Transpiration and evaporation in a Californian oak-grass savanna: Field measurements and partitioning model results

Abstract

As the eddy-covariance technique enables intensive measurements of evapotranspiration (ET) at the ecosystem level, the interest in further partitioning ET into two main process-based components transpiration (T) and surface evaporation (E) □ is increasing. Although models for partitioning tower-measured ET have been developed, their reliability for different types of ecosystems still requires extensive validations. From 2001 to 2019, we measured CO2 and H2O vapor fluxes over an oak-grass savanna landscape from three eddy-covariance towers (i.e., one over an oak woodland; the other two over annual grasslands under tree canopy and in open area). Annual ET (± standard deviation) from the oak woodland, understory grassland, and open grassland was 419±85 mm, 167±36 mm, 324±43 mm, respectively. The differences between the above- and below-canopy ET indicated that oak canopy transpiration (Toak) was 281±48 mm year−1, accounting for 67±8% of the total ET of the woodland. The Toak/ET ratio varied in seasons, similar to the pattern of oak's leaf area index but opposite to that of soil moisture. We then tested two ET-partitioning models: Scott's long-term-regression-interception (LTRI) model (Scott and Biederman, 2017) and Zhou's quantile-regression-maximum-slope (QRMS) model (Zhou et al., 2016). Even though we expected that the two models would give divergent results since theiremo working principles, both models captured reasonable magnitudes and seasonal patterns of the T/ET ratio, as suggested by tower measurements. The study confirms that the LTRI and QRMS models are applicable for savanna ecosystems, but some modifications are necessary for tree dominated areas. In combination with field and modeling approaches, this study improves our understanding on the contributions of transpiration and evaporation to total ET from ecosystems with vertical vegetation layers.