In this study, high-spatial, low-temporal scale visible remote sensing data were used to calibrate an ecosystem model (EM) for semiarid perennial grasslands. The model was driven by daily meteorological data and simulated plant growth and water budget on the same time step. The model was coupled with a canopy reflectance model to yield the time course of shortwave radiometric profiles. Landsat Thematic Mapper (TM) and Enhanced TM Plus (ETM+) images from 10 consecutive years were used to refine the model on a spatially distributed basis. A calibration procedure, which minimized the difference between the normalized difference vegetation index (NDVI) simulated from the coupled model and measured by the TM and ETM+ sensors, yielded the spatial distribution of an unknown parameter and initial condition. Accuracy of model products, such as daily aboveground biomass, leaf area index (LAI) and soil water content, was assessed by comparing them with field measurements. The promising results suggest that this approach could provide spatially distributed information about both vegetation and soil conditions for day-to-day grassland management.