Because of the ability of an ensemble Kalman filter (ENKF) to generate 3D covariance structures, it is a promising method to properly spread surface observations vertically through any relevant physical structures. In this experiment the Weather Research and Forecasting (WRF) model is used to simulate an idealized convective PBL in the X-Z plane. In a control simulation the depth of the PBL is approximately 1 km. The horizontal grid spacing is 2 km so that the convection is not resolved, and the eddies are parameterized by the so-called MRF scheme. An ensemble is generated around the control solution by varying the surface heating, resulting in a range of convective velocities and PBL depths.
When the PBL is in a statistically-steady state, we examine the convergence of the ensemble toward the control solution under three scenarios. In the first, the surface heating in each ensemble member is set to be equal to the control simulation, effectively correcting the heating. This experiment involves no assimilation of observations, but simply measures the physical time scale over which the solution approaches a steady state for a given heating. In the second, the heating is corrected and surface observations of temperature and wind sampled from the control are assimilated via Newtonian relaxation (nudging). The third approach corrects the surface heating and the same observations are assimilated via an ENKF.
Results will be presented with emphasis on the time scale and mechanism for ensemble convergence. Implications for short-range forecasting and 3D diagnostic applications will be discussed.
Supplementary URL: http://