Data based model to estimate subgrade moisture variation case study: Low volume pavement in North Texas

Abstract

Variations in moisture content significantly alter the geotechnical characteristics of unsaturated soils, including resilient modulus, shear strength, permeability and volumetric deformation. Unpredicted variations of soil parameters can shorten the service life of infrastructures and induce major distresses on buried or soil supported structures, including low volume pavements. Estimating patterns of moisture variation can provide better understanding of soil properties variation and long term serviceability of the pavements. In the current paper, a case study was performed to develop a real time data based model and estimate subgrade moisture variation as a function of rainfall. A low volume rural hot mix asphalt pavement in North Texas was selected, investigated and instrumented with moisture sensors and a rain gauge. The sensors were installed under one lane of the pavements as deep as 4.5m in lines representing pavement centerline, inner wheel path, outer wheel path and pavement edge. The site was monitored for over 2years and hourly records of precipitation and volumetric moisture content at different soil depths were collected. The obtained data were non-parametrically analyzed in Matlab and a model was developed for moisture variation in real time. Based on the results, it was concluded that the pattern of moisture variation can be best described as a combined seasonal and temporal variation. Amplitude of seasonal moisture variation is fairly limited and tolerates within ±5% of an average value. On the other hand, temporary increase in moisture content due to rainfall can be as much as 12%. Since major changes in parameters of unsaturated soil happens within moisture contents close to saturation, this sudden increase in volumetric moisture content can lead to significant soil recharacterization. The developed model was able to capture both seasonal and temporary variation of moisture content and the outputs were within 90% confidence band of measured moisture contents. Comparison of model outputs with existing predictive models showed that the developed model has the capability to significantly increase the accuracy of the estimations. The developed model can be used to estimate moisture variation around buried structures, foundations, slopes and within pavement systems.