At present, most studies on airport runways focus on icing thickness prediction and temperature prediction,to further improve the safety and efficiency of airport operations, a prediction model for snow thickness on airport runways based on a parallel architecture of Transformer and long short-term memory (LSTM) networks was proposed by integrating meteorological data and runway surface condition monitoring data. The LSTM branch was adopted to extracted the local temporal dynamics of meteorological data and capture short-term contextual relationships, while the Transformer branch was used to accurately captured the nonlinear coupling relationships among meteorological factors through the self-attention mechanism, with suitable weights assigned to different meteorological factors under different conditions. Finally, the different features captured by the two branches are weighted and fused through an adaptive weight module to complete the prediction. Comparative results show that the parallel Transformer-LSTM model achieves satisfactory prediction performance in both single-step and multi-step predictions on two different datasets. Accordingly, snow removal scheduling arrangements can be made by airports based on the prediction results.