Evapotranspiration components determined by stable isotope, sap flow and eddy covariance techniques

Understanding and modeling water exchange in arid and semiarid ecosystems is complicated by the very heterogeneous distribution of vegetation and moisture inputs, and the difficulty of measuring and validating component fluxes at a common scale. We combined eddy covariance (EC), sap flow, and stable isotope techniques to investigate the responses of transpiration and soil evaporation to an irrigation event in an olive (Olea europaea L.) orchard in Marrakech, Morocco. The primary goal was to evaluate the usefulness of stable isotope measurements of water vapor in the turbulent boundary layer for partitioning evapotranspiration under such dynamic conditions. The concentration and deuterium isotope composition (δ2H) of water vapor was collected from different heights within the ecosystem boundary layer of the olive canopy before and over several days following a 100 mm surface irrigation. 'Keeling plots' (isotope turbulent mixing relationships) were generated from these data to estimate the fractions of evaporation and transpiration contributing to the total evapotranspiration (ET) flux. Transpiration accounted for 100% of total ET prior to irrigation, but only 69-86% of ET during peak midday fluxes over the 5-day period following irrigation. The rate of soil evaporation and plant transpiration at the stand level was calculated from eddy covariance measurements and the evaporation and transpiration fractions from isotope measurements. Soil evaporation rate was positively correlated with daily atmospheric vapor pressure deficit (D), but transpiration was not. Component fluxes estimated from the isotope technique were then compared to those obtained from scaled sap flow measurements. Sap flow in multiple-stemmed trees increased following the irrigation, but large single-stemmed trees did not. We matched the source area for eddy covariance estimates of total ET fluxes with scaled sap flow estimates developed for the different tree types. Soil evaporation was determined from the difference between total ET and the scaled sap flow. Ecosystem-level transpiration and soil evaporation estimated by the isotope approach were within 4 and 15% of those estimated by scaled sap flow, respectively, for periods of peak fluxes at midday. Our data illustrate the utility of the isotope 'Keeling plot' approach for partitioning ET at the ecosystem scale on short time steps and the importance of accurate spatial representation of scaled sap flow for comparison with eddy covariance measurements of ET. © 2004 Elsevier B.V. All rights reserved.