Arid and semi-arid rangelands are characterised by strong seasonality, inter-annual rainfall variation and periodic droughts. Understanding the relationships between spatial and temporal variation in rainfall, forage availability and the dynamics of herbivore populations is crucial to inform management. We monitored plant productivity and composition in the Richtersveld National Park (RNP), as well as body condition and demographic rate data on 500 tagged female goats, to disentangle the effects of environmental stochasticity and density-dependent processes. During the four-year study period, annual rainfall varied nearly four-fold from 230 mm in 2006 to 60 mm in 2009, with considerable spatial variation across the RNP. Goats utilize areas of seasonally green Succulent Karoo vegetation around stock posts in the wet winter months, and evergreen woody vegetation in the riparian zone of the Orange River during the dry summer. Plant productivity in the wet season range responded strongly to rainfall, whereas it was more stable in the riparian zone. Grazing impacts on the wet season vegetation were evident in the vicinity of stock posts, where perennial plant cover had decreased and became dominated by unpalatable species. Composition and productivity of annuals responded primarily to rainfall, with no evidence of increased production of annuals near stock posts where perennial cover had decreased. Goat populations showed a density-dependent seasonal decline in body condition in response to depletion of the dry-season forage in the riparian zone. Loss in body condition reduced pregnancy rates, litter sizes, and pre-weaning survival. Survival was lowest following the most severe dry season and for juveniles. Population growth was determined by the length of the dry season and population size in the previous year. Within the riparian zone, browse was depleted more rapidly lower down in the canopy and where goat densities were greater, producing a distinct browse line that moved higher as the dry season progressed. Smaller goats thus experienced more acute forage deficits earlier in the dry season, leading to asymmetric effects on individuals of different sizes. Our research shows that herbivore populations in arid and stochastic environments are not in disequilibrium with their forage resource, but that consumer-resource coupling is largely restricted to a subset of resources in the seasonal cycle. In the case of the RNP, the goat population is in dynamic equilibrium with their dry season forage resource, and the riparian zone thus constitutes the key resource of the population. This has a number of management implications in the RNP and other arid areas. The long-term carrying capacity of the area as a whole is largely determined by the availability and stability of forage in the key resource. Identifying a population's key resource and ensuring access to it is thus crucial for maintaining the population trajectories of wild and domestic herbivores. Larger, more productive and more stable key resource areas support higher herbivore populations, which influences the intensity of herbivory on the wet season resource. Understanding the spatial and temporal dynamics between herbivores and their wet-and dry-season resources is thus important for managing herbivore populations and their impacts.

Biome boundaries offer valuable opportunities to monitor climate or land-use driven vegetation change to detect shifts in vegetation regimes. Reports of change in historically over-utilised natural rangelands in South Africa have suggested that under persistent livestock overgrazing, the drier dwarf shrublands have encroached further eastwards into the grasslands. Historical vegetation surveys and landscape photographs from the mid-20th century collected from sites spanning the Nama-Karoo – Grasslands Biome boundary have provided the opportunity to monitor these environments, thereby revisiting the 'expanding Karoo hypothesis' through repeated observation in time and space. The aim of this study was to identify the dominant vegetation changes in terms of i) species composition and growth form abundance, ii) rangeland condition (based on veld condition score) and iii) grazing capacity (ha/Large Stock Unit) of the sampled area, at 27 study sites spread across an east-west gradient of decreasing precipitation. The descending-point method was used to record initial plant species counts from 1000 points per transect at sites between 1956 and 1962 and the same method was used in 2017–18. Total canopy cover per species from each survey was used to calculate the percent change in all growth forms over time. Non-metric multidimensional scaling analysis using the Bray-Curtis distance measure was also employed to examine site dissimilarities and trajectories of vegetation change in relation to the preceding 20-year mean annual precipitation and mean annual temperature at Karoo (K), Grassland (G) and Escarpment (E) sites. Overall, mean percent change in plant cover has increased significantly at all sites (p < 0.05; K = 11.21% ± 3.29; G = 7.81% ± 2.48; E = 5.19% ± 1.88). This increase is characterised by an increase in perennial grasses (K: = 31%; G = 35%; E = 22%) and perennial shrubs (K = 22%; G = 2%; E = 4%). Comparisons of site dissimilarity between years indicate that sites have followed a similar change of trajectory between K, G and E vegetation units and that there is a greater distinction between the cool, temperate E sites, and the K and G sites. Veld condition scores indicate an improvement at all K (1960 = 55.54; 2018 = 295.60), G (1960 = 190.98; 2018 = 503.24) and E (1960 = 222.28; 2018 = 342.08) sites. Comparably, grazing capacity scores of surveyed areas have improved over time (K: 1960 = 105.81 ha/LSU, 2018 = 20.05 ha/LSU; G: 1960 = 26.49 ha/LSU, 2018 = 9.67 ha/LSU; E: 1960 = 29.48 ha/LSU, 2018 = 14.32 ha/LSU), indicating that more area was required to graze one LSU historically, than is needed today as a result of the improvement in veld condition. These findings are at odds with changes observed in most other shrubland-grassland ecotones where shrubs have generally replaced grasses in recent decades. Due to a complex mix of interacting climatic and land use changes and feedbacks, rangelands in the eastern grassy Karoo have experienced a regime shift to grassland-dominant vegetation, with an overall improvement in condition. 

The influence of rainfall and grazing in the Karoo are well documented. Good rains allow increased plant growth in general, and increased grass growth in particular. Droughts can kill both grasses and dwarf shrubs, with the former being more susceptible to water stress. Historically, most grazing effects arose from sporadic large migrations of springbok. In the late 1800s, livestock were introduced and numbers increased to cause novel levels of herbivory. Concomitantly the incidence of grazing by springbok declined. Dwarf shrubs are primarily of the C3 photosynthetic pathway, and can grow when water is available nearly throughout the year. Grasses, which are predominantly C4, grow mainly during the warm midsummer months if sufficient moisture is available. Accordingly, grasses can be damaged if grazed in summer, and be largely unaffected by winter grazing as they are dormant then. The influence of summer grazing on dwarf shrubs is much more variable owing to variations in palatability, growth form, and patterns of deciduousness. The objective of this research was to understand the influences of season of grazing and of rainfall on vegetation composition, and on grass:shrub balances in the eastern Karoo. Plant compositional data from long-term trials situated at the Grootfontein Agricultural Development Institute and collected from the mid-1940s to 1970 were used. Grazing treatments varied in the proportion of grazing that took place during summer (October–March) and winter (April–September) respectively. Annual rainfall decreased over the period, with exceptionally dry periods occurring around 1950 and in the late 1960s. Eighty-six plant species from twelve growth-forms were encountered, with the most common perennial species from the most common growth forms being Chrysocoma ciliata and Pentzia incana (dwarf shrubs) and Eragrostis lehmanniana and Tragus koelerioides (grasses), and the most common short-lived species was the grass Aristida congesta. In general, grasses, in particular E. lehmanniana and A. diffusa, responded well to winter grazing, while summer grazing was associated with the shrubs P. incana, C. ciliata and Phymaspermum parvifolium. Creeping grasses showed differing responses to rainfall trends, with Cynodon incompletus decreasing and T. koelerioides increasing in abundance over time. Short-lived grasses were highly responsive to rainfall and were negatively impacted by summer grazing. Perennial grasses and dwarf shrubs declined over time as rainfall decreased. Droughts induced rapid declines in grass populations, which then recovered. Dwarf shrub populations collapsed following a minor drought that had followed a severe drought. Plant response to drought was largely unrelated to grazing. It was concluded that season of grazing and long-term rainfall trend strongly influenced perennial vegetation composition, and that season of grazing was a potent determinant of grass cover for most grass species, and influenced shrub cover. Furthermore, recent rainfall influenced mainly short- to medium-lived grasses. Long-term rainfall trend was associated strongly with cover of most growth forms and species, but the mechanisms of this remained unexplained.