Improving the Prediction of Building Façade Degradation Using Quantile Regression: Revealing the Heterogeneity of Influencing Factors

The durability of building façades is critical to sustainable construction because it affects maintenance demand, safety, and long-term service performance. As building stocks age, especially in rapidly urbanizing countries such as China, reliable prediction of façade degradation becomes increasingly important for service-life planning and maintenance decision-making. However, conventional service-life prediction methods are commonly based on ordinary least squares (OLS) regression, which mainly estimates the conditional mean and may therefore fail to represent the heterogeneity of degradation processes. Using visual inspection data from 375 painted façade samples in Ningbo, China, this study applies quantile regression (QR) to model façade degradation and predict service life. Degradation was quantified using an overall degradation level (ODL) index that integrates defects related to aesthetic deterioration, loss of integrity, and loss of adhesion. The results show that façade degradation follows heterogeneous rather than uniform trajectories, and that the effects of key variables vary across degradation levels. In particular, pollution exposure and water ingress become markedly more influential at higher quantiles, while the effect of routine maintenance weakens in severely degraded façades. After 5-fold cross-validation, the median quantile model reduced MAE by approximately 5.3% relative to the OLS benchmark (0.0537 vs. 0.0567), and the fitted quantiles showed good calibration, with empirical coverage deviations not exceeding 0.007. The QR framework predicted a service-life range of 4.3–31.8 years, substantially wider than the 8.8–20.2 years obtained from the MLR model, indicating a stronger ability to represent uncertainty and high-risk degradation paths. These results show that QR provides a more informative basis for risk-based inspection planning and façade service-life assessment in existing buildings.