Can we really trust traditional compression index correlations? Lessons from a global soil database

Abstract

The compression index (Cc) is a key parameter for predicting consolidation settlement in fine-grained soils. In geotechnical practice, empirical correlations based on simple index properties are widely used as a fast and economical alternative to laboratory oedometer tests. However, most existing formulations were developed under site-specific conditions, raising concerns about their reliability when applied beyond their original scope. In this study, the predictive performance of 20 widely used empirical correlations for Cc estimation is evaluated using a comprehensive global database comprising 1008 soil samples compiled from multiple countries and geological settings worldwide. The dataset spans wide ranges of liquid limit (17.1–199.0%), plasticity index (2.0–82.0%), initial void ratio (0.279–7.114), natural water content (8– 244.1%), and Cc (0.013–2.2). Correlations are assessed for the complete dataset and for different compressibility ranges using multiple performance metrics, with Theil’s inequality coefficient adopted as the main ranking criterion. Results show that most traditional correlations exhibit large prediction errors and high dispersion when applied globally, particularly for low-compressibility soils. A simple correlation recently proposed by the authors, based solely on natural water content, demonstrates competitive performance, negligible bias, and improved robustness, offering a practical alternative for preliminary Cc estimation in data-scarce conditions.