DEVELOPMENT OF STOCHASTICALLY PERTURBED PARAMETERIZATION SCHEME FOR THE SURFACE VARIABLES IN WRF BY OPTIMIZING THE RANDOM FORCING PARAMETERS USING THE MICRO-GENETIC ALGORITHM

• Published in: 18th Annual Meeting of the Asia Oceania Geosciences Society pp. 10 - 12
• Main Authors: Lim, S., Park, S. K., Cassardo, C.
• Format: Book Chapter Conference Proceeding
• Language: English
• Published: WORLD SCIENTIFIC 01.04.2022
• Subjects: AS - Atmospheric Sciences; stochastic perturbation; model error; land surface model; optimization; Ensemble data assimilation
• ISBN: 9811260095; 9789811260117; 9811260109; 9789811260100; 9789811260094; 9811260117
• DOI: 10.1142/9789811260100_0004

The ensemble data assimilation system expresses the model uncertainties by ensemble spread, that is, the standard deviation of ensemble background error covariance. The ensemble spread generally suffers from underestimation problems due to the limited ensemble size, sampling errors, model errors, etc. To solve this problem in terms of model errors, recent studies proposed stochastic perturbation schemes to increase the ensemble spreads by adding random forcing to model tendencies or variables. In this study, we focus on the near-surface uncertainties which are affected by interactions between the land surface (LS) and atmospheric processes. Although the LS variables are crucial to provide various fluxes as the lower boundary condition to the atmosphere, their uncertainties were not much addressed yet. We employed the stochastically perturbed parameterization (SPP) scheme for the LS variables within the Weather Research and Forecasting (WRF) ensemble system. As a testbed experiment with the newly developed WRF ensemble-SPP system, we perturbed soil temperature and moisture in the Noah Land Surface Model to improve the performance of low-level atmospheric variables. To determine the optimal random forcing parameters used in perturbation, we employed a global optimization algorithm - the micro-genetic algorithm. Our results depicted positive impacts on the ensemble spread.