One of the essential challenge for atmospheric CO₂ forecasting is predicting CO₂ flux variation on synoptic timescale. For CAMS CO₂ forecast, a process-based vegetation model is used. In this research we evaluate another type of model (i.e., the light-use-efficiency model VPRM), which is a data-driven approach and thus ideal for realistic estimation, on its ability of flux prediction. Errors from different sources are assessed, and overall the model is capable of CO₂ flux prediction.