Evapotranspiration assessment of a mixed temperate forest by four methods: Eddy covariance, soil water budget, analytical and model

- Evapotranspiration fluxes (ET) were calculated by 4 methods in a temperate forest.
- ET fluxes partitioning was measured and modeled for beech and Douglas-fir.
- Estimates of errors on each method were developed based on Taylor propagation laws.
- Each method is necessary in more or less long term to characterize the forest water balance.

In forest ecosystems, assessment of evapotranspiration fluxes (ET) and distinction of its components, i.e. tree transpiration (T), rainfall interception (I), and soil plus understory evapotranspiration (ETu), are a major issue. At the Vielsalm Terrestrial Observatory in Belgium (VTO, Integrated Carbon Observation System network, ICOS), covered by a mixed forest of broadleaved (mainly Fagus sylvatica L.) and coniferous species (mainly Pseudotsuga menziesii [Mirb.] Franco), the water vapor fluxes have been continuously measured by eddy covariance since 1996, without distinction of its components. Widely validated for CO2 fluxes, water vapor fluxes measured by eddy covariance still lack validation, particularly in mixed and/or heterogeneous stands. During 2010-2011, three other methods to assess ET were implemented, in order to inter-validate them and to disentangle species specific ET into its components. These methods were: (i) the analytical method which relies on measurement of each elementary flux, i.e. tree transpiration, interception loss and evapotranspiration of soil plus understory (ETA); (ii) the soil water budget method (ETS) and (iii) modeling of stand ET (ETM).Our results showed that during dry foliage conditions, stand T + ETu estimated by the analytical method or model were in very good agreement with eddy covariance ET measurements. Interception loss measured was 14% and 30% of rainfall (P) for beech and Douglas-fir respectively, leading to 19% for the whole stand. Beech interception being quite low, ETA for the stand is probably slightly underestimated. Over 2010 and 2011, the mean seasonal value of ET was 1.9 mm d−1, considering all the methods. The four methods gave close estimates, with maximum deviations from the mean being of 12%, and uncertainties, while remaining acceptable (maximum 26% with analytical method) overlap between methods. Regardless of methods, flux measurements during rainy conditions were more complex to characterize. A slight underestimation of ET by the eddy covariance method was observed at different time step, in comparison to the others implemented methods. On an annual scale, the forest water balance model estimates (BILJOU©; ETM) were similar in beech (68% of potential evapotranspiration, PET) and Douglas-fir sub-stands (72% of PET), despite their differences in phenology and water use patterns. Low differences were also observed between species in the water flux partitioning, with modeled ET composed of ca. 58%, 33% and 9% of T, I and ETu respectively on the annual timescale (2010-2011).