Grazing management of Australian native woody regeneration as an effective nature-based climate-change solution

Major drivers of woody plant encroachment

In this review, we define woody plant encroachment (WPE) as an increase in the abundance of indigenous woody vegetation (Van Auken 2009). WPE is a defining feature not only of arid and semi-arid environments in Australia (Harrington 1979; Noble 1997), but also elsewhere across the globe (Scholes and Archer 1997; Archer et al. 2017). Where it occurs, WPE is associated with ‘land degradation’ reflected in reduced pastoral productivity (Noble 1997) and declining landscape function (Archer 2009). Although WPE is widely associated with overgrazing by domestic livestock (Archer et al. 2017), it is recognised as one of several factors driving woody recruitment and successional processes. It is therefore important to understand the role of grazing land management in the context of other drivers.

Climate is recognised as a primary driver of WPE, but reduced fire frequency and climate change also play important roles (Harrington 1979; Noble 1997; Eldridge et al. 2011; Archer et al. 2017). Witt (2013), using local knowledge from land managers in south-western Queensland, identified climatic extremes that lead to episodic recruitment and mass mortality events, with total grazing pressure and the absence of fire as the most important factors in WPE. Whereas there may be some agreement between the science and practice in the major influencing factors of WPE, they can often be interacting (Scholes and Archer 1997; Throop and Archer 2007; Witt 2013), resulting in debate over the relative importance of each, including grazing.

Multi-decadal Australian climatic extremes have been shown to result in mass tree and shrub recruitment events during periods of high rainfall (Noble 1997; Watson et al. 1997; Witt 2013; Kenny and Moxham 2022) and mass mortality during protracted droughts (Fensham et al. 2019; Godfree et al. 2019). While episodic seedling recruitment and mortality driven by wet and dry cycles may appear spectacular, ongoing ‘background’ recruitment continually occurs for many woody species, with some recruits entering the population even in dry years, but most spectacularly following heavy rains at the appropriate time (Brown 1985; Gardiner 1986; Burrows et al. 1988; Booth et al. 1996; see also Brown and Archer 1999). This is likely a ‘hedge-betting’ adaptive strategy of woody species in Australian arid (Watson et al. 1997) and semi-arid environments (Kenny and Moxham 2022).

As future climate scenarios predict an increase in the frequency of Australian droughts by as much as −6 to +64% (Jenkins and Warren 2015), the requirement for grazing management to be adaptive, responding to recruitment opportunities during post-drought recovery periods and to minimise negative impacts during periods of drought, is emphasised.

Apart from an increased frequency and intensity of climatic extremes, climate change has also resulted in increased concentration of carbon dioxide in the atmosphere. The role of elevated CO₂ (eCO₂) in driving enhanced photosynthesis and ‘greening’ trends in semi-arid areas has been highlighted (Poulter et al. 2014; Rifai et al. 2022). However it is unclear whether these trends reflect changes in woody or grass cover (Rifai et al. 2022). Van Auken (2009) noted that the period of greatest WPE in North American grasslands (1850–1870) does not coincide with the higher concentrations of eCO₂ found today and argued that eCO₂ concentrations are therefore unlikely to play a major role in the stimulation of WPE. Free-air CO₂ enrichment (FACE) studies have provided empirical evidence of the impacts of eCO₂ on tree growth, with increases of 23–30% in net primary productivity being reported for temperate forests (Norby et al. 2005; McCarthy et al. 2010). More recent FACE experiments relevant to Australian environments have shown that the magnitude of this effect is dependent on nutrient availability and can be dampened when nitrogen (Norby et al. 2010) or phosphorus (Ellsworth et al. 2017) availability is limited, a situation which is typical of Australian soils. Therefore, the magnitude of eCO₂ on tree growth in Australia remains uncertain, but is likely to have some enhancement effect.

Changes in fire regimes have fundamentally altered the structure and composition of arid and semi-arid rangelands (Harrington 1979; Noble 1997; Scholes and Archer 1997). Disturbance such as grazing-induced loss in perennial grass cover, reduces fuel loads and fire frequency, resulting in an altered competitive balance between tree and grass in favour of woody plants (Hodgkinson and Harrington 1985; Archer et al. 2017). In altering the tree–grass competition, grazing rather than being a primary cause of WPE, appears to play a modifying role (Van Auken 2009).

It is difficult to separate the effects of seasonal conditions from those resulting from grazing herbivores, either directly as a loss of perennial grass cover or indirectly through WPE (Silcock and Fensham 2013; Zhu et al. 2016; Archer et al. 2017). This has resulted in a general acceptance that processes and drivers of WPE are likely to be a result of multiple, interacting factors (Eldridge et al. 2011; Archer et al. 2017) where site condition, grazing management histories, soil, topography and landform may each modify these drivers at a local scale. For example, edaphic characteristics can control the effectiveness of rainfall where topography (slope) influences the lateral distribution of water by its influence on runoff and run-on (Wu and Archer 2005). Where grazing impacts have been assessed, this has often been undertaken without consideration of confounding effects of local abiotic and biotic factors (Milchunas and Lauenroth 1993; Maestre et al. 2022). This has led to a general acceptance that ‘robust generalisations’ around the causes of WPE are elusive (Archer et al. 2017; Mochi et al. 2022) with climate, elevated CO₂ concentrations, fire regime as well as grazing being implicated and associated with positive, neutral and negative effects on WPE (Scholes and Archer 1997; Archer et al. 2017).

Although rates of woody recruitment, dynamics and patterns of successional processes may be dependent on interacting factors, location-specific factors related to climate, fire frequency, grazing intensity and edaphic site characteristics (Eldridge et al. 2011; Muňoz-Robles et al. 2011a), as well as plant functional traits, may also dictate species-specific, local responses to the major drivers of WPE (Brown and Archer 1999; Eggemeyer and Schwinning 2009; Archer et al. 2017). This has been illustrated in contrasting results from several studies. Kenny and Moxham (2022) highlighted the importance of mass recruitment in above-average rainfall seasons where it coincides with low grazing pressure, suggesting that it is a combination of both the ideal climate conditions for recruitment and a lack of grazing pressure that set the conditions for enhanced recruitment. Watson et al. (1997) reported that low rates of background, continuous recruitment were at least as important as larger episodic recruitment events in driving population dynamics for both Eremophila maitlandii and E. forestii in arid environments.

Grazing can reduce seedling densities of woody plants directly through herbivory, which may be significant enough to change the population dynamics of the woody plants (Travers et al. 2019). Contrasts of pastoral and conservation land use (where herbivores have been excluded) have shown a 1.1% increase in non-photosynthetic vegetation (woody component as well as senescent vegetation) under conservation, providing clear evidence of a management effect in arid rangelands (Retallack et al. 2024). In this latter study, vegetation community composition and condition as well as position in landscape were both identified as important factors in determining grazing impact. Recognising and accounting for the local factors influencing woody plant regeneration is thus of central importance in understanding how grazing may influence the trajectory of woody regeneration.

Herbivore dietary selectivity

The extent to which herbivores directly affect trees and shrubs in attempting to meet their daily intake requirements will depend on the dietary preference of the herbivore, which is largely driven by the palatability of species. However, herbivore impact needs to be considered in the context of the relative availability of alternative forage, which will change with vegetation community type, season and climate variability. Pahl (2020) illustrated this well, providing a comprehensive assessment of the sequential utilisation of forage for different herbivores under contrasting seasonal conditions for major vegetation communities in southern Australian rangelands.

Pahl (2020) emphasised that many past studies of herbivore dietary composition reported the abundance of plant species in herbivore diets without accounting for the relative abundance of alternative forage, making any dietary preference hierarchy qualitative rather than quantitative. An adaptation of dietary preference hierarchy under wet and dry seasonal conditions for popular box–mulga woodlands is provided in Supplementary Fig. S1. Here, the proportion of woody vegetation in the diets of sheep and cattle may increase under increasingly dry conditions. Where palatable woody species such as mulga (Acacia aneura) are a dominant component of the vegetation, cattle diets will include moderate levels of woody vegetation (Pahl 2020). A radiocarbon dating study by Witt et al. (2000), assessing sheep diet from the mid-1930s to mid-1990s by using deposits of sheep dung from shearing sheds, showed that sheep dung contained 10–30% of Acacia (likely A.aneura and A. cambagei) over all these years, suggesting a significant intake of mulga by sheep.

Goats will ordinarily consume a greater proportion and a wider range of tree and shrub species than do either sheep or cattle, with Acacia aneura and Dodonea viscosa contributing up to 30% and 83%, of goat diets respectively (Squires 1980). However, when pasture biomass falls below 180 kg ha^–1, shrubs and trees may also contribute a large proportion of sheep (22%) and cattle (34%) diets (Squires 1980). A ~4-year grazing study in south-western Queensland also showed that during dry conditions sheep diets contained a higher proportion of mulga (35%), compared with <6% usually present in sheep diets (McMeniman et al. 1986). In the study of Kenny and Moxham (2022), arid woody species considered as ordinarily unpalatable were exposed to herbivory by rabbits when alternative forage was unavailable. The impact of rabbits and goats on woody vegetation can be exacerbated by selective grazing behaviours that suppress the seedling growth of some woody species in a targeted manner. For instance, Munro et al. (2009) found that even low densities of rabbits suppressed mulga establishment, a finding replicated in several other studies of arid-zone woody species. The selective browsing behaviour of goats in targeting the seedlings of palatable species such as mulga and cypress pine is evident in the field but undocumented; however, goat browsing of unpalatable shrubs during dry conditions has been reported (Harrington 1986). Kangaroos are likely to pose a direct threat only to woody plant regeneration in the seedling stage by trampling or herbivory (Travers et al. 2019) and only under extreme dry conditions when there is little alternative feed on offer (Pahl 2020). The consequence of selective grazing behaviour is that near total exclusion of goats and sheep (and potentially Dorper sheep, see below) may be necessary to achieve establishment of palatable woody species as even low numbers may suppress growth. The requirement for exclusion of cattle is less clear (also see below).

When there is an abundance of feed, herbivores may exercise their greatest dietary selectivity, although goats, sheep, cattle and macropods will overlap in ephemeral forbs and green perennial grasses (Pahl 2020). However, as seasonal conditions become dry, goats, sheep, rabbits and cattle will have an increasing proportion of browse in their diets (Dawson and Ellis 1994; Pahl 2020). This highlights the importance of accounting for the prevailing local seasonal conditions when considering the impacts of grazing on both woody regeneration and growth.

Herbivore grazing behaviour

Grazing impact will be determined by how far herbivores can travel to water and how often they need to water. This varies among types of herbivore, recognising that water demands are driven by temperature, animal physiology and body condition. However, the most important factor will be the availability of forage (James et al. 1999; Fukuda et al. 2009).

Waterpoints act as a focal point for herbivores. The piosphere effect reflects the radial grazing pattern created where grazing impact is greatest close to water and decreases with distance from water (James et al. 1999). Studies have shown high grazing intensity near water points results in vegetation compositional shifts from palatable perennial species to increased annual species (Friedel et al. 2003; Landsberg et al. 2003; Hendricks et al. 2005), accompanied by decreased species diversity (Hendricks et al. 2005; Todd 2006) and increased soil erosion (Tongway et al. 2003; Tabeni et al. 2014). However, results can be dependent on the size of the paddock relative to the number of water points as well as the number and type of herbivores and the distribution of land types.

The distance travelled by cattle to water can vary up to 20 km (James et al. 1999), but most time (90%) will be spent grazing within a 3 km radius (Hunt et al. 2007; Crowley 2015). Others have reported cattle grazing within a radius of ~5 km (NSW DPI (Department of Primary Industries) 2014) and up to 10 km, with 80% of the grazing occurring within 2 km of water (MLA 2024). This contrasts with sheep and goats that can forage 3 km from water (James et al. 1999; Letnic et al. 2015). Domestic livestock and feral goats are more dependent on water than are kangaroos. Red kangaroos and Euros have been reported to have a much larger range than domestic livestock and goats (James et al. 1999). Other factors will alter or distort the pattern of the piosphere effect such as the placement of fences, dietary preferences, and the amount of shade available to herbivores (James et al. 1999).

Total grazing pressure

A defining feature of the study region is that pastoralism, through grazing management practices that persistently remove ground cover, combined with changes in the density of water points, has increased total grazing pressure (TGP) from managed (domestic livestock) and unmanaged native and feral herbivores (such as feral goats, kangaroos and rabbits) substantially altering the competitive balance between trees and grass (discussed above). TGP is a defining issue in southern Australian rangelands where substantial numbers of native and feral animal populations co-exist with livestock (Fisher et al. 2004; Hacker et al 2019); at times, the grazing pressure of non-domestic herbivores can represent almost twice that exerted by livestock alone (Waters et al. 2019). The impact of total grazing pressure is not just confined to pastoral areas but also conservation areas where both introduced and native herbivores can negatively affect tree and shrub recruitment (Bond and Keeley 2005; Prowse et al. 2019). In western NSW, a dingo/wild dog barrier fence excludes this predator and, consequently, the region experiences some of the highest kangaroo and goat densities in southern Australian rangelands (Waters et al. 2019).

Recently introduced ‘exclusion’ fencing, designed primarily to provide protection from dingos, is also effective in restraining unmanaged goats and kangaroos. Such fencing is sometimes erected as a boundary around several adjacent properties forming a ‘cluster’, although in western NSW exclusion fencing is more usually erected on a property-by-property basis. Within properties, particularly those without boundary exclusion fencing, less expensive ‘TGP fencing’ is commonly erected to control the movement of unmanaged goats and, to a lesser degree, kangaroos.

Where adequate control of feral goats cannot be achieved by fencing infrastructure, traps at watering points, or self-mustering facilities established for livestock handling, are an economically attractive means of reducing total grazing pressure and are considered a socially acceptable form of control (Sinclair et al. 2019). However, the presence of kangaroos may increase TGP to the point where goats or sheep are more likely to browse establishing woody plants.

Unmanaged or feral goats, kangaroos and rabbits, all have the potential to directly affect the stages of woody plant regeneration. Effects can be expected in relation to landscape function, seedling survival, and growth of regenerating species. Grazing management aimed at increasing woody regeneration will therefore need to consider the TGP to which a site is exposed, the non-domestic component of which may be greater than the livestock component (Atkinson et al. 2019).

Hardier sheep breeds

While TGP management is a defining feature of southern Australian rangelands, the increased prevalence of hardier (sheep) breeds such as the Dorper (Alemseged and Hacker 2014), and managed goat enterprises in lieu of opportunistic harvesting of feral populations (NSW DPI (Department of Primary Industries) 2018), has resulted in relatively recent changes in the enterprise mix of these areas. Hardier sheep breeds will have dietary preferences and feed requirements different from the traditional sheep and cattle, potentially modifying grazing patterns and the ability to manipulate woody growth.

In a review of largely South African studies, Brand (2000) compared Dorper, Merino and goat grazing behaviour and dietary preferences and found that Dorpers have less selective diets, utilising a greater number of plants than do Merinos. However, Dorpers have been reported to select more shrubs in their diets (35.7%) than do Merinos (13.9%) whereas Merinos selected more grass (86.1%) than did Dorpers (63.7%) (Roux 1992, cited in Brand 2000).

In a more recent Australian study, scat analysis employing DNA barcoding found diets of Dorpers to be more diverse than those of Merinos, but reported no statistical differences in diet selectivity between the two species because of the large variation in diet selection among individual Dorpers (Lee et al. 2018). These authors also reported goat diets to be less diverse than Merino and Dorper diets; however, they noted that the study coincided with good seasonal conditions and the abundance of feed may have allowed Merinos to access a wide variety of preferred feed. Brand (2000) also reported that Dorpers utilise a wide variety of herbage and browse, which enables them to walk shorter distances to feed than do Merinos, along with consuming less herbage mass than do Merinos (on the basis of equivalent metabolic weight), which may have implications for the spatial patterns of grazing but is yet untested.

Flowering and seed set

Mulga produces mature seed pods if sufficient rainfalls occur at the right times, with peak flowering resulting from good summer or winter rain. In a 1-year irrigation experiment involving 30 mulga trees in western NSW, Preece (1971a) found that most mulga (63–100%) flowered when winter (November) or late summer (March) rains exceeded 100 mm irrespective of irrigation, although light flowering occurred at any time. Seed production and specifically maturation occurs only with summer rains. In the same experiment, Preece (1971a) found that seed pods mature only following summer (January–March) rains, provided winter rainfall (May–August) in the same year exceeds 25 mm (Davies and Kenny 2013). This flowering–seed maturation cycle may take up to 10 months (Preece 1971a; Simmons 1987) and examination of historical rainfall records since 1890 suggests that seed set could be expected once every 6 years in western NSW (Preece 1971a).

Thus, mulga appears to have relatively specific seasonal requirements for seed production. The strong climatic influence on mulga’s seed production is consistent with another study from NT, which concluded that the proportion of flowering or fruiting mulga trees is explained primarily by 6–7-month soil moisture (along with day length and temperature) (Friedel et al. 1994). Further, there is evidence of masting (profuse flowering and seeding) in mulga occurring from extremely high rainfall events (Wright and Zuur 2014). This dependence on rainfall contrasts with other regionally important woody species (Eremophila, Dodonaea and Senna), which have more seasonal flowering patterns and seed set in the winter–spring–summer period (Hodgkinson 1979; Cunningham et al. 1981).

Fire can have a profound effect on seed production and thus on the size of the seed pool available to initiate a regeneration process. Individuals that re-sprout after a fire may not produce seed for 5–10 years (Hodgkinson and Harrington 1985). Massive reductions in shrub seed production per hectare have been recorded immediately following a grass fire, with effects on seed production evident for at least 6 years (Harrington 1986). The fire history of an area being considered for a HIR project should be considered in relation to the likely size of the seed pool available to initiate the regeneration process.

Some evidence suggests that under heavy sheep stocking rates, fruit set can be affected directly through herbivory for unpalatable species, for example, Eremophila gilesii (Burrows 1973). However, the experimental conditions of this latter study are unlikely to be feasibly achieved in extensive pastoral systems and it is improbable that grazing could directly influence flowering and seed set of woody species, and thus the seed supply. It is likely that indirect effects of grazing could influence the size of the seed pool. Grazing of young regrowth may have the effect of producing smaller plants (Brown 1985), delaying the onset of flowering or the amount of seed set. However, we are not aware of published evidence that directly supports this. An indirect effect of grazing on the seed pool could also be exercised via an impact on seed predation. Mochi et al. (2022) found that grazing reduced predation on seed of the invasive woody species Vachellia caven by half in an Argentinian savanna, possibly owing to reduced refuge availability for granivores afforded by the lower biomass of grazed plots. In central Australia, seed predation by granivores reduced a massive seed pool pulse produced by a mass seeding event of slender mulga (Acacia aptaneura) in 2011, to very low pre-mast levels within 18 months (Wright and Zuur 2014). We are unaware of any published studies of the effect of grazing on seed predation, or seed pool size, for woody species in western NSW.

Germination and establishment

While rainfall can promote large seed pools in mulga, germination-ready seed pools can be considerably smaller. The initial quantity of mulga seeds depends on the amount of rainfall and is independent of tree size (diameter or height) (Wright and Fensham 2017). In a study site in the NT, a 651 mm annual rainfall (in an otherwise 280 mm MAR area) increased seed pools from 6.8 seeds per m² to 1520 seeds per m² (Wright and Zuur 2014). Seed pools may also be quickly depleted by horizontal and vertical seed movement and seed predation despite Hodgkinson (1979) suggesting it was unlikely that seed predation would have much effect on the size of future shrub populations.

A 7-year seed bank study in the NT showed that the high rainfall-mediated seed pools in the top 2 cm of soil persisted for only 18 months owing to these horizontal and vertical movements (Wright and Zuur 2014). In non-mast years, seeds collected in NSW had an average viability of 20% (Preece 1971a). Burrows et al. (1988) recommended a minimum of 40 seed trees per hectare to ensure that seed rain can cover the entire landscape, which suggests that past clearing patterns may be important in understanding the site regeneration potential.

Background germination rates in mulga average ~2%, even following a period of drought (Preece 1971b). However, any additional germination is strongly dependent on rainfall and temperatures. For example, during a drought year, <0.003 seedlings per shrub, compared with 0.8 seedlings per shrub following a wet year, have been reported (Wright and Fensham 2017; Wright et al. 2023). Mass germination events can occur following heavy rains (Hall et al. 1964). This dependence on rainfall is further supported by Anderson and Hodgkinson (1997) who found more mulga seedlings upslope (40–60%) than downslope (20–35%), reasoning that upslope areas retained more soil moisture. From observations of Preece (1971b), it is evident that follow-up rain is required within 3 months of seeding for successful mulga germination. Using historic climatic records from NSW and NT, Preece (1971b) concluded that successful germination may occur once in 8–9 years.

Fire can be a stimulant for mulga germination, but is not required for germination (Hodgkinson and Harrington 1985). In a fire-burn experiment in the NT, mulga germination rates were highest at temperatures of 70–100°C, irrespective of the heat duration (5–60 min) (Wright et al. 2016).

Average mulga establishment rates (proportion of surviving germinants) are low, ~50%; however, rates can be highly responsive to rainfall. Witt et al. (2011), in a series of old enclosure sites in south-western QLD, demonstrated that tree regeneration (<15 cm diameter) was reduced (4–7 individuals) at sites with lower mean rainfall (280–320 mm) than at those (2–182 individuals) with higher mean rainfall (420–480 mm). In a mulga shrubland in NSW, seedling densities increased from 1700 to 4200 per ha between March 1983 and September 1984, following a period of above-average rainfall (Hodgkinson and Harrington 1985). The establishment of woody seedlings (individuals <30 cm in height) in semi-arid and arid zones therefore appears to be a continuous process, with some recruits entering the population even in dry years, but is most spectacular following heavy rains at the appropriate time (Brown 1985; Gardiner 1986; Burrows et al. 1988; Booth et al. 1996; see also Brown and Archer 1999).

Although mulga establishment is primarily climate-driven, grazing can markedly reduce establishment rates. Livestock grazing may directly (herbivory or trampling) limit or prevent the recruitment and establishment of mulga and several other woody species in semi-arid areas. This has been documented in ‘several studies (Harrington 1979; Fensham et al. 2011; Witt et al. 2011). In another example, a 390-ha livestock and rabbit-proof reserve in SA showed no seedlings outside the reserve, whereas seedlings within the reserve survived and persisted for 30 years (Hall et al. 1964). Young mulga densities (<15 cm diameter) were much higher in ungrazed areas, averaging 1650 trees per hectare, than in macropod-only grazed areas with 196 trees per hectare and all herbivore grazed areas with 100 trees per hectare across three sites (Fensham et al. 2011). The effect of removal of all grazing pressure (full exclosure) on regeneration of mulga in south-western QLD was also demonstrated by Witt et al. (2011) at a series of old exclosure sites. Here, tree regeneration was reduced at sites with lower mean rainfall (290–310 mm) compared with those with higher mean rainfall (368–469 mm); however, the comparisons is complicated by the fact that two of the three lower-rainfall sites were fenced to exclude stock only rather than all herbivores.² Moore et al. (2001) concluded on the basis of a modelling exercise that suppressing fire (which reduces biomass) and grazing can let mulga develop into dense stands.

Any direct impact of grazing on woody seedlings will depend on herbivore type. Sheep (Merinos) and goats are considered to be likely to affect woody seedlings severely. The effects of Dorper sheep are probably equally as severe as those of Merinos, perhaps more so, because their dietary habits are more akin to those of goats (Alemseged and Hacker 2014). Burrows et al. (1988) recommended heavy grazing by sheep in winter to control mulga seedlings in cleared areas, implying that sheep grazing directly affects mulga seedling cohorts. Pressland (1976) reported a much greater impact of sheep than of cattle on the density and height of young regenerating mulga. Cattle appear to be relatively benign in terms of their impact on woody seedlings (although Mochi et al. (2022) observed that their impact on woody seedlings in Argentina was sufficient to offset the benefit of grazing in reducing seed predation). Fensham et al. (2012) observed a marked fence-line contrast between dense mulga and open country in the Quilpie district in south-western QLD, which may have been attributed to differences in grazing histories. Here, currently dense mulga had been subjected to moderate grazing by cattle over the course of a recruitment event in the 1970s, whereas the now open country was continuously grazed by sheep. Grazing by cattle may therefore be an appropriate management practice, depending on seasonal conditions, to reduce competing growth and lower mortality of a woody seedling cohort; however, no local studies have specifically tested this hypothesis. Evidence of no impact from grazing has also been reported. In a study at Cobar, NSW, two 30.5 m² plots were monitored four times between 1964 and 1972, one inside and one outside livestock exclosures (with sheep density of 1.7–2.8 per ha). Both plots were recorded with similar seedlings counts, two inside and one or two outside (Cunningham and Walker 1973).

Native and feral herbivores may also strongly influence establishment rates. The capacity of feral goats to destroy mulga seedlings, noted by Peeters and Butler (2014), probably extends to other woody species as well. Anecdotal evidence is available (Peeters and Butler 2014) to indicate that kangaroos may also destroy mulga seedlings, although their impact may be minimal (Travers et al. 2019), with little if any impact on larger plants (Griffiths and Baker 1966). Wilson et al. (1992) noted that rabbits ‘consume almost all seedlings of these [semi-arid zone tree] species’ (p. 64). Crisp (1978) observed that almost no regeneration of mulga occurs in the presence of rabbits. This high impact by rabbits is prevalent in other species. For example, Whipp et al. (2012) noted that in many regions of eastern Australia, Callitris recruited abundantly in the late 1800s before rabbit infestation, livestock grazing and drought, which disrupted recruitment until the 1950s when wetter conditions returned and control of rabbits by myxomatosis was achieved. Preece (1971b) concluded that germination of mulga should have occurred approximately once every 9 years in parts of western NSW in which mulga regeneration is absent. No explanation of this situation was provided but it would be reasonable to assume that herbivory by rabbits was at least partially responsible. Although rabbit abundance is currently low in the case study region due to biological control agents, vigilance against any increase in the population would need to be ongoing.

Peeters and Butler (2014) noted that plant establishment and growth may be retarded once soils have lost their A-horizon through erosion, and that restoration of mulga vegetation may be more problematic where there are large areas of bare ground. More generally, the issue concerns the extent to which landscape functionality limits the capacity of woody (or other) vegetation to germinate and successfully establish. This capacity will be limited in dysfunctional landscapes in which resources, particularly water and nutrients, are not retained within the local system. Grazing management has a central role to play in preserving functional landscapes or attempting to restore dysfunctional ones. For example, the maintenance or improvement of ground cover will be fundamental to the control of run-off and sediment production (Muňoz-Robles et al. 2011a, 2011b). Bean et al. (2015) demonstrated that landscape functionality was more important than (grass) seed availability in determining the regeneration of a degraded mulga landscape in western NSW.

Following germination, usually in the summer–autumn period, most woody seedlings can survive the first winter even if rainfall is not high. The fate of seedling cohorts, and recruitment into the established population, is primarily determined by the conditions experienced in the following summer (Booth 1986; Booth et al. 1996; Harrington 1991). If rainfall in the first summer is low, survival of seedlings may be dependent on the biomass of ground storey species competing for limited soil moisture. Alternatively, if the first summer is wet, woody seedlings may survive even in the presence of a competing ground storey. Competition from perennial grasses appears to be particularly important and was able to eliminate the entire seedling cohort of the woody narrow-leaf hopbush (Dodonaea attenuata) growing in relatively intact woollybutt (Eragrostis eriopoda) grassland (Harrington (1991), although the relative seedling drought tolerance of different woody species would probably influence this interaction. Burrows (1973), for example, demonstrated the greater drought tolerance of mulga seedlings than that of Eremophila gilesii (turkey bush). Harrington (1979) noted that removal of all herbaceous vegetation (equivalent to the destruction of perennial grasses by historic grazing as well as removal of ephemeral growth) markedly improved the survival of hopbush seedlings over summer, and regular trimming of the ground layer (equivalent to a high level of utilisation of both perennial grasses and ephemerals) significantly improved seedling survival for plots irrigated in late spring, although not for plots irrigated in early spring or autumn.

Although perennial grasses provide the strongest competition, growth of annuals in the winter following germination could also influence survival by reducing the amount of soil moisture available in the subsequent (dry) summer. In this situation, successful establishment of a woody seedling cohort may depend primarily on survival over the first summer and may be largely independent of the density of the established population. Moore et al. (2001), following Burrows (1973), did model mulga recruitment as a function of adult tree density, but noted that perennial grasses tend to out-compete mulga seedlings <0.3 m in height, so the relevance of competition from adult trees for establishment seems doubtful. This suggests that grazing can have indirect effects on germination and establishment by regulating the level of competition experienced by woody seedlings in the first summer, provided the seedlings themselves are not consumed by herbivores.

Growth

Growth in mulga can be slow. Pot experiments have shown that seedlings grow at 59.5 mg g⁻¹ day⁻¹ (Atkin et al. 1999). In another SA study, the height of mulga seedlings inside a 390-ha livestock and rabbit-proof fence was found to increase at only 1.2 cm per year (Crisp 1978). Given these slow growth rates, reproductive maturity will be attained slowly. Although there is no consistency in the records, Beadle (1948; cited in Preece 1971a) reported that it takes 10 years for mulga to reach reproductive age, whereas Preece (1971a) noted that under glasshouse conditions, reproductive maturity occurred at 18–20 months. Munro et al. (2009) reported a 190 cm mulga flowering in a livestock-proof enclosure in SA.

Water availability can accelerate growth rates, in terms of height and diameter. In a study at Wanaaring, NSW, Cunningham and Walker (1973), monitoring mulga over 7.5 years, reported that 58% of height growth was explained by rainfall. In an artificial irrigation experiment with 30 trees in western NSW, Preece (1971a) found that control trees grew 15% of their original height, whereas irrigated trees grew at a mean of 35% of their original height, with this increase being directly proportional to the frequency of irrigation. In the NT, stem-elongation rates measured over 3 years at fortnightly intervals showed that when rainfall was <50 mm, stem-elongation rates were <0.1 mm per day, but when rainfall was >50 mm, the rates increased to 0.1–0.4 mm per day (Winkworth 1973). Contrary to the expectation that elevated CO₂ concentration will increase growth rates, glasshouse experiments have shown no such response in mulga seedlings (Atkin et al. 1999; Evans et al. 2000).

Although growth is largely climate (rainfall)-driven, grazing can strikingly suppress growth rates in mulga. Within the study area, some studies have reported the effect of grazing on the growth of woody plants (Table 1). At Cobar, NSW, grazing reduced the height of mulga and pine to 53% and 39% respectively, of the growth of ungrazed plants, and maintained most of the initial population below the grazing height of sheep. For mulga, only 9 of 28 trees were able to grow beyond the ‘grazing height’ of 1 m after 7.5 years, whereas all 44 ungrazed mulgas were able to do so (Cunningham and Walker 1973). For pine, the corresponding figures were 5 of 11 for the grazed treatment and 15 of 18 for the ungrazed treatment. Sheep grazing had little impact on the unpalatable species, turpentine. In another study, ungrazed mulga at Arabella, south-western Queensland, had a growth rate comparable to that at Cobar, but the effect of grazing in suppressing height growth was even more marked; even at the lightest grazing rate (20% pasture utilisation), none of the seedlings grew in height, whereas ungrazed seedlings grew at 22 cm per year (Brown 1985). In the same study, stem diameters increased by approximately 300% in ungrazed plots but only by an average of approximately 25% under grazing. Over 90% of ungrazed trees had grown above sheep grazing height (considered to be 120 cm) after 4 years. Both goats and cattle have the capacity to access vegetative material beyond the reach of sheep and may therefore reduce growth, even if regenerating cohorts have grown beyond 120 cm. Cattle, in particular, may break down branches to access ‘leaf’ material.

At a regional scale, Fensham et al. (2012) found that mean annual rainfall (MAR) was overwhelmingly important in determining above-ground biomass of mulga dry forest, with contributing effects from mean maximum temperature (negative) and soil depth (positive). Distance to the nearest watering point, used as an index of long-term grazing pressure, was not significant. The data set included several sites, at which grazing management had clearly affected above-ground woody biomass, and these probably accounted in part for the residual variability of the model. Although it is likely that mean annual rainfall would be found to be the dominant factor influencing above-ground biomass at regional scales even if a more sensitive measure of grazing management had been available, the result is not inconsistent with other published data (Table 1) indicating considerable potential for grazing management to influence growth and biomass accumulation at the site level. A feature of all these observations is the wide variation in growth rates among individual plants, with some individuals sometimes exhibiting a decrease in height when the average growth was positive.

Mortality

Background mortality rates of mulga appear to be low, between ~0.3% and 0.58% (Crisp 1978; Davies and Kenny 2013), whereas drought-induced mortality may result in as much as ~50% mortality (Fensham et al. 2019; Godfree et al. 2019). Episodic, widespread drought-induced death of mulga and other woody species has been commonly reported for the semi-arid areas of eastern Australia (Condon 1949; Fensham and Holman 1999; Fensham et al. 2011, 2012, 2019; Godfree et al. 2019). Although climate may be a primary driver of mortality, local scale features such as position in the landscape can also modify mortality rates (Anderson and Hodgkinson 1997; Rossi et al. 2018). Here, downslope positions had greater mulga mortality rates than did upslope, particularly when grasses were present (Anderson and Hodgkinson 1997); however, both the orientation of the mulga vegetation patch and the relative proximity of bare patches appear to be important in modulating water availability (Rossi et al. 2018).

The role of grazing in adult mulga mortality is less clear. Fensham et al. (2012) found that recent tree mortality across the mulga lands of south-western QLD was related (positively or negatively) to measures of drought (3-yearly rainfall deficit), maximum temperature, soil depth and coarse sand content, but not to a grazing index represented by the distance of sites to the nearest watering point. Brown (1985), at Arabella, found no significant difference in mortality of non-seedling individuals between grazed and ungrazed populations across a wide range of grazing intensities.

Grazing can exacerbate drought-induced mortalities, especially of mulga saplings. In the study of Brown (1985), grazing had no significant direct effect on mortality but in one period of drought-related deaths, 22 of 23 deaths were recorded in grazed paddocks in which mulga growth had been strongly suppressed (Table 1). Brown considered that this suppression of growth may have contributed to mortality in dry periods by maintaining plants in a drought-susceptible state. Several studies have observed higher rates of mortality among smaller (juvenile) than larger (adult) plants. Cunningham and Walker (1973), in a study lasting 175 months at Wanaaring (within the study region), recorded higher mortality rates for plants <1 m (~2–44%) than for those >1 m (~0–17%).

Other factors that influence mortality are fire and patterns of clearing. It is repeatedly noted that mature mulga has a low fire tolerance (Cunningham et al. 1981; Latz et al. 1995) and reported rates of mortality are as high as 80–100% for adults in NSW (Hodgkinson and Harrington 1985). For susceptible species, intense fire is not required to produce tree death; less intense fires that result in scorching of all leaves are sufficient (Hodgkinson and Harrington 1985).

An indirect effect of grazing could arise through density-dependent mortality. Fensham et al. (2012) found no evidence of density-dependent effects at the regional scale in their analysis of drought-related tree deaths in south-western Queensland. However, they noted ‘widespread death’ in the dense mulga stand at Quilpie, described above. In western NSW, Cunningham and Walker (1973) found higher, drought-related mortality at their Wanaaring site than at Cobar, where tree density was much lower, and considered that ‘catastrophic death’ would appear to be a feature of more dense stands in dryer areas. Fensham et al. (2012) concluded that ‘while density-dependent mortality can occur in mulga it did not emerge as a dominant influence (at regional scales) and is probably less important than in eucalypt woodlands’ (p. 905). However, in local situations, grazing management could have the effect of allowing the development of dense stands that would be susceptible to subsequent drought.

Effect of fire on the demography of woody species

Seedlings of all woody species are susceptible to fire and the seasonal conditions that result in large germination events are likely also to produce the fuel loads capable of carrying an effective fire (900–1200 kg ha^–1, Campbell and Hacker 2000) that could destroy most of the seedlings (Hodgkinson and Harrington 1985). Management of this fuel load in a way that reduces or eliminates the risk posed by fire to the initiation of a regeneration event, while preserving much of the seedling cohort, is an essential element in the management of regeneration. The potential for use of cattle for this purpose has been discussed previously.

Once past the seedling stage, the susceptibility of woody species to fire varies greatly and has been summarised in Nolan et al. (2019); however, as noted above, mature mulga has a low fire tolerance. For susceptible species, intense fire is not required to produce tree death. Death can be expected provided all leaves are scorched (Hodgkinson and Harrington 1985).

Fig. 2.

An overview or the grazing impacts on regenerating mulga at critical stages.

Land manager skills and capability

The success of grazing management will depend on land manager skills, capacity and motivation to implement science-based strategies and can be expected to have a broad range of outcomes. For example, an evaluation of groundcover incentive programs conducted in the study area found wide variation in on-ground outcomes where land managers were financially supported to implement self-guided, improved grazing management through infrastructure upgrades and/or grazing management plans. Program evaluation that included quantification of outcomes through remotely sensed ground-cover estimates showed that of 170 projects where total grazing-pressure fencing was funded, 40% had distinctly positive outcomes (increasing trends in ground cover), 23% had mixed outcomes, 21% had neutral outcomes and 16% had poor outcomes (Local Land Services, unpubl. data, 2018). The positive outcomes demonstrated that infrastructure and grazing management could improve groundcover. The mixed or neutral outcomes potentially represented situations where management systems were inappropriate or inconsistently implemented. The negative outcomes possibly reflected situations where landholders failed to adopt management suitable for improving groundcover. This suggests that it is likely that relying on land managers to implement grazing management to facilitate woody regeneration may have uncertain outcomes. Better outcomes require ongoing support and relationship building from technical providers, flexibility in the implementation of plans and involvement of land manager in planning and evaluation processes (Swann and Richards 2016).

A regional benchmarking survey conducted in 2017 (Local Land Services, unpubl. data, 2017) showed that 60% of landholders had a ‘professional’ farming style, carefully considering alternative practices in their property management. Approximately, 50% of landholders recognised improved grazing management as a reason to make property operational changes; 35% indicated that they had undertaken recent (past 3 years) grazing/land management training. However, Ison and Russell (2000) broadly considered that land managers in the study region generally place a low value on knowledge derived off-farm. In relation to woody regeneration activities, this suggests that although some landholders remain open to grazing management training, and to changing their operations, there remains a large proportion of land managers that may be resistant to the adoption of new strategies.

Landholder capacity to adopt alternative grazing management strategies is determined by access to physical, human, psychological, financial, natural and social resources. Of these, the factors affecting an important component such as control of TGP include the availability of adequate finance for fencing, labour for construction and maintenance, the support of neighbours and contractors, good markets and income, as well as favourable seasonal conditions (Local Land Services, unpubl. data, 2017).

In reviewing the outcome of incentive funding processes, Swann and Richards (2016) considered that to achieve long-term outcomes in grazing management programs, specific emphasis must be placed on keeping landholders motivated beyond the initial ‘honeymoon’ period of project implementation. Landholders need to be involved from project planning to project evaluation stages, and be provided with relevant and engaging training, as well as ongoing advisory, information and extension support services.

Infrastructure factors

Because most parts of the study area are subject to the influence of unmanaged herbivores, investment in adequate infrastructure development is necessary to apply differential grazing management to carbon project areas. Traditional property infrastructure, such as plain wire fencing and open-access waterpoints, fails to provide control of grazing pressure from macropods, meat sheep and unmanaged goats, resulting in mediocre land condition and a susceptibility to drought, widespread in the region (Waters et al. 2019; Hacker and McDonald 2021). Achieving a suitable level of containment or exclusion to control grazing pressure according to vegetative growth requires the following: