The role of grass-tree Xanthorrhoea semiplana (Asphodelaceae) canopies in temperature regulation and waterproofing for ground-dwelling wildlife

Study sites

Xanthorrhoea semiplana represents a prevalent mid-stratum habitat in the Mount Lofty Ranges of South Australia (Armstrong et al. 2003), where this study took place in 2005 (Fig. 2). We worked at four sites: (1) Deep Creek National Park (Deep Creek); (2) Scott Creek Conservation Park (Scott Creek); (3) Para Wirra Conservation Park (Para Wirra); and (4) Warren Conservation Park (Warren). Deep Creek (35°39′S, 138°46′E) is located at the southern end of the Fleurieu Peninsula, 100 km south-west of Adelaide. It consists of 4554 ha of sclerophyll open forest and shrubland associations. Two open forest formations are dominated by Eucalyptus obliqua and Eucalyptus fasciculosa; common understorey species include Banksia marginata, Acacia spp., Lepidosperma semiteres, Pultenaea spp, Hibbertia spp, and X. semiplana (DENR 1997). Scott Creek (35°01′S, 138°71′E,) is located 30 km south-east of Adelaide, and supports 706 ha of native vegetation. Eucalyptus obliqua, Eucalyptus cosmophylla, and Eucalyptus baxteri are the dominant overstorey (Schram 1986); common understorey species are similar to those detailed for Deep Creek. Para Wirra (34°43′S, 138°48′E) is located 40 km north-east of Adelaide and covers 1409 ha. Nearby, Warren (34°44′S, 138°54′E) covers 365 ha. Both Para Wirra and Warren support low open E. obliqua forests over common understorey plants of Acacia pycnantha, X. semiplana, Hibbertia exutiacies, and Pultenaea largiflorens (DEP 1991). All study sites are characterised by a Mediterranean climate with warm dry summers and cool wet winters (Table 1).

Fig. 2.

Location of the study sites in South Australia.

Temperature under the canopy of grass-trees

At all four study sites, we measured temperature on the ground under single grass-trees (n = 12), which had canopies that were dense enough to cover at least 50% of the ground and were all located on the same slope and aspect. In winter at Warren n = 10 and at Scott Creek n = 11 because some data loggers failed to record. We also measured ambient temperature in the shade and the ground temperature 2 m away from each grass-tree, in a random direction. These points away from the grass-trees were either on bare ground or under other understorey shelter.

We sampled temperature once each in summer and winter, when expected diurnal temperatures in summer were above 35°C, and below 13°C in winter. Temperature was recorded every 2 h over a 24-h period, in time intervals beginning at 08:30 hours. At each grass-tree in summer, we placed an alcohol thermometer (−10°C to 60°C; accuracy ± 0.5°C) on the ground under a dense section of the grass-tree canopy that appeared to provide good cover for vertebrate wildlife, and another at a random point 2 m away selected by spinning a compass dial. We also recorded ambient temperature in the shade (other than that provided by a grass-tree) during the same period. In winter, we used Hygrochron iButtons DS1923 (Dallas Semiconductor Corp., range from −20°C to 85°C, accuracy ± 0.5°C) instead of thermometers, but followed the same procedures. We calculated the mean temperatures for the 12 (or 10 or 11) grass-trees at each time interval. We also calculated the largest difference in mean temperature between ambient and grass-tree, and 2 m away and grass-tree for each site, the temperature range for grass-trees, ambient, and 2 m away, and plotted the temperature data. As a result of the loss of most of the raw data files after compilations of the means, we are unable to present the standard errors other than for Scott Creek in winter. However, these Scott Creek data show very small standard errors. Such small standard errors also appear in Keppel et al. (2017) for grass-trees in summer and Waudby and Petit (2017) for soil cracks in summer, indicating that similar shade types are associated with similar temperatures in the same environment at the same time. We are thus confident that means are fair representations of temperatures. To determine how grass-trees may buffer animals against cold in winter and heat in summer, we compared the maximum differences between the mean temperatures under grass-trees and ambient temperatures at the four sites each season, and did the same with 2-m away temperatures (night in winter, day in summer; see Supplementary Table S1). We used a paired t-test because this test is robust at extremely small sample sizes (de Winter 2013). SPSS (IBM Corp. 2021) was used for all statistical analyses, and we calculated effect size as recommended by Field (2013). The ranges of temperatures in the three different conditions were compared with repeated-measure ANOVA and Bonferroni post-hoc tests, with partial eta squared (ηp²) indicating effect size. Temperature stability increases as the range in temperature decreases.

Soil examination during and after rainfall under grass-tree skirts

We examined the soil under X. semiplana ssp. semiplana at Scott Creek Conservation Park in June 2005. It was the wettest June between 2001 and 2021 at Longwood Alert station (023108), 5.5 km away. It rained a total of 227.4 mm for the month compared to a June mean of 127.3 mm over the 20 years of operation of the station (Bureau of Meteorology 2022c). The study took place after it had rained 141.8 mm over 13 days and during heavy rainfall (Bureau of Meteorology 2005). East of the Bandicoot Track (Gate 3), we walked along a transect line, stopping every 10 m to sample the nearest X. semiplana until we had sampled 60 individuals, 20 in each canopy cover category: 0 (<50% of the ground covered by the canopy), 1 (50–90% of the ground covered), and 2 (>90% of the ground covered; after Frazer and Petit 2007). If we had already sampled 20 grass-trees of a particular canopy cover, and the nearest grass-tree to be measured at the 10-m point had the same canopy cover, we measured the next closest grass-tree in another canopy cover category. At each grass-tree, we parted the canopy and gave the soil under the canopy a score of 0 (dry, will sift through fingers when scooped; no moist clump), 1 (partially dry with wet areas; wet clumps present when soil is scooped), or 2 (saturated). We tested the independence of canopy cover and soil dryness with a two-way chi-squared test; Cramer’s V indicated effect size. We also scored the soil at a randomly selected point 5 m away.

Summer temperatures over 24-h periods under grass-trees

Grass-trees had a large mitigating impact over hot summer temperatures. Over the 24-h period, the mean temperature range under grass-trees was smaller at Warren, Scott Creek, Deep Creek, and Para Wirra, respectively (4.6, 2.4, 1.8, and 3.8°C; overall = 3.1°C ± s.e. 0.6) than it was for 2 m away (27.3, 11.7, 16.9, and 27.3°C; overall = 20.8°C ± s.e. 3.9) and ambient (16.0, 20.0, 14.0, and 19.0°C; overall = 17.3°C ± s.e. 1.4; F_2,6 = 16.35, P = 0.004, ηp² = 0.845, Bonferroni difference only between grass-trees and 2 m away, P = 0.041 and grass-trees and ambient, P = 0.007; Figs 3 and 4). The largest mean temperature differences between 2 m away and grass-trees, and ambient and grass-trees, respectively, were: 19.8 and 8.9°C (Warren), 10.9 and 19.7°C (Scott Creek), 15.2 and 14.1°C (Deep Creek), and 19.0 and 13.7°C (Para Wirra). The largest mean difference between grass-tree canopy and ambient temperatures in summer was significant (mean = −14.1°C ± s.e. 2.2; t₃ = −6.38, one-tailed P = 0.004), with a strong effect size of 0.965. The largest mean difference between temperatures under grass-tree and 2 m away was even more dramatic (mean = −16.2 ± s.e. 2.0; t₃ = −7.96, one tailed P = 0.002; effect size = 0.975).

Fig. 3.

Mean temperatures under grass-trees (n = 12, except for Warren in winter (n = 10) and Scott Creek (n = 11)), at random positions 2 m away, and in the shade (ambient) for 24-h periods in winter and summer at Warren Conservation Park, Scott Creek Conservation Park, Deep Creek Conservation Park, and Para Wirra Recreation Park. Standard errors are shown only for Scott Creek in winter (see text).

Fig. 4.

Mean temperature ranges at the four parks studied (Warren Conservation Park, Scott Creek Conservation Park, Deep Creek Conservation Park, and Para Wirra Recreation Park) under the canopy of grass-trees Xanthorrhoea semiplana ssp. semiplana, at random locations 2 m away, and in the shade (ambient) in (a) summer and (b) winter. Although the differences among locations are significant overall for each season, pairwise comparisons are not for winter (see text).

Winter temperatures over 24-h periods under grass-trees

In winter, grass-tree canopies also provided more stable temperatures than did other areas, but the differences were not as large as they were in summer. At Warren, Scott Creek, Deep Creek, and Para Wirra, respectively we measured mean temperature ranges under grass-trees of 0.7, 4.9, 3.9, and 2.1°C (overall = 2.9°C ± s.e. 0.9), compared to 1.4, 9.7, 10.2, and 5.4°C for 2 m away (overall = 6.7°C ± s.e. 2.1), and 1.5, 8.0, 7.0, and 3.0°C (overall = 4.9°C ± s.e. 1.6) for ambient (Figs 3 and 4). Although the difference among conditions is significant (F_2,6 = 9.13, P = 0.015, ηp² = 0.753), pairwise comparisons are not, possibly because of the conservativeness of the Bonferroni test and the small sample size. At these sites, the largest mean temperature differences between grass-trees and 2 m away were 1.2, 2.5, 3.0, and 3.0°C, and they were 0.8, 3.4, 2.5, and 2.6°C between grass-trees and ambient. Conditions were warmer under grass-trees during the night. The largest difference between mean temperature under grass-trees and ambient temperature was significant (mean = 2.3°C ± s.e. 0.5; t₃ = 4.24, one-tailed P = 0.012; effect size = 0.926), as was that between temperature under grass-trees and temperature 2 m away (mean = 1.8°C ± s.e. 0.4; t₃ = 4.35, one-tailed P = 0.011; effect size = 0.929).

Sampling of soil during and after heavy rainfall

Grass-tree canopy cover and soil dryness were strongly associated (χ24 = 67.88, P < 0.001; Cramer’s V = 0.752), with z-scores indicating the significant superiority of canopy cover 2 (>90%) in keeping the soil dry, and the significantly better performance of canopy cover 1 over canopy cover 0 as well. During and after heavy rainfall, 80% of X. semiplana individuals with a canopy cover of 2 provided completely dry conditions under the trees, where the soil was dry to touch. In contrast, grass-trees that had a canopy cover of 0 provided no protection from rainfall and the soil was always saturated at the base of the tree. As for grass-trees with a canopy cover of 1, the majority (65%) had partially dry soil; the soil was driest close to the trunk and became more saturated at the edge of the canopy. Another 25% of these kept the soil completely dry, and 10% had saturated soil. Soils at the location 5 m away were predominantly saturated despite the presence of other understorey vegetation; the 3% recorded as partially dry were under large specimens of wire rapier-sedge (Lepidosperma semiteres) (Table 2).

Microclimate under grass-trees

This study demonstrated that the area under a thick X. semiplana ssp. semiplana canopy provides a remarkably stable temperatures in both winter and summer (Figs 3 and 4). In winter, grass-trees moderated cold extremes, with a smaller range of mean temperatures than found in other conditions, as determined also by Swinburn et al. (2007) for Xanthorrhoea preissii. The mean of the largest difference in temperature between grass-tree and ambient for all sites was 2.3°C, representing a significant difference, which may be consequential to determine animal activity. The difference was lower between grass-trees and 2 m away, since there was a chance that some of the 2-m away temperatures could have been taken under some protection from vegetation, but it still showed a slightly higher temperature under grass-trees (1.8°C on average). Torpor below and above minimal body temperature in animals is costly (Geiser and Broome 1993), and it is possible that the stability in temperature offered by grass-trees facilitates the maintenance of torpor in some species. Certainly, the protection of animals in torpor from predators would be enhanced by grass-trees. Species that do not use torpor could also benefit from the cover of thick skirts, the effect of which could be studied in the field with a warm animal body mimic.

The buffering role of grass-trees was strong in summer, with a maximum range for mean temperatures of 4.6°C (Warren), in comparison to random locations (27.3°C: Warren, Para Wirra) and ambient conditions in the shade (19.7°C: Scott Creek). The fact that the mean temperature at the hottest time of day for 12 grass-trees could be ~20°C lower under the grass-trees than in other shade points at their extraordinary value as shelters. For the times of largest mean differences between grass-tree and ambient temperatures for all sites, we found a mean of – 14.10°C, which demonstrates the likely contribution of grass-trees to the survival of several animal species. This difference calculated for 2 m away reached – 16.23°C, with mean potentially lethal temperatures at Warren and Para Wirra of 41.2 and 42.6°C, respectively, whereas mean temperatures under grass-trees at those sites were 21.4 and 23.6°C. Likewise, Keppel et al. (2017) found a significant impact of grass-trees in moderating high temperatures, compared to ground-below-canopy sensors. Our results cannot be compared directly to theirs because methods differed, and although they did not specify the subspecies of grass-tree, it appears from a figure that they were using Xanthorrhoea semiplana ssp. tateana. It would be interesting to compare the performance of the two subspecies, which vary in growth form, with X. semiplana ssp. tateana growing elongated stems as they age.

The Adelaide Hills and Mount Lofty Ranges are sometimes subjected to temperatures >40°C in summer. For example, Mount Crawford Bureau of Meteorology Station 023878, close to Para Wirra and Warren, the temperature reached 43.7°C in January 2019 and 42.0°C in December 2019 (Bureau of Meteorology 2022b). An increase above thermoneutral temperature can cause cognitive impairment (Soravia et al. 2021) and be lethal to ectotherms (e.g. Spellerberg 1972) and endotherms (e.g. Ikonomopoulou and Rose 2003). Maximum body temperatures in the few small marsupial species studied were in the 38–40°C range (McKechnie and Wolf 2019). Some Australian mammals avoid high ambient temperatures by sheltering in burrows during the heat of the day (Collins 1973; Walsberg 2000; Waudby and Petit 2017), but selecting appropriate shelters to escape hot weather extremes is particularly important for small mammals that do not normally burrow, such as Peramelidae (bandicoots and bilbies), Portoridae (potoroos and bettongs), and short-beaked echidnas (Tachyglossus aculeatus). We observed the entrance of vertebrate burrows hidden under the skirts of grass-trees. Echidnas radio-tracked at Warren Conservation Park sought shelter under grass-trees and in rabbit warrens (Feuerherdt 2002; L. Feuerherdt and S. Petit, unpubl. data). In Robinson’s (1954) experiments, the echidna was unable to thermoregulate at a temperature of 35°C and the southern long-nosed bandicoot (Perameles nasuta) could not endure prolonged periods at 40°C, although Brice et al. (2002) showed that in Western Queensland echidnas tolerated ambient temperatures of 35–40°C in the field and Barker et al. (2016) confirmed evaporative water loss above thermoneutrality in Tachyglossus aculeatus acanthion. The eastern barred bandicoot (Perameles gunnii) had difficulty thermoregulating above 30°C (Ikonomopoulou and Rose 2003). Seven out of eight bandicoot species have become extinct in South Australia, leaving the last one as the focus of great conservation concern. The Endangered southern brown bandicoot was observed to nest in a cluster of grass-trees at Scott Creek Conservation Park. It also sought shelter from the radio-tracker under grass-trees and foraged in a creek line with dense grass-trees (Frazer 2005). Although these observations took place in winter, they indicated that bandicoots used grass-tree habitat there, as they did in other studies (Paull 1993; Haby et al. 2013; Robinson et al. 2018).

We did not establish the impacts of grass-trees on humidity in summer, but X. preissii with skirts offered a higher and more stable relative humidity than did recently burnt ones in winter (Swinburn et al. 2007). Grass-trees may affect evaporative water loss as do burrows (e.g. Bulova 2002), and additional research would need to establish how grass-trees may minimise this loss in summer for different local species. Keppel et al. (2017) showed that at a time when relative humidity was <10% away from a grass-tree on the Fleurieu Peninsula in summer, it remained above 40% under the grass-tree, although Walsberg (2000) indicated that absolute humidity is a better index, because it sets the gradient for evaporative cooling. The risk of desiccation is associated with spatiotemporal patterns of air humidity (Rozen-Rechels et al. 2019), and humidity should be considered in future studies that examine the role of shelter plants. Air moisture under grass-trees may be critical for some animals to survive very high temperatures, since it reduces water loss in animals, which normally benefit from humidity inside burrows (Schmidt-Nielsen and Schmidt-Nielsen 1950).

Protection from rainfall

A crucial role of grass-trees in winter comes from the extraordinary capacity of the large specimens to keep the soil conditions completely dry in torrential rainfall, and undoubtedly protect animals from the wind as well. The largest, and presumably oldest, grass-trees with canopy covers of 2 provided completely dry shelter in 80% of cases, and performed significantly better than did the other categories of 0 and 1. Experiments conducted on cold and wet sheep showed a dramatic increase in plasma cortisol concentration before the final phase of hypothermia, accompanied by cardiac insufficiency (Panaretto and Vickery 1970, 1971). Such experiments have been mostly conducted on livestock, and we assume that small wildlife will seek shelter during downpours, considering potential impact of wetting on thermal protection (Webb and King 1984). The disappearance of grass-trees could limit dramatically the protection from wetness normally afforded to small ground-dwelling animals.

Grass-tree conservation in view of their role as shelters

Many vertebrates and invertebrates will use soil cracks, burrows, and other structures such as caves and rock crevices to avoid lethal temperatures or heavy rains. However, few plants other than those bearing hollows (e.g. Gibbons et al. 2002), some sedges (e.g. Happold 1976), and other dome-shaped plants (e.g. Triodia spp., Bell et al. 2021) can perform the same functions as those of grass-trees, and deliver with their foliage the remarkable stability of temperatures and waterproofing quality demonstrated for X. semiplana ssp. semiplana. Grass-trees with trunks and dense skirts also offer the benefit of a more stable microclimate than may be found away from the canopy (Swinburn et al. 2007; Keppel et al. 2017).

The availability of thermally appropriate microhabitats combined with concealment is potentially a major limiting factor affecting animal survival in degraded ecosystems, such as the Hills and Fleurieu region of Adelaide, where 41% of mammal species have already become extinct (Landscapes South Australia Hills and Fleurieu 2022). Considering the extremely high value of grass-trees as shelters for wildlife, it is likely that their availability may also affect foraging times of animals, with implications for predation.

The slow growth rate of grass-trees (e.g. Lamont et al. 2004 cited stem growth rates from 5 mm to 50 mm per year depending on species and condition) and the increasing value of the shelter they provide as they age make them particularly important to conserve. Although the resilience of some grass-tree species to bushfires is well known (e.g. Lamont et al. 2000), it may be overestimated by some. Starch loss may reduce resprouting success after fires depending on season (Korczynskyj and Lamont 2005). Tozer and Keith (2012) demonstrated that a single fire interval of 17 years resulted in the local extinction of Xanthorrhoea resinosa. Curtis (1998) showed that prescribed burning, which is now widely undertaken in South Australia, had a long-term deleterious impact on large X. australis. It also reduces tree hollow availability (Flanagan-Moodie et al. 2018) and could compound habitat loss. Southern brown bandicoots, vulnerable to predation after fires, established nests in dense eucalypt regrowth in New South Wales, when Xanthorrhoea spp. had been burnt and were unavailable (MacGregor et al. 2023); although the comparison of grass-tree and eucalypt regrowth canopy was not made, the latter is more open and likely to be inferior to that of grass-trees. The re-establishment of large skirts of dry leaves is a very slow process for Xanthorrhoea species. Skirt burning of X. semiplana ssp. tateana is undertaken by some landholders on Kangaroo Island because they fear that the grass-trees will carry bushfires. Although some invertebrates may persist even in burning grass-trees (Brennan et al. 2011), this practice is likely to have a severe impact on local animal diversity, and we suggest examining how different fire intensities modify the skirts of grass-trees. Swinburn et al. (2007) demonstrated the loss of microclimate-stabilising properties associated with the burning of grass-tree skirts.

Long-term negative impacts of Phytophthora cinnamomi, a soil fungal pathogen to which grass-trees are particularly sensitive, can be dramatic for both plant and animal communities (Wilson et al. 2020). It has resulted in significant declines in grass-trees and associated biodiversity, with little evidence of recovery post infestation (Aberton et al. 2001; Laidlaw and Wilson 2003; Lamont et al. 2004; Casey 2022). Our recent observations of the spread of the pathogen on Kangaroo Island after the 2019–2020 bushfires are alarming. Soil compaction and its negative impacts on X. preissii (Ward et al. 2011) also indicate that special care must be taken in silviculture environments.

We recommend evaluating the relevance of grass-tree microclimate, including humidity, to specific animal taxa. The effect of grass-tree skirt availability on foraging times of animals should be examined. We also recommend the study of population dynamics of X. semiplana to facilitate its conservation, particularly in relation to the highly detrimental impacts of plant losses due to P. cinammomi, and those of fire on the temporal availability of protective foliage.