The study of temperature regionalization in China using complex networks

Abstract Temperature regionalization is important for guiding the development of agriculture and forestry in a region. However, few existing studies are able to objectively regionalize temperature in terms of both the temperature change and temperature difference. In both cases, we use the ERA5 2 m temperature reanalysis dataset to construct the temperature correlation network by traversing the winter anomalies with higher amplitudes for all possible pairs of grid points. On the basis of the temperature correlation network, the first and second‐grade regionalization is performed using the spin glass algorithm and agglomerative hierarchical clustering, respectively. Our results derived from the network topology analysis indicate that there are a few teleconnections in the network and that the nodes are mainly connected to neighbouring nodes. Some nodes in Tibet, Qinghai and western Sichuan Province have independent temperature change, and some nodes in Guizhou and Jiangxi Provinces tend to connect more nodes with an average distance of >500 km. The results of temperature regionalization show that the boundaries of the first‐grade regionalization are consistent with the relief and the distribution of major mountains in China, and the spatial distribution coincides with the four core regions with the highest convergence coefficient of temperature change (TCC) of nodes and 10 regions with relatively high TCC. The second‐grade regionalization is mainly located in the different temperature zones. In contrast to previous works on temperature regionalization, this study provides a new integration framework that further considers both the convergence in the temperature change and the closeness of the temperature between regions and finally objectively regionalizes temperature in China.