An online system for calculating and delivering long-term carrying capacity information for Queensland grazing properties. Part 1: background and development

Keywords: safe pasture utilisation, livestock carrying capacity, stocking rates, land condition, rainfall, pasture growth, buffel grass.

Across global and Australian rangelands, the calculation of grazing capacity or long-term carrying capacity (LTCC) has been a major challenge for grazing and resource managers (Holechek 1988; de Leeuw and Tothill 1990; Zhang et al. 2014; Meshesha et al. 2019). In general terms, LTCC is the number of animals that can be grazed without causing long-term (e.g. at multi-generational human timescales) loss of productivity of pasture and land resources. The paper describes the development of calculating and applying LTCC over the past 30 years for native pastures in northern Australia, and in particular Queensland, Australia.

Global rangelands vary in many attributes such as land tenure, infrastructure (water points, fencing), type of livestock enterprise, topography, soils, vegetation, social and regulatory control, and available land and pasture research. The online system described here is specific to Queensland’s grazing industries. Nevertheless, the general goals, structure and components of the calculation of LTCC (namely pasture production, safe pasture utilisation rate, animal intake and the application to individual users) are common to most rangeland situations; hence, we use this commonality as the basis for the paper.

Since early land settlement in the 1860s, the grazing lands of Queensland have been divided into individual holdings (i.e. properties), predominantly for agricultural and livestock production. An important feature of Queensland grazing lands, compared with other global rangeland regions, is the relatively high year-to-year and multi-year variability in rainfall (e.g. Fatichi et al. 2012). There is also substantial spatial variation across Queensland’s grazing lands with average annual rainfall decreasing from coastal to inland regions (in contrast to temporal rainfall variability increasing). Despite the resulting high temporal variability in pasture growth, Queensland grazing enterprises generally include a large nucleus/core of self-replacing flocks or herds, as well as varying degrees of trading (i.e. buying and selling) of livestock. The flexibility that grazing managers have to respond to seasonal and yearly variability in forage supply depends on: (1) long-term rainfall variability (Ash et al. 2000; Pahl et al. 2015); (2) attitude to risk (e.g. responsiveness to market fluctuations; McIvor 2010); and (3) the dependence of the enterprise on maintaining the herd/flock breeding nucleus/core (e.g. Foran and Stafford Smith 1991; McIvor 2010).

The estimation of LTCC has been an important component of grazing land settlement in Queensland since the late 19th century (Heathcote 1965). However, early government land administrators assessing viable property size and livestock managers were understandably unaware of the importance of high rainfall variability in multi-year wet and dry periods (e.g. McKeon et al. 2004, 2021; Stone 2004). This climatic variability has had major impacts on livestock carrying capacity, herd and flock dynamics and enterprise performance (e.g. Miller et al. 1973) and has led to deterioration of land and pasture condition (Weston et al. 1981; Tothill and Gillies 1992; McKeon et al. 2004). Knowledge of LTCC is important to manage for temporal rainfall variability, so as to achieve ecological sustainability (Stone 2004; O’Reagain et al. 2018) and financial viability (Passmore and Brown 1992). However, it is not suggested that LTCC is used as a continuous stocking rate value; but as a guide around which stock numbers can be varied seasonally/annually in order to match variability in available forage. For more risk-averse grazing managers, LTCC provides a guide that minimises the frequency of impacts of dry periods. However, a climatic feature of the grazing lands of Queensland is that extreme widespread multi-year drought conditions do occur (Day and McKeon 2018; Stone et al. 2019; Irvine 2021; McKeon et al. 2021), which still necessitate stock reduction to very low levels, even for conservatively managed properties.

Spatially, grazing for livestock is the major land use in Queensland (~85%; QDAF 2017). Extensive grazing occurs in a wide range of climatic zones (dry monsoonal, arid, semiarid, sub-tropical). Vegetation types range from treeless grasslands (Mitchell grass, Astrebla spp.) to open woodlands (Eucalyptus spp.) and edible shrublands (e.g. Mulga, Acacia aneura) – similarly, soils also vary considerably in texture and fertility. These native pastures also vary in land condition as a result of overgrazing in some regions (Tothill and Gillies 1992). An additional source of complexity is that there have been large areas of native pastures converted to sown grasses and legumes, with variable histories of nitrogen disturbance (e.g. tree clearing, deep ripping, blade ploughing).

As a consequence of these sources of variability, estimation of pasture productivity and associated LTCC for all locations in Queensland has been a major challenge. Over the past 30 years, this challenge has been addressed by a combination of field trials, pasture monitoring, documentation of grazier experience and regional expert consensus (Johnston et al. 1996a, 1996b). Since the 1940s, field experiments (i.e. grazing trials) have compared different levels of pasture utilisation or stocking rates for particular locations. The trials have demonstrated the value of conservative pasture use in managing for temporal climatic variability, and have aligned with the experience of graziers (pastoralists) with properties in benchmark condition (i.e. exhibiting stability of good land condition). Given the documented cases of grazing land degradation in Queensland (Weston et al. 1981; Gardener et al. 1990; Tothill and Gillies 1992; Ash et al. 2002; McKeon et al. 2004), the challenge has been to extrapolate scientific findings and successful grazier experiences to properties across Queensland. To be useful to individual graziers and their advisors, the calculation of LTCC needs to be accessible, transparent and specific to their own particular situation.

This paper (Part 1) describes the evolution of LTCC science in Queensland over the past 30 years and the development of a ‘prototype’ new online system for the rapid delivery of the FORAGE Long-Term Carrying Capacity report. This capability has been supported by advances in computing speed and capacity, and availability of spatial data (i.e. cadastral, climate, soil, grazing land types, and remote sensing of woody vegetation and ground cover). The technical detail of the online system that calculates and provides LTCC information is described in Part 2 (Zhang et al. 2021).

The objectives of this paper (Part 1) are to describe:

1. the background to livestock carrying capacity science and application in Queensland;

2. the development, improvements and limitations of the online LTCC calculation system;

3. the FORAGE Long-Term Carrying Capacity report;

4. the report applications and feedback; and

5. future challenges and solutions for calculating LTCC across Queensland.

We describe the development of LTCC in Queensland through three stages since the 1990s, highlighting how the online system builds on these previous advances. We compare improvements in the components of the calculation procedure and delivery (Table 1). The major developments have occurred in a technical environment of increasing computer and remote sensing capability, but a declining availability of personnel to support on-property delivery.

Table 1.  Components of long-term carrying capacity (LTCC) calculation and the developments through each stage over the past 30 years

The ‘cycles’ of multi-year wet and dry periods in Australia’s rangelands (McKeon et al. 2004, 2021) led to increased livestock numbers during wet periods (e.g. early 1950s and 1970s in south-west Queensland), followed by heavy grazing pressure when livestock have been retained in the subsequent average or dry period. Over the past 100 years, episodes of heavy grazing pressure across northern Australia resulted in native pasture deterioration and degradation (e.g. 1960s and 1980s, Miles 1990; Tothill and Gillies 1992; Mott and Tothill 1993). As a consequence, the grazing industry, resource scientists and government departments began to develop objective estimates of long-term livestock carrying capacity in order to aid graziers in managing for multi-year rainfall variability. For example, in 1988, members of the United Graziers Association (located in south-western Queensland) ‘called for a critical re-evaluation of the Queensland Department of Lands rated carrying capacities: seeing the figures as subjective judgements derived in the good run of seasons in the 1950s and leading to an over-expectation of the land’ (Garrad and Johnston 1998).

In developing an approach to estimate LTCC for sheep and wool enterprises in south-western Queensland, Johnston et al. (1996a) reviewed international methods for determining ‘grazing capacities’. Johnston et al. (1996a) defined a ‘safe’ grazing capacity (i.e. LTCC) as:

‘…the number of dry sheep equivalents (DSE) that can be carried on a land system, paddock or property in the long term without any decrease in pasture condition and without accelerated soil erosion’.

Johnston et al. (1996a, 1996b) described the development of a system for the calculation of LTCC and application for over 200 properties in south-west Queensland. The system had several key components:

• estimation of LTCC from benchmark properties of the region (i.e. exhibiting stability of good land condition; Johnston et al. 1996a);

• accurate description of climate, land resources (i.e. soil, vegetation and landforms that are easily recognised by land managers in a region), soils and woody vegetation cover at property scale;

• pasture and soil parameters for simulation of long-term ‘median’ or ‘average’ pasture growth with the pasture model (GRASP; GRASs Production; McKeon et al. 2000; Rickert et al. 2000);

• estimates of safe pasture utilisation derived from benchmark properties, grazing trials and consensus from experienced graziers and land resource officers;

• use of on-property mapping of soils and attributes including a grazier’s knowledge; and

• explanation of the LTCC calculation procedure and the results to property owners by grazier consultants.

In south-western Queensland, 217 properties were assessed from 1994 to 1998. Calculated LTCC was compared with graziers’ own LTCC values: 66% of property calculated LTCCs were within ± 10% of grazier estimates; for 23% of properties, calculated LTCC was 10% (or more) below grazier estimates; and for 11% of properties, calculated LTCC was more than 10% above grazier estimates (Johnston and Garrad 1999; Johnston et al. 2000).

Using similar methods, approaches for estimating LTCC were developed for beef cattle enterprises utilising native pastures in north-eastern Queensland (Scanlan et al. 1994) and south-eastern Queensland (Day et al. 1997b). When the three studies (i.e. including Johnston et al. 1996a) were combined, the average pasture utilisation was ~20% of average annual pasture growth, with a suggested range of 15–25% (Hall et al. 1998). Hunt (2008) reported a similar range of values in a review of Australian pasture utilisation studies. As described below, Walsh and Cowley (2011), also found similar values for beef cattle grazing a range of pasture communities in the Northern Territory (NT).

Following Johnston et al. (1996a), the formula for calculation of LTCC is:

where area is the grazed area of the property, land parcel, paddock of interest (ha); discount is the proportion of the area that is grazed considering the limitations of access to water and topography (0–1); pasture growth is median pasture growth calculated from long-term climate records (e.g. 40 years), using the pasture growth simulated from the GRASP model (kg dry matter ha⁻¹ year⁻¹). The terms ‘median’ and ‘average’ have been used interchangeably in popular discourse; however, they are distinctly different mathematically, particularly for variables such as rainfall and pasture growth; safe pasture utilisation is the proportion of long-term median pasture growth that is consumed (0–100%). The value of safe pasture utilisation is derived from grazier estimates of LTCC, as well as grazing trials and regional expert opinion; and animal intake is pasture dry matter intake per head per year. For sheep, a Dry Sheep Equivalent (DSE) consumes 400 kg DM year⁻¹ (i.e. 1.1 kg day⁻¹; Johnston et al. 1996a); for cattle, an adult equivalent (AE) is a 450 kg dry animal consuming 2920 kg DM year⁻¹ (i.e. 8 kg DM day⁻¹; Walsh and Cowley 2011; McLennan et al. 2020). We recognise that a range of annual beef cattle intake values have been used in past studies for the calculation of safe utilisation rates, representing to some extent, the impact of pasture nutrition on animal consumption (e.g. 2700 kg DM AE⁻¹ Hall et al. 1998, 3000 kg DM AE⁻¹ Ash et al. 2002, 2920 kg DM AE⁻¹ Walsh and Cowley 2011, 3650 kg DM AE⁻¹ GLM rule of thumb). We follow the work of Walsh and Cowley (2011) and McLennan et al. 2020 to use 2920 kg DM AE⁻¹ as a reasonable compromise of this uncertainty. In addition, LTCC animal units can be commonly expressed in several different forms depending on location and user (i.e. adult equivalents per paddock/property; adult equivalents per hectare; adult equivalents per 100 ha; hectares per adult equivalent).

In the late 1990s, Meat and Livestock Australia (MLA; a producer-owned company, which in conjunction with the Australian government and livestock producers, delivers marketing and research programs for Australia’s red meat producers), supported the production of an education package for ‘grazing land management’ (GLM; Anon 2002; Chilcott et al. 2005c; Quirk 2006). In 1998, MLA funded research into the types of information and package structure that potential end-users would prefer. There were also similar investigations into extension packages for nutrition and animal breeding for grazing enterprises. Following recommendations on the structure and content of a GLM training package (in 2000), courses were subsequently developed by Queensland Department of Primary Industries (QDPI) and Commonwealth Scientific Industry Research Organisation (CSIRO) staff (MLA 2002). The first GLM workshop (delivered in 2003), included a module on LTCC using the components described here as the basis for the derivation of LTCC information.

Participants in the GLM workshops were shown how to calculate LTCC from their own property description. A key component of the property description is the mapping of ‘land types’ for grazing. A land type has characteristic patterns of soil, vegetation and landform that are easily recognised by landholders in a region (https://futurebeef.com.au/knowledge-centre/land-types-of-queensland/). Estimation of LTCC for properties across Queensland first required identification of the land types within a region. Land types are currently derived from regional ecosystem mapping, which is largely based on overstorey floristics and aspects of soils, geology and landform (Neldner et al. 2020). The identification and description of regional land types involved the combination of experienced graziers and land resource scientists with relevant land resource mapping materials (e.g. Land management manuals, WARLUS, CSIRO Land Research Series; Christian and Stewart 1968; Dawson 1974).

Land type parameter sets of biophysical attributes for pasture growth modelling (Day et al. 1997a; McKeon et al. 2000) were developed in accordance with expert-derived estimates of LTCC (Anon. 2007; The GRASP Modelling Team 2008). The GRASP model (McKeon et al. 2000; Rickert et al. 2000) is a biophysical simulation model of soil–water balance (including evapotranspiration, runoff and drainage) and pasture dry matter flow (growth, senescence, detachment and trampling, litter decomposition and animal intake). Regional land type pasture growth estimates were then provided to participants in GLM workshops to enable them to estimate LTCC for their own properties for comparison with their own experience. Between 2002 and 2020, 110 workshops were held in over 73 locations across northern Australia, with ~1200 attendees, representing more than 700 businesses (C. Paton, unpubl. data).

As part of the GLM initiative, over 230 land type parameter sets for the GRASP model were developed for Queensland, with a further 50 currently under development for the NT (State of Queensland 2019). These land type parameter sets have enabled the simulation of long-term median pasture growth required for the calculation of LTCC at GLM workshops (Allen 2011–2019, various issues).

Since the LTCCs of properties managed by graziers were considered private, no formal assessment of comparison between the calculated LTCC and grazier experience was sought for recording purposes. Nevertheless, GLM workshop facilitators noted that participants were enthusiastic about conducting the calculation for themselves while at the workshops, which indicated the need for the further development of the LTCC system described below and in Zhang et al. (2021). The GLM workshop presentations and exercises also demonstrated to grazier attendees the importance of both LTCC estimates and grazing management practices such as forage budgeting.

In 2015, the capability of combining pasture growth modelling and Geographical Information Systems (GIS) with digital spatial data such as property infrastructure, woody vegetation cover and land type mapping was demonstrated for 20 grazing properties in Queensland (Whish et al. 2016). Whish and Holloway (2016) found that estimates of LTCC could be improved through changes in model parameters to account for on-property information. Five of the 20 properties from this study provided LTCC comparisons that have been used in testing the online LTCC system (Zhang et al. 2021). Thus, the innovative combination of the GRASP model and GIS to calculate LTCC (Whish and Holloway 2016) provided a proof of concept for the online system described below.

Building on the above progress, the prototype online LTCC system (Zhang et al. 2021) is a rapid computation procedure to obtain property-level LTCC estimates for native pastures throughout Queensland. The FORAGE Long-Term Carrying Capacity report (FORAGE LTCC report, described below) provides LTCC calculations with the most recently available property description (i.e. the most recent woody vegetation cover and land type mapping). The processes, inputs and outputs, and issues to be considered are described in detail in the FORAGE LTCC report (shown below), and in Zhang et al. (2021).

The calculation of LTCC and its application at property level combines a very wide range of scientific and geographical endeavour by graziers, advisors and agricultural/environmental scientists. Table 1 describes the components of LTCC calculation and the developments through each stage over the past 30 years. The three biophysical components of the calculation: LTCC = [(1) (long-term median pasture growth (kg ha⁻¹ year⁻¹); (2) safe pasture utilisation (%); and (3) animal intake (kg AE⁻¹ year⁻¹)] have provided the focus for continual improvement through:

• field research (Hassett et al. 2000; Orr 2005; Phelps 2012; O’Reagain et al. 2014; O’Reagain et al. 2018; McLennan et al. 2020);

• development of pasture modelling (e.g. Carter et al. 2011; Owens et al. 2019) and pasture growth parameter sets for land types (Whish 2011; Whish et al. 2016; Zhang et al. 2021);

• assimilation of digital spatial datasets: land type mapping (Irvine and Holloway 2019); woody vegetation cover (Zhang et al. 2021); historical climate data (Jeffrey et al. 2001); and

• on-property evaluation (Whish et al. 2016; Zhang et al. 2021).

The major achievement of these improvements has been the capability to expand the coverage of LTCC calculation to the whole of Queensland. This expansion has been supported by the rapid advances in computing capability (from personal computers to ‘high performance’ computing) in combination with the databases listed above.

An important factor in all three stages in the evolution of LTCC has been the use of the GRASP biophysical model to simulate pasture growth from historical daily climate data. The model has been under continual development and testing since the early 1980s (Rickert et al. 2000) and has been supported by field methodologies (i.e. GUNSYNpD and SWIFTSYNpD; Day and Philp 1997) to measure key parameters such as: available soil water range; potential annual nitrogen uptake; nitrogen use efficiency; potential pasture regrowth; grass basal area; and the relationship between green or total cover (Day et al. 1997a). In Queensland, the GRASP model has been calibrated for 183 native pasture field sites, as well as 10 grazing trials (Day et al. 1997a; Mayer 2013).

The capability of GRASP to simulate pasture growth for a wide range of soils and communities has been further demonstrated at an Australia-wide level (i.e. AussieGRASS; Carter et al. 2000, 2011). Across Australia’s rangelands, the AussieGRASS version of GRASP has been parameterised for 185 native pasture communities and also for specific field locations in New South Wales, central Australia and Western Australia (McKeon et al. 2004). Thus, despite the complexity of soils, climate and vegetation across Queensland grazing lands, the GRASP model and its parameter sets have provided a logical basis to extrapolate to land types that have yet to be covered by field experimentation.

As discussed below, sown pastures remain a difficult issue for pasture growth estimation due to the variation in fertility (i.e. nitrogen availability) over time since initial pasture establishment. Additionally, an issue that is yet to be resolved is that only 20% (to date) of the 225 native pasture land types have supporting pasture growth field data (G. Fraser, unpubl. data). For these land types where field data are not available, land type parameter sets have been developed from the ranking of land types in terms of productivity by regional expert consensus.

Woody vegetation (i.e. trees and shrubs) is an important element of rangelands and Queensland’s grazing lands (Burrows et al. 1988). Woody vegetation has both beneficial and competitive effects on pasture growth, as well as contributing directly to animal diets (e.g. through browse and leaf litter; Johnston et al. 1996a). In the mulga woodlands, sheep and cattle consume some mulga leaf throughout the year. Johnston et al. (1996a) included a proportion of leaf fall and leaf litter in the calculation of LTCC (after Beale 1975). In the online LTCC system, this approach has been implemented for mulga land types. Other land types where browse and leaf litter make a significant component to livestock diet are yet to be identified.

For the wider application across Queensland, several approaches were considered to quantify the competitive relationship between pasture and woody vegetation. Field studies had been used to develop empirical equations between tree/shrub density (measured as tree basal area (TBA), m² ha⁻¹) and annual pasture growth (Beale 1973; Carter and Johnston 1986; Burrows et al. 1990; Scanlan and Burrows 1990). Most of these equations are strongly curvilinear, reflecting the strong competitive effects of woody vegetation for water and nutrients. To address the variability in these relationships, Scanlan and Burrows (1990) combined Queensland and international studies to create a general empirical equation across a range of tree and shrub communities to calculate average annual pasture growth as a function of TBA.

The GRASP model also includes equations that represent competition between pasture and woody vegetation for water and nutrients at a daily time step (e.g. Scanlan and McKeon 1993). The model accurately simulated pasture growth at treed and cleared sites at two locations in northern Australian savannas (Day et al. 1997a; Cafe et al. 1999). However, wider application across Queensland land types is uncertain because of the difficulty in measuring/estimating important attributes such as: rooting depth; available water holding capacity below 1 m; size of pools of free water above impermeable layers; and variation between tree species in rooting patterns. As a consequence, a more practical approach has been used to represent the general equation of Scanlan and Burrows (1990) within the GRASP model. Thus, the current parameterisation of GRASP is consistent with the more general view of tree competition across a wider range of pasture communities.

In Stage 3, advances in remote sensing (Carter et al. 2011) have resulted in foliage projective cover (FPC) replacing TBA as a more useful ecophysiological measure of woody vegetation (Owens et al. 2019). The major source of variation in tree effects at a paddock scale is likely to be the spatial fragmentation of woody vegetation cover. To some extent, this issue is addressed in the online system by remote sensing at a high spatial resolution (i.e. 30 × 30 m) measurement of woody vegetation cover, with classification into 13 categories of FPC (Zhang et al. 2021).

Sub-models have also been developed in GRASP representing the beneficial effects of trees on understorey microclimate, as well as including the larger paddock scale effects of tree strips on climate elements such as wind and temperature. These developments remain the subject of current research.

Published literature (e.g. Scanlan et al. 1994; Johnston et al. 1996a; Day et al. 1997b; Chilcott et al. 2005a, 2005b) and other studies (D. Phelps and P. Jones, unpubl. data), together with regional expert-derived estimates, were used to determine safe pasture utilisation levels for all land types of Queensland (G. Whish, unpubl. data). Safe pasture utilisation rates of annual pasture growth for native pastures in Queensland range between 10–30% (State of Queensland 2019), with higher rates for the more productive, resilient pastures that occur in wetter areas on fertile soils. Safe pasture utilisation rates may be higher where buffel grass (Cenchrus ciliaris) is dominant in pastures (State of Queensland 2019), as discussed below.

Walsh and Cowley (2011; p. 137) reviewed safe pasture utilisation rates for Queensland and their own studies in the NT. For example, they found that an average annual pasture utilisation rate of up to 25% was considered safe for uniform Mitchell grass pastures in good condition on cracking clay soils within the Barkly Tableland region of the NT. However, most importantly, they noted that such a rate impacted adversely on preferred areas where there was a mix of land types. They also observed that safe pasture utilisation rates (i.e. <15%) applied for land types of lower fertility and less resilience in other regions of the NT. Thus, their findings support the general range of 15–25% (e.g. Hall et al. 1998) with fertility, mix of land types, and temporal variability in rainfall being important sources of variation in estimating safe pasture utilisation rate.

The goal of calculating LTCC within the FORAGE report is to provide an LTCC estimate that would help grazing managers maintain land and pastures in the best productive state (‘A’ condition) for livestock through a wide range of temporal variation in rainfall. The symptoms of declining land condition (e.g. McIvor et al. 1995) vary with land types. Nevertheless, graziers and their advisors may be able to estimate the relative land condition of their paddocks/property from several indicators including botanical condition, ground cover, surface soil condition and animal performance. To provide a more general approach in GLM workshops, land condition is ranked from ‘A’ (good) to ‘D’ (very poor), with discounts being applied to ‘A’ condition pasture growth (i.e. ‘B’ (75%); ‘C’ (45%); and ‘D’ (20%), to show the effect on carrying capacity (after McIvor 2010).

In Stage 1 (south-west Queensland), the direct impact of woody vegetation on pasture growth was calculated; however, other aspects of declining land condition were not included in the report to graziers. In Stage 2 (i.e. GLM workshops) and the online system, woody vegetation effects are accounted for in the GRASP model through the inclusion of TBA and FPC respectively. In Stage 3, the likely effects of declining land condition are also calculated in the FORAGE LTCC report.

In the GLM workshops, there were no formal comparisons between calculated LTCC and graziers’ values (as stated above). Nevertheless, the participants were very engaged in using the capability to calculate the relative benefits of improving land condition on current carrying capacity; e.g. changes from ‘C’ to ‘A’ land condition (C. Paton, unpubl. data). When possible increases in current carrying capacity were linked to the financial benefits, the impact of improving land condition took on a different perspective. For example, if a paddock’s current carrying capacity was 45 AE due it being in ‘C’ land condition, improving the paddock to ‘A’ condition could increase the current carrying capacity to 100 AE. Workshop participants acknowledged that the calculations of LTCC and discounts for the effects of varying land condition were worthwhile; the potential benefits of improving land condition could then be performed through economic analysis (e.g. McLean et al. 2020). This observation supports the value of performing LTCC calculations, especially for estimating the relative effects of improving pasture condition; despite the known uncertainties and difficulties in the overall calculation process.

The online LTCC system necessarily has less human involvement in the delivery and explanation of the calculation of LTCC in comparison to previous stages (Table 1). The success in south-west Queensland and in the GLM workshops was in the ability of the consultants and presenters to interact with participants to explain the results and receive feedback, and hence improve components of the calculation (e.g. Whish et al. 2016).

Johnston et al. (1996b) put considerable effort into testing their LTCC model on individual properties, using grazier consultants who were involved in detailed training and assessing their own properties to develop skills and confidence. From their review and feedback, refinements were made to account for factors affecting pasture production, including woody vegetation cover and ‘run-on’ on flood plains. Grazier consultants reported that more accurate land system/land unit (see Table 1) mapping at property scale would improve the accuracy of LTCC estimates.

The online LTCC system provides objective property-level information rapidly. However, the number of personnel previously required to support on-property evaluation and model improvements in applications are not readily available for the Queensland-wide approach. Nevertheless, the automated system enables on-going improvement of modelling framework through additional field data and user feedback and evaluation. The users of the prototype FORAGE LTCC report include land managers, extension providers, Natural Resource Management (NRM) groups, consultant services and educational facilities, who have also been providing feedback to improve the system.

The previous GLM workshops (over 3 days) provided a structured interaction between presenters and participants, including a worked example of LTCC for each individual’s property. A finding from the GLM workshops was that LTCC calculations were unlikely to be made by participants without the support of the workshop environment. Nevertheless, participants reported that the workshop did influence subsequent management and financial decisions (C. Paton, unpubl. data). The online system would support many of the educational activities in GLM workshops, and additionally, would provide the opportunity for evaluation and improvement to the LTCC system.

In Zhang et al. (2021, fig. 6) there was reasonable agreement reported between modelled and grazier-estimated LTCCs, where 28 of the modelled LTCC estimates (out of the 43 properties) were within ± 25% of owner-provided LTCC values. The low number of LTCC estimates reflects the difficulty of collecting this information, given the issues of privacy, confidentiality and commercial-in-confidence. There are also several uncertainties that are yet to be quantified, including climatic conditions associated with the grazier estimate period, pasture resource condition, land types and land type parameters. In addition, the collection of both pasture growth measurements and benchmark estimates of LTCC to calculate safe pasture utilisation is an expensive and time-consuming process. To address these uncertainties, the engagement of the grazing community and their advisors remains a critical component of the continuing process and ownership of the LTCC methodology described here. The evaluation of user feedback will provide the opportunity to enlarge the much-needed databases of benchmark LTCCs, improved land type maps and GRASP modelling parameter sets.

The current online system draws upon existing broadscale spatial databases for property/paddock attributes. However, more accurate estimates of LTCC require user-defined knowledge of property description. To address this issue, the new ‘My FORAGE Map web program’ interface is being developed (see Zhang et al. 2021). The interface provides the capability to describe the property and/or paddocks in detail with essential attributes that affect pasture growth and grazing management (e.g. land types, pasture species, woody cover, fencing and water points). In particular, this capability will allow the assess of the effect of grazing distribution/preference on estimates of LTCC (Walsh and Cowley 2011).

The major output of the online system is a comprehensive report described below covering many of the aspects that the previous three stages of development showed to be necessary to provide understanding and explanations in depth. At the time of writing (August 2021), the online system is necessarily a prototype with users’ feedback providing an important component to improve the elements of the LTCC calculation (Table 1). At present, the report provides the opportunity for advisors to undertake a one-on-one dialogue with users. Thus, the online system forms the basis for discussion of the implications of carrying capacity for land condition and financial performance.

The FORAGE Long-Term Carrying Capacity report is the major output of the online system, including a five-page PDF-format report and two accessory Excel spreadsheet files (see Supplementary materials, available at the journal’s website). The report is requested through the FORAGE page of the Long Paddock website (https://www.longpaddock.qld.gov.au/), which provides a range of information for grazing land management (Zhang and Carter 2018). The LTCC estimates in the report can be used as a starting point for discussion with land managers and to some extent, can be customised based on their knowledge of property attributes (including land types and land condition). Requests submitted for LTCC reports are emailed to the user usually within 10–20 hours.

The structure of the FORAGE LTCC report (see Supplementary material) is to first provide the overall summary of LTCC (page 1) and then continues with increasingly more detailed information to assist the user with explanations of the components that have been used to make the calculation. The report provides the following information:

• Page 1 of the report provides a location map and a ‘Summary of estimated LTCC’. The summary displays both total livestock (in AE) and stocking density (ha AE⁻¹) values for a range of land condition classes (‘A’, ‘B’, ‘C’, ‘D’).

• Page 2 presents three time series: (1) modelled historical annual safe stocking rates and estimated LTCC for the property. As indicated above, annual stocking rates are likely to fluctuate around an estimated LTCC in response to year-to-year variation in pasture growth (time series ‘3’ below). The stocking rate time series shows a practical approach to changing seasonal/annual stocking rates in response to variable pasture growth; (2) the historical annual and summer season rainfall graph for the requested property derived from the SILO database (www.longpaddock.qld.gov.au/silo/); and (3) the historical 12-month pasture growth (kg DM ha⁻¹) for April–March;

• Page 3 shows the LTCC calculated for each land parcel (QDNRME 2013), or paddocks if supplied as a shapefile, and land type with the four land condition classes (‘A’, ‘B’, ‘C’, ‘D’);

• Page 4 shows the median annual pasture growth for each representative land type showing the variation in pasture productivity along with the four land condition classes; and

• Page 5 provides the estimated FPC map obtained from the latest satellite-derived woody vegetation cover data for the selected area.

Each of the above components are described in further detail in Zhang et al. (2021).

On requesting an LTCC report, the user will also receive two accessory Excel spreadsheet files by email (in addition to the PDF report). The accessory spreadsheet files deliver detailed information at a paddock/land parcel scale as a link between the LTCC data and the user’s own knowledge. This information provides a basis for the user to adjust land condition proportions and the opportunity to provide feedback from their assessment.

In many cases, the FORAGE LTCC report will be the first contact between the user (grazier/landholder) and advisors (e.g. extension providers, consultants) regarding grazing management discussions. The online system by itself necessarily lacks this important human interaction. For example, a likely discussion point is the comparison between the estimated LTCC and the grazier’s long-term property livestock numbers. The variation in the mixture of animal size, gender and age needs to be accounted for, so as to convert actual livestock numbers to total AEs. The calculation of total AEs for various herd/flock structures has been formalised in Queensland Department of Agriculture and Fisheries (QDAF) extension products such as the ‘Stocktake GLM’ application to facilitate this comparison.

Explanatory information including ‘frequently asked questions’ is also provided on the Forage page of the Long Paddock website, along with a two-page ‘quick guide’, and a more detailed LTCC section within the online ‘Guide to using FORAGE’. In addition, the report is supported by an ‘awareness’ video, which is designed to provide information and guidance of the report components in ‘conversational’ language. Feedback and enquiries are encouraged, as they are a necessary component for improving the online system (an email address is provided on The Long Paddock website). It is envisaged that users of the FORAGE LTCC report will be surveyed after a period of time of receiving a report (e.g. 6–12 months). The goals of the survey will be three-fold: (1) evaluate the accuracy of the report components (e.g. maps of land types and FPC); (2) compare estimated LTCC to owner-defined LTCC; and (3) determine whether the content of the report assisted users to better manage year-to-year stock numbers.

In developing the online system and designing the report for wider application across Queensland, we have identified the future challenges of: (1) parameterising land types where field data are not available; (2) parameterising sown and naturalised pastures where changes in productivity are occurring relatively rapidly (5–10 years); and (3) increasing the number of benchmark properties to provide better estimates of safe pasture utilisation rates.

The three stages of the evolution of LTCC calculations (discussed above) have concentrated on native pastures, primarily because of the apparent impact of overgrazing identified in the 1980s (Weston et al. 1981; Tothill and Gillies 1992). Since the 1900s, there has been an increase in many non-native grasses and sown legumes (Peck et al. 2011); and since the 1960s, there has been a rapid increase in sown and naturalised areas of buffel grass (Walker and Weston 1990). Peck et al. (2011, 2017) estimated that buffel grass was ‘dominant’ in 5.8 million ha (including over 30 land types) and ‘common’ in a further 25.9 million ha. The botanical and agronomic attributes of buffel grass contribute to its potential to invade (i.e. naturalise) existing native pastures (Martin et al. 2015). Compared with native pasture, buffel grass pastures are highly productive, especially after woody vegetation clearing or subsequent renovation (e.g. blade ploughing). However, rundown of available soil nitrogen after clearing and overgrazing have led to a widespread decline in productivity and carrying capacity (Walker and Weston 1990; Peck et al. 2011, 2017). The methodology used to measure/calibrate GRASP parameters (Day et al. 1997a) was designed for native pastures, with relatively low year-to-year variability in fertility (e.g. potential nitrogen uptake). The application of this field approach to buffel grass represents a major challenge, given that a wide range of locations and management histories have to be considered, including land type, age since initial clearing and subsequent renovations, and possible over-sowing with legume species (including Leucaena; Peck et al. 2011).

While buffel grass is recognised as the major sown grass species, it is also recognised as an invasive species in native pastures (Fensham et al. 2013; Martin et al. 2015). The invasive attributes of buffel grass are likely to be exacerbated by: (1) the interaction of grazing cycles of drought and high rainfall (Fensham et al. 2013); and (2) by temperature increase (i.e. global warming; Martin et al. 2015). Thus buffel grass is a current and emerging challenge in estimating LTCC for both sown and native pasture communities. The issue has been addressed in the online LTCC system with the addition of a ‘buffel grass pasture’ option at the point of request. For 14 land types where buffel grass is most likely to affect LTCC, parameters have been altered to provide higher pasture growth and safe utilisation, resulting in increased LTCC values (see Zhang et al. 2021 for further detail).

To a lesser extent, similar issues exist with native pasture land types where woody vegetation clearing has occurred in the past 5–10 years. The online LTCC system provides the basis for a future approach to combine current field research, expert opinion, and user evaluation/feedback to detect and represent such changes in pasture productivity and carrying capacity.

There have been recent developments in field and remote sensing technologies for improving the measurements of pasture cover and biomass, and associated pasture modelling. These advances could aid the improvement of land type parameter sets, including the estimation safe utilisation rates. For example, the operational AussieGRASS model (based on GRASP; Carter et al. 2000, 2011) assimilates a wide range of spatial and temporal datasets (climate data, remote sensing of woody vegetation and ground cover, rapid mobile pasture assessments, pasture nitrogen concentrations, pasture biomass from historical grazing trials and exclosures). This approach has allowed parameterisation of the major pasture communities (46) and soil types across Queensland. The AussieGRASS model is used to simulate the effects of long-term climatic variability (i.e. 130 years) to evaluate the ranking of current conditions (relative to past years) and seasonal climatic risk assessment.

AussieGRASS has provided the capability to demonstrate the value of new developments and to include them in the online LTCC system. However, there are insufficient measurements of pasture growth to develop a similar capability at the finer spatial scale of land types (225 in Queensland). At present, it is unlikely that there will be sufficient measurements of pasture growth and other parameters to build an equivalent system as AussieGRASS at the land type scale. Consequentially, AussieGRASS will have to continue as the ‘testbed’ for demonstrating how new information sources (e.g. remote sensing of green and ground cover, surface soil moisture, pasture nitrogen concentration, pasture standing dry matter) can be integrated into pasture growth modelling. After these new information sources have been tested in AussieGRASS, they are likely to be included in the LTCC modelling system, and hence, result in more accurate estimations of pasture growth at the paddock scale.

Examples of emerging technological developments that will improve the capability of the online LTCC system to meet the above challenges are:

• Owens et al. (2019) showed that flux tower measurements of evapotranspiration and remote sensing of green ground cover could be used to better parameterise the soil water sub model in GRASP for a central Australian mulga woodland;

• similarly, Liu et al. (2019) showed how estimations of rangeland forage production (e.g. 1 ha scale) could be improved by a small unmanned aerial system combined with remote sensing of pasture cover; and

• non-destructive sampling methods using advances in measurement of yield and chemical analysis will allow scientists to capture more sites with greater frequency, which will improve parameterisation of the GRASP model (e.g. potential nitrogen uptake and attributes of land condition; Barnetson et al. 2020; Paton et al. 2021). The parameterisation of land condition classes (McKeon et al. 2000) will unlock the information already available in remotely-sensed timeseries of green, ground and total cover.

The challenges listed above share the common problem of pasture growth and nitrogen uptake measurements. Field sampling with GUNSYNpD, SWIFTSYNpD methodologies (Day and Philp 1997) is expensive, time consuming and labour intensive. The next stages of development of the LTCC system are likely to benefit from the technological advances currently being demonstrated.

The prevention of degradation of the rangeland resource and the loss of livestock productivity has been a global issue for centuries (Zerga 2015). For government agencies responsible for the sustainability of the grazed resource, the major issues are first, how to monitor land and pasture condition; and second, how to separate the effects of grazing management (e.g. stocking rates) from climate variability (e.g. multi-year wet and dry periods). The use of research grazing trials and the development and application of objective systems to estimate LTCC have supported/recommended more conservative use of the pasture resource. To what extent industry has adopted ‘this advice’ has always been difficult to measure (i.e. in terms of land condition). Nevertheless, from the viewpoint of resource outcomes, the developments in remote sensing of vegetation (woody and ground cover) and erosion features (e.g. gullies and sediment transport) are providing a comprehensive high-spatial assessment resolution of the grazed resource. This assessment supports on-ground monitoring (e.g. Land Condition Assessment Tool; Hassett 2021), as well as other tools designed to provide individual property managers assistance in decision-making (e.g. FORAGE online reports; Stocktake GLM application). However, the estimation of grazing management effects in contrast to the impacts of climate (mainly rainfall variability) remains a challenge. Simulations studies (McKeon et al. 2000; Scanlan et al. 2011; Scanlan et al. 2013) have provided one approach to addressing this challenge, by representing the impact of grazing on management on pasture composition, pasture cover and pasture production.

Estimates of LTCC are an important component of property management decisions, especially with regard to the impact of multi-year wet and dry periods on grazing enterprises and resources. The LTCC estimates provide a guide for graziers around which stock numbers can be varied to match forage availability, and for more conservative managers, concentrating on reducing the impact of dry years in order to maintain or improve property land condition. Historically, government resource managers, graziers and their advisors have had difficulty in estimating LTCC, unless they have had access to successful long-term experiences. Since the 1990s, procedures and models have been developed to extrapolate the successful experience from benchmark properties. This paper has described how the on-property methods used in the 1990s in south-west Queensland have been developed for a Queensland wide application, using the latest scientific methods of pasture modelling, remote sensing technologies and online delivery system (i.e. FORAGE; Zhang et al. 2021).

In reviewing the basis of the online LTCC system, we found that the simple formula (pasture growth kg DM ha⁻¹ year⁻¹ × safe pasture utilisation %) ÷ animal intake kg head⁻¹ year⁻¹) has provided a very useful approach, despite the large spatial variability in climate, pasture communities, soils and woody vegetation cover across Queensland. The calculation of the formula’s components has been underpinned by the combination of field research, grazier experience and expert opinion. The estimation of pasture growth (using the GRASP model) has been improved by field research and the development of land type parameter sets of key biophysical variables (e.g. potential nitrogen uptake, available soil water range). Improved models of competition (for water and nutrients) between pasture and woody vegetation have been developed using empirical relationships derived from field studies and physical measurements of soil water balance. The combination of data from grazing trials, benchmark properties and regional experts has allowed the estimation of safe pasture utilisation rates across Queensland. Thus, the online LTCC system represents the continuing accumulation of grazing land science and grazier experience over the last 30 years.

We found that technological advances in computing and remote sensing data acquisition have allowed the online system to overcome some of the major limitations of previous approaches. In addition, we found that previous collation of pasture growth measurements and land type parameter sets has provided the necessary base for the modelling components for the online system. The major improvement in the online system is the rapidity and ease that property data (i.e. cadastral, land type and woody cover mapping) can be analysed and included in a report, including the calculation of LTCC. Although the online system has only been operational since November 2020, it is already proving to be a logical basis for discussion support with grazier managers and their advisors.

However, knowledge of LTCC is only one component of grazing land management; tactical response to climatic extremes, especially severe droughts, will always be necessary including management of land condition recovery. The necessity and capability to change stock numbers (year-to-year) are likely to vary with region (i.e. long-term rainfall variability) and enterprise flexibility (e.g. breeding component). Livestock enterprises that are flexible enough to match the fluctuations of the market and pasture availability have the capability to vary stock numbers around LTCC. To meet this tactical need, the online LTCC system is also supported by products such as pasture budgeting (e.g. the QDAF Stocktake GLM application) and other FORAGE online reports (e.g. Ground Cover and Pasture Growth Alert reports), to allow for more informed real-time stocking rate decisions.

The calculation of LTCC for Queensland properties remains a challenging task, given the large spatial variation in topography, soils and vegetation communities, and high temporal variability in climate (especially rainfall). Thus, the online system represents an important stage in the evolution of the calculation of grazing capacity to achieve ecological sustainability and financial viability.

Ms Giselle Whish is a Guest Associate Editor of this issue of The Rangeland Journal and is also a co-author of this paper. Ms Whish was blinded from the peer-review process for this manuscript before its acceptance for publication.