Multi-Variable Regression Approach for Pavement Surface Distress Forecasting in Nepal

Abstract

Accurate prediction of pavement surface deterioration is crucial for effective pavement management and efficient allocation of limited maintenance resources. In Nepal, pavement condition evaluation and maintenance prioritization mainly depend on the Surface Distress Index (SDI); however, robust SDI-based deterioration models aligned with national practices are still limited. This study develops and validates a multiple linear regression (MLR) model to predict SDI across Nepal’s national highway network using nationally available pavement, traffic, and climatic data. The model links SDI to six explanatory variables: initial SDI, initial International Roughness Index (IRI), pavement age, total annual rainfall, temperature variation, and equivalent single axle loads (ESAL). Data from 157 highway sections, totaling 790 observations collected between 2012 and 2022, were used. Model development and validation involved an 80:20 data split, followed by a thorough out-of-sample forecast assessment with independent 2023 data from 125 highway sections. The MLR model shows strong explanatory and predictive performance, with coefficients of determination (R²) of 0.724, 0.730, and 0.727 for the development, validation, and overall datasets, respectively. Error metrics reflect satisfactory accuracy, with mean absolute error (MAE) values between 0.292 and 0.358 and mean squared error (MSE) between 0.145 and 0.197. Sensitivity analysis highlights initial SDI, initial IRI, pavement age, and rainfall as the most influential factors in surface distress progression. Under true out-of-sample conditions, the model achieves an R² of 0.723, MAE of 0.359, RMSE of 0.460, and mean absolute percentage error (MAPE) of 16.93%, confirming its robustness and generalization ability. These results demonstrate that a well-specified deterministic regression model can reliably forecast SDI at the network level and serve as a practical, data-driven tool integrated into Nepal’s Pavement Management System, assisting maintenance prioritization, budgeting, and long-term asset managemen

Keywords: Surface Distress Index, International Roughness Index, Multiple Linear Regression, National Highway Network