**4. Discussion**

In this study, we found that the temperature rose 1.53 ◦C with an average rising rate of 0.465 ◦C/10 years during the study period, which was higher than the global average rate. The result was consistent with previous studies [37–39]. It confirms the regional differences in climate change. In addition, it means that the temperature was not only affected by global warming, but also affected by various driving factors within the region. In addition, the temperature rose quickly in the dense areas of population and urban, and the temperature rose slowly in the sparse areas of population and urban. It reflected the urban-rural differences in temperature distribution from the side.

The climate system was an open system with external forcing and nonlinear dissipation [40], and fractal theory was one of the effective methods to quantitatively describe the nonlinear evolution process of climate and its self-similar structural features. Numerous studies [41–45] had shown that fractal analysis could calculate its fractal dimension from a seemingly chaotic climate sequence, confirming the fractal information of the climate system. Temperature was an element of the climate system and also had nonlinear characteristics. Especially in the YRD, the temperature was more complicated due to the influence of human activities. By calculating the CDs on the daily, seasonal, and annual scales of the YRD, we confirmed that the temperature in the YRD was a chaotic dynamic system with nonlinear characteristics. We found that temperature on the annual scale was more complicated than on the daily scale in the YRD. It was because the annual average temperature was the average of the daily temperature, which was the macroscopic performance of the daily temperature and influenced by many factors [46,47], so it showed greater complexity on the whole. Xu et al. [17] found that the temperature process on the daily scale was more complicated than the temperature process on the annual scale in Xinjiang. It was contrary to the YRD, indicating the complexity of the temperature process had regional differences. In the spatial distribution, whether in the daily, seasonal, or annual scale, the high CD values were mainly distributed in the sparse areas of population and urban, while the low CD values were mainly distributed in the dense areas of population and urban. In the dense areas of population and urban, due to the density of cities and people, industrial and urbanization were developing rapidly, and the temperature was mainly a ffected by the rapid development of cities, showing an upward trend [38]. While in the sparse areas of population and urban, the temperature changes were mainly a ffected by natural factors and socioeconomic factors together, so the temperature changes more complicated.

The e ffects of five driving factors on the TS were quantitatively investigated by using the Geogdetector method. UD was the most important factor a ffecting TS. From Figures 2c and 8, we can see that the spatial distribution of UD was similar to the spatial distribution of TS, that is, decreasing toward the periphery with Shanghai as the center. The UD reflected the intensity of the city. In Shanghai and its surrounding areas, cities were dense and urbanization was high. One of the most striking features of this was that the impervious surface of the city increased rapidly [48,49]. The impervious surface of the city had strong heat storage, poor water storage capacity, and hindering airflow transmission, which seriously a ffected the city's surface hydrological cycle [50], energy distribution [49] and urban microclimate [51], resulting in an urban heat island e ffect, which causes the temperature in dense areas of urban to rise quickly. The impact of urban impervious surface on surface temperature had been verified in di fferent regions and a certain consensus had been reached [52–55]. The contribution rate of GDP to TS was second. From Figures 2 and 6, we can see that the spatial distribution of GDP was similar to the spatial distribution of TS. The development of GDP inevitably consumed a large amount of energy, which would emit a large amount of greenhouse gases, resulting in a quick increase in temperature. The GDP in Shanghai and its surrounding areas was increasing rapidly, so the TS in this area was high. The contribution of NL was similar to the contribution rate of GDP, and the spatial distribution of NL was similar to the TS. NL reflected the level of GDP and energy consumption from the side [56,57], so it had a high contribution rate to TS. The NDVI was the smallest in Shanghai and its surrounding area, while the TS was largest in this area, which meant that the vegetation coverage rate played an important role in suppressing the increase of temperature, but it was not enough only to rely on the vegetation coverage rate. Most of the YRD was plain and the fluctuation of terrain was small, so the contribution rate of AT to the TS was small and can be ignored. Each driving factor had an e ffect on TS; they did not work alone, but di fferent driving factors interacted with each other and had an enhanced influence on temperature.

Our main purpose is to explore which factors or interactions between factors contribute the most to temperature. The paper only analyzed the complexity of the temperature process and the contribution rates of driving factors to temperature, but the mechanism behind it remains to be studied further. In addition, through the analysis of the driving factors, some policy opinions to mitigate the temperature rise need to be proposed in the next study.
