Modelling Seasonal and Inter-annual Variations in Carbon and Water Fluxes in an Arid-Zone Acacia Savanna Woodland, 1981-2012

Changes in climatic characteristics such as seasonal and inter-annual variability may affect ecosystem structure and function, hence alter carbon and water budgets of ecosystems. Studies of modelling combined with field experiments can provide essential information to investigate interactions between carbon and water cycles and climate. Here we present a first attempt to investigate the long-term climate controls on seasonal patterns and inter-annual variations in water and carbon exchanges in an arid-zone savanna-woodland ecosystem using a detailed mechanistic soil-plant-atmosphere model (SPA), driven by leaf area index (LAI) simulated by an ecohydrological model (WAVES) and observed climate data during 1981-2012. The SPA was tested against almost 3 years of eddy covariance flux measurements in terms of gross primary productivity (GPP) and evapotranspiration (ET). The model was able to explain 80 and 71% of the variability of observed daily GPP and ET, respectively. Long-term simulations showed that carbon accumulation rates and ET ranged from 20.6 g C m(-2) mon(-1) in the late dry season to 45.8 g C m(-2) mon(-1) in the late wet season, respectively, primarily driven by seasonal variations in LAI and soil moisture. Large climate variations resulted in large seasonal variation in ecosystem water-use efficiency (eWUE). Simulated annual GPP varied between 146.4 and 604.7 g C m(-2) y(-1). Variations in annual ET coincided with that of GPP, ranging from 110.2 to 625.8 mm y(-1). Annual variations in GPP and ET were driven by the annual variations in precipitation and vapour pressure deficit (VPD) but not temperature. The linear coupling of simulated annual GPP and ET resulted in eWUE having relatively small year-to-year variation.