The central west plains (CWP) grasslands of New South Wales have adapted over millions of years to climate and rainfall variability. Understanding how these grasslands function offers insight into the response of agricultural production systems attuned to the unique characteristics of the CWP. This thesis examines the structure and function of CWP native grazed grasslands addressing the hypothesis that these grasslands have an underlying hill slope patch structure that results in higher biomass production. Water from rainfall events when redistributed through this mosaic (structure) of resource patches shedding (source) and resource capturing patches (sink), results in greater growth than would otherwise occur if rainfall were evenly distributed. The research was conducted at Myola, Trundle (average annual rainfall 490 mm), on the CWP of NSW. The principal research approach was the use of a chronosequence substitution of spacefor-time with six grassland-monitoring sites. Each site had five permanent 30-m transects and represented a different period of recovery from cropping, 4 to 40 years. Data was collected at the end of the winter and summer growing seasons over a 4-year period commencing in 2010.

Chapter 3 addressed whether CWP grasslands are patchy and, if so, what patch types can be distinguished and what soil features best discriminate between these patches? Decision tree partitioning, based on the eleven Soil Surface Assessment (SSA) indicators developed by Tongway and Hindley (2004), was undertaken to aggregate the 15 a priori patch types into six a posteriori patch types – bare sealed (Bs), annual (Als), crusted perennial (Cp), perennial (P), high cover and high surface roughness (CovR) and low resistance to disturbance (Lrd). These patch types were significantly different in five of the 11 SSA indicators – crusting, perenniality, litter score, biological soil crust (BSC) cover and surface roughness. These five SSA indicators were used in a framework for the rapid field assessment of the six different patch types.

Chapter 4 examined how the six patch types differ from each other in terms of Landscape Function Analysis indices, infiltration, soil moisture, nutrient availability and soil microbial characteristics, plant species presence and biomass production. Large and significant differences were found between the different patch types in biomass production and species composition. Biomass production in CovR (6235 ± 263 kg/ha) and Lrd (6683 ± 93 kg/ha) patch types was nearly double that of other patch types. Patches also had significant differences in sorptivity, infiltration, depth of wetting, hydrophobicity, BSC composition and abundance, nutrient availability, patch size and patch position in the landscape. It was evident that CovR and Lrd patch types behave as sink patches most of the time, as do P patches some of the time depending on rainfall. Patches immediately downslope of CovR patches had a lower wetting front depth and less topsoil moisture, which suggested that CovR behave as sink patches. Bs, Cp and Als patches had significantly less depth of wetting and topsoil moisture than patches immediately downslope, providing evidence that these patches behave as source patches.

Chapter 5 examined the spatial structure and juxtaposition of patch types and related this to biomass production. The chapter also looked at the relationship between the dominant late post-disturbance recovery patch types (P and CovR) and how patch structure and rain event characteristics – amount, timing, duration and intensity – interact to influence biomass production. The monitoring site transects were examined for patch structure (i.e. the changing proportion of Bs, Als, Cp, P, CovR and Lrd patches) at different topographic positions down the hill slope. Biomass production was highest when the hill slope contained 10–30% source patches, 50–80% P patches and 30–50% CovR patches. Patch structure varied spatially by both hill slope position and site disturbance history. Bs and Als patch types were more abundant in early post-crop recovery sites while CovR patches type were significantly more frequent in sites 20 and 40 years post-crop than at 4 and 6 years post-crop. Patch type juxtaposition was clearly defined. For example, CovR patches were nested within larger P patches and were preceded and followed by P patches on 89% and 92% of occasions, respectively. CovR biomass decreased from 6900 kg/ha in quadrats near to upslope source patches (Bs, Als or Cp) to 5600 kg/ha in quadrats 16 m from these upslope patch types. CovR patch biomass was highest when transect P patch proportion was in the range of 40–80%. A generalised linear model found high rainfall-intensity best explained CovR patch type soil moisture at both 30 and 75 cm depth but rainfall event duration provided a better explanation of soil moisture at 30 and 75 cm depth in other patch types. Mean subsoil moisture (75 cm depth) carryover from winter to spring was nearly 45% greater in CovR than any other patch types. Mean patch type aggregate seasonal rainfall use efficiency (RUE) was 9.1 kg/ha/mm. CovR patch type RUE was 14.0 kg/ha/mm in winter and 20.4 kg/ha/mm in summer. There was a summer season rainfall threshold of about 70 mm, below which no biomass production occurred. It was concluded that biomass production of CWP grazed grasslands is influenced by spatial patch structure and rainfall characteristics – amount, duration, intensity frequency and timing. Individual patches types form a mosaic of source and sink patches and the resulting redistribution of rainfall results in large and significant differences in patch type biomass production. Spatial patch structure affects the amount and location of CWP grassland biomass.

Chapter 6 studied changes in patch structure over time and examined the drivers of patch transition from a less productive patch type to a more productive patch type or vice versa. The research questions examined the long-term (>50-years recovery from disturbance) dynamics of sink patches and the influence of disturbance intensity and duration and seasonal influences. Species transitions were examined as the likely cause of patch transitions, and disturbance intensity and duration were examined as a likely influence on species composition. Substitution of space-for-time was used and specific patches followed over a 5-year period. Patch type progressed from Als dominance immediately following disturbance to dominance of P and CovR patch types in the range of 20–60% of each, respectively, depending on antecedent seasonal conditions. This progression was first evident in lower-slope positions and moved upslope over time. Over 60% of transitions from one patch type to another occurred at the edge of patches. Species growth and litter characteristics were more influential than rainfall event characteristics in explaining transitions. Changes in litter score were observed in a high proportion (60–90%) of patch types that had progressed to a more functional patch type (litter score trend increasing) or regressed to a less functional patch type (litter score trend decreasing). Extrapolation of patch transition probabilities derived from Bayesian Belief Network (BBN) analysis indicated that patch-type progression and regression resulted in a basin of attraction for patch composition, given Trundle rainfall characteristics, of 25% CovR, 47% P, 18% Cp and 10% Als, 15–20 years post-cropping disturbance. In the absence of grazing, patch type composition was similar after 30 years of recovery from disturbance regardless of disturbance type. However species composition in the 30-year post disturbance sites under grazing differed with disturbance history. This indicates that seed availability and disturbance history (grazing vs cropping and subsequent recovery) can modify species composition trajectory over time, but that succession in patch type composition follows a similar path regardless of differences in within-patch type species composition. Disturbance intensity (degree and duration) was a more important influence on recovery than disturbance type. In conclusion, grassland recovery from disturbance is characterised by progression from low litter cover patch types (Bs, Als, Cp) where surface sealing under raindrop impact restricts water infiltration to high litter cover patch types (P, CovR, Lrd) with higher water infiltration. In the absence of disturbance, CovR patch types in the range of 50– 80% dominate CWP grasslands. Patch progression in grasslands subject to grazing restricts CovR proportion to the range of 20–30%.

This study extended the understanding of grassland heterogeneity in general and the functioning of CWP grasslands in particular. Patch dynamics were explored – how different patches form and patch types transition over time forming the underlying patch mosaic driving grassland productivity, stability and resilience. The interaction of this patch mosaic and rainfall (amount, timing, duration and intensity) and the resulting effects on CWP grassland productivity were examined. In summary, CWP grassland hill slopes comprise a mosaic of sink and source patches, which differ in water infiltration, species composition and biomass production. These patches form in response to complex feedback dynamics between grass species, physical and biological crust formation, herbivore off take, plant–soil biology associations and nutrient availability, collectively called vegetation-driven spatial heterogeneity.

This research has provided the basis for the further examination of stability and resilience of CWP grasslands. Sufficient insight has been gained to develop and validate an agent-based model (ABM) of these grasslands. ABMs are used to examine emergent behaviour in complex adaptive systems, such as CWP grasslands, and can be used in simulated experiments, for example, to determine optimal patch structure for CWP grassland production stability.

Two percent organic carbon (OC) (grams C per gram soil) provides a threshold for soil stability; below this threshold, soils become highly erodible as macroaggregates slake to form microaggregates. However, it is a long-held belief that a degree of slaking upon rapid wetting of aggregates is an inherent trait of vertosols, regardless of OC content. This is attributed to their physico-chemical attributes (e.g. high clay content, shrink-swell capacity, cation exchange capacity and pH). Studies investigating the erodibility of vertosols have concentrated on cropping soils, usually with low OC content (≤2%). Therefore, the importance of OC in maintaining structural stability and minimising erosion in vertosols is often dismissed. This study examined vertosol macroaggregate and microaggregate stability in natural ecosystems where OC can be >2%. We found a positive relationship between macroaggregate stability and OC content in vertosols, especially when OC was ≥3.5% in the surface soil (0-5 cm). Microaggregate stability was attributed to the dominance of Ca2+ over Na+ on clay-exchange sites. OC was positively correlated with Ca2+ and negatively correlated with Na+ and ESP. OC may play a role in microaggregate stabilisation through its capacity to lower soil pH and increase the availability of Ca2+. We demonstrate that OC can stabilise vertosol aggregates, and is therefore important in preventing erosion on this soil type.