ZHAO De-Ming nd WU Jin

aCAS Key Laboratory of Regional Climate-Environment for Temperate East Asia, Institute of Atmospheric Physics, Chinese Academy of Sciences,Beijing, China; bDepartment of Atmospheric Science, Yunnan University, Kunming, China

ABSTRACT To detect the impacts of urban surface expansion on surface air temperature at 2-m (SAT) in Shanghai, China, nested numerical integrations based on satellite-derived urban data between the 1980s and 2010s were performed using the Weather Research and Forecasting (WRF) model.Urban surface expansion induced an annual-averaged warming of 0.31 °C from 1980 to 2016 across the whole of Shanghai, showing the greatest intensity between 2010 and 2016. The values were 0.36, 0.78, and 0.75 °C over grids that were classified as urban in both time periods (U2U), landuse grids that changed from non-urban to urban (N2U), and urban areas (including U2U and N2U),respectively, and revealed weak warming over the inner-ring areas because the urban surfaces had been there since the 1980s, whereas warming areas were coincident with the outward expansion of the urban surface. Meanwhile, marked seasonal variations could be detected, which were greater in spring and summer but less in autumn and winter. The approximately homogenously distributed SAT maximum (weaker) and heterogeneously SAT minimum (stronger) contributed to the decreased diurnal temperature range. Regional warming induced by urban surface expansion was approximately 0.12 °C per decade, which accounted for 19% of the overall warming across the whole of Shanghai. The values were 0.11 °C per decade and 0.39 °C per decade over U2U and N2U,which accounted for approximately 17% and 42% of the overall warming, respectively, and resulted in approximately 41% of the overall warming over urban areas.

KEYWORDS Urban surface expansion;surface air temperature;regional warming; nested integration; warming trend

1. Introduction

As Shanghai is the national economic, financial, and commercial center of China, the gross domestic product(GDP) of Shanghai ranks first among Chinese cities (China Statistical Yearbook 2016). The Yangtze River Delta (YRD)area, including Shanghai, Jiangsu, Zhejiang, and Anhui provinces, has become one of the six largest city clusters in the world (Tu 2015). In 1980, the GDP was 31.2 billion Yuan and the resident population was 11.5 million in Shanghai(Shanghai Statistical Yearbook 2016). With the implementation of reforms and opening-up policies since the 1980s,Shanghai has experienced rapid economic development accompanied by GPD improvement and greater population concentration. In the years of 1990, 2000, 2010, and 2016, the GDP was 78.2, 477.1, 1716.6, and 2746.6 billion Yuan, respectively, and the population increased to 13.3,16.1, 23.0, and 24.2 million (Shanghai Statistical Yearbook 2016; Shanghai Municipal Bureau of Statistics 2017), which were accompanied by the expansion of urban areas. Peng et al. (2013) noted that the urban heat island intensity is mostly influenced by urban surface expansion in Shanghai.Therefore, it is necessary to evaluate the influences of the expanded urban areas on surface air temperature at 2-m(SAT) in Shanghai.

Several approaches have been developed and adopted to explore the impacts of urbanization on SAT. One is the observation-minus-reanalysis (OMR) method (Kalnay and Cai 2003), which has been applied in evaluations of the effects of urbanization over eastern China (Zhou et al. 2004; Zhang et al. 2005; Yang, Hou, and Chen 2011).However, the results may contain substantial uncertainties due to the coarse resolution of the reanalysis data when a single city, such as Shanghai, with an area of 6340 km2, is considered. Another method is the SAT difference between urban and rural meteorological stations. However, there are difficulties in choosing the rural stations because the observational stations over eastern China, most of which are located near cities, are to varying degrees affected by the urbanization (Ren, Ren, and Zhang 2010). Indeed, some stations have had to be relocated due to the urban expansion. Results based on 11 observational stations from 1980 to 2006 revealed the rate of increase ranged from 0.56 to 1.05 °C per decade in Shanghai (Hou et al. 2008), and the impacts of urban surface expansion on the SAT differed depending on the rural stations selected (Cao et al. 2008;Zhang et al. 2010). The third approach is fine-resolution numerical integration with a regional climate model (RCM),which has also been widely used over eastern China (Zhao and Wu 2017a); however, most experiments of this kind have concentrated on a single city or city cluster (Liu et al.2005; Du et al. 2010), which might cause them to underestimate the interactions between different cities or city clusters (Zhang, Shou, and Dickerson 2009). Meanwhile,some experiments have been performed based on virtual land use data (Shao, Song, and Ma 2013), which might lead to overestimated or underestimated results.

Here, nested integrations with reconstructed annual urban data based on five satellite-derived land-use images in 1980, 1990, 2000, 2010, and 2016, in China (U1980,U1990, U2000, U2010, and U2016, respectively) (Hu et al. 2015; Jia et al. 2014), which could represent the urban surface expansion in the past several decades, were performed to explore the influences of the increase in the urban area on the SAT using the Weather Research and Forecasting (WRF) model (Skamarock et al. 2008).

2. Experimental design and data

The experimental design and data were the same as those employed in Zhao and Wu (2017b), except that their study was focused on urban-related warming in Beijing. Here, the regional warming induced by urban surface expansion in Shanghai was explored. The model domain and terrain elevation with nested domains (BTH, Beijing–Tianjin–Hebei;YRD, Yangtze River Delta; PRD, Pearl River Delta), the terrain elevation in the YRD region (including Shanghai city), and the terrain elevation over Shanghai city, are displayed in Figure 1(a)–(c).

In order to detect the impacts of urban surface expansion on regional warming between 1980 and 2016, two long-term integrations driven by NCEP–NCAR reanalysis data (1979–2016; Kanamitsu et al. 2002) were performed.For the first experiment (EX1), the urban data of U1980 were used throughout the integrations; whereas, the reconstructed annual urban data from 1980 to 2016 based on the five satellite-derived images (Hu et al. 2015; Jia et al. 2014) were adopted in the second experiment (EX2).The two numerical experiments shared the same integration backgrounds, with the exception of the urban grid distribution across the whole of China at different spatial resolutions (30 km resolution across the whole of China,10 km resolution over eastern China, and 3.3 km resolution in the three city clusters (BTH, YRD, and PRD)), in which the reconstructed annual urban data were used to replace the default values for both coarse and fine meshes in China.Compared to continuous long-term integrations, systematic errors of RCMs can be mitigated with consecutive reinitializations (Qian, Seth, and Zebiak 2003). Meanwhile,an unrealistic discontinuity of interannual urban surface data can induce spurious values during long-term integrations. Therefore, the numerical experiments consisted of a series of restarts for individual years with the reconstructed annual urban data from 1 July of the previous year, the first six months of which were used as ‘spin-up’ time and only the integrated results for the present year were analyzed.The dominant land use category was determined based on Guo and Chen (1994). Detailed information on the experimental design and data can be found in Sections 1 and 2 of the Supplemental Material, respectively.

3. Results

3.1. Urban surface expansion in the past 37 years

The fixed-in-time urban data in the commonly used RCMs(such as MM5 (U1992) and WRF (U2001)) could not capture the rapid urban surface expansion across the whole of Shanghai. Meanwhile, a marked difference could be detected for the numbers of urban areas in Shanghai between U1992 and U2001, which included 16 and 253 urban grid cells with the greatest fraction for urban and built-up areas, respectively. The reconstructed urban data displayed changes in the number of urban grid cells, which represented urban surface expansion in Shanghai in the last 37 years (Figure 1(d)–(h)). There were only 12 urban grid cells in 1980. However, the number of urban grid cells was 71, 134, 202, and 264 in 1990, 2000, 2010, and 2016, respectively. Long-term integrations with the reconstructed land use data (EX1 and EX2) exhibited good performance in SAT simulation over China, which is discussed in Zhao and Wu (2017b). The bias in Shanghai was ?1.29 °C under EX1, but ?0.98 °C under EX2, as evaluated with the 0.25° × 0.25° resolution SAT data from 2400 China meteorological observation stations throughout the whole of China (Wu and Gao 2013).

Figure 1. (a) Model domain and terrain elevation (units: m) with nested domains (BTH, Beijing–Tianjin–Hebei; YRD, Yangtze River Delta;PRD, Pearl River Delta). (b) Terrain elevation in the YRD region (units: m), including Shanghai city. (c) Terrain elevation over Shanghai city(units: m). (d–h) Spatial distribution of different land use categories in Shanghai under (d) U1980, (e) U1990, (f) U2000, (g) U2010, and(h) U2016, with red coloring denoting urban surface grid cells. Land use categories: (1) evergreen needleleaf forest; (2) mixed forest; (3)closed shrubland; (4) open shrubland; (5) woody savanna; (6) grassland; (7) permanent wetland; (8) cropland; (9) urban and built-up areas; (10) cropland/natural vegetation mosaic; (11) barren or sparsely vegetated areas; (12) water. (i–m) Spatial distribution of annualaveraged urban-related warming (units: °C), (i) between 1980 and 1989, (j) between 1990 and 1999, (k) between 2000 and 2009, (l)between 2010 and 2016, and (m) between 1980 and 2016, in Shanghai ((i), (j) and (k) depict 10-yr averages; (l) shows 7-yr averages; (m)shows 37-yr averages; the shaded areas passed the t-test at the 90% confidence level).

3.2. Urban surface expansion induced warming over different time periods

The impacts of urban surface expansion–induced warming exhibited varied intensities over different time periods(Figure 1(i)–(m)). The annual-averaged values of the warming across the whole Shanghai area were 0.12, 0.30*, 0.43***,and 0.45 °C*** (*, **, ***, and **** denote passing the t-test at the 80%, 90%, 95%, and 99% confidence level, respectively) for the periods 1980–1989, 1990–1999, 2000–2009,and 2010–2016, respectively (10-yr-averaged for the former three time periods, but 7-yr-averaged for the last one;hereafter, we employ the same definition for multiple-year averaged values), and were much stronger for the period 2010–2016 but much weaker for the period 1980–1989. As a whole, the value for the period 1980–1916 was 0.31 °C**.When Chongming Island (Figure 1(c)) was excluded, the warming was 0.35 °C**, which was anomalously high compared to the value across the whole of Shanghai.

For the spatial distributions over different time periods,similar distributions could be detected, which showed weak warming over the inner-ring areas, as the urban areas had been there since the 1980s, whereas greater warming could be detected outside the inner-ring areas with the urban areas’ outward expansions. The values were 0.17, 0.39***, 0.46***, and 0.45 °C*** over grids that were classified as urban in both time periods (U2U) for the time periods of 1980–1989, 1990–1999, 2000–2009, and 2010–2016. For land use grids that changed from non-urban to urban (N2U), the values were 0.21, 0.68****, 1.13****,and 1.24 °C****. This meant that the warming was much greater over N2U than over U2U. The values over urban areas (including U2U and N2U), which were 0.21, 0.66****,1.09****, and 1.19 °C****, were close to those over N2U due to the intense urban surface expansion. As a result, the 37-yr annual-averaged values were 0.36***, 0.78****, and 0.75 °C**** over U2U, N2U, and urban areas, respectively.

3.3. Seasonal characteristics for urban surface expansion–induced warming

Urban surface expansion–induced warming across the whole of Shanghai exhibited marked seasonal variation (Figure 2), in that the warming was greater during spring (March–April–May, 0.42 °C***) and summer (June–July–August (JJA), 0.38 °C***) but weaker in autumn(September–October–November, 0.29 °C*) and winter(December–January–February, 0.15 °C), which is consistent with the results from Kang (2014) using observed data between 1970 and 2009 in the YRD region.

3.4. Spatial characteristics of SAT maximum/minimum and diurnal temperature range

The spatial characteristics of the JJA SAT maximum/minimum and diurnal temperature range (DTR) showed that nearly homogenously equal changes in the SAT maximum and heterogeneously distributed changes in the SAT minimum contributed to the similar spatial characteristics of the DTR as those of the SAT minimum(Figure 3(a)–(c)).

Figure 2. Spatial distribution of seasonal urban-related warming for 37-yr averages in Shanghai between 1980 and 2016 (units: °C): (a)spring; (b) summer; (c) autumn; (d) winter.Note: Shaded areas passed the t-test at the 90% confidence-level.

3.5. Impacts on the DTR

To further disclose the impacts of urban surface expansion on SAT, the daily SAT maximum/minimum, daily mean SAT and DTR in summer imposed by the probability density functions were adopted (Figure 3(d)–(f) and Figure S1).Across the whole of Shanghai, a larger increase in the SAT minimum (0.61 °C) and a smaller increase in the SAT maximum (0.28 °C) resulted in a decrease in the DTR(?0.33 °C). For the U2U and N2U areas, the DTR decreased by ?0.17 and ?0.96 °C, respectively, i.e. the effect was much stronger over N2U. The differences in the DTR between U2U and N2U could be explained by the different values for the SAT minimum changes (0.54 and 1.41 °C), while the SAT maximum changes were nearly the same (0.37 and 0.45 °C). Changes in the DTR over urban areas (?0.91 °C)were close to the values over N2U due to the similar intensity of the SAT maximum (0.45 and 0.44 °C) and minimum(1.41 and 1.35 °C) changes between N2U and urban areas.

3.6. Changes in warming rates

Figure 3. (a–c) Changes in JJA-averaged values of the (a) SAT maximum (Tmax), (b) SAT minimum (Tmin), and (c) diurnal temperature range(DTR) for 37-yr averages in Shanghai between 1980 and 2016. (d–f) Changes in the probability density function for Tmax/Tmin, daily mean SAT (Tmean) and DTR in the summer (d) across all areas and (e, f) in subregions ((e) U2U; (f) urban areas) of Shanghai between EX1 and EX2.

Time series of regional warming induced by urban surface expansion were further explored across the whole of Shanghai. The rate of warming for the 37-yr-averaged SAT was 0.54 °C per decade from EX1. When urban surface expansion was considered (EX2), the rate of warming was 0.66 °C per decade. Regional warming induced by urban surface expansion was approximately 0.12 °C per decade, which accounted for 19% of the overall warming(Figure 4(a)).

The rates of SAT increase over U2U under EX1 and EX2 were 0.52 °C per decade and 0.63 °C per decade, respectively (Figure 4(b)). However, the corresponding values over N2U were 0.53 °C per decade and 0.92 °C per decade, respectively (Figure 4(c)). Regional warming induced by urban surface expansion was 0.11 °C per decade and 0.39 °C per decade over U2U and N2U, which accounted for approximately 17% and 42% of the overall warming,respectively. The warming rates were 0.53 °C per decade and 0.90 °C per decade from EX1 and EX2 over urban areas,respectively, which accounted for approximately 41% of the overall warming (Figure 4(d)), and was comparable with the values (44%) using the OMR method over a large metropolis in eastern China (Yang, Hou, and Chen 2011).

4. Conclusions and discussion

With reconstructed annual land use data between 1980 and 2016, it was possible to simulate the rapid urban surface expansion in China, especially in the three city clusters of BTH, YRD and PRD. Urban surface expansion–induced warming across the whole of Shanghai (within the YRD region) was explored using the WRF model, which expressed considerable impacts on the SAT for the period 1980–2016.

The 37-yr annual-averaged urban surface expansion–induced warming was 0.31 °C across the whole of Shanghai. However, the value over N2U areas (0.78 °C) was much greater than that over U2U (0.36 °C), the latter of which corresponded to the inner-ring areas in Shanghai.Meanwhile, due to the rapid urban surface expansion and intense increase in urban grid cell numbers, the warming over urban areas was similar to that over N2U areas.

Urban surface expansion–induced warming showed varied intensities under different time periods with different urban surface expansion backgrounds. Meanwhile, marked seasonal characteristics could be detected for the impacts on SAT, which were much weaker in autumn and winter.

The homogenously distributed SAT maximum and the heterogeneously distributed SAT minimum, both of which increased, but more so the latter, contributed to the decreased DTR, especially over urban areas.

When exploring regional warming induced by urban surface expansion from 1980 to 2016, marked spatial differences could be detected. In the case of the whole Shanghai area, the contribution was approximately 19%.Meanwhile, the corresponding values were 17% and 42%over the U2U and N2U areas, respectively, which resulted in an approximate 41% contribution over urban areas.

Figure 4. Time series of annual averages in surface air temperature and the trends between 1980 and 2016 under EX1 and EX2 (a) across all areas and (b, c) in subregions ((b) U2U; (c) N2U; (d) urban areas) of Shanghai.

The intense urbanization from 1980 to 2016 contributed to an increase in SAT in both spatial (Figure 1(i)–(m)) and warming rates (Figure 3(a)–(d)) (Zhao and Wu 2017a). Meanwhile, the increased SAT maximum (less)and minimum (greater) induced a decreased DTR (Figure 2(a)–(c)), which could be attributed to the differences for the impacted surface energy budget between the daytime and nighttime (Table 1). In the daytime, changes in the radiation budget mainly resulted from the weakened upward shortwave flux due to the urbanization-induced decreased albedo, and resulted in a positive radiative forcing, with the exception of negative radiative forcing over N2U. The Bowen ratio increased, showing an increase in sensible heat flux but a decrease in latent heat flux (due to the decreased near-surface wind speed and increased impervious surfaces inducing weakened upward moisture). The ground heat flux increased in the daytime, which in turn induced an increase in upward longwave radiation flux and sensible heat flux at nighttime and contributed to an increase in the SAT minimum, and thus a decreasein DTR, especially over urban areas. Meanwhile, a weak increase in DTR could be detected over suburbs due to the increased SAT minimum being mainly located over urban areas (while much less so in the suburbs) at nighttime.Whereas, in the daytime, the increase in SAT over urban areas could have impacts on SAT changes over the suburbs through urbanization-induced intensified turbulence exchanges in the planetary boundary layer, which induced a homogenous spatial distribution for the SAT maximum.

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

Acknowledgments

The authors thank the two anonymous reviewers for their numerous valuable comments that helped to improve the original manuscript.

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 number 41775087]; National Key R&D Program of China [grant number 2016YFA0600403]; the Chinese Academy of Sciences Strategic Priority Program [grant number XDA05090206]; the National Natural Science Foundation of China [grant number 41675149]; and the Jiangsu Collaborative Innovation Center for Climatic Change.