Evapotranspiration partition using the multiple energy balance version of the ISBA-A-gs land surface model over two irrigated crops in a semi-arid Mediterranean region (Marrakech, Morocco)

The main objective of this work is to question the representation of the energy budget in soil-vegetation-atmosphere transfer (SVAT) models for the prediction of the turbulent fluxes in the case of irrigated crops with a complex structure (row) and under strong transient hydric regimes due to irrigation. To this end, the Interaction between Soil, Biosphere, and Atmosphere (ISBA-A-g(s)) is evaluated at a complex open olive orchard and, for the purposes of comparison, on a winter wheat field taken as an example of a homogeneous canopy. The initial version of ISBA-A-g(s) based on a composite energy budget (hereafter ISBA-1P for one patch), is compared to the new multiple energy balance (MEB) version of ISBA that represents a double source arising from the vegetation located above the soil layer. In addition, a patch representation corresponding to two adjacent, uncoupled source schemes (hereafter ISBA-2P for two patches) is also considered for the olive orchard. Continuous observations of evapotranspiration (ET), with an eddy covariance system and plant transpiration (T-r) with sap flow and isotopic methods were used to evaluate the three representations. A preliminary sensitivity analyses showed a strong sensitivity to the parameters related to turbulence in the canopy introduced in the new ISBA-MEB version. For wheat, the ability of the single- and dual-source configuration to reproduce the composite soil-vegetation heat fluxes was very similar; the root mean square error (RMSE) differences between ISBA-1P, ISBA-2P and ISBA-MEB did not exceed 10 W M-2 for the latent heat flux. These results showed that a composite energy balance in homogeneous covers is sufficient to reproduce the total convective fluxes. The two configurations are also fairly close to the isotopic observations of transpiration in spite of a light underestimation (overestimation) of ISBA- 1P (ISBA-MEB). At the olive orchard, contrasting results are obtained. The dual-source configurations, including both the uncoupled (ISBA-2P) and the coupled (ISBA-MEB) representations, outperformed the single-source version (ISBA-1P), with slightly better results for ISBA-MEB in predicting both total heat fluxes and evapotranspiration partition. Concerning plant transpiration in particular, the coupled approach ISBA-MEB provides better results than ISBA-1P and, to a lesser extent, ISBA-2P with RMSEs of 1.60, 0.90, and 0.70 mm d(-1) and R-2 of 0.43, 0.69, and 0.70 for ISBA-1P, ISBA-2P and ISBA-MEB, respectively. In addition, it is shown that the acceptable predictions of composite convective fluxes by ISBA-2P for the olive orchard are obtained for the wrong reasons as neither of the two patches is in agreement with the observations because of a bad spatial distribution of the roots and a lack of incoming radiation screening for the bare soil patch. This work shows that composite convection fluxes predicted by the SURFace EXternalisee (SURFEX) platform and the partition of evapotranspiration in a highly transient regime due to irrigation is improved for moderately open tree canopies by the new coupled dualsource ISBA-MEB model. It also points out the need for further local-scale evaluations on different crops of various geometry (more open rainfed agriculture or a denser, intensive olive orchard) to provide adequate parameterisation to global database, such as ECOCLIMAP-II, in the view of a global application of the ISBA-MEB model.