Accurate estimations of forest evapotranspiration (ET) and its components, transpiration (T) and evaporation (E), are important for deep understanding and predicting the responses of forest water cycles to climate change. In this study, the improved Shuttleworth-Wallace model (SWH) was applied to estimate ET, T, and E during 2003–2014 in a subtropical planation, and the modeled results were verified using in situ measurements by the eddy covariance technique, sap flow, and micro-lysimeter method. The study aimed to clarify whether it is feasible and reliable to use the SWH model to estimate and partition ET in forests. In addition, depending on the long-term data, the specific performances in modeling ET under different climatic backgrounds were investigated, and the underlying mechanisms were explored. The results verified that the SWH performed relatively well in the subtropical forest, and the modeled ET, T and E could track the seasonal variations, although overestimations were found in the peak seasons. However, the model was relatively weaker in estimating the interannual variabilities. It performed well in modeling ET in normal years but showed larger model residuals in years with obvious climatic anomalies. In the severe summer-drought (2003) and cold-spring (2005) years, the model greatly overestimated ET. It also overestimated ET in summer since 2010, which may be ascribed to the less dependency of ET on VPD induced by the more humid microclimate in forest accompanied with forest development. For the ET partitioning results, the modeled and measured E and T values were all in reasonable ranges. The possible reasons for underestimations (overestimations) of E and T by measurements (SWH model) were discussed. In this study, the data obtained using different methods and from different scales matched each other and could be cross validated, and the discussion on discrepancies would be beneficial for understanding the advantages and flaws of different methods and could be the basis for optimizing the measurement and model methods. In sum, this study verified that it is feasible to use the SWH model in forests and provided a basis for further improving and optimizing the modeled results under different climate backgrounds.

No abstract