Prediction of Soaked and Unsoaked California Bearing Ratio from Basic Geotechnical Properties of Lateritic Soils Along Zaria–Kano Road, Nigeria

Abstract: The California Bearing Ratio (CBR) test is one of the most widely adopted pavement subgrade evaluation methods globally, yet it is time-consuming and costly when applied to large numbers of samples. This study developed multiple linear regression analysis (MLRA) models to predict soaked and unsoaked CBR values from seven basic geotechnical index properties — maximum dry density (MDD), optimum moisture content (OMC), grain size fractions passing sieves No. 7, No. 36, and No. 200 (N7, N36, N200), liquid limit (LL), and plasticity index (PI) — for lateritic soils along the Zaria–Kano Road corridor in Kaduna and Kano States, Nigeria. Secondary data comprising 80 datasets were obtained, of which 65 were used for model training and 15 for independent validation. MATLAB R2019b was the primary modelling environment, and Minitab 17 was used for corroborative analysis. Five soaked CBR models (M1–M5) and five unsoaked CBR models (M6–M10) were developed by progressively eliminating statistically insignificant predictors. Pearson correlation analysis identified MDD and N200 as the dominant predictors for soaked CBR, while MDD and N200 were equally dominant for unsoaked CBR. The best-fit soaked CBR model (M1) incorporating all seven predictors achieved R² = 0.766 and R = 0.875, while the best unsoaked model (M6) yielded R² = 0.715 and R = 0.846. Cross-validation against 15 independent samples confirmed model reliability for both conditions. Comparison of MATLAB and Minitab outputs demonstrated consistency across software platforms. The study confirms that CBR values of lateritic soils along this corridor can be reliably predicted from index properties, significantly reducing laboratory time and costs associated with the 96-hour soaking protocol.