Macro-texture is crucial for its impact on tire-road friction, road safety, noise, and even greenhouse gas emissions. However, the substantial variability and uncertainty in surface texture pose challenges in optimizing the design and analysis system of asphalt pavement. Therefore, this study aims to identify the sources of macro-texture variability by detecting and quantifying meso-structural features and correlating them with asphalt pavement surface characterization. To achieve this objective, two typical macro-textural parameters, mean texture depth (MTD) and mean profile depth (MPD), were redefined in the three-dimensional domain using digital image processing techniques after eliminating the wall effect. Subsequently, thirty-nine meso-structural indicators were counted, grouped, and filtered based on Pearson correlation coefficients. Later, texture prediction/analysis models were developed and optimized through stepwise regression analysis and under-sampling techniques. The research results revealed significant variability in MTD and MPD, with coefficients of variation around 20%. The determination coefficients (R2) of the final models for MTD and MPD were 0.916 and 0.920, respectively, indicating the practicality and reliability of the proposed models. On this basis, these models provide a robust theoretical and methodological foundation for managing uncertainty and variability in macro-texture within asphalt pavement. Notably, aggregate-based variables, particularly those closely related to skeleton stability, were found to have a more significant impact on texture depth than void-structure variables. Therefore, optimizing pavement surface characteristics through adjusted aggregate-related design processes is more recommended.
