Specific surface area of soils with different clay mineralogy can be estimated from a single hygroscopic water content

The soil specific surface area (SSA) is an important variable for soil science and geoenvironmental engineering applications, but traditional measurement methods are difficult and time-consuming. Regression models or pedotransfer functions are often used to estimate SSA from other soil properties (e.g., clay content and cation exchange capacity), but these models do not consider the impact of clay mineralogy. Hygroscopic water content (w_h) is intimately linked to these soil properties, which suggests that w_h may be a better parameter for SSA estimation. This study (i) proposes regression models that estimate SSA from w_h at different relative humidity values (5 to 90%) for kaolinite-rich samples (KA), illite-rich or mixed clay samples (IL/MC), montmorillonite-rich samples (ML), and a combination of all samples (ALL) and (ii) compares the performance of the w_h models to other published models that comprise clay, silt and soil organic carbon contents and cation exchange capacity. We found that the sample-specific w_h regression models accurately estimated SSA for KA, IL/MC and ML samples. For KA and IL/MC samples, the performance of the KA model (e.g., for adsorption, average RMSE = 10.5 m²/g) and IL/MC model (average RMSE = 21.3 m²/g) were better than the ALL-calibration model (KA: average RMSE = 18.7 m²/g; ML: average RMSE = 22.4 m²/g). For ML samples, similar model performance between the ML-calibration model (average RMSE = 41.4 m²/g) and the ALL-calibration model (average RMSE = 41.1 m²/g) was observed. In addition, the model performance of regression models based on w_h was superior to models published in the literature that are based on clay, silt and soil organic carbon contents and cation exchange capacity. Overall, this study confirms that a single measure of w_h can provide reliable estimates of the SSA while revealing a significant impact of clay mineralogy on model performance.