Recently, Mathis et al. (2013) developed a model for predicting the instantaneous fluctuations of the wall shear-stress in turbulent boundary layers. This model is based on an inner-outer scale interaction mechanism, incorporating superposition, and amplitude-modulation effects, and the only input required for the model is a time series measurement of the streamwise velocity signal taken in the logarithmic region of the flow. The present study applies this new approach for the first time to environmental flows, for which the near-bed information is typically inaccessible. The data used here are acoustic Doppler velocimeter time series measurements from a shallow tidal channel (Suisun Slough in North San Francisco Bay). We first extract segments of data sharing properties with canonical turbulent boundary layers. The wall (bed) shearstress model is then applied to these selected data. Statistical and spectral analysis demonstrates that the field data predictions are consistent with laboratory and DNS results. The model is also applied to the whole available data set to demonstrate, even for situations far from the canonical boundary layer case, its ability to preserve the overall Reynolds number trend. The predicted instantaneous bed stress is highly skewed and amplitude modulated with the variations in the large-scale streamwise velocity. Finally, the model is compared to conventional methods employed to predict the bed shear-stress. A large disparity is observed, but the present model is the only one able to predict both the correct spectral content and the probability density function.