利用2018—2020年汛期（5—9月）京藏高速青海省公路沿线交通自动监测站逐小时气象观测资料，研究路面水膜厚度变化特征，构建水膜厚度与气象因子的预报模型。结果表明：（1） 京藏高速公路高庙桥站和汉庄村站逐小时路面水膜厚度主要分布均在0.0~0.2 mm之间，频率分别为66.0%和63.0%，大于0.5 mm以上的路面水膜厚度频率均较小（10.0%），2站均属于强变异性地区。（2） 采用相对阈值法统计分析，发现2站路面水膜厚度在0.1 mm以内的比例分别为33.8%和36.3%，路面水膜厚度在0.1~0.6 mm之间的比例分别为59.2%和56.0%，路面水膜厚度大于0.7 mm以上即易发生水滑，引起车辆失稳、失控等危险的比例分别为7.0%和7.6%。（3） 路面水膜厚度日变化和月变化特征明显。高庙桥站和汉庄村站水膜厚度的月变化均呈弱双峰性，2站的月变化趋势不完全一致。高庙桥站的日变化峰值出现在02:00—06:00，低谷出现在14:00—16:00，汉庄村站日变化峰值出现在06:00，低谷出现在16:00。（4） 随着降水强度的增加，平均水膜厚度均遵循幂函数关系迅速增加；在降水强度0.00~1.75 mm·h^-1之间时，平均水膜厚度增加趋势明显，降水强度大于1.76 mm·h^-1平均水膜厚度变化有增有减。（5） 采用多元回归统计方法建立依据气象因子和不同降水强度下分别构建的水膜厚度模型具有较好的使用价值，可在实际业务工作中推广应用。（6） 不同降水强度下构建的水膜厚度模型计算值明显高于季天剑模型和罗京模型，本文模型与罗京模型变化趋势较为一致，水膜厚度随降雨强度增加增长趋势明显，季天剑模型水膜厚度计算值随降雨强度增加增长趋势缓慢。研究成果可应用于高原环境雨天车速管理和路面交通安全管理，能够为公路设计人员或运营管理人员提供辅助决策的依据。

The research uses the hourly measured data from the traffic automatic monitoring station along the Qinghai Provincial Highway of the Beijing-Tibet Expressway during the flood season (May-September) of 2018—2020, study the change feature of the variety of the pavement water film thickness, and finally build the forecast model between the water film thickness and meteorological factors. The results showed that: (1) The hourly pavement water film thickness at Gaomiaoqiao station and Hanzhuangcun station of Beijing-Tibet Expressway was mainly distributed between 0.0 mm and 0.2 mm, and the frequencies were 66.0% and 63.0%, respectively. There would be a small frequency (10.0%) when the pavement water film thickness over 0.5 mm, and both stations belong to strong variability areas. (2) The relative threshold method was used for statistical analysis, and it was found that the proportions of the pavement water film thickness within 0.1 mm at the two stations were 33.8% and 36.3%, respectively. The proportions of the pavement water film thickness between 0.1 mm and 0.6 mm were 59.2% and 56.0%, respectively. When the pavement water film thickness was over 0.7 mm, it is easy to slip, 7.0% and 7.6% of the vehicles were unstable and out of control respectively. (3) The daily and monthly variation characteristics of pavement water film thickness were obvious. The monthly variation of water film thickness at Gaomiaoqiao station and Hanzhuangcun station both showed a weak bimodal character, and the monthly variation trend of the two stations was not completely consistent. The peak of daily variation appeared at 02:00—06:00, and the low point appeared at 14:00—16:00, the peak appeared at 06:00 and the low point appeared at 16:00 at Hanzhuang Village station. (4) With the increase of precipitation intensity, the average water film thickness increased rapidly following the power function; when the precipitation intensity was between 0.00 mm·h^-1 and 1.75 mm·h^-1, the precipitation intensity was greater than 1.76 mm·h^-1, and the average water film thickness increased and decreased. (5) The model of water film thickness based on meteorological factors and different precipitation intensities was established by using multiple regression statistical method, which has good operational value and can be popularized in practical work. (6) The calculated values of the water film thickness model under different precipitation intensities are significantly higher than those of the Ji Tianjian model and the Luo Jing model. The variation trend of the model presented in this article is more consistent with the Luo Jing model, the water film thickness increases obviously with the increase of rainfall intensity, and the calculated value of water film thickness increases slowly with the increase of rainfall intensity.