A coupled atmospheric/land-surface model covering the conterminous United States with an associated 1-hour atmospheric data assimilation cycle, the Mesoscale Analysis and Prediction System (MAPS), has been improved to include a snow accumulation/melting scheme and also parameterization of processes in frozen soil. The new aspects of the land-surface model are described in this paper, along with detailed one-dimensional (1-D) tests using an 18-year observation data set from Valday, Russia. These tests show that the MAPS 1-D soil/vegetation/snow model is capable of providing accurate simulations over multilyear periods at locations with significant snow cover and frozen soil. A statistical analysis of the tests shows the expected improvement in snow depth, skin temperature, and especially in runoff from inclusion of these additional surface processes during the spring melting season. This performance in 1-D tests is a necessary prerequisite for robust long-term behavior of soil temperature and moisture fields and other components of the hydrological cycle in the 3-D MAPS coupled assimilation cycle. ¿ 2000 American Geophysical Union