The large-scale forcing data diagnosed from the European Center for Medium Range Weather Forecast (ECMWF) model for driving Single-Column Models (SCMs) and Cloud System Resolving Models (CSRMs) are compared with forcing data derived using the objective variational analysis constrained by observations collected at the Atmospheric Radiation Measurement program (ARM) Southern Great Plains (SGP) site. The comparison covers the following three different synoptic conditions: a strong precipitation period dominated by subgrid scale processes during the ARM summer 1997 Intensive Operational Period (IOP), a moderate precipitation period dominated by synoptic scale processes during the spring 2000 IOP, and a nonprecipitation period during the fall 2000 IOP. In the study we demonstrate that the differences between the two forcing data sets are considerably large during the strong convective precipitation period, while they are much less during the moderate and nonprecipitation periods. By analyzing the column-integrated heat and moisture budgets we show that errors in the ECMWF-model-derived forcing are closely associated with errors in the model-predicted surface precipitation, which largely reflect deficiencies of model parameterizations. In SCM tests we show that SCM simulations are sensitive to the prescribed large-scale forcing data. The simulation errors are not well correlated between the SCM runs with the two different forcing data sets for all the three cases. Some important SCM simulated fields, such as surface precipitation, tend to follow the ECMWF model simulations rather than the observations when it is forced with the ECMWF forcing, especially for the summer case.