Quantification of Post-Rainfall Moisture Content in Pavement Unbound Layers Using Long-Term Pavement Performance Data

Abstract

For developing a pavement resilience framework, it is critical to understand and predict the moisture content variation of the unbound layers, which significantly affects pavement response and performance. Utilizing data from the Long-Term Pavement Performance (LTPP) program, two models were developed and compared: a linear regression model and a random forest model. The linear regression model explained 59.5% of moisture content variation, identifying key factors such as maximum single-day precipitation, time since maximum rainfall, and depth of the water table, as well as surface and unbound layers’ material types. The random forest model demonstrated superior performance, explaining 92.3% of moisture content variation. A case study for a LTPP section in Texas demonstrated the models’ ability to simulate moisture distribution over time and depth after a significant rainfall event, providing insights into the drainage behavior of different pavement layers and subgrade materials. The case study also demonstrated the random forest model’s capability to capture different moisture behaviors after precipitation, which the linear model fails to account for. While detailed trends in subterranean layers’ moisture content level can be site-specific, coarser materials tend to handle the excessive water from rainfall better than finer ones, as they experience a shorter duration with elevated moisture concentration. While this was expected, the model allows this aspect to be quantified.

Topics & Concepts