Developing pedotransfer functions to estimate the S-index for indicating soil quality

Indicating soil quality usually requires many soil properties of which the measurements are time consuming. Therefore, it is desirable to developing simple and effective indices for reflecting soil quality based on soil properties that can be readily obtained. The soil physical quality index, S-index, derived from the slope at the inflection point of the water retention curve (particularly the Van-Genuchten equation), is a comprehensive index for indicating soil properties. By comparing the S-index with a widely used soil quality index (SW, this study used 298 samples to determine soil chemical and physical properties for calculating SQI, and found that the correlation coefficient between the S-index and SQL was 0.88, indicating that the S-index can represent soil quality well. An artificial neural network (ANN) model and a linear regression (LR) model were proposed for estimating S-index. Results showed that the ANN model was better than LR model in estimating S-index. Particularly, the ANN model with the soil bulk density and soil organic carbon (scenario Al) as inputs, had the highest R-2 of 0.807, while the LR model get the highest R-2 (predicted v.s. observed) of 0.75 with the combination of soil organic carbon, soil bulk density, total nitrogen and available nitrogen. This study is helpful for extending the applications of S-index.