Rigid pavements are strongly influenced by nonlinear temperature gradients induced by environmental variability, which generate equivalent linear temperature differentials and eigenstresses that contribute to fatigue damage. Current Indian design guidelines (IRC:58) assume a single uniform temperature differential over large regions and neglect eigenstress effects, which can lead to both conservative and nonconservative designs depending on location and failure mode. This study proposes a Machine Learning (ML)-based thermal microzonation framework to capture spatial variability in pavement thermal behavior and enable zone-specific design recommendations. Hourly meteorological data from the ERA5 reanalysis dataset covering two 30-year periods (1961–1991 and 1992–2022) were used to simulate pavement temperature profiles at 126 land-based grid points across a climatically diverse state in western India using the ILLI-THERM model. Simulations were conducted for three slab thicknesses (200, 250, and 300 mm) and two surface albedo values (0.30 and 0.50). Four thermal performance indicators—bottom-up and top-down linear temperature differentials (ΔTL,BU, ΔTL,TD) and the corresponding critical eigenstress ratios (ESRc,BU, ESRc,TD)—were extracted at the 90th percentile level. The resulting 48-dimensional thermal dataset was reduced to 15 components using principal component analysis (PCA), retaining over 95% of the total variance, followed by clustering using k-Means, DBSCAN, and Agglomerative Hierarchical methods. Thereafter, the application of an ensemble-based centroid approach resulted in five spatially coherent thermal microzones. Substantial spatial variability was observed, with ΔTL,BU values ranging from around 16.3°C to 17.2°C across zones. ESRc,TD values demonstrated a narrower distribution varying approximately 3%–6% across microzones, although these differences were systematic and persisted across slab thicknesses, highlighting spatially distinct eigenstress accumulation characteristics. Comparisons with IRC:58 code of practice illustrated notable deviations from uniform design assumptions and underscore the omission of the eigenstress perspective. The proposed framework is transferable and can support the development of climate-resilient, region-specific rigid pavement design practices in other regions of the world as well.
Nandi et al. (Sun,) studied this question.