Long-term water immersion can substantially alter the hydro-mechanical behavior of water-sensitive geomaterials used in highway subgrades. This study investigated the responses of highly weathered phyllite (HWP) and 2% cement-improved HWP (CIHWP) using a full-scale subgrade model subjected to prolonged water immersion and intermittent static surcharge loading. The degree of compaction, California bearing ratio (CBR), and dynamic deformation modulus ($E_{vd}$) were measured during construction. After construction, earth pressure, pore water pressure, volumetric water content, and layered settlement were monitored. Both materials met the required compaction criteria. However, CIHWP exhibited better compactability and higher CBR and $E_{vd}$ values. Following long-term immersion, water content increased across approximately 88% of the HWP profile, while the saturated area accounted for approximately 28% of the profile. Progressive wetting induced aggregate softening, stress redistribution, and increased deformation, with the final surface settlement reaching 9.3 mm under the applied static surcharge. In contrast, only approximately 12% of the CIHWP profile showed moisture changes, and its final settlement was 6.5 mm. These findings provide full-scale experimental evidence for understanding and improving the hydro-mechanical performance of water-sensitive geomaterials in transportation infrastructure, with direct implications for applied geotechnical engineering and subgrade stabilization.

Objective: To examine the differences between quota-based and field-derived CO₂ emissions from highway subgrade earthworks while assessing the robustness of procedure attribution. Methods: Quota workdays were compared with equipment CO₂ concentrations and exhaust-flow observations from selected operations in one Chinese expressway project. The productivity-duration factor η and emission-intensity factor λ were used to decompose, retrospectively, the reported field-to-quota emission ratio. Nineteen procedures were described by four mutually exclusive η–λ quadrants. Two-class Jenks thresholds were recalculated, and their stability was explored by leave-one-out analysis and 5, 000 procedure-level bootstrap resamples. Results: Soft-rock excavation generated 360.91 kg CO₂ during a representative day (0.435 kg/m³), versus 448.19 kg (0.351 kg/m³) during moderately hard-rock excavation. The workday reductions for rock compaction, soft-rock crushing, and soil grading were 68.7%, 64.4%, and 62.2%, corresponding to productivity increases of approximately 219%, 181%, and 165%, respectively. The four nominal quadrants contain 6, 6, 4, and 3 procedures. Recomputed Jenks midpoints were η = 0.668 and λ = 1.300, differing from the original heuristic boundaries of 0.65 and 1.15. One crushing-procedure record is internally inconsistent, and the evidence is insufficient to verify how all 19 procedure totals were extrapolated. Conclusions: The two-factor identity is useful for retrospective accounting diagnostics but is not independently predictive. Classification thresholds and procedure-level totals must be regarded as exploratory until the monitoring-to-inventory links, source data, and cross-project calibration have been validated.