The effects of human movements on urban climate over Eastern China

We live on an urban planet with unprecedented human mobility. In this study, we collectively analyze ten large cities over densely populated Eastern China, and detect that mass people outflows during Spring Festival (SF) holiday significantly cool down urban climate expressed as urban heat island intensity variation particularly at the nighttime after minimizing the effects of other factors. We estimate that the average nighttime cooling effects of the ten large cities over Eastern China during the SF holiday relative to the nearby background period are 0.63 °C stronger during the 2000s than during the 1990s. The attribution analysis points to that the urban cooling effects are primarily caused by mass people outflows during the SF holiday. Our findings help to better understand the complex interactions of human population dynamics, urban development, and the environment, and may have important implications for promoting sustainable, people-centered, and resilient development of our urban planet.


INTRODUCTION
The number of global human population has reached up to around 7.7 billion in 2019 with a further expected increase to nearby 10 billion by 2050 1 . In the increasingly populous, interlinked and urbanized world, people mobility is reshaping the human society and the urban planet. Mainly due to rapidly growing movements from rural communities to urban areas, 55% of the total population in the world-4.2 billion people-resided in cities in 2018 2 . As the most noticeable rural-to-urban migration continues, more 2.5 billion urban dwellers are expected to add by 2050, with 35% of the growth occurring in just three countries including India, China, and Nigeria 2 . Cities are at the center of social and economic development of the world, and also become main drivers of global energy consumption, greenhouse gas emissions and climate change [3][4][5][6][7][8][9] . Managing urban growth with unprecedented human mobility to promote the benefits and opportunities of the agglomeration, while reduce adverse impacts on the environment and our society is key to sustainable development of the world now and in the future [10][11][12][13][14][15][16][17] . However, it remains very challenging to assess what effects human mobility can have on urban climate and other aspects of the environment.
China is the largest country in terms of population size, and has witnessed unprecedented urbanization development in the human history during the past decades [44][45][46][47][48] . China's total population reached 1.4 billion with around 900 million people living in urban areas in 2020, and the urban dwellers are expected to surpass 1 billion by 2030. Spring Festival (SF) is the most important holiday in China, and each year mass national population movements take place during the SF holiday. The number of passenger trips and journeys during the SF holiday was estimated to run into 1 billion in the early 1990s, and increased rapidly to 3 billion or more in the 2010s. The mass outflows of urban people from large cities lead to abruptly decreased human activities associated with largely reduced anthropogenic heat release and other affected processes, and may thus be noticeably influential to urban climate 49,50 . Several recent case studies indeed demonstrated observational evidence for the SF human movement impacts on urban temperature and relative humidity of Chinese large cities [51][52][53][54][55][56] .
The majority of floating people which are the main trigger of the SF mass population movements live in large cities over densely populated and economically developed Eastern China, especially concentrating in municipalities and provincial capital cities 49,57 . Here, we systematically assess the effects of people movements on urban climate expressed as UHI intensity (UHII) variation during the SF holiday over Eastern China. Ten municipalities and provincial capital cities including Beijing, Tianjin, Shijiazhuang, Taiyuan, Zhengzhou, Shanghai, Hangzhou, Nanchang, Changsha, and Fuzhou are selected and analyzed for the period of 1990-2014, taking into account carefully the availability, quality, and reliability of station observations (see "Methods" section). In 2020, the ten large cities are home to over 126 million population, and account for above 15% of China's total gross domestic product (GDP) (Supplementary Table 1).
We detect collective effects of population outflows on urban surface air temperature relative to the nearby surroundings at the ten large cities over Eastern China by minimizing the roles of other 1 processes (see "Methods" section). The SF mass human movements are driven primarily by the floating people, which have increased rapidly in recent years (Supplementary Table 2). Thus, we take the 1990s as the reference period with the least likely impacts, and calculate their differences with those during 2000s and 2005-2014. Finally, we perform attribution analysis of urban coolings to mass people outflows during the SF holiday.

RESULTS
The effects of SF human movements on urban climate Large cities over Eastern China have increasingly attracted the floating people from rural areas and smaller cities in the past decades. Most of the floating people or migrant workers in these large cities tend to leave for their native place during the SF holiday, which is the most precious family reunion time. We define the SF day to Lantern Festival as the SF holiday period, and choose 60 days including the 45th days to the 16th days before the SF day, and the 25th days to the 54th days after as the non-SF background period ( Table 1). The UHII was calculated as surface air temperature difference between urban and non-urban reference stations for each large city ( Fig. 1 and Supplementary Table 3). Then, the daily UHII anomalies were obtained by removing the daily UHII climatological values of all years. The trends in the differences of UHII anomalies between the SF holiday and the non-SF background period are used to assess the effects of human population movements on urban temperature. Details of data sources and statistical analysis are provided in the "Methods" section.
We find that the UHII anomalies during the SF holiday with regard to the nearby non-SF background period all show the cooling trends for the ten large cities over Eastern China, with magnitudes ranging from −0.1 to −0.56°C per decade for the study period in terms of mean surface air temperature (Supplementary Fig. 1 and Supplementary Table 4). Especially, the cooling trends have stronger magnitudes of −0.14 to −0.70°C per decade during the night-time.
Collectively, the averaged cooling trends of urban daily mean, maximum, and minimum surface air temperatures in the SF holiday are −0.30, −0.14, and −0.43°C per decade for the ten large cities over Eastern China during 1990China during -2009, and are all significant at the 95% confidence level (Supplementary Table 4). In comparison, the cooling effect is much stronger during the nighttime than the daytime. Among the ten large cities, the urban station of Shijiazhuang and the urban and non-urban stations of Taiyuan have relatively large relocations (>5 km) during the 2010-2014 (Supplementary Table 3). We excluded the two cities and perform the trend analysis for other eight large cities to assess the urban cooling effects of people outflows during 1990-2014. The cooling trends in the SF holiday are −0.28 and −0.40°C per decade for urban daily mean and minimum surface air temperatures, both significant at the 99% confidence level (Supplementary Table 4). Ten-year moving averages to reduce the effects of interannual variations are further conducted, and the results show that the cooling effects of people outflows are much larger in recent years over Eastern China, especially for the night-time ( Supplementary Fig. 2).

The urban cooling effects during different periods
China's population was urbanizing rapidly during the study period, with urbanization rate increasing from 26.41% in 1990 to 54.77% in 2014 (Supplementary Table 2). Comparatively, large cities have experienced accelerating urbanization since around 2000 according to China's urbanization policies 47 . The average annual rate of increased urban population was estimated at 1.33% during 2000-2014, which was more than 0.4% higher than that of the 1990s (Supplementary Table 2 Table 2). As a consequence, human population movements were much greater during the 2000s and 2010s than during the 1990s. Therefore, we take the 1990s as the reference period, which represents the least likely impacts of human movements on urban climate during the SF holiday, and further  Table 5).
The averaged cooling effects characterized as urban daily mean surface air temperature during the SF holiday relative to the non-SF background period in the 2000s are much stronger than those in the 1990s for all ten large cities, ranging from −0.25 to −0.79°C. The nighttime cooling effects during the 2000s with respect to the 1990s vary at magnitudes of −0.17 to −1.06°C for ten large cities.
Collectively, the averaged urban cooling effects during the 2000s relative to the 1990s baseline in terms of daily mean, maximum and minimum temperatures for the ten large cities over Eastern China are −0.44, −0.19, and −0.63°C, and they are all significant at the 95% confidence level. During the cold season, energy consumptions for heating are much higher over Northern China than over Southern China. The averaged cooling effects during the 2000s relative to the 1990s are −0.45, −0.26, and −0.65°C for five cities over Northern China (Beijing, Tianjin, Shijiazhuang, Taiyuan, and Zhengzhou) for daily mean, maximum, and minimum temperatures, while they are −0.43°C, −0.13°C, and −0.62°C for five cities of Southern China (Shanghai, Hangzhou, Nanchang, Changsha, and Fuzhou). We also examine the averaged cooling effects during the 2005-2014 relative to the 1990s for eight large cities, which have no relocations or small relocations during the whole study period of 1990-2014. The averaged urban cooling effects during the 2005-2014 relative to the 1990s in terms of the daily mean, maximum, and minimum temperatures are −0.52°C, −0.22°C, and −0.73°C, and they are significant at the 99%, 95%, and 99% confidence levels, respectively (Supplementary Table 6). Together, these results highlight the significant cooling effects of population outflows on urban climate during the SF holiday over Eastern China.
The dependence of urban coolings on the floating population Beijing, Tianjin, and Shanghai municipalities rank the top three most densely populated cities in China, and have the floating population data we can access (Supplementary Table 7). Here we further take these three large cities to explore the dependence of the urban coolings on the floating people, which are the main trigger of mass human movements during the SF holiday (Fig. 3). The results show that the floating people all had rapidly increased in recent years, accompanying with continuously remarkable decreases in urban temperature during the SF holiday over the three selected large cities. The R 2 values of the linear fits for urban  (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)) for ten large cities over Eastern China: a daily mean surface air temperature (T mean ). b Daily maximum surface air temperature (T max ). c daily minimum surface air temperature (T min ). d T mean , T max , and T min averaged over ten large cities. Urban temperature variations are characterized by the differences in the UHII anomalies between the SF holiday period and the non-SF background period. The gray and black stars in d represent the 95 and 99% confidence levels, respectively. The ±0.5 standard deviations are also shown in d. Map was generated using ArcGIS 10.2 (www.esri.com/software/arcgis).
daily mean temperature variations to the floating people are all high ( Table 2, 79-93%), pointing to that the urban coolings during the SF holiday are primarily induced by mass people outflows. The tight couplings between urban temperature variations during the SF holiday and the floating people are also evident via the linear fits for daily maximum and minimum temperatures. Using logarithmic fits, we also detect the strong couplings of urban temperature decreases during the SF holiday with the floating people (Table 2).
Abruptly decreased human activities induced by mass people outflows can lead to large reductions of energy consumptions and anthropogenic heat releases, thus causing the remarkable urban   L. Wu and J. Zhang coolings 49,50,55 . A recent study demonstrated that sudden and large drops of energy consumptions and associated CO 2 emissions occurred during the SF holiday over China 58 . Based on these analyses above together with our and other previous studies, the notable urban coolings of large cities over Eastern China during the SF holiday can be attributed primarily to the effects of mass people outflows via reducing the anthropogenic heat emissions and modulating the other processes.

DISCUSSION
Mass migration towards cities has reshaped our planet into the urban era. Deepening scientific understanding and enhancing effective managements of accelerated people movements, and their impacts play a particularly vital role in transforming the urban planet to a world of sustainable development in the Anthropocene 59,60 . However, little attention has been paid to the role that human mobility can play in reshaping urban climate.
Taking the peculiar opportunity of the greatest periodic SF human movements, we reveal the remarkable cooling effects of mass population outflows during the SF holiday relative to the non-SF background period in ten typical large cities over Eastern China. The cooling effects averaged over the ten large cities during the 2000s relative to the 1990s are −0.44, −0.19, and −0.63°C for daily mean, maximum, and minimum temperatures, which are all significant at the 95% confidence level. The much larger cooling effects of mass people outflows during the nighttime than during the daytime are expected, since the UHIs usually have remarkably stronger magnitudes at night. The averaged cooling effect for five large cities located over Northern China are larger than those of five cities over Southern China for all three temperature variables, which is in line with more energy consumptions for heating over Northern China. The urban coolings are well coupled with the floating population, and can be attributed primarily to people movement effects.
Collectively, our findings highlight the vital importance of human mobility to urban climate. Additionally, mass human movements to cities can worsen air quality, rise GHG emissions, and generate more wastes and noise among others, posing unprecedented environmental pressures on urban systems [61][62][63][64][65] . These issues are all vital and need to be addressed in the future.
Meanwhile, human population dynamics, urban systems, and other specific local and regional contexts are highly complex and diverse. Further efforts are needed to reduce the uncertainties of our estimates and better elucidate the underlying mechanisms for the people movement effects on urban climate. For example, it is very difficult to choose appropriate urban and non-urban stations for Chinese large cities due to rapid urbanization over the past decades, and the non-urban surroundings also experienced the people outflows to some degree during the SF holiday. The observations from dense automatic weather stations should be used in future, and may be able to reduce the biases of our estimates.
A portfolio of strategies and approaches such as increasing green space and ecological corridors, improving energy efficiency, and deploying clean and renewable energy can produce multiple co-benefits such as reduced UHI effects, improved air quality and decreased greenhouse gas emissions 10,11,66 . In the coming decades, the number, areas, and population of urban settlements will continue to grow rapidly, particularly in the global south 13 . Special attentions should be paid to the collaborative development of large, medium, and small cities, taking into consideration local, regional and global contexts in an integrated and systematic way 9,67 .
We have entered a new era of unprecedentedly growing human mobility. Except for rural-to-urban movements, the international migrants have risen from 153 million in 1990 to around 272 million in 2019 68 . In Europe and Northern America, international migrants currently account for more than 10% of the total population. With easier and faster transportation, the international tourist arrivals have grown noticeably from 278 million in 1980 to 1451 million in 2018 69 . Better understanding the complex interactions of human movements, urban systems, and the environment across various space and time scales provides critical scientific information for policy makers, private sectors, and other actors to successfully implement and achieve the New Urban Agenda and several key Sustainable Development Goals of the United Nations 70,71 .

METHODS Data
In this analysis, the daily mean, maximum, and minimum surface air temperature data were obtained from the China Meteorological Administration (http://data.cma.cn/[2020-01-24]).
The data of population, GDP, and percentage of tertiary industry in Supplementary Table 1

Spring Festival holiday period and the non-Spring Festival background period
In the lunar calendar, the first day of the first month is called SF in China, and it falls on a different date each year in the solar calendar (Table 1). Traditionally, people celebrate the SF holiday from the SF day to Lantern Festival (the first day to the fifteenth day of the first month in a Chinese year). We therefore define the fifteen days as the SF holiday period. The annual mass people migration generally occurs during 40 days with starting at the 15th days before and ending on the 24th days after the SF day-Spring Festival Transportation (SFT) period. After excluding the SFT period, we select 60 days as the non-SF background period which includes the 45th days to the 16th days before the SF day, and the 25th days to the 54th days after (Table 1).

Ten large cities over Eastern China
The definitions and classification of cities vary across countries and regions. In 2014, Chinese government defines a city with the population living in core urban areas over 1 million as a large city (http://www.gov.cn/zhengce/ content/2014-11/20/content_9225.htm/[2021-06-20]). The floating people which are the main trigger of SF mass human movements are mainly concentrated in densely populated and economically developed Eastern China, especially in municipalities and provincial capital cities 49,57 . There are 15 municipalities and provincial capital cities including Beijing, Tianjin, Shijiazhuang, Taiyuan, Zhengzhou, Jinan, Nanjing, Hefei, Shanghai, Wuhan, Hangzhou, Nanchang, Changsha, Fuzhou, and Guangzhou over Eastern China. According to the definitions and criteria established by Chinese government, the 15 cities are all classified as large cities.
For each large city, a pair of urban and non-urban reference stations were chosen according to the following criteria: (1) The two stations have no any relocations or only have small relocations (<5 km); (2) the two stations have consistent or similar climate background: the distance between two stations is shorter than 100 km, and the elevation difference is smaller than 120 m; (3) the two stations have obvious contrast in population condition: the population density difference between two stations is larger than 2000 person per km in 2000. The population density was calculated by averaging the values within 5 km circling each station. We used the tools of Buffer, Extract by Mask, and Zonal Statistics as Table  in software of ArcGIS 10.2 to convert the population density data to the tabulated data (Supplementary Table 8). Note that the elevation difference of 120 m between urban and surrounding reference stations may cause a bias of above 0.5°C in estimating the UHII. Such an effect can be largely removed in this study since we use relative change of the UHII.
Carefully taking into consideration the availability, quality, and reliability of the observational data judged by the above three criteria, ten large cities including Beijing, Tianjin, Shijiazhuang, Taiyuan, Zhengzhou, Shanghai, Hangzhou, Nanchang, Changsha, and Fuzhou were selected and analyzed among 15 municipalities and provincial capital cities over Eastern China for the period of 1990-2014 in this study. Figure 1 and Supplementary Table 3 present the locations and population densities of these urban and nearby non-urban reference stations of the selected ten large cities. The remaining five large cities (Jinan, Hefei, Nanjing, Wuhan, and Guangzhou) were excluded mainly due to stations relocations caused by such as rapid urbanization and land use management.
For each one of all the selected ten large cities over Eastern China, the amounts of the total and urban population are above 5 million and 4 million, respectively (Supplementary Table 1). The three municipalities of Beijing, Tianjin, and Shanghai all have urban population of exceeding 10 million. The ten large cities accounted for over 11% of China's total urban population, and over 15% of China's total GDP. The economies of the ten large cities are dominated by the tertiary industry.

Analysis
We assess the effects of mass human movements on urban climate of ten large cities (all for municipalities and provincial capital cities) over Eastern China during the SF holiday after minimizing the impacts of other factors 55 . First, we calculated the surface air temperature difference (ΔT) between urban and nearby non-urban reference stations or UHII for each large city to separate the local urban effects because the UHII can minimize the impacts of large-scale or regional atmospheric circulation 55,73 . Second, to reduce the impacts of the variability, the daily UHII anomalies were produced by subtracting the daily UHII climatological values of all years over a given period. Then, the differences in UHII anomalies (ΔT SF-BK ) between the SF holiday period and the non-SF background period were calculated, and we examined the trends of ΔT SF-BK to detect urban temperature variations induced by mass people outflows over Eastern China (Supplementary Fig. 1 and Supplementary Table 4).
Since the human movements are much smaller during the first 10 years (1990s, 1990-1999) of the study period than during the following years, we estimated urban temperature changes induced by more mass human movements during the 2000s for ten large cities and during 2005-2014 for eight large cities by computing the differences with the reference period of 1990s ( Fig. 2 and Supplementary Tables 5-6). Note that the resulting variations in urban temperature variables including daily mean, maximum, and minimum temperatures are relative to the surrounding non-urban values. We also calculated the 10-year moving values of ΔT SF-BK relative to those for the 1990s baseline period (Supplementary Fig. 2). Finally, we performed attribution analysis by examining the dependence of urban temperature variations on the floating population, as the primary driver of mass people movements in three selected large cities including Beijing, Shanghai, and Tianjin municipalities using linear and logarithmic fits (Fig. 3 and Table 2).

DATA AVAILABILITY
The data used in this study, including population, GDP, and percentage of tertiary industry were all obtained from publicly available data with free accesses. The population density in 2000 and 2010 over Eastern China were taken from Shen et al. 72 . The surface air temperature data were obtained from the China Meteorological Administration.