AUTHOR=Li Xiaotong , Chen Yudong , Lv Guanghui , Wang Jinlong , Jiang Lamei , Wang Hengfang , Yang Xiaodong 

TITLE=Predicting spatial variability of species diversity with the minimum data set of soil properties in an arid desert riparian forest

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.1014643

DOI=10.3389/fpls.2022.1014643

ISSN=1664-462X

ABSTRACT=Species diversity has quite spatial heterogeneity in ecological communities, especially in desert areas. Despite abundance of studies having shown the important influence of soil fertility properties on species diversity, many studies have neglected the spatial variability in the relationship between plant species diversity and soil fertility properties. Sampling plots were set up on the vertical transect of Aqikesu river. The minimum data set (MDS) of soil fertility properties (SSC, STP, SAP, SOC and SNN) that could represent the total data set (TDS) was established by principal component analysis (PCA) and the norm value method (R2=0.622). Spatial variability in species diversity was obtained by geostatistics. Also, we investigated the applications of random forest (RF) methods to predict the species diversity and spatial distribution of three diversity indices comparing with multiple linear regression (MLR) models, and visualised the predictions of both models using ordinary kriging. The results showed that all three species diversity indices had a high spatial dependence (C0/(C0+C) < 25%). Ordinary kriged distribution maps showed that the spatial distribution pattern of species diversity predicted by RF is closer to the true distribution. The RF models exhibited better performance than the MLR models in estimating the species diversity due to its relatively lower error indices and higher explained variance (56%, 49% and 36%). RF results suggested that SSC had significant effect on spatial distribution of Shannon–Wiener, Simpson and Pielou indices, followed by SOC. However, the STP had little effect on the spatial distribution of all three indices. Our work provided a research case for studying the spatial relationship between soil fertility properties and plant species diversity.