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

Validation Boundary’s accuracy. The boundaries extracted by this study were divided into two groups: the broad-leaved forest and the coniferous forest (Fig 5a and 5b, respectively). The plant community survey data were used as a reference to represent the real distribution of veg- etation types. Compared with these data, the SVM results were highly consistent with the plant community survey data. Specifically, as shown in Fig 5a, the number of typical subtropical broad-leaved vegetation types (Platycarya strobilacea forest) increases gradually from the south to the boundary. At the same time, there is no Platycarya strobilacea forests are not observed north of the boundary. The same pattern is evident in the eastern and central regions of the QDM. This finding confirms that the boundary is consistent with the real distribution of the broad-leaved forest. In addition, as shown in Fig 5b, the number of Pinus tabuliformis forests gradually decrease north of the coniferous boundary, whereas the number of Pinus massoniana forests gradually increase from the coniferous forest boundary to the south. More- over, the numbers of both forests near the boundary are low. These findings indicate that the delimitation of the coniferous forest boundary is reasonable and consistent with the real distri- bution characteristics of the coniferous forest. What’s more, the location is approximately the same as the previous climatic boundaries (Renmeie, Luokaifu Houxueyu) (Fig 5c). Is indicate that the delimitation of transition zone is reasonable and consistent with the previous studies.

Characteristics of the transition zone As described in the above sections, q-statistics can assess the spatial stratified heterogeneity. Higher q-statistics value indicate increased reliability of the delimitation of the transitional zone. As shown in Fig 5c, the QDM are divided into three regions, namely, the temperate zone, the subtropical zone, and the transition zone. According to the regionalization indexes of all the regionalization schemes, we chose seven of the most commonly used indexes to ana- lyze the characteristics of the transition zone [7–10, 17, 18, 44]. Based on the Geodetector model, a total of 500 random samples were chosen in the three above zones, and the average value and the quantitative q-statistics of the seven indexes were generated (shown in Table 2). On one hand, the values of the seven indexes of the three zones exhibited visible gradient changes. The values of accumulated temperature, annual average temperature, and annual average precipitation gradually increase from the warm temperate zone to the subtropical zone and reach the median value in the transitional zone. The transitional zone acts as a buffer zone; the climate in this region exhibits a distinct transitional character. For example, in the transition zone, the coldest month average temperature is -1.4175˚C, the warmest month aver- age temperature is 20.812˚C, and the annual precipitation is 803.257 mm. On the other hand, statistically significant spatial heterogeneity is noted amount the seven

indexes, and their q-statistics are generally greater than 0.35. For example, the q-statistic of the average annual precipitation is 0.7054. The average q-statistic of the previous regionalization is approximately 0.33 [36], demonstrating that the transition zone in this study is reasonably Table 2. Characteristics analysis result of the transition zone. meantem07 meantem01 dem tadem pa aat10dem aat0dem q 0.35012 0.467286 0.314 0.3778 0.7054 0.351184 0.37433 Temperate zone 16.1323 -5.61467 2323 49.963 6327.66 17805.49 24530.28 Transitional zone 20.812 -1.4175 1422 91.155 8032.57 30051.34 35703.27 Subtropical zone 23.443 1.675667 963 118.12 10035.1 37481.14 43698.89 https://doi.org/10.1371/journal.pone.0241047.t002 PLOS ONE Climate regionalization based on Geodetector-SVM model PLOS ONE | https://doi.org/10.1371/journal.pone.0241047 November 2, 2020 12 / 16