In this study, a zero-dimensional land surface model based on a vertically averaged representation of soil moisture in the root-zone is coupled with grass dynamics represented by live and dead biomass pools. Five variants of the model combined the different time steps (daily vs interstorm), representations of potential evapotranspiration (PET), and constant versus variable water use efficiency (WUE). Model results were compared with field observations of soil moisture, live and dead biomass, and leaf area index (LAI) from a long-term experiment in the Nebraska Sand Hills (NSH), USA. Results show that the simplest model configuration using a constant WUE term, PET represented from a cosine function, and an inter-storm model time step can capture grass dynamics nearly as well as daily model simulations with variable WUE. The model is then applied across the NSH to investigate the performance of the model across a climate gradient, and identify parameters that can be calibrated to improve regional predictions. For this purpose a senescence parameter that controls the decay of live aboveground vegetation is further calibrated at 14 different locations across NSH. Model calibrations revealed the dependence of the senescence parameter to regional variations in annual precipitation. Relating the senescence coefficient to annual precipitation using an empirical equation improved the model predictions at the regional scale, and revealed a strong climatic control on grass LAI. Finally, the model is tested against some existing empirical equations that relate annual grassland productivity to annual precipitation and evapotranspiration. Copyright © 2011 John Wiley & Sons, Ltd.