Verification of an Atmospheric Mixed-Layer Model for a Coastal Region

Abstract

An evaluation of the ability of an integrated (slab) marine atmospheric boundary-layer (MABL) model to predict changes in the inversion and mixed-layer temperature and humidity using data from the Los Angeles-San Diego Basin is described. The model microphysics and initialization methods are evaluated separately. The Stage and Businger stratocumulus entrainment closure formulation is used. Standard radiative flux approximations (e.g., delta-Eddington) are employed with up-to-date cloud microphysical parameterizations. The assumption of well-mixed properties is relaxed to permit a constant vertical gradient that is a function of the surface flux and the entrainment rate. Initialization of the subsidence rate receives considerable attention and is analyzed using data from suitably spaced multiple stations and from a single station. Two cases, a cloud covered and a clear sky period, are examined. In both cases the island and shoreline data are from regularly reporting locations and from a research ship which moved around the region. In one case an instrumented aircraft also provided vertical temperature profiles. Evaluation of the prediction for the cloudy sky case concentrates on the model physics. In this case, the external forcing (subsidence, surface wind and sea-surface temperature) is based on observations during the prediction period and updated every 6 h. Comparison of observations and model results illustrates the important role of both long- and shortwave radiation, and the validity of the Stage and Businger entrainment closure. The agreement is quite good for mixed-layer parameters, mixed-layer depth and the cloud base. Evaluation of the prediction for the clear sky case emphasizes the initialization problem. Entrainment and radiative flux divergences are roughly an order of magnitude smaller in the cloud-free situation. External forcing for this case is based on data available prior to the prediction (versus updates during the prediction). Since subsidence was large during the period, the initialization was well-tested. A period when the subsidence rate was well-established from radar and acoustic remote sensing showed excellent agreement between observed and predicted values for more than 18 h of a 24 h forecast. Results during a period when subsidence was based on single-station-derived information showed reasonable agreement only during the first 12 h of a 24 h forecast. The influence of very near coastline effects is evident in the comparison of mixed-layer temperatures and humidities at the land stations. It is concluded that existing integrated mixed-layer predictive models can, with caution, be applied to coastal prediction problems on the basis of multiple- or single-station data. Specification of the subsidence and the effects of near-coastal circulations are critical.