Effects of vegetation processes on water resources at global and continental scales

Global environmental change is expected to alter the spatial and temporal distribution of water resources but quantifying its effects on the terrestrial water balance remains a challenge. Evapotranspiration (ET) is a key ecosystem variable linking water, energy and carbon cycles, but although global aggregate ET is expected to increase in a warming climate, regional changes in ET have remained particularly poorly constrained due in part to difficulty in measuring ET. The effects of vegetation and increasing atmospheric CO₂ represent a further complication. Experiments have shown that elevated atmospheric CO₂ affects vegetation productivity and water use, but it remains uncertain whether these processes have led to detectable changes in ET or runoff at ecosystem scales. This thesis aims to quantify large-scale variations in ET, and, in particular, to better constrain the effects of elevated CO₂ on catchment-scale hydrology. The approach relies on observations to the greatest extent possible, and the water-balance method (the difference of precipitation and streamflow at the river catchment scale) is used to quantify ET throughout, as it remains the most firmly observationally based measure of ET. The first data chapter of the thesis attributes causes of water-balance ET trends globally by considering all key drivers, including climate, land use and vegetation effects. The study statistically accounts for ET trends in energy-limited "wet" and water-limited "dry" river catchments and finds ET is primarily controlled by precipitation in both environments (with precipitation explaining 45-55% and 80-95% of interannual variability and trends in ET in wet and dry catchments, respectively). There is some evidence for vegetation effects based on model simulations, but the detection and attribution of ET trends is hindered by uncertainties in the data available for global analysis and the lack of direct observations of vegetation properties covering the study period 1960-1999. Making use of high quality streamflow observations for Australia, the second chapter quantifies recent CO₂ effects on runoff, ET and vegetation cover across Australian river catchments in varying climates. The CO₂ effect was quantified by cncurrently analysing vegetation cover (Normalised Difference Vegetation Index) and water-balance ET, relying solely on observations. Contrary to common expectation, the analysis shows that vegetation is not only greening, but also consuming more water in sub-humid and semi-arid climates, leading to significant (24-28%) reductions in runoff over the period 1982-2010. The analysis pointed to increased runoff in the wet and arid climates but the results were not statistically significant and it was thus not possible to detect a CO₂ effect in the wettest or driest climates based on the measurements. Finally, future water resources in Australia are projected using bias-corrected state-of-the-art climate model projections to drive the Land Surface Processes and eXchanges dynamic global vegetation model (LPX). LPX explicitly simulates vegetation CO₂ responses and is shown to capture observed CO₂ effects on ET. LPX results are contrasted with projections from the empirical Budyko framework, which accounts for only the climatic effects on ET. Future precipitation patterns remain highly uncertain across much of Australia, but predicted vegetation effects are robust. Vegetation is shown to buffer the effects of climate change, alleviating water stress in highly poplated regions with robust declines in precipitation due to CO₂-induced water savings. In northern Australia, CO₂ fertilisation reduces runoff but is accompanied by increased vegetation productivity. The findings of this thesis highlight the importance of considering vegetation in studies of water resources. Recent increases in atmospheric CO₂ are shown to have left a detectable imprint on Australian ecosystems, and the impacts are projected to continue to the end of the 21st century and beyond.