Changes in vegetation and soil properties across 12 years after afforestation in the hilly-gully region of the Loess Plateau

Vegetation restoration is the primary method used to prevent soil erosion and improve the ecological environment. The objective of this study was to explore how vegetation and soil properties changed over time in natural and artificial vegetation types. We conducted vegetation surveys in four permanent plots over 12 years (2007–2018) and measured soil properties in 2007, 2013, and 2016. The results showed that the change in species composition in natural vegetation was a continuous and progressive process. After approximately ten years of natural herb restoration, trees began to appear and formed sparse forests. In artificial vegetation, the species composition of the undergrowth varied under Robinia pseudoacacia forests during restoration years, and was dominated by Poaceae and annual plants that could tolerate shade and drought. Species composition was stable under Caragana korshinskii shrubs, where the undergrowth was dominated by perennial plants that were similar to natural vegetation. In different vegetation types, species richness, diversity, and herb community cover fluctuated, while the succession rate decreased with the number of restoration years. The soil moisture condition was better in natural vegetation than in artificial vegetation, and could be maintained near the stable field capacity. The soil moisture was supplemented to 280 cm in R. pseudoacacia forests and 320 cm in C. korshinskii shrubs in a year with extreme rainfall, but the dry soil layer returned in the following years. Soil nutrients improved over time, especially in artificial vegetation, which appears to be better than natural vegetation. In general, natural vegetation improved plant community and soil properties. Artificial vegetation only improved soil nutrients, but caused soil desiccation which is difficult to recover. Therefore, the natural vegetation restoration model of sparse forest grassland was better than artificial vegetation model.