Data assimilation of soil moisture measurements in land surface simulations to study the impact on evapotranspiration estimates in European forests

Land surface models are important tools to improve our understanding of interacting ecosystem processes and for the prediction of future risks of droughts and fires. However, such predictions are associated with uncertainties related to model forcings, parameters and process simplifications. Therefore, the increasing availability of high-quality observations should be used to improve the accuracy of land surface model predictions. In this study, we use the Ensemble Kalman Filter for the fusion of in-situ soil moisture observations from different observation networks across Europe (e.g. eLTER, FLUXNET, TERENO, ICOS) into the Community Land Model 5.0 (CLM5). The sites selected for this study cover different regional climate zones and forest types and feature in-situ soil moisture as well as evapotranspiration observations from eddy covariance towers for the period from 2009 to 2019. In this study, we specifically focus on European forested study sites where both in-situ soil moisture and evapotranspiration observations are available for the period from 2009 to 2019. CLM5 simulates a broad variety of important land surface processes including water and energy partitioning, surface runoff, subsurface runoff, photosynthesis and carbon and nitrogen storage in vegetation and soil. Here, we focus on improving the accuracy of model predictions by updating soil moisture dynamics and related soil hydraulic parameters by coupling CLM5 to the Parallel Data Assimilation Framework (PDAF) to assimilate soil moisture data into CLM5 during simulation runtime. Additionally, we implemented a new and more direct approach to update the hydraulic parameters compared to previous versions of the CLM5-PDAF coupling and show the effects of this implementation.We demonstrate the value and limitation of assimilating soil moisture data for simulating evapotranspiration focusing on recent drought events in 2018 and 2019. We found that soil moisture dynamics were better characterized by data assimilation, but this did not result in improved estimation of evapotranspiration for the different sites during both wet and dry periods.