A simple process-based model was applied to a tall Eucalyptus forest site over consecutive wet and dry years to examine the importance of different mechanisms linking productivity and water availability. Measured soil moisture, gas flux (CO2, H2O) and meteorological records for the site were used. Similar levels of simulated H2O flux in 'wet' and 'dry' years were achieved when water availability was not confined to the first 1.20 m of the soil profile, but was allowed to exceed it. Although the simulated effects of low soil and atmospheric water content on CO2 flux, presumably via reduction in stomatal aperture, also acted on transpiration, they were offset in the dry year by a higher vapour-pressure deficit. A sensitivity analysis identified the processes that were important in wet versus dry years, and on an intra-annual timeframe. Light-limited productivity dominated in both years, except for the driest period in the dry year. Vapour-pressure deficit affected productivity across more of each year than soil moisture, but both effects were larger in the dry year. The introduction of a reduced leaf area tended to decrease sensitivity in the dry year. Plant hydraulic architecture that increases plant available water, maximises productivity per unit water use and achieves lower sensitivity to low soil moisture levels should minimise production losses during dry conditions.