Use of Four-Dimensional Data Assimilation in a Limited-Area Mesoscale Model Part II: Effects of Data Assimilation within the Planetary Boundary Layer

A four-dimensional data assimilation (FDDA) scheme based on Newtonian relaxation or nudging has been developed and evaluated in the Pennsylvania State University/National Center for Atmospheric Research (PSU/NCAR) Limited-Area Mesoscale Model. It was shown in Part I of this study that continuous assimilation of standard-resolution rawinsonde observations throughout a model integration, rather than at only the initial time, can successfully limit large model error growth (amplitude and phase errors) while the model maintains intervariable consistency and generates realistic mesoscale structures not resolved by the data. The purpose of this paper is to further refine the previously reported FDDA strategy used to produce “dynamic analyses” of the atmosphere by investigating the effects of data assimilation within the planetary boundary layer (PBL). The data used for assimilation include conventional synoptic-scale rawinsonde data and mesoalpha-scale surface data. The main objective of this study is to determine how to effectively utilize the combined strength of these two simple data systems while avoiding their individual weaknesses. Ten experiments, which use a 15-layer version of the model, are evaluated for two midlatitude, real-data cases. It is found that the homogenizing effect of vertical mixing during free convective conditions allows the three-hourly surface-layer wind and mixing ratio observations to be applied throughout the model PBL according to an idealized conceptual model of boundary-layer structure. Single-level surface temperature observations, however, are often poorly representative of the boundary layer as a whole (e.g., shallow superadiabatic layers, nocturnal inversions), and are more attractive for FDDA applications when additional vertical profile information is available. Assimilation of surface wind and moisture data throughout the model PBL generally showed a positive impact on the simulated precipitation by better reserving the low-level structure and movement of systems (e.g., cyclones, fronts) during the 12-h periods bracketed by the standard rawinsonde data. Improved precipitation simulations due to assimilation of surface data are also possible even in cases with weak large-scale forcing, because a significant portion of the vertically integrated moisture convergence often occurs in the boundary layer. Overall, the best dynamic analyses of precipitation, PBL depth, surface-layer temperature and tropospheric mass and wind fields were obtained by nudging to analyses of rawinsonde wind, temperature, and moisture above the model PBL and to analyses of surface-layer wind and moisture within the model PBL.