**1. Introduction**

The climate system is a complex system, which influence on ecosystems and human society has attracted more and more attention from scholars, and temperature is one of the most important factors in the climate system.

A number of studies [1–7] have indicated that the spatiotemporal variation of temperature and its driving factors had regional differences. Sharma et al. [8] analyzed the temperature changes in eastern India, and the results showed that the average temperature in central, southern and western was decreasing, while the average temperature in the northeast, west and southeast was on the rise. Salman et al. [9] conducted a hybrid model to select the climate models for simulating spatiotemporal changes in temperature of Iraq. They found that temperature would increase during the period of 2070–2099 and temperatures in the north and northeast had increased significantly. Kenawy et al. [10] pointed out that temperatures in northeastern Spain showed an upward trend during the 1960–2006 period, and the Eastern Atlantic (EA), the Scandinavian (SCA), and the Western Mediterranean Oscillation (WeMO) patterns had a significant impact on temperature changes. Iqbal et al. [11] found temperature had di fferent correlations with North Atlantic Oscillation (NAO), Arctic Oscillation (AO), El Niño-Southern Oscillation (ENSO), and North Sea Caspian Pattern (NCP) in di fferent months in Pakistan. It can be seen that the temperature change is complicated. Therefore, further understanding the mechanisms for spatio-temporal variation of temperature and its driving factors are highly desired.

In order to reveal the complexity of temperature, many methods had been proposed, such as wavelet analysis [12,13], ensemble experience mode decomposition [14], spectrum analysis [15], Mann-Kendall trend test [16], and correlation dimension [17]. All of these methods had explored the complexity of temperature from di fferent perspectives and go<sup>t</sup> some achievements. On the other hand, there are many studies about the driving factors of temperature change. The main driving factors are atmospheric circulation [2], land use changes [3], greenhouse gas emissions [18], urbanization development [19], and so on. However, under the global warming, coupled with rapid economic development, population growth, and urbanization, the temperature and its driving factors became more and more complicated. In addition, the contribution rate of natural and socioeconomic factors and their interactions on temperature variation were rarely studied and remained one of main gaps in our current knowledge.

Due to the regional di fferences, it is necessary to conduct an in-depth analysis of temperature variations in some key areas, especially those that play an important role in national development. The Yangtze River Delta (YRD), one of China's most developed, dynamic, densely populated and concentrated industrial areas, is growing into an influential world-class metropolitan area. However, the developed industries and frequent human activities have led to an increasingly serious urban heat island phenomenon in this region, forming a strong regional heat island, leading to the temperature presenting a significant warming trend over the past 50 years and extremely high temperatures occuring frequently [20]. Property, economic losses, and social impacts caused by extremely temperature events in this region are often enormous. In addition, extreme changes in temperature can also have an impact on the environment and endanger human health [21]. Therefore, it is of grea<sup>t</sup> significance to study the temperature changes in this region, to find the reasons that a ffect its changes, and to try to reduce losses.

Therefore, we attempt to explore the spatiotemporal dynamics of temperature in the YRD, and assess the influences of factors and their interactions on temperature. Based on observed temperature data at 68 meteorological stations during the period of 1980–2012, we first investigated the spatiotemporal complexity of temperature by using the Correlation Dimension (CD) method; and then we analyzed the individual contribution rates and interactional contribution rates of driving factors to temperature slope (TS) by using the Geogdetector method. Our main purpose is to explore which factors or interactions between factors contribute the most to temperature.

#### **2. Materials and Methods**

#### *2.1. Study Area and Data*

The study area includes four regions: Jiangsu Province, Anhui Province, Zhejiang Province, and Shanghai (Figure 1). The study area lies between 114◦54 –122◦42 E and 27◦12 –35◦20 N, and has an area of approximately 344.03 10<sup>3</sup> km2, accounting for 3.58% of China's total land area. The area is under a monsoon climate regime, with hot and humid summer and cold and dry winter. The annual precipitation is about 1000 mm, of which the precipitation in summer accounts for two-thirds of the total precipitation [22]. The average temperature is close to 30 ◦C in July and August, and the maximum temperature recently exceeded 40 ◦C in Shanghai [23]. The high terrain is in the north and south and low terrain in the middle, which is dominated by plains and hills. In addition, the YRD is one of the most developed regions in China, with dense population, convenient transportation, and developed tertiary industry.

**Figure 1.** The study area and spatial distribution of 68 meteorological stations.

On a global scale, the temperature is mainly affected by factors such as atmospheric circulation, volcanic eruptions, sunspots, and so on. However, on the regional scale, the temperature is mainly affected by surface properties and human activities. According to previous studies [24–26], the altitude (AT), normalized difference vegetation index (NDVI), urban density (UD), gross domestic product (GDP), and night light (NL) datasets were selected. The first two can be seen as natural factors and the last three can be seen as socioeconomic factors. The daily temperature of 68 meteorological stations from 1980 to 2012 is from the China Meteorological Data Service Center (http://data.cma.com). We analyzed data from the period from 1980 to 2012 because we couldn't ge<sup>t</sup> station data of temperature for 2013–2018. AT and UD data are provided by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC) (http://www.resdc.cn), and the UD data is from 1990 to 2010. NDVI is from the Geospatial data cloud (http://www.gscloud.cn/), and its period is from 1989 to 2012. NL is from the National Centers for Environmental Information (https://www.ngdc.noaa.gov/), and its period is from 1992 to 2012. GDP is from the "Shanghai Statistical Yearbook", "Anhui Statistical Yearbook", "Jiangsu Statistical Yearbook", "Zhejiang Statistical Yearbook", and "China Regional Economic Statistics Yearbook" and other statistics, and its period is from 1980–2012. To ensure a consistent data format, a 0.5 km by 0.5 km grid for the whole area in ArcGIS 10.5 software (Manufacturer, City, US State abbrev. if applicable, Country) was built, assigned values to each grid, and deleted the outliers by using a box-plot analysis method. According to different standards, all factors were divided into different strata using ArcGIS 10.5 software. The division of the results is shown in Figure 2 below.

**Figure 2.** The distribution of driving factors: (**a**) altitude (AT); (**b**) normalized difference vegetation index (NDVI); (**c**) urban density (UD); (**d**) gross domestic product (GDP) change rate; (**e**) night light (NL).
