To examine the impact of land surface parameterization on the National Centers for Environmental Prediction (NCEP)'s Climate Forecast System (CFS) where the Noah land model is employed to compute surface fluxes, summer season T126 CFS experiments are carried out for selected 9 years with 4 ensembles whose initial conditions are from early May. The nine years are composed of three ENSO cold, warm and neutral years. The first set of CFS experiments is conducted using the current land surface parameterizations where the thermal and momentum roughness lengths are independent of vegetation condition. A second companion run is then conducted using a vegetation–dependent parameterization scheme. Comparisons between the two runs are made with precipitation and 2-meter temperature prediction skills, surface fluxes, and their diurnal characteristics. The results show that there are noticeable differences, indicating that a careful and improved treatment for the turbulent flux parameters in the Noah land model is important to seasonal predictions.