Edge- and mid-span thermal-response signatures in continuous box-girder bridges

Routine bridge-monitoring systems often provide much less information than research deployments. For routine monitoring of many in-service continuous box-girder bridges, the available record is limited to sparse temperature channels and one strain channel at each monitored section. This paper treats that sparse setting as the working condition and shows that such records still carry usable thermal-response information once lag, slow drift, and warming–cooling asymmetry are handled separately. Using 4 to 5 years of field data from six monitored sections of two cold-region prestressed concrete continuous box-girder bridges, branch-separated thermal response analysis (BSTRA) is developed to estimate a section-level thermal-response slope b . The procedure applies section-specific lag alignment, 12-h differencing, and separate warming/cooling branch fitting in differential space. Across the six monitored sections, the mean annual coefficient of variation of b falls from 105.8% in raw-space branch splitting to 6.0%, while the mean branch-fit R 2 increases from 0.182 to 0.738. After this stabilization, the edge spans show a cold-season-high, warm-season-low pattern. The two mid-span sections depart from this edge-span pattern: both show larger seasonal amplitudes, and one shows an opposite cold–warm seasonal direction. These results support the construction of span-region-specific monthly reference bands in the two bridge cases instead of a single pooled seasonal baseline.