This dataset consists of two aspects. One is the original plant functional traits dataset collected in this study that were used for machine learning models, which was named by Data file used for machine learning models.csv. The other is functional trait maps for China. The spatial resolution of eight functional trait maps is 1 km in a GeoTIFF format. These trait datasets were generated in combination with field observations, environmental variables, and vegetation indices data for use in an ensemble modelling approach based on two machine learning methods (i.e., random forest and boosted regression trees).
Plant trait data is mainly obtained and collected through two sources. The first source is the public trait database, including the TRX database and the Chinese plant trait database. The second source is literature. To ensure data quality and comparability, we only include trait observation data that meet the following five criteria. (1) Measurement data must be obtained from natural land fields to minimize the impact of management interference, therefore observation data from cultivated land, aquatic habitats, control experiments, and gardens are not included. (2) According to the quality ratio hypothesis, the impact of plant species on ecosystem function largely depends on the traits and functional diversity of dominant species, while being relatively insensitive to the richness of dependent species. Therefore, we only included studies that measured plant trait observations of all species or dominant species in the community. (3) In order to consider intraspecific trait variation, when the same species appears at the same sampling point in different studies, we included all raw observation data from different studies, rather than the average value at the species level. (4) Observation of plant traits must be targeted at mature and healthy plant individuals, thus excluding certain growth stages (such as seedlings) and size grades (such as saplings) to reduce the confounding effects of ontology development. (5) We only included studies with clear geographic coordinates to match predictor variables. The dataset also includes sampling locations and sampling times. The sampling time is mostly concentrated during the growing season of the year (i.e. May to October), which ensures relative consistency in sampling time and minimizes the impact of seasonality.
The drawing of the spatial distribution map of plant functional traits in China is based on machine learning methods, which are trained on a large amount of field measured data, environmental variables, and vegetation indices.
The data quality is good.
| # | number | name | type |
| 1 | 41991234 | National Natural Science Foundation of China |
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.FIGSHARE.DB6662.2024
10.12072/ncdc.figshare.db6662.2024