Exploring on the climate regionalization of Qinling-Daba mountains based on Geodetector-SVM model

average value with each 1-km pixel. All data were aggregated into one data product. A total of 1 image is available for each variable. Elevation model (DEM). SRTM (Shuttle Radar Topography Mission) data were used as elevation data. SRTM data are jointly measured by NASA and the National Bureau of Survey- ing and Mapping (NIMA) with a resolution of 90 m2. These data are obtained from the Geos- patial Data Cloud (GDC) via downloading from http://www.gscloud.cn/. The DEM was then resampled to 1-km2 resolution by calculating the average elevation within each 1-km pixel. Soil data. The spatial distribution data of soil texture in China were collected from the Resource and Environment Data Cloud Platform (http://www.resdc.cn/). The data included the spatial distribution of organic content, nitrogen content, phosphorus content, sand con- tent, silt content, clay content, and soil type with a resolution of 1 km2. The data were gener- ated by scanning and digitizing based on “The Second National Soil Survey” and “1:1,000,000 Soil Type Map in the People’s Republic of China”. Plant community survey data. The plant community survey data consisting of 121 quad- rats were independently collected in QDM by the “China’s North-South Transitional Zone Comprehensive Scientific Investigation” project (20 m×20 m quadrat area). In total, 71 quad- rats were located east of the study area, and 50 quadrats were located in the middle. The attri- butes of each quadrat included vegetation types, the number of each vegetation type, height,

and breast diameter size. Analysis of the real distribution of representative vegetation from the quadrats data can provide effective validation support for our research results. Method The spatial distribution pattern of zonal vegetation is a natural indicator of long-term environ- mental characteristics [29, 30]. Therefore, the spatial distribution patterns of typical subtropi- cal vegetation and typical temperate vegetation are consistent with the spatial distribution of subtropical and temperate climates. To extract the boundary between the subtropical and tem- perate zone, in this paper, the suitable distribution of different vegetation types was calculated based on the Geodetector-SVM model. The spatial distribution of various vegetation types was comprehensively analyzed to delimit the subtropical northern boundary and identify the loca- tion of the transition zone. The identifying process of the transition zone was divided into four steps (As shown in Fig 3). The first step was data preprocessing. The spatial distribution of six representative vegeta- tion types were extracted from the Chinese vegetation map (1:1,000,000). The coordinate sys- tem of covariates (climatic factors, dem and soil data) was unified to WGS-1984, and the data were resampled to a 1-km2 resolution by calculating the average elevation within each 1-km pixel. The second step involved establishing multiple index systems for different vegetation types. The main regionalization indexes of previous regionalization schemes were as follows:

the accumulated temperature with �0 ˚C (aat0dem), the accumulated temperature with �10 ˚C (aat10dem), the annual average precipitation (pa), the annual average temperature (tadem), the coldest month average air temperature (meantem01), and the warmest month average air temperature (meantem07) [7–10, 17, 18, 44]. However, in this study, considering the complex topography of QDM and the sensitivity of vegetation to soil, topographic and soil factors were also added into the index system, including DEM, soil organic content, soil nitrogen content, soil phosphorus content, sand content, silt content, clay content, and soil type. Then, to remove redundant information and make full use of potential information, the optimal index system was established using the Geodetector model. The model evaluated the influence of every single factor using q-statistics, which can measure the implication of a single factor on the spatial distribution of the paired vegetation (the paired vegetation refers to the typical PLOS ONE Climate regionalization based on Geodetector-SVM model PLOS ONE | https://doi.org/10.1371/journal.pone.0241047 November 2, 2020 6 / 16