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

mostly used to group databases by the degree of similarity in variability based on temperature- and precipitation-related indicators. Cluster analysis has been employed to climatic regionali- zation in CONUS [25] and the Turkey [26]. Principal component analysis (PCA) was used by [11, 27] for the construction of homogeneous regions, whereas Sadri and Burn (2011) used Fuzzy C-Means (FCMs) along with L-moments [28]. Each technique has its own advantages and shortcomings [14]. All of these methods have the ability to recognize the subtropical/tem- perate boundary. However, none of these methods can objectively identify the transition zone. Moreover, climatic regionalization typically corresponds to vegetation distribution in the sense that each climate type is dominated by a vegetation zone or ecological region [29–32]. None of the above methods takes into account the impact of vegetation. It is possible to use the conjugate relationship between vegetation and climate to indicate the regional zone attributes [33]. Moreover, the sensitivity of various vegetation types to cli- mate differs, and the boundary of typical subtropical/temperate vegetation will be slightly dif- ferent. For example, the typical subtropical/temperate vegetation boundary of coniferous forest and broad-leaved forest in same area will be different. These differences are the manifes- tations of the gradual transition characteristics of the transition zone. It is reasonable to delimit the transition based on the boundaries of zonal vegetation of different types (coniferous forest

and, broad-leaved forest). Therefore, to identify the location of the transition zone of the QDM, this paper focuses on the delimitation of the boundaries between typical zonal vegeta- tion (typical subtropical vegetation and typical temperate vegetation). To delimit the transitional zone of subtropical/temperate boundaries, a combination of the Geodetector model and the support vector machine (SVM) was used to identify multiple boundaries between typical zonal vegetations. First, the Geodetector model is a novel tool to measure the degree of spatial stratified heterogeneity [34, 35]. Similarly, the basis of compre- hensive physical regionalization is spatial stratified heterogeneity. Xu (2018) used this model to evaluate the indicators of China’s subtropical/tropical zone [36]. The indicator system of covariates for different vegetation types can be built using this model. Second, the goal of our workflow is to identify the boundary between two zonal vegetation type. The SVM model is ideal classifier to solve this problem. Because the SVM model was initially designed for binary classification [37], it can effectively distinguish between two zonal vegetation types. By analyz- ing the spatial characteristics of numerous boundaries of zonal vegetation, the composition of boundaries can be seen as the transitional zone. Such research can provide a new perspective on the comprehensive physical regionalization of China and reveal the ecological significance of the QDM for China. Materials and methods Study area

QDM, which is situated in the central part of China (30˚-36˚N, 101˚-114˚E), is the transitional zone of the subtropical zone and temperate zone (shown in Fig 1). It stretches across six prov- inces, including Shaanxi, Gansu, Szechwan, Chongqing, Hubei, and Henan, with a land area of 300,000 km2 and greater than 1000 km long [38]. Influenced by both the monsoon climate and the continental monsoon climate, its weather is warm and humid, and its environment has a high level of biological diversity [39, 40]. Moreover, characterized by high mountains alternating with deep gorges, the Qinling Mountains serve as the watershed of the Yangtze River and the Yellow River and are regarded as the natural boundary of geography and climate between northern and southern China. Given the integrated influence of horizontal zonality, vertical zonality, aspect and slope, soil environment and even human activities, the environ- ment of QDM is highly complex, diversity and heterogeneous. The spatial distribution of PLOS ONE Climate regionalization based on Geodetector-SVM model PLOS ONE | https://doi.org/10.1371/journal.pone.0241047 November 2, 2020 3 / 16