Riparian vegetation varies according to hydrogeomorphic processes operating across different scales over two didmensions: transversely (across-stream) and longitudinally (parallel to stream). We tested the hypothesis that vegetation patterns reveal the scale and direction of underlying processes. We correlated patterns of dominant woody vegetation with environmental variables at 28 sites located within four geomorphologically distinct regions along the length of the South Alligator River catchment of Kakadu National Park, northern Australia. Across the catchment there existed a strong transverse boundary between upland savanna vegetation and two zones of riparian vegetation: Melaleuca-spp.-dominated closed-forest vegetation along stream channels and mixed open-woodland vegetation adjacent to closed forest. We surmise that there is hierarchic constraint on smaller-scale catchment processes due to fire incursion into the riparian zone and access to water during the dry season. Within the closed-forest zone, vegetation did not vary transversely, but did longitudinally. Riparian woodlands also varied longitudinally, but in the upper reaches varied independently of stream variables. By contrast, in the lower reaches woodland was strongly correlated with stream variables. The observed pattern of weak transverse linkages in headwaters but strong linkages in lower reaches is analogous to models developed for in-stream patterns and processes, particularly the river continuum and flood-pulse concepts.

A modelling approach to simulating vegetation dynamics is described, incorporating critical processes of carbon sequestration, growth, mortality and distribution. The model has been developed to investigate the responses of vegetation to environmental change, at time scales from days to centuries and from the local to the global scale. The model is outlined and subsequent tests, against independent data sources, are relatively successful, from the small scale to the global scale. Tests against eddy covariance observations of carbon exchange by vegetation indicated significant differences between measured and simulated net ecosystem production (NEP). NEP is the net of large fluxes due to gross primary production and respiration, which are not directly measured and so there is some uncertainty in explaining differences between observations and simulations. In addition it was noted that closer agreement of fluxes was achieved for natural, or long-lived managed vegetation than for recently managed vegetation. The discrepancies appear to be most closely related to respiratory carbon losses from the soil, but this area needs further exploration. The differences do not scale up to the global scale, where simulated and measured global net biome production were similar, indicating that fluxes measured at the managed observed sites are not typical globally.

The model (the Sheffield Dynamic Global Vegetation Model, SDGVM) has been applied to contemporary vegetation dynamics and indicates a significant CO2 fertilisation effect on the sequestration of atmospheric CO2. The terrestrial carbon sink for the 20th century is simulated to be widespread between latitudes 40° S and 65° N, but is greatest between 10° S and 6° N, excluding the effects of human deforestation. The mean maximum sink capacity over the 20th century is small, at 25 gC m−2 year−1, or approximately 1% of gross primary production.

Simulations of vegetation dynamics under a scenario of future global warming indicate a gradual decline in the terrestrial carbon sink, with the capacity to absorb human emissions of CO2 being reduced from 20% in 2000 to approximately 2% between 2075 and 2100. The responses of carbon sequestration and vegetation structure and distribution to stabilisation of climate and CO2 may extend for up to 50 years after stabilisation has occurred.

During the twentieth century the level of the Caspian Sea dropped from -26 m (1930) to -29 m (1977) below global sea level and subsequently rose again to -26.66 m in 1996. We aimed to describe responses of the vegetation in the lower Volga Delta to these substantial sea-level changes using an analysis of historic vegetation maps produced by aerial photography and satellite imagery.

The sea level drop in the earlier part of the century was followed by rapid progression of the vegetation. The subsequent rapid sea-level rise in the 1980s did however not result in similarly rapid regression of the vegetation. This partial irreversibility of the vegetation response to sea-level change is explained by the wide flooding tolerance of the major emergent species, namely Phragmites australis. Floating vegetation increased in extent, most likely due to the increased availability of more favourable conditions, particularly for Nelumbo nucifera, a tropical plant reaching its northernmost distribution in the Volga Delta. This species increased in distribution from 3.5 ha in the 1930s throughout the entire Volga Delta to several thousands of hectares in the Astrakhanskiy Biosphere Reserve alone in the 1980s. The reported sea-level changes swept the ecosystems in the Astrakhanskiy Biosphere Reserve back and forth within the Reserve boundaries. At longer time scales, ten-fold greater sea-level change has been reported. The ecosystems for which the Reserve is renowned might be pushed completely out of the Reserve under these conditions. We therefore question whether the current Reserve will be sufficiently large to guarantee conservation of the biota in the lower Volga Delta at longer time scales.