Evaluation of root water uptake in the ISBA-A-gs land surface model using agricultural yield statistics over France

摘要

The simulation of root water uptake in land surface models is affected by large uncertainties. Thedifficulty in mapping soil depth and in describing the capacity of plants todevelop a rooting system is a major obstacle to the simulation of theterrestrial water cycle and to the representation of the impacts of drought.In this study, long time series of agricultural statistics are used toevaluate and constrain root water uptake models. The inter-annual variabilityof cereal grain yield and permanent grassland dry matter yield is simulatedover France by the Interactions between Soil, Biosphere and Atmosphere,CO₂-reactive (ISBA-A-gs) generic land surface model (LSM). The two soilprofile schemes available in the model are used to simulate the above-groundbiomass (B_ag) of cereals and grasslands: a two-layerforce–restore (FR-2L) bulk reservoir model and a multi-layer diffusion (DIF)model. The DIF model is implemented with or without deep soil layers belowthe root zone. The evaluation of the various root water uptake models isachieved by using the French agricultural statistics of Agreste over the1994–2010 period at 45 cropland and 48 grassland départements, for arange of rooting depths. The number of départements where the simulatedannual maximum B_ag presents a significant correlation with theyield observations is used as a metric to benchmark the root water uptakemodels. Significant correlations (p value < 0.01) are found for up to29 and 77% of the départements for cereals and grasslands,respectively. A rather neutral impact of the most refined versions of themodel is found with respect to the simplified soil hydrology scheme. Thisshows that efforts should be made in future studies to reduce other sourcesof uncertainty, e.g. by using a more detailed soil and root density profiledescription together with satellite vegetation products. It is found thatmodelling additional subroot-zone base flow soil layers does not improve (andmay even degrade) the representation of the inter-annual variability of thevegetation above-ground biomass. These results are particularly robust forgrasslands, as calibrated simulations are able to represent the extreme 2003and 2007 years corresponding to unfavourable and favourable fodderproduction, respectively.