Objective analysis transforms information from randomly spaced observing sites into data at regularly spaced grid-points. Specification of forecast error covariance is required by most analysis schemes. It is common practice to simplify the form of the forecast error covariance. Homogeneity and isotropy of the background error correlations are most often assumed.

In complex terrain, surface observation sites are usually located in valleys. The analysis of surface data is made complicated by orographic features. Isotropy is a particularly questionable assumption.

This work develops a technique to try to utilize anisotropy and heterogeneity for objective analysis at the terrain-following lowest model level in mountainous regions. The goal is to improve the meso-scale details of lowest level analysis in complex terrain, with an ultimate goal of providing better input-fields for higher resolution numerical weather prediction model in such regions.

The technique developed here uses a mother-daughter approach to account for terrain effects in the analysis of surface data. The mother-daughter approach is incorporated into the analysis tool (the ARPS data assimilation system - ADAS) used in this study. The proposed approach and preliminary results and comparisons with other existing methods is presented.