Application of Particle Filter to Land Surface Model Using in Site Soil Moisture Observation in Semi-arid Region

It has been developed that land surface data assimilation system which is based on particle filtering system applied to land surface model. In the developed system, Sequential Importance Sampling (SIS) and Sequential Importance Resampling (SIR) has been applied and land surface model SiBUC has been utilized. The developed system was proposed to enhance the accuracy of soil moisture and land surface fluxes estimation. In this study, SIS/SIR particle filtering system was applied not directly to update state variables but to estimate model parameter. The proposed system was applied to in site observation for determining appropriately soil parameter to land surface model. The assimilation simulation results show the improvement of not only soil moisture estimation at 1st layer but that at 2nd layer even if 1st layer soil moisture was used as observation for assimilation. Moreover, the improvement of accuracy for the surface fluxes can be shown, too. It is assumed that proposed assimilation technique works effectively.