Quantifying the partitioning of evapotranspiration (ET) and its controls are particularly important for
accurate prediction of the climatic response of ecosystem carbon, water, and energy budgets. In this
study, we employed the Shuttleworth–Wallacemodel to partition ET into soil water evaporation (E) and
vegetation transpiration (T) at four grassland ecosystems in China. Two to three years (2003–2005) of
continuous measurements of ET with the eddy covariance technique were used to test the long-term
performance of the model. Monte Carlo simulations were performed to estimate the key parameters in
the model and to evaluate the accuracy in model partitioning (i.e. E/ET). Results indicated that the
simulated ET at the four ecosystems was in good agreement with the measurements at both the diurnal
and seasonal timescales, but the model tended to underestimate ET by 3–11% on rainy days, probably
due to the lack of model representation of rainfall interception. In general, E accounted for a large
proportion of ET at these grasslands. The monthly E/ET ranged from 12% to 56% in the peak growing
seasons and the annual E/ET ranged from 51% to 67% across the four ecosystems. Canopy stomatal
conductance controlled E/ET at the diurnal timescale, and the variations andmagnitude of leaf area index
(LAI) explained most of the seasonal, annual, and site-to-site variations in E/ET. A simple linear
relationship between growing season LAI and E/ET explained ca. 80% of the variation observed at the four
sites for the 10 modeled site-years. Our work indicated that the daily E/ET decreased to aminimum value
of ca. 10% for values of LAI greater than 3 m2 m2 at the ecosystem with a dense canopy. The sensitivities
of E/ET to changes in LAI increased with the decline in water and vegetation conditions at both the
seasonal and the annual time scales, i.e., the variations in LAI could cause stronger effects on E/ET in the
sparse-canopy ecosystems than in the dense-canopy ecosystems. It implies that the hydrological
processes and vegetation productivity for ecosystems in arid environmentsmight bemore vulnerable to
projected climate change than those in humid environments.