Data for project: Spatiotemporal prediction of plant species richness using the hybrid framework of machine learning and spatiotemporal geostatistics

This dataset supports the study “Spatiotemporal prediction of plant species richness using the hybrid framework of machine learning and spatiotemporal geostatistics”. The data were processed from the long-term ecological monitoring dataset of typical Chinese ecosystems covering 1998–2010. The source dataset was compiled from observations of the Chinese Ecosystem Research Network (CERN) and is described by Zhang et al. (2025), Dynamic dataset of environmental elements, species richness and biomass of long-term observation plots of typical ecosystems in China from 1998 to 2010 (DOI: 10.17521/cjpe.2025.0001). The original dataset is available from Science Data Bank (DOI: 10.57760/sciencedb.17718). This study used the forest-station subset of the source dataset. After data cleaning, harmonization, and matching by station, plot, and year, the analytical dataset contained 235 site–plot–year observations from ten forest ecosystem monitoring stations. Plant species richness was used as the response variable. The main variables include geographical information, plot characteristics, annual soil properties, temporal information, and disturbance-related records. Definitions of the main variables are as follows: site: forest monitoring station identifier; plot: monitoring plot identifier; lon/lat: longitude and latitude in decimal degrees; altitude: elevation above sea level (m); year: observation year; pH: soil pH; SOM: soil organic matter; TN: soil total nitrogen; TP: soil total phosphorus; vegetation: vegetation-type information; human_activity: human-activity record; animal_activity: animal-activity record; disaster: disturbance/disaster record; richness: observed plant species richness.