Assimilation of soil moisture in LPJ-DGVM

Abstract

Process-oriented dynamic vegetation models are effective tools to assess carbon and water exchanges between vegetation and environment for different scales. Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ-DGVM) is one of the well-established, process-oriented dynamic vegetation models. It can simulate seasonal trends of EvapoTranspiration (ET) and Net Ecosystem Exchange (NEE) forced by weather data. In this study, LPJ-DGVM was employed to simulate the ET and NEE in Yingke (YK) oasis station and A'Rou (AR) freeze/thaw observation station. The results indicate that LPJ-DGVM could not make good estimations in both YK station and AR station. The simulation results were validated with the water and CO2 flux observation from Eddy Covariance (EC). The freeze-thaw phenomenon and irrigation have great impacts on soil water content dynamic in arid region, but they are not considered in LPJ-DGVM. In order to improve the simulation accuracy, a soil water content data assimilation scheme was designed. The observed soil water content was assimilated into LPJ-DGVM with Ensemble Kalman Filter (EnKF) algorithm. The simulation accuracy of LPJ-DGVM was improved obviously when soil water content was assimilated into LPJ-DGVM. The EnKF is effective for assimilating in situ observation.