Distribution of biocrusts under the canopy of vascular plants in the Loess Plateau of China and random forest prediction

Biological soil crusts (biocrusts) are important for maintaining the health and stability of dryland ecosystems. The distribution and rapid measurement of cover are practical issues that are worth exploring. However, it is difficult to rapidly measure the cover of underlying biocrusts because of the shelter provided by the plant canopy. This study conducted field surveys across five regions in the Loess Plateau, spanning aridity indices of 2.05-13.36, quantifying canopy traits for 500 vascular plants representing nine species. We analyzed the distribution patterns of biocrusts beneath vascular plant canopies and investigated the relationship between canopy traits and biocrusts. Subsequently, a random forest (RF) model was employed to construct prediction models for biocrusts. The results indicated that biocrust coverage was significantly greater under northwest than southeast canopies (p<0.05). The cyanobacteria crust increased from the inner to the outer canopy zone. The biocrust coverage under the canopies of vascular plants increases with an increase in the aridity index. Compared with coniferous plants and arbors, the canopies of broad-leaved plants and shrubs are more conducive to the growth of biocrusts, especially cyanobacteria crust. Canopy area was the key factor affecting biocrusts, with a 0.66 correlation coefficient and 25.33% prediction importance. The RF model performed well in predicting biocrusts based on the canopy traits of vascular plants, with a prediction accuracy of 0.83 and a root mean square error of 0.86 m2. Overall, this study offers a reliable and convenient approach for predicting biocrusts area based on plant canopy traits in dryland ecosystems worldwide.