To begin integrations, global weather and climate models need accurate and reliable initial conditions, which includes land surface soil moisture. Due to the lack of observationally-based global analyses, the global estimates of the soil moisture state need to be generated using a land surface scheme (LSS) driven by observed external forcing, which generally comes from global operational analyses or reanalyses. For a typical LSS, those external forcing variables consist of precipitation, near-surface air temperature, relative humidity, pressure, incoming long-wave and short-wave radiation. For each field, there usually exists several options. In this study, the sensitivity of soil moisture to the choice of external forcing data sets was examined with SSiB land surface scheme through a suite of experiments within the GSWP framework. In comparison of simulated soil moisture with observations, the quality of various external forcing products is indirectly evaluated.