Prediction of the Compression Index of Alluvial Soil Using Index Properties

<title>Abstract</title> <p> Laboratory evaluation of the compression index (Cc) is vital for calculating the settlement of fine-grained soils, but it requires significant time and high expenses to extract and test high quality undisturbed samples. This study develops and evaluates single and multi-variable regression models for predicting C _c using basic index properties such as Liquid Limit (LL), Plastic Limit (PL), Plasticity Index (PI), and initial void ratio (e ₀ ), for central Iraqi alluvial plain soils. The single variable baseline models (Cc vs LL, R ²  = 0.4158) and (Cc vs PI, R ²  = 0.6817) revealed that substituting the LL with the PI improves the model’s accuracy, as PI acts as a superior proxy for mineralogical activity by filtering the non-plastic fractions introduced by the localized depositional environment of the soil. A comparative analysis of the single variable models reveals distinct behavioral patterns across global and regional models. To overcome the limitations of single variable models, two multi-linear frameworks were established to incorporate <italic>e</italic> ₀ as a structural state variable with LL and with PI. The second Model (C _c with PI, e ₀ ) demonstrated the highest statistical precision, achieving an R ² of 0.824 (p &lt; 0.001), a Mean Absolute Error (MAE) of 0.00865, and limiting the Mean Absolute Percentage Error (MAPE) to 4.943%. These results indicated that while simple index properties capture the mineralogical trends, incorporating volumetric state variables within localized multi-linear frameworks is essential to ensure reliable and cost-effective predictive models of compressibility of heterogeneous alluvial soils. </p>