Accurate initialization of land states, namely, soil moisture, soil temperature, and snowpack, is critical in numerical weather and climate prediction systems because of their regulation of simulated water and energy fluxes between the land surface and atmosphere over a variety of time scales. Currently, land states used in prediction systems are often found to have substantial errors owing to bias in the land surface forcing of the coupled systems, mainly in precipitation and surface radiation. A research project is ongoing in close collaboration between NASA and NOAA using the NASA Land Information System (LIS) to generate alternative land initial conditions for the NOAA NCEP Global Forecast System (GFS). The goal is to investigate the influence of enhanced land initial conditions on the prediction system. A LIS infrastructure has been built on the NCEP supercomputer where the operational GFS is executed. This system is configured identically to the seasonal forecast mode of GFS including the same soil and vegetation specifications. A baseline experimental GFS simulation that uses the Noah land surface model will be executed as the control run also providing the baseline atmospheric forcing. The same version of Noah will be used in this LIS simulation. Observation based, both in-situ and satellite-driven, precipitation and surface radiation are used as alternative forcing options. With the advantage of its high performance parallel computing technique, LIS is able to generate offline uncoupled land initial conditions ensemble in a prompt manner.

In this paper, the simulated land states and fluxes corresponding to the alternative land surface forcing options will be compared to the control GFS forecast fields. The translation from the perturbation in the forcing to that in the resulting land states will be quantified. Selected observations from the Coordinated Enhanced Observing Period (CEOP) reference sites will be used for evaluation.