Soil Quality Assessment in Loess Hilly Areas: Comparing Evaluation Methods Under Coupled Vegetation and Rainfall Conditions

ABSTRACT

Understanding the interactive effects of rainfall and vegetation on soil quality and functioning is critical for sustainable land management. This study investigated three typical vegetation types—arboreal forests, shrublands and grasslands—across different rainfall zones in the Loess Hilly Region in China. We measured 22 soil physical, chemical and biological indicators, and two mainstream soil quality assessment approaches, Principal Component Analysis (PCA) and Network Analysis (NA), were employed to construct a Minimum Data Set (MDS). A total of 12 Soil Quality Index (SQI) models (2 × 2 × 3) were developed by combining additive and weighted integration methods with three scoring functions: linear, non‐linear and membership function. The SQI results derived from each model were systematically compared. The key findings are: (1) Soil properties and SQI values differed markedly among vegetation types along the rainfall gradient. Lower rainfall areas generally exhibited poorer soil quality, though shrublands maintained relatively higher SQI values. With increasing rainfall, overall SQI improved, and arboreal forests gradually demonstrated superior soil quality. In high‐rainfall zones, a distinct soil quality hierarchy emerged: arboreal forests > shrublands > grasslands. (2) Compared to PCA, the MDS derived from NA contained fewer indicators while still capturing the core parameters essential for soil quality evaluation. (3) The soil sensitivity index calculated using NA (1.45–6.81) was higher than that obtained from PCA (1.34–4.03). Furthermore, the linear scoring function yielded higher sensitivity indices than the non‐linear and membership function approaches. In conclusion, as a straightforward and robust tool, NA can effectively identify key soil indicators under varying rainfall regimes when constructing the MDS. It also demonstrates higher sensitivity than PCA in computing SQI, thereby better discriminating spatiotemporal variations in precipitation patterns. The NA based models with linear scoring functions (SQINAla and SQINAlw) are recommended as a suitable framework for soil quality assessment in the loess hilly region, providing a scientific basis for regional land management and ecological restoration.