To design a wireless sensor network for farms, it is necessary to understand and predict the effect of vegetation. In this study, the change in the propagation loss characteristics in 920-MHz band is examined during the growth of mulberry bushes. The received signal strength indicator (RSSI) is measured as a function of the distance between the transmitting antenna (Tx) and the receiving antenna (Rx) in a 50×50m mulberry field. The Tx and Rx are placed at a height of 1.5m. Moreover, the horizontal and vertical polarizations are measured and the differences are shown. Three empirical vegetation attenuation models are introduced, and the measured data have been fitted to each model. The results show that the non-zero gradient model is the best model at predicting the vegetation attenuation in a mulberry farm regardless of the polarization or mulberry growth. It is found that the attenuation dependence on the plant height is linear. Furthermore, the results have revealed that the horizontal polarization had about 1.5 times as large an effect on the vegetation attenuation as the vertical polarization.