ZHAO Deming,ZHA Jinlinnd WU JinCAS Key Lortory of Regionl Climte-Environment for Temperte Est Asi,Institute of Atmospheric Physics,Chinese Acdemy of Sciences,Beijing,Chin;Deprtment of Atmospheric Science,Yunnn University,Kunming,Chin

ABSTRACT Urban-related warming in two first-tier cities(Guangzhou and Shenzhen)in southern China with similar large-scale climatic backgrounds was compared using the nested weather research and forecasting regional climate model.The default urban data in the model were replaced by reconstructed annual urban data retrieved from satellite-based images for both coarse-(including all of China)and fine-resolution domains(eastern China and three city clusters in China:Beijing–Tianjin–Hebei(BTH),the Yangtze River Delta(YRD),and the Pearl River Delta(PRD)),which reproduced urban surface expansion during the past few decades.The results showed that the 37-year(1980–2016)area-averaged annual urban-related warming was similar(0.69°C/0.64°C)between the urban areas of Guangzhou/Shenzhen;however,the values across the entire area of the two cities varied(0.21°C/0.45°C).Seasonal characteristics could be detected for mean surface air temperatures(SAT)at 2 m,SAT maximum and minimum,and diurnal temperature range(DTR).Both the SAT maximum and minimum generally increased,especially over urban areas;however,changes in the SAT minimum were larger,which induced a decrease in DTR.The DTR in summer decreased by-0.25°C/-0.86°C across the entire area of the two cities and decreased by-0.93°C/-1.15°C over urban areas.The contributions of urban surface expansion to regional warming across the entire area of the two cities were approximately 17%/35%of the overall warming and much greater over Shenzhen.However,the values over urban areas were much closer to the values from total warming(35%/44%).

KEYWORDS Urban surface expansion;Guangzhou;Shenzhen;urban-related warming;numerical simulation

1.Introduction

China has experienced rapid urbanization and economic growth since reform and open policies started in late 1978(Zhou et al.2004),accompanied by population aggregation,urban area expansion,and pollution in cities.However,the increasing extent in population and urban areas might vary among different cities,which can result in different impacts on,and contributions to,regional climate change.Guangzhou,located in southern China,was already a metropolis in the early 1980s,with a population size of nearly 5.00 million.The population subsequently increased to 8.83(1990),9.95(2000),12.71 (2010),and 13.50 (2015) million.Meanwhile,the gross domestic product(GDP)was 5.75 billion Yuan in 1980 and grew to 31.96(1990),249.27(2000),1074.83(2010),and 1810.04(2015)billion Yuan,showing a rapid increase(GMBS 2016).However,Shenzhen,which is also located in southern China and is not far from Guangzhou in terms of its large-scale climatic background,was only a small town with a population of 0.33 million population in the early 1980s,but has since developed into one of the four first-tier cities in China(the other three are Beijing,Shanghai,and Guangzhou).The population in Shenzhen increased over time to 1.68(1990),7.01(2000), 10.37(2010),and 11.38(2015)million people;meanwhile,the GDP increased to 0.27(1980),17.16(1990),218.75 (2000),977.33 (2010),and 1750.28(2016)billion Yuan(SMSB 2016).These statistics show a typical urbanization process(Zheng et al.2011),which is further revealed by the urban surface expansion between 1980 and 2016.The geographic location and large-scale climatic background were quite similar;however,the increasing extent of the population and urban areas varied,which may have induced different impacts on surface air temperature at 2 meters(SAT)both spatially and temporally.As a result,urban-related warming comparisons were performed to reveal the impacts on regional climate in Shenzhen and Guangzhou.

Many previous studies have been performed to detect the impacts of urbanization on SAT(Fan et al.2006;Xiong et al.2012)and their corresponding trend changes using the observation-minus-reanalysis(OMR)method(Kalnay and Cai 2003;Zhou et al.2004;Kalnay et al.2006),the urban-minus-rural(UMR) method(Hua,Ma,and Guo 2008),and numerical simulations(Wang et al.2016;Zhang et al.2016;Zhao and Wu 2017a;b).However,results have varied and shown many uncertainties.For example,Si et al.(2010)used the UMR method and found that urban-related warming between 1979 and 2005 was 0.243°C per decade,accounting for 36.3% of the total local warming.However,when the OMR method was adopted,urban related warming was approximately 0.315°C,which accounted for 47.1%of total warming.

The OMR method can be used to investigate the impacts of urbanization on SAT,but errors from different sources in the reanalysis data remain unavoidable and can be falsely attributed to the effects of urbanization.Furthermore,the OMR method is difficult to use to evaluate small-scale influences on land use changes due to the coarse resolution of the reanalysis data and the heterogeneity of the observation stations(Wu et al.2017).Therefore,the uncertainty of the estimation based on the OMR method might be non-negligible(Vose et al.2004).For the UMR method,most Chinese meteorological observation stations are located in or near cities,with only a few actually in mountainous or remote regions,which can be regarded as true rural areas.Furthermore,due to rapid urbanization during the past three decades,a rural station might change into a city station and relocation was made.The continuous expansion in population and in city areas makes the classification of urban and rural stations much more difficult(Zhou et al.2004).Therefore,it is difficult to find truly rural observation stations for most cities,especially in southeastern China.Numerical simulations with fine-resolution regional climate models have also been widely used to investigate the impacts of urbanization on regional climate and are regarded as an effective approach when evaluating the urbanization effect on SAT,which can provide homogenous and area-averaged values for spatial and time series.However,many previous studies were performed at local scales(Zhang et al.2016),and these studies failed to include the impacts of urbanization from large-scale circulation against climatic backgrounds.

Here,reconstructed annual urban data derived from satellite-based images(Jia et al.2014;Hu et al.2015)across China(30-km resolution),with an emphasis on eastern China(10-km resolution)and three city clusters(3.3-km resolution in Beijing–Tianjin–Hebei(BTH),the Yangtze River Delta(YRD),and the Pearl River Delta(PRD))were used to perform nested numerical experiments using the Weather Research and Forecasting(WRF;Skamarock et al.2008)regional climate model to detect the impacts of urban surface expansion on SAT across China.Two cities (Guangzhou and Shenzhen)in southern China,against similar largescale climatic backgrounds,were chosen to make comparisons.

2.Data and experimental design

Detailed information on the data used and the numerical experimental design can be found in Zhao and Wu(2017b),which were used to perform an analysis on the impacts of urban surface expansion on SAT and its trends in Beijing(please refer to section 1 and 2 of the Supplemental Material,and Figure S1).Here,we transferred our focus to two cities(Guangzhou and Shenzhen)in southern China,which have experienced different urban area increases over the past few decades.

3.Results

3.1.Changes in urban surface distributions over the last 37 years

Land use and land cover changes(including urban surface changes)are usually not covered in reginal climate models,so land use data are kept at fixed values.However,the urban surface distribution of land use in Guangzhou/Shenzhen from the United States Geological Survey (10/0 urban cells)and the International Geosphere Biosphere Programme Moderate Resolution Imaging Spectroradiometer(240/107 urban cells)were quite different.With satellite-based derived data,changes in urban surface distributions were clearly detected over the past 37 years(Figure 1(a)–(l)).For Guangzhou city,the urban grid cells were 10(1980),63(1990),113(2000),210(2010),and 283(2016),and the corresponding values were 0(1980),31(1990),73(2000), 120(2010),and 143(2016)in Shenzhen.Marked differences could be found between the two cities in the 1980s,because no urban surface existed in Shenzhen since it was only a small town.With the reconstructed urban data,the model displayed a good performance in terms of SAT simulations,and the simulated SATs were close to the observed data over China,as shown in Zhao and Wu(2017b).

Figure 1.Spatial distributions of different land use categories in(a–f)Guangzhou and(g–l)Shenzhen under(a,g)U80,(b,h)U90,(c,i)U00,(d,j)U01,(e,k)U10,and(f,l)U16,where red denotes urban surface grid cells.Land use categories are:(1)evergreen needleleaf forest,(2)evergreen broadleaf forest,(3)deciduous broadleaf forest,(4)mixed forests,(5)closed shrublands,(6)open shrublands,(7)woody savannas,(8)grasslands,(9)permanent wetlands,(10)croplands,(11)urban and built-up areas,(12)cropland/natural vegetation mosaic,(13)barren or sparsely vegetated areas,and(14)water.

3.2.Urban-related warming during different time periods

The 37-year annual and area-averaged urban-related warming was 0.21°C*/0.45°C***(*,**,***,and****denote passing the 80%,90%,95%,and 99%confidence level based on the t-test,respectively)for the areas of Guangzhou/Shenzhen(GMBS 2016;SMSB 2016),as shown in Figure 2(e)and(j).The corresponding warming was 0.69°C***/0.64°C***in urban areas,including grids that were classified as urban for both time periods(U2U)and grids that changed from non-urban to urban(N2U).Warming over urban areas was similar,however,and contributions to areas of Guangzhou and Shenzhen varied,in which warming for Shenzhen was twice as much as that of Guangzhou.These results can be expressed as the percentage of urban areas compared to the whole area of the two cities in 1980 and 2016,which was much greater in Shenzhen.Furthermore,urban-related warming in U2U(0.36°C**)over Guangzhou was much smaller than inN2U(0.70°C***),which was close to the values in urban areas(0.69°C***).However,there were no U2U areas over the Shenzhen area.

Figure 2.Spatial distributions of annual averaged urban-related warming(units:°C)between(a,f)1980 and 1989,(b,g)1990 and 1999,(c,h)2000 and 2009,(d,i)2010 and 2016,and(e,j)1980 and 2016,in(a–e)Guangzhou and(f–k)Shenzhen.(a,b),(f,g)and(c,h)depict 10-yr averages;(d,i)shows 7-yr averages;(e,j)shows 37-yr averages;shaded areas pass the 90%confidence-level based on the t-test.

Due to the varying degree of urban surface expansion over different time periods,area-averaged urban related warming displayed a variation in intensity:0.018°C/0.10°C,0.17°C/0.33°C*,0.39°C**/0.68°C***,and 0.30°C*/0.79°C*** over the periods of 1980–89,1990–99,2000–09,and 2010–16,respectively(a 7-year mean was calculated for the period 2010–2016,while a 10-year mean was calculated for the other periods)in Guangzhou/Shenzhen (Figure 2(a)–(d)and 2(f)–(i)).The corresponding values over urban areas were greater:0.12°C/0.14°C,0.54°C***/0.49°C***,1.14°C****/0.96°C****,and 1.14°C****/1.13°C****,respectively.

Urban-related warming displayed marked seasonal differences(see Figure S2),which were lower in winter and much stronger in other seasons.For the areas of Guangzhou/Shenzhen,the values were 0.24°C*/0.46°C***in spring,0.33°C*/0.46°C**in summer,0.23°C*/0.54°C***in autumn,and 0.046°C/0.32°C*in winter.However,over urban areas,the values were 0.70°C***/0.66°C***,0.87°C***/0.66°C***,0.79°C***/0.77°C***,and 0.40°C**/0.48°C**.Urban-related warming was bigger in summer in Guangzhou,whereas the intensity was greater in autumn in Shenzhen.These could be further revealed by the influence on the radiation budget due to urban surface expansion,for which surface radiative forcings were 2.70 W m-2(summer)and 6.74 W m-2(autumn)over urban areas of Guangzhou,whereas the values were 1.89 W m-2(summer)and 7.90 W m-2(autumn)over urban areas of Shenzhen.The radiative forcing in summer over urban areas of Guangzhou was bigger than that of Shenzhen,while opposite results were detected in autumn.

3.3.Impacts on diurnal temperature range

The impacts on daily SAT maxima and minima,which further influence the diurnal temperature range(DTR),have been shown to vary based on observed data(Zhou et al.2004).Simulated annual results showed an increase in maximum and minimum SAT,which induced a decrease in DTR.For the areas of Guangzhou and Shenzhen,changes in the SAT maximum,minimum,and DTR were 0.0078°C/0.026°C,0.33°C/0.88°C,and-0.32°C/-0.85°C,respectively;however,the differences in values were 0.12°C/0.013°C,1.19°C/1.23°C,and-1.08°C/-1.21°C over urban areas(Figure S3).The DTR was generally greater in Shenzhen than in Guangzhou,especially over the whole area of both cities,which can be attributed to a greater decrease in the SAT minimum due to urban surface expansion in Shenzhen.The DTR also expressed marked seasonal variations,which were-0.23°C/-0.87°C,-0.25°C/-0.86°C,-0.49°C/-1.09°C,and-0.30°C/-0.82°C from spring to winter over the Guangzhou/Shenzhen area,and-0.96°C/-1.18°C,-0.93°C/-1.15°C,-1.37°C/-1.44°C,and-1.06°C/-1.08°C over the urban areas.

Further analysis of the summer data using probability density functions(Figure 3)showed that both the SAT maximum and minimum generally increased,especially over urban areas;however,changes in the SAT minimum were bigger,which induced a decrease in DTR.

Over the Guangzhou area,the impacts on the SAT maximum and minimum showed that changes in the annual SAT maximum were close between U2U(0.12°C)and N2U(0.12°C);however,the values for the annual SAT minimum(0.45°C and 1.23°C)were markedly different,which induced a smaller DTR in U2U(-0.33°C)and a bigger DTR in N2U(-1.11°C),which more closely resembled urban areas(-1.08°C).

3.4.Relative contribution of urban-related warming

Time series for the 37-year averaged SAT showed increasing trends a cross the Guangzhou/Shenzhen area and were 0.55°C/0.47°C per decade in the experiment without urban surface expansion(EX1),which were comparable to the results from observational data(Si et al.2010;Feng and Pan 2011;Zheng and Yang 2016).The increasing trends were 0.67°C/0.73°C per decade when urban surface expansion was considered(EX2).The relative contribution of urban surface expansion to regional warming was approximately 0.12°C/0.26°C per decade,which was approximately 17% and 35% of the total warming,respectively(Figure 4(a)and(c)).

When only urban areas were considered,the increasing trends were 0.54°C/0.48°C per decade for EX1 and 0.94°C/0.85°C for EX2.The contributions from urban surface expansion were approximately 35%and 44%,respectively.The evaluated contributions were comparable to the values using the UMR or OMR methods,which were 44%(UMR method)and 56%(OMR method)between 1979 and 2010 in the PRD(Chen 2013),whereas 36%(UMR method)and 47%(OMR method)between 1979 and 2005 in Shenzhen(Si et al.2010).

Over Guangzhou areas,contributions of urban related warming in U2U(24%)were smaller than in N2U(43%),for which the latter were closer to those in urban areas(Figure S4).

Figure 3.Changes in the probability density function for the daily SAT maximum/minimum(Tmax/Tmin),daily mean SAT(Tmean)and diurnal temperature range(DTR)in summer across(a,c)the whole areas and(b,d)urban areas of(a,b)Guangzhou and(c,d)Shenzhen for EX1 and EX2.

4.Conclusions and discussion

Area-averaged urban-related warming of two cities(Guangzhou and Shenzhen)in southern China was analyzed with reconstructed satellite-retrieved urban surface images over China at three spatial resolutions(30,10,and 3.3 km)using the nested WRF regional climate model. Comparisons of urban-related warming between Guangzhou and Shenzhen were performed since the two cities are geographically close and are controlled by similar large-scale climatic backgrounds.However,differences were detected for relative urban areas near the two cities.

Though the 37-year,annual and area-averaged urban-related warming over the urban are as of Guangzhou and Shenzhen were similar(0.69°C/0.64°C),values across the entire two cities were considerably different(0.21°C/0.45°C),which was expressed by the percentage of urban areas compared to the entire area of the two cities in 1980 and 2016.Meanwhile,urban related warming in U2U(0.36°C)over Guangzhou was much smaller than in N2U(0.70°C)due to the intense expansion of urban surface areas between 1980 and 2016,and the value in N2U was much closer to the urban area value(0.69°C).These impacts also display varying intensities during different time periods,which are highly connected with urban areas and the degree of expansion.

Marked seasonal characteristics can be detected for the mean SAT,maximum,and minimum SAT,as well as the corresponding DTR.Further analysis of characteristics in summer using the probability density functions showed that both the SAT maximum and minimum generally increased,especially over urban areas;however,changes in the SAT minimum were larger,which induced a decrease in DTR.The DTR in summer decreased by-0.25°C/-0.86°C across the entire area of the two cities,and by-0.93°C/-1.15°C over the corresponding urban areas;these results were similar to the results over the urban areas but had quite different intensities for the area of the two cities.

Figure 4.Time series of annual averages in surface air temperature and the trends for EX1 and EX2 for(a,c)the whole areas,(b,d)urban areas(including U2U and N2U)of(a,b)Guangzhou and(c,d)Shenzhen.

Relative contributions from urban surface expansion to urban-related warming across the whole area of the two cities were approximately 17%/35%of the overall warming between 1980 and 2016 and were much greater over Shenzhen.However,the values over urban areas had more similar values,of 35%/44%.

The impacts of urban surface expansion on the SAT maximum and minimum in summer can be explained by changes in the energy budget during the daytime and nighttime(Table 1).Surface radiative forcings were all positive during the daytime,at 5.40 W m-2/12.33 W m-2over the entire area of Guangzhou/Shenzhen,and 11.28 W m-2/28.99 W m-2over the urban areas.However,values were negative at nighttime(-1.58 W m-2/-1.86 W m-2over the entire area of Guangzhou/Shenzhen,and-4.69 W m-2/-2.71 W m-2over the urban areas).Precipitation increased during both daytime and nighttime over the entire area of the two cities and the urban areas,which was consistent with the increased cloud cover at different levels(with the exception ofdecreased low-cloud cover over the Guangzhou area at nighttime).The increased cloud cover further contributed to a weakened downward shortwave flux(daytime only)and an intensified downward longwave flux.Meanwhile,with the urban surface expansion,albedo decreased,which induced a weakened upward shortwave flux.This decreased downward shortwave flux(smaller)and increased upward shortwave flux(larger)induced a positive shortwave flux.

Table 1.Changes in the energy budget,cloud fraction,and near-surface wind speed in the summer under EX1 and EX2 between 1980 and 2016.

In the daytime,changes in net radiative forcing mainly resulted from the impact of the shortwave flux,primarily from the decreased upward shortwave flux due to a decreased albedo over urban areas.In contrast,at nighttime,the corresponding changes were only induced by the impacting longwave flux.The intensified ground heat flux in the daytime contributed to an increased upward longwave radiation flux and sensible heat flux at nighttime.The upward longwave flux at nighttime was stronger than that during the daytime,and more energy was used to heat the atmosphere,which contributed to an increased SAT minimum at nighttime.Meanwhile,the Bowen ratio generally increased,showing an increased sensible heat flux with a decreased latent heat flux,which can also be attributed to the increase in impervious surfaces and weakened near-surface wind speed,even though precipitation and soil moisture increased.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the National Natural Science Foundation of China(Grant Nos.41775087 and 41675149),the National Key Research and Development Program of China(Grant No.2016YFA0600403),the Chinese Academy of Sciences Strategic Priority Program(Grant No.XDA05090206),and the Jiangsu Collaborative Innovation Center for Climatic Change.