermosols). In the latter part of the 20th century, the Domain was inhabited by native possums (Trichosurus vulpecula (Kerr)) and Pseudocheirus peregrinus (Winge)). There were no signs or sightings of macropods and small digging marsupials during sampling between 1974 and 1994. By 2022, the Domain had been occupied by the Bennett’s wallaby (Notamacropus rufogriseus (Desmarest)), the pademelon (Thylogale billardierii (Desmarest)), the brown bandicoot (Isoodon obesulus (Shaw)) and the nationally endangered eastern barred bandicoot (Parameles gunnii (Gray)), with all being observed during sampling, as were widespread scats and diggings. This putative return to the remnant was possibly as a result of dog control in suburbia, as many people used to let their dogs roam freely at all times of day and night even in the late 20th century, a practice that has now ceased (J. B. Kirkpatrick, pers. obs.).

Many of the nature conservation values most dependent on the Domain are those associated with the nationally listed Tasmanian Themeda triandra tussock grassland. To be regarded as being in good condition, this grassland is required to have an intertussock native herb layer, as well as less than 10% tree projective foliage cover. The Domain has populations of many herbs that are listed as rare or threatened. Some occur in grasslands, many in the zone of root suppression beneath woodland and exotic trees and others in wetlands and sheoak thickets (Kirkpatrick 2004).

The Domain was grazed by domestic stock between the British invasion of 1803 and the mid-20th century. In the same period there was much illegal harvesting of trees. Trees and understoreys became less dense (Sorensen and Kirkpatrick 2021). The release from stock grazing and wood hooking in the second half of the 20th century resulted in woody thickening and a consequent increase in fire incidence (Kirkpatrick 1986, 2004; Sorensen and Kirkpatrick 2021). The Hobart City Council currently manages the vegetation of the Domain by mowing parts of it, planned burning other parts, thinning sheoaks where they have formed thickets and by widespread weeding of non-natives, particularly woody non-natives. Recently, the managers have also attempted to reintroduce native herb species, initially in cages, to parts of the Domain from which they had disappeared, but this gardening activity is so far experimental.

Data collection

In 1974, one of us (JBK) used a random number table to locate 75 quadrats within the vegetation dominated by native species on the Domain. Sixty-nine of these quadrats avoided being developed for recreation or roads by the last sampling in 2022/23 (Fig. 1). Approximate relocation in 1984, 1994 and 2000 was achieved by navigating using an aerial photograph on which all trees were visible with pinholes for each quadrat location. In 2000, photographs were taken of each quadrat and a Garmin GPS, accurate to 10 m, was used to obtain locations. The geocoordinates and photographs facilitated relocation in 2022.

The quadrats were 1 × 15 m, laid parallel to the contours. Observable vascular plant species above or in the quadrat were recorded in December. The senior author was an observer at all times. The top right-hand corner of the quadrat was used as the point in point-centred-quarter sampling (Mueller-Dombois and Ellenberg 1974) of the nearest tree in each quarter. Quarters were oriented with the quadrat. The circumference at breast height (1.3 m) and species of each tree were recorded, along with distance from the central point. The senior author was an observer at all times.

The slope at each point was measured using a clinometer and the aspect was measured using a compass, and later converted into a scale representing topographically controlled moistness (1 = north-west (the driest aspect); 2 = north and west; 3 = north-east and south-west; 4 = east and south; 5 = south-east (the moistest aspect)). Equinoxial solar radiation was calculated using the formula of Nunez (1983). Altitude was obtained from topographic maps. Whether the site was mown or not was recorded at each time. Previously published fire mapping (Kirkpatrick 1986, 2004) and the fire history layer in ListMap were used to determine the number of fires in the previous decade, fire incidence in the previous 2 years and the number of years since the last fire for each quadrat.

Data analysis

Minitab21.4. (Minitab, LLC, 2023, https://www.minitab.com) was used for all analyses. Chi-squared test was used to determine the significance of relationships among class variables. Year of sampling was related to the following: fire in the past 2 years; stems per hectare of species and tree-size classes; and the presence/absence of individual taxa. Mowing presence/absence was related to the presence/absence of individual taxa, as was fire in the past 2 years. One way ANOVA with Tukey’s test was used to determine the significance of relationships between class and continuous variables. Year of sampling was related to number of fires in the past decade, years since the last fire, and species richness variables. Linear models were created in the ‘general linear model’ routine to explain richness variables. The variable with the weakest explanatory power was successively removed until only significant variables remained. The predictor variables included in the modelling process were as follows: year; mown (yes/no), burned in the past 2 years (yes/no); years since fire; number of fires in the past decade; equinoxial radiation (MJ m⁻² day⁻¹); aspect (1–5); slope (°); and altitude (m). Year of sampling was treated as a random variable, whereas all others were fixed. The normality of the residuals was examined to determine the suitability of the models.

Fire history

The number of fires increased between 1974 and 2000. The average number of fires in the previous 10 years to sampling in 1974 was 0.6, leaping to 2.06 in 2022 (Table 1). There was much greater differentiation between years in the incidence of fires in the 2 years previous to sampling (χ² = 51.8, d.f. = 4, P < 0.001), with a range from zero in 2001 to approximately half of the plots in 1984 (Table 1). The mean number of years since the last fire was greater in 2022 than in 1974 and 1984. The apparent contradiction between this result and the high decadal fire frequency for 2022 is resolved by the increasing spread of values between 1974 and 2022 (Fig. 3).

Fig. 3.

Dot plot of time since fire at the sampling years. The blue line is the mean.

Changes in tree structure and dominance

The total number of native stems per hectare increased steadily between 1974, when it was 98, and 2023, when it was 254 (Table 2). Allocasuarina verticillata was the strongest proportionate increaser in density, followed by Bursaria spinosa, Acacia mearnsii and Eucalyptus viminalis. Eucalyptus globulus and E. pulchella did not significantly vary in density through time (Table 2). The density of stems <10 cm in diameter increased until 2000 and then decreased between 2000 and 2023 because plant growth, rather than plant death, resulted in individuals being reclassified into the 10–30 cm diameter class (Table 2). The density of stems 10–30 cm in diameter increased steadily over the period, whereas the density of those larger than 30 cm increased markedly only between 2000 and 2023 (Table 2).

Changes in species richness

Native species richness was higher than for 1974 in the period 1984–2000 (Table 3). There was no variation in the percentage of native species among years (Table 3). In the linear model for native species richness, it was depressed by mowing and was higher in moist, sheltered situations, on steeper slopes and at high altitudes. In the linear model for percentage of native species, the value was depressed by mowing and time elapsed since the last fire and higher in places burned in the previous 2 years, on steep slopes and at higher altitude (Table 4).

Non-native species richness was higher in 1984 and 1994 than in 1974 (Table 3). Mowing, time since fire, gentle slopes and moister topographic positions all resulted in higher values for non-native species richness (Table 4).

Native tree species richness increased through the period of observation (Table 3). It was negatively affected by mowing and positively affected by high fire frequencies at both the decade and 2-year time scales. It was also higher on steeper slopes and in moist topographic situations (Table 4).

Changes at the species level

Allocasuarina verticillata was the only significantly varying consistent increaser of the 67 most frequent taxa (Table 5). High frequency of A. verticillata was associated with dry and steep sites at higher altitudes, frequent fire and a lack of mowing (Table 6). The native perennial herbs Acaena echinata and Leptorhynchos squamatus and the non-native shrubby weed Chrysanthemoides monilifera were the only significantly varying consistent decreasers (Table 5). Acaena echinata was associated with moist aspects and higher altitudes and decreased with time elapsed since the last fire (Table 6). Leptorhynchos squamatus was associated with gentle slopes and moister situations and was favoured by mowing and infrequent fire (Table 6). Chrysanthemoides monilifera was associated with steep slopes and a lack of mowing (Table 6).

In total, 24 (36%) of the 67 taxa with sufficient data for chi-squared analysis did not change in frequency between years (Table 5). The number of significantly varying taxa with peak values in each year was not different between years (χ² = 5.26, d.f. = 4, P > 0.05), although there was a range from 5 in 1974 to 13 in 1984.

Most of the more common taxa responded to mowing, with the numbers approximately equal between those negatively and positively affected (Table 6). Most, but not all, native species were reduced by mowing, the reverse of non-native species. A similar pattern was observed for fire, which tended to favour more natives than non-natives (Table 6). Most taxa were concentrated in the moister areas as indicated by aspect and solar radiation incidence.

At the reserve scale, many of the rare and threatened native species that occurred on the Domain have become rarer or apparently locally extinct between 2000 and 2022. We did not relocate any populations of Lepidium hyssopifolium or Eryngium ovinum. Vittadinia cuneata has been reduced from three populations to one population. Vittadinia muelleri has been reduced to one population from a much larger number in 2000. Carex tasmanica has continued to decline, as has Scleranthus fasciculatus. Lepidium pseudotasmanicum does not appear to have changed. In contrast, Hydrocotyle laxiflora, which is found only on the Domain in Tasmania, has increased its number of populations where felled Allocasuarina verticillata has been burned.

Stability and conservation?

There were very few instances of consistent increases or decreases for the full 1974–2022 period in the variables we examined. A large proportion of the steady increases related to A. verticillata, which increased in stems per hectare, in frequency in quadrats, in stems of 10–30 cm diameter per hectare, and in total stems, indicating a dramatic change towards closed-forest and bigger individuals over much of the study area. The only other tree species that steadily increased in its stem density was Bursaria spinosa, a species that has also occupied the understorey in Cumberland Plain Woodland where fire is infrequent (Watson et al. 2009; Watson and Morris 2020). Both trees resprout after fire from their bases and are eaten by native animals (Hazeldine and Kirkpatrick 2015). Rabbits can also eliminate A. verticillata seedlings and resprouts (Cooke 1988), but the low numbers of rabbits on the Domain have not evinced such an effect. The combination of fire with either subsequent intense grazing by stock and/or natives (Kirkpatrick et al. 2007) or mechanical stem thinning (Kirkpatrick and Jenkinson 2022) appears to be necessary in eastern Tasmania to prevent woody thickening. The increase of A. verticillata and B. spinosa on the Domain occurred after stock grazing was removed (Sorensen and Kirkpatrick 2021) and before a population of macropods re-established in numbers commensurate with similar environments outside the city.

A very recent mechanical thinning program run by the Hobart City Council (Jeffery and Smith 2019) has been effective in reversing thickening where implemented (Kirkpatrick and Jenkinson 2022), but not in reversing thickening back to its 2000 levels on the Domain as a whole. In the absence of mowing or thinning, the succession that is taking place is towards a closed-forest largely consisting of A. verticillata, except in the topographically moistest areas, where Dodonaea viscosa appears likely to become the dominant in the type of closed-forest sometimes labelled as ‘dry rainforest’ (Pollard 2006), as it has done in a long unburned quarry reserve (Fig. 4). The understorey dominance of these small tree species is often followed by dieback of the overstorey eucalypts (Kirkpatrick and Marks 1985; Fig. 4).

Fig. 4.

Vegetation of fenced quarry on western slopes of the Domain (bottom right-hand (SE) corner 94/55 5253704N, 526342E), an area unburned for at least 60 years. Dodonaea viscosa closed-forest is bottle green. Allocasuarina verticillata closed-forest is greyish-green. The dead upper branches of dead eucalypts are visible throughout.

Given that both D. viscosa and A. verticillata will regenerate in small tree-fall gaps when stands are mature (Kirkpatrick 1995; Pollard 2006), they fit one of the major criteria for rainforest, namely, self-replacement without exogenous disturbance (Bowman 2008). Thus, if this transition to closed-forest were to be allowed, the vegetation would become much more stable, but probably also much more native-species poor, given that A. verticillata excludes most ground-stratum species when in its thicket phase. The same phenomenon of understorey species depletion occurs with rainforest invasion in Queensland (Baker et al. 2020) and Leptospermum scoparium invasion of grassy woodland in the Grampians, Victoria (Price and Morgan 2008). Understorey thickets of A. verticillata are also associated with dieback of eucalypts, particularly in drought conditions (Kirkpatrick 1997). Eucalypts cannot regenerate in dry rainforest gaps (Pollard 2006).

The three variables that steadily decreased in magnitude between 1974 and 2022 were all species frequencies. Two of the species, Leptorhynchos squamatus and Acaena echinata, are common lowland grassy woodland inter-tussock herbs in Tasmania (Kirkpatrick et al. 1988). They do not survive the dense shade of Allocasuarina thickets, along with the four other perennial herbs and six perennial grasses, with their lowest frequencies occurring in 2022. The only significantly varying perennial grass that had its highest frequency in 2022 was Microlaena stipoides, which grows well in shade (Magcalemacandog and Whalley 1991). The decrease in frequency of Themeda triandra between 2000 and 2022 may have implications for non-native species invasion, as T. triandra takes up the nitrates that many non-natives thrive on (Prober and Lunt 2009). However, there are few of the non-native species that prosper with shade and intense root competition.

The third steadily decreasing species was a woody weed Chrysanthemoides monilifera, which once seemed to threaten the native understorey vegetation of the Domain (Paterson and Volframs 1976; Kirkpatrick 1986). Its numbers have been reduced by extensive weeding and more frequent burning. The recent occupation of the Domain by the pademelons and wallabies that had been absent for many decades could also have contributed to the decline of C. monilifera, because they selectively browse its leaves and small stems (Scurr et al. 2008).

Overall, there is no indication in the data that non-native species as a whole were increasing or decreasing between 1974 and 2022. Variation in non-native richness relates to recent fire history and recent climatic conditions, with non-natives seemingly favoured by longer time periods after fire as was also observed by Watson and Morris (2020) in Cumberland Plain Woodland. The observation time with the highest non-native richness was 1984 after a long drought. Non-natives also appear to have been promoted by mowing, but have attained constancy in the mown areas, which in themselves have largely been constant in location. The observations of Kirkpatrick (1986) and Fensham (1989) suggest that non-natives can co-exist in Tasmanian grassy woodland without affecting the native plants.

Management implications

Our analyses suggested that the Domain supports potentially stable synthetic vegetation sensuBridgewater (1988). The major way of realising this stability is to stabilise the abundance of small trees, particularly Allocasuarina verticillata, as declines in native herbs and grasses appear consequent on their thickening. The Hobart City Council is well on the way to stabilising the small trees with a combination of thinning and burning (Kirkpatrick and Jenkinson 2022).

Continuation of active control of woody weeds, such as Chrysanthemoides monilifera, Acacia pycnantha, Rosa rubiginosa, Cotoneaster spp. and Ulex europaeus, will also be necessary to achieve stability, as they have all demonstrated the capability to dominate the understorey on parts or all of the Domain. Urospermum dalechampii, an introduced herbaceous daisy with a fleshy root stock, has been subjected to herbiciding since 2021. Its widespread distribution on the drier parts of the Domain has not changed statistically since 1974, suggesting that it may not be as appropriate a target as the invasive non-native shrubs.

Broad floristic and structural stability may be achieved by managing the density of Allocasuarina verticillata in a variety of ways. However, the decline in listed and uncommon native herbs may require specific actions to maintain rare environmental conditions or to provide a specific set of disturbances where disseminules are still available. The unanticipated increase in Hydrocotyle laxiflora as a result of the Allocasuarina verticillata thinning program provides an educational example of unexpected outcomes.