A two-year integration of SSiB forced with ISLSCP Initiative I data (as part of the Global Soil Wetness Project, GSWP, evaluation and intercomparison) produced generally realistic land surface fluxes and hydrology. Nevertheless, the evaluation also helped to identify some of the deficiencies of the current version of SSiB. The simulated snowmelt was delayed in most of the regions, along with too high runoff and too dry soil. The origin of these deficiencies was: 1)excessive cooling of the snow and ground, and 2) deep frozen soil disallowing snowmelt infiltration. The problem was most severe in regions that experience very cold winters. In SSiB, snow was treated as a unified layer with the first soil layer, causing soil and snow to cool together in the winter months, as opposed to snow cover acting as an insulator. In the spring season, a large amount of heat is required to thaw a hard frozen snow plus soil layer, delaying snowmelt and causing snowmelt water to runoff over the frozen soil rather than infiltrate. A new snow-physics scheme was written to solve the above problem. The snow layer was separated from the soil, with its own energy budget and temperature. Solar energy at the top of the snowpack is reflected by or absorbed by the snow, as well as transmitted to the ground. Surface fluxes occur at the top of the snow, while transfer of thermal energy between the snow layer and the ground is achieved by joint contributions from conduction, convection, and radiation. All of this allows the ground to remain insulated under a deep snowpack, conserving soil heat during winter. In turn, this leads to earlier thawing, larger snowmelt infiltration, and reduced runoff. Comparison to observations of both simulated soil moisture and runoff show a significant improvement in the performance of SSiB, particularly in regions with large winter snow cover. The new snow-physics scheme produced more realistic initial soilmoisture (ISM) with ISLSCP data. Several June-July-August (JJA) simulations for 1987 and 1988 were performed with the GEOS II GCM (with the new snow scheme) using the new ISM. The GCM produces higher precipitation (as observed) in northern regions that have large snowmelt. The new snow-physics scheme also leads to more realistic land surface hydrology in many regions. However, in some areas, particularly in those areas with high sub-grid topography, the scheme produces an over correction. This is currently being improved by modifying the surface runoff model