Accurate field-scale crop yield prediction is essential for sustainable agricultural management under increasing climate variability. Process-based crop growth models such as WOFOST provide a physically consistent framework for simulating crop development and biomass accumulation, yet their predictive performance is often constrained by uncertainties in parameterization, inputs, and canopy representation. To address these limitations, this study integrates satellite-derived biophysical parameters into WOFOST using an Ensemble Kalman Filter (EnKF) to improve canopy dynamics and yield estimation for green maize and winter barley in the Netherlands. Sentinel-2–retrieved Leaf Area Index (LAI), Canopy Chlorophyll Content (CCC), and Canopy Water Content (CWC) were individually assimilated at the parcel scale during the 2022 growing season. Model performance was validated against independent provincial statistics. Results show that LAI assimilation significantly enhanced model performance for maize, reducing the mean yield bias from -3.78 t ha⁻¹ (-27.6%) to -0.11 t ha⁻¹ (-0.81%) and lowering RMSE from 3.89 t ha⁻¹ to 0.99 t ha⁻¹. CCC and CWC contributed less due to their dependence on fixed biochemical coefficients but highlighted challenges and opportunities for incorporating physiological constraints. Sensitivity analyses emphasized the importance of observation density, timing, and uncertainty, with the greatest impact observed during canopy expansion and peak growth. Overall, this study demonstrates the potential of remote sensing data assimilation to enhance process-based yield prediction and provides a foundation for advancing operational crop monitoring under diverse agricultural conditions.