In mesoscale data assimilation, it is important to analyze narrow zone boundary layer convergence (or “boundaries”) for thunderstorm development and evolution. Since these nonlinear boundaries are usually absent from a background field, observations become a main source for analyzing these features in a data assimilation system. When observation networks are increasing, it is critical for a data assimilation approach to provide analyses of these boundaries when they can be well observed. Therefore, a data assimilation process can be divided into two steps: retrieving resolvable information and statistical data assimilation. In this presentation, a sequential variational analysis approach is examined for retrieving the resolvable information from dense surface observations. It is a 3DVAR analysis of horizontal surface plus time instead of height and is called Space and Time Mesoscale Analysis System (STMAS. This analysis approach greatly improves the variational analysis on nonlinear mesoscale features. Numerical experiments are presented here to show its improvement from a single 3DVAR. The numerical results demonstrate its ability of handling multi-scale nonlinear boundaries, such as cold front or bore.