DATA-DRIVEN ASSESSMENT OF PAVEMENT PERFORMANCE AND RESILIENCE UNDER EXTREME WEATHER USING THE LONG-TERM PAVEMENT PERFORMANCE PROGRAM
Keywords:
Pavement performance, Climate resilience, Extreme weather, Machine learning, Explainable AI, SHAP, LTPP, Freeze–thaw, Infrastructure asset managementAbstract
Extreme weather driven by a changing climate is placing unprecedented stress on road infrastructure, yet the quantitative link between weather extremes and observed pavement deterioration—and the resilience of pavements to that stress—remains poorly characterized at scale. This paper presents an end-to-end, reproducible framework that couples the U.S. Federal Highway Administration Long-Term Pavement Performance (LTPP) database with interpretable machine learning and a novel pavement resilience formulation. From 376 asphalt-concrete test sections monitored between 1988 and 2016, we construct a panel of 2,656 section-year records enriched with engineered extreme-weather indicators (freeze–thaw cycles, extreme-heat and freezing days, intense-precipitation days, and thermal range). A variance decomposition shows that 48% of the variance in surface cracking is attributable to unobserved site-specific factors, explaining why absolute distress in unseen sections is intrinsically hard to predict from climate alone (transfer R² ≈ 0). A linear mixed-effects model that controls for site and pavement age isolates intense-precipitation days (standardized β = 1.10, p < 0.001) and extreme-heat days (β = 0.70, p = 0.043) as the statistically significant weather accelerators of cracking, whereas a condition-informed forecasting model reaches R² = 0.53 with prior condition dominating the SHAP attribution. We introduce a Climate Stress Index (CSI) and a two-dimensional stress–impact positioning that yields a Demonstrated Resilience Score, separating climate-driven vulnerability from intrinsic fragility. The framework offers actionable screening for climate-adaptive pavement asset management.References
[1] Underwood BS, Guido Z, Gudipudi P, et al. Increased costs to US pavement infrastructure from future temperature rise. Nature Climate Change, 2017, 7(10): 704-707.
[2] IPCC. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report. Cambridge, UK: Cambridge University Press, 2021.
[3] Mills B, Tighe SL, Andrey J, et al. Climate change implications for flexible pavement design and performance in southern Canada. Journal of Transportation Engineering, 2009, 135(10): 773-782.
[4] Meagher W, Daniel JS, Jacobs J, et al. Method for evaluating implications of climate change for design and performance of flexible pavements. Transportation Research Record, 2012, 2305: 111-120.
[5] Qiao Y, Dawson AR, Parry T, et al. Evaluating the effects of climate change on road maintenance intervention strategies and life-cycle costs. Transportation Research Part D: Transport and Environment, 2015, 41: 492-503.
[6] Tighe SL, Smith J, Mills B, et al. Evaluating climate change impact on low-volume roads in southern Canada. Transportation Research Record, 2008, 2053: 9-16.
[7] Stoner AMK, Daniel JS, Jacobs JM, et al. Quantifying the impact of climate change on flexible pavement performance and lifetime in the United States. Transportation Research Record, 2019, 2673(1): 110-120.
[8] Gong H, Sun Y, Shu X, et al. Use of random forests regression for predicting IRI of asphalt pavements. Construction and Building Materials, 2018, 189: 890-897.
[9] Gong H, Sun Y, Hu W, et al. Investigating impacts of asphalt mixture properties on pavement performance using LTPP data through random forests. Construction and Building Materials, 2019, 204: 203-212.
[10] Damirchilo F, Hosseini A, Mellat Parast M, et al. Machine learning approach to predict international roughness index using long-term pavement performance data. Journal of Transportation Engineering, Part B: Pavements, 2021, 147(4): 04021058.
[11] Guo R, Fu D, Sollazzo G. An ensemble learning model for asphalt pavement performance prediction based on gradient boosting decision tree. International Journal of Pavement Engineering, 2022, 23(10): 3633-3646.
[12] Marcelino P, de Lurdes Antunes M, Fortunato E, et al. Machine learning approach for pavement performance prediction. International Journal of Pavement Engineering, 2021, 22(3): 341-354.
[13] Georgiou P, Plati C, Loizos A. Soft computing models to predict pavement roughness: A comparative study. Advances in Civil Engineering, 2018, 2018: 5939806.
[14] Ker HW, Lee YH, Wu PH. Development of fatigue cracking prediction models using long-term pavement performance database. Journal of Transportation Engineering, 2008, 134(11): 477-482.
[15] Titus-Glover L. Reassessment of climate zones for high-level pavement analysis using machine learning algorithms and NASA MERRA-2 data. Advanced Engineering Informatics, 2021, 50: 101435.
[16] Bruneau M, EERI M, Chang SE, et al. A framework to quantitatively assess and enhance the seismic resilience of communities. Earthquake Spectra, 2003, 19(4): 733-752.
[17] Qiao Y, Flintsch GW, Dawson AR, et al. Examining effects of climatic factors on flexible pavement performance and service life. Transportation Research Record, 2013, 2349: 100-107.
[18] Qiao Y, Guo Y, Stoner AMK, et al. Impacts of future climate change on flexible road pavement economics: A life cycle costs analysis of 24 case studies across the United States. Sustainable Cities and Society, 2022, 80: 103773.
[19] Lu D, Tighe S, Xie WC. Pavement risk assessment for future extreme precipitation events under climate change. Transportation Research Record, 2018, 2672: 122-131.
[20] Federal Highway Administration. Long-Term Pavement Performance (LTPP) InfoPave. U.S. Department of Transportation. https://infopave.fhwa.dot.gov/.
[21] Chen T, Guestrin C. XGBoost: A scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016: 785-794.
[22] Friedman JH. Greedy function approximation: A gradient boosting machine. Annals of Statistics, 2001, 29(5): 1189-1232.
[23] Hajiha M, Liu X. Processed Long-Term Pavement Performance (LTPP) data. Department of Industrial Engineering, University of Arkansas, 2019. https://github.com/dnncode/LTPP-Data.
[24] Breiman L. Random forests. Machine Learning, 2001, 45(1): 5-32.
[25] Lundberg SM, Lee SI. A unified approach to interpreting model predictions. In: Advances in Neural Information Processing Systems (NeurIPS), 2017, 30: 4765-4777.
[26] Lundberg SM, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees. Nature Machine Intelligence, 2020, 2(1): 56-67.
[27] Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 2011, 12: 2825-2830.
[28] Mallick R, Radzicki M, Daniel JS, et al. Use of system dynamics to understand long-term impact of climate change on pavement performance and maintenance cost. Transportation Research Record, 2014, 2455: 1-9.
[29] Transportation Research Board. Potential Impacts of Climate Change on U.S. Transportation, Special Report 290. Washington, DC: TRB, 2008.