Examination of Statistical Downscaling Methods for Pavement Climate Modelling Applications

There has been increased interest in utilizing future climate projections to study the effects of climate change on pavement infrastructure. Statistical downscaling techniques from climatology research to have been applied to downscale atmosphere-ocean general circulation model (AOGCM) outputs to climate observations, therefore producing future climate projections at a spatial resolution appropriate for pavement applications. The research question was: Is there a difference in the predictive capacity between statistical downscaling procedures typically used for pavement applications? Subsequently, this research investigated this question using two procedures: the Asynchronous Regional Regression Model (ARRM) and the Delta Method. The statistical downscaling methods were used to produce future climate projections for daily maximum and minimum temperature 2 m above the ground surface for eight sites in the Seasonal Monitoring Program (SMP) of the Long-Term Pavement Performance Program (LTPP) across the United States. The resultant projections were truncated to 1 year target year period 45 years after the latest continuous year of SMP analysis site thermistor temperature data. Non-parametric two sample Kolmogorov–Smirnov (KS) testing was conducted to determine if the samples from the two methods represent the same underlying distribution on a site-by-site basis. There were few cases where the testing results indicated that the temperature samples were derived from the same distributions, therefore the conclusion was that the Delta Method is not interchangeable with the more complex ARRM procedure. This paper is a milestone to compare atmospheric temperature and precipitation future climate projections from different statistical downscaling methodologies to estimate temperature and moisture at depth through pavement structures for a larger population of SMP sites.