Prediction of clay content from water vapour sorption isotherms considering hysteresis and soil organic matter content

Soil texture, in particular the clay fraction, governs numerous environmental, agricultural and engineering soil processes. Traditional measurement methods for clay content are laborious and impractical for large‐scale soil surveys. Consequently, clay prediction models that are based on water vapour sorption, which can be measured within a shorter period of time, have recently been developed. Such models are often based on single‐point measurements of water adsorption and do not account for sorption hysteresis or organic matter content. The present study introduces regression relationships for estimating clay content from hygroscopic water at different relative humidity (RH) levels while considering hysteresis and organic matter content. Continuous adsorption/desorption vapour sorption isotherm loops were measured for 150 differently textured soils with a state‐of‐the‐art vapour sorption analyser within a RH range from 3 to 93%. The clay contents, which ranged between 1 and 56%, were measured with a combination of sieving and sedimentation methods. Two regression models were developed for both adsorption and desorption at 10 RH levels (5, 10, 20, 30, 40, 50, 60, 70, 80 and 90%). While the first model encompasses all 150 soils regardless of organic carbon (OC) content, the second model considers only soils with OC<2.4%. Independent validation of the proposed regression models at 50, 60 and 90% RH using literature data for water vapour adsorption showed reasonably accurate (average RMSE = 5.0%, ME = 2.4%) prediction of clay contents. However, the model for soils with small OC contents showed only minor improvement when compared with recently published models. Three main sources of prediction errors, namely large OC and silt contents, and a prevalence of 1:1 clay minerals were identified for both the proposed and published models. To compensate for large OC content, an OC‐corrected model was developed and compared with the other models. The corrected model markedly improved clay prediction accuracy for OC‐rich soils when compared with all other models considered.