Wind and temperature have been shown to be the critical sources causing changes in the modal properties of large-scale bridges. While the individual effects of wind and temperature on modal variability have been widely studied, the investigation about the effects of multiple environmental factors on structural modal properties was scarcely reported. This paper addresses the modeling of the simultaneous effects of wind and temperature on the modal frequencies of an instrumented cable-stayed bridge. Making use of the long-term monitoring data from anemometers, temperature sensors and accelerometers, a neural network model is formulated to correlate the modal frequency of each vibration mode with wind speed and temperature simultaneously. Research efforts have been made on enhancing the prediction capability of the neural network model through optimal selection of the number of hidden nodes and an analysis of relative strength of effect (RSE) for input reconstruction. The generalization performance of the formulated model is verified with a set of new testing data that have not been used in formulating the model. It is shown that using the significant components of wind speeds and temperatures rather than the whole measurement components as input to neural network can enhance the prediction capability. For the fundamental mode of the bridge investigated, wind and temperature together apply an overall negative action on the modal frequency, and the change in wind condition contributes less to the modal variability than the change in temperature.