- Email: [email protected]
Discrepant impacts of land use and land cover on urban heat islands: A case study of Shanghai, China
Ecological Indicators 47 (2014) 171–178
Contents lists available at ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
Discrepant impacts of land use and land cover on urban heat islands: A case study of Shanghai, China Weifeng Li a , Yang Bai b , Qiuwen Chen a,c,∗ , Kate He e , Xiaohua Ji a,d , Chunmeng Han a,d a
RCEES, Chinese Academy of Science, Shuangqing Road 18, Beijing 100085, China Shanghai Academy of Environmental Sciences, Shunghai 200233, China CEER, Nanjing Hydraulic Research Institute, Nanjing 210029, China d China University of Geosciences, Beijing 100083, China e Department of Biological Sciences, Murray State University, Kentucky 4207, USA b c
a r t i c l e
i n f o
Article history: Received 4 November 2013 Received in revised form 31 July 2014 Accepted 1 August 2014 Keywords: Land cover Landscape pattern Anthropogenic activities Land surface temperature Shanghai
a b s t r a c t The most visible aspect of urbanization is that more and more natural landscape is replaced by anthropogenic land cover/land use, which is the driving force of many ecological and environmental consequences such as urban heat islands. However, the difference between land use and land cover and their implications in ecology is often overlooked. The impacts of urban land cover composition and configuration on land surface temperature (LST) have been extensively investigated, but few studies have explored the relation between LST and land use category. This research takes the inner city of Shanghai as a case and comprehensively investigates the discrepant impacts of land use and land cover on LST. Land use and land cover data are derived respectively from aerial photography and high-resolution satellite imagery (ALOS), and the LST is estimated from Landsat TM images. There are five dominant land use types (new residential, old residential, villas, industrial, and institutional land use) and two major land cover types (vegetated and impervious land cover) in the study area. For most land use types, the land cover composition and configuration are varied. By contrast, no statistical difference is observed among old residential, industrial and institutional land uses for LST. The mean LST of new residential and industrial land use is significantly different, although their land cover compositions and configurations are quite similar. These results indicate that the key factors affecting urban LST are not only land cover patterns, but also other anthropogenic forces. Therefore, the explanation of urban LST by land cover alone is inadequate. Especially at fine spatial scales, information on land use is more meaningful than that of land cover to indicate the impacts of urbanization on ecosystems. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction It is well known that urbanization is one of the most powerful and visible anthropogenic forces on Earth (Cohen, 2004; Chen et al., 2010; Angel et al., 2011). The most evident aspect of urbanization is that more and more natural landscape is replaced by anthropogenic land cover/land use, which causes many ecological and environmental problems, such as the phenomenon of urban heat islands (Chen et al., 2007; Wiedmann et al., 2013). Therefore, better understanding of the driving forces of urban land cover/land use change and their impacts on ecosystems is critical for urban
∗ Corresponding author at: RCEES, Chinese Academy of Science, Shuangqing Road 18, Beijing 100085, China. Tel.: +86 10 62849326; fax: +86 10 62849326. E-mail address: [email protected] (Q. Chen). http://dx.doi.org/10.1016/j.ecolind.2014.08.015 1470-160X/© 2014 Elsevier Ltd. All rights reserved.
ecosystem research and sustainable urban planning (Grove, 1997; Breuste et al., 2008). Notably, the difference between land use and land cover and their implications in ecology is often overlooked. Land use generally refers to how people use the land in terms of social-economic functions, whereas land cover defines the physical pattern of land surfaces. Thus, it is important to distinguish between the two surface landscape indicators, which can reveal the linkage between biophysical features such as land cover and anthropogenic features such as social-economic activities (Hubacek and Sun, 2001; Chen et al., 2006; Chen and Chen, 2010; Han et al., 2014). Numerous studies have investigated urban land cover patterns using multi-resolution remote sensing imagery (Griffiths et al., 2010; Redo and Millington, 2011; Taubenbock et al., 2012; Schneider, 2012; Wu and Zhang, 2012; Ramachandra et al., 2012; Lasanta and Vicente-Serrano, 2012; Sexton et al., 2013), but few studies have quantified urban land use distribution at fine scales.
172
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Urban heat islands, as the most evident negative impact of urbanization, have been widely attributed to land cover/land use change (Jones et al., 1990; Grimm et al., 2008). Generally, urban heat islands can be characterized by air and land surface temperature (LST). Because of the limitation of in situ air temperature monitoring across large spatial-temporal scales, remote sensing provides a powerful way to quantify LST. Many studies have examined the relationship between LST and land cover composition and configuration on multiple scales (Zhang et al., 2009; Buyantuyev and Wu, 2010; Zhou et al., 2011; Li et al., 2011, 2013; Connor et al., 2013; Lazzarini et al., 2013). Most studies found the consistent result that urban vegetation could decrease LST, while impervious land cover increases LST. By contrast, the correlations between LST and land cover configuration were found inconsistent or contradictory in many studies (Zhou et al., 2011; Buyantuyev and Wu, 2010; Li et al., 2013; Connor et al., 2013), and were mostly explained by the scale effect (Zhou et al., 2007; Li et al., 2013). However, due to the notable effects of urban socio-economic activities on LST, the use of land cover pattern as the single driving factor for LST might be inadequate. In recent years, some studies have begun to explore the impacts of anthropogenic factors such as population density and night light on LST at the city or multi-city spatial scale over a region (Diamond and Hodge, 2007; Smith et al., 2009; Sailor, 2011; Peng et al., 2012; Zhou et al., 2012). Yet, a comprehensive study on the influence of anthropogenic factors on LST at fine scale is still lacking. This study takes the inner city of Shanghai, the largest metropolis in China, as a case to comprehensively examine the relations of urban land use and land cover with urban heat islands. The specific objectives are to: (1) quantify LST distribution among different land use categories; (2) investigate whether there are discrepant relations of LST with land use and land cover; (3) analyze the driving forces of anthropogenic factors on LST. This study expects to gain a better understanding of urban LST by analyzing the interactions of the biophysical and socio-economic drivers in term of land use and land cover. Results from this study can support urban land developers and managers in taking concrete measures to regulate urban social-economic activities and landscape patterns to mitigate urban heat islands.
2. Material and methods 2.1. Study area Shanghai is one of the largest and most important industrial centers of China. It is located in East China, between latitudes 30◦ 82 30 N and 31◦ 82 70 N, and longitudes 120◦ 85 20 E and 121◦ 84 50 E, surrounded by the Yangtze River estuary to the north, the East Sea to the east, and the Hangzhou Bay to the south. Shanghai belongs to a subtropical monsoon climate. The average total precipitation is 1067 mm per year, and average monthly temperature ranges from 2 to 27 ◦ C. The total area of Shanghai is about 6340.5 km2 and the total population of the city is about 23.0 million (Shanghai Municipal Statistics Bureau, 2011). Shanghai has experienced rapid urbanization since the implementation of the Reform and Open Policy in 1978. The urban area of Shanghai has increased from 149.85 km2 in 1982 to 998.8 km2 in 2010 (Shanghai Municipal Statistics Bureau, 1982, 2010). The study was conducted at the urban core of Shanghai, which is the origin and heart of this mega city (Fig. 1). The inner city is located within the out-ring road of Shanghai, covering 670 km2 , with very intensive and concentrated urban human activities. The landscape pattern is highly heterogeneous and the impacts of various human activities incur many ecological and environmental problems, such
Table 1 Land cover and land use category of the study area. Categories of urban landscape
Land cover
Impervious land cover
Vegetated land cover Bare soil Water New residential Land use
Old residential
Villas
Industrial
Institutional
Description Artificial lands such as roads, roofs and parking lots paved by impenetrable materials such as asphalt, concrete, bricks, etc. Natural and artificial lands paved by penetrable vegetation such as arbor, shrub and grasses Bare soil or sand not covered by vegetation or impervious lands Land covered by water body such as river, ponds and lakes Lands used by families for private residences or dwellings, with the age less than 15∼20 years old, high-rise apartment buildings and some open space Lands used by families for private residences or dwellings, with the age older than 15∼20 years, dense low-rise apartment buildings and less open space Lands used by single or multiple families for private residences or dwellings, with free-standing homes and greater open space Lands for industrial purposes, usually with multi-builds for different industrial activities, such as workspace, factories or warehouses and associated infrastructure Lands for schools, colleges, universities and research institutes, and associated infrastructure
as urban heat islands, water and air pollution (Li et al., 2009, 2012; Ball et al., 2009; Wang et al., 2013). 2.2. Land cover and land use pattern Land cover and land use have quite different meanings in urban ecosystems, as explained above. This study extracted the main land cover and land use types of the site by applying high resolution Advanced Land Observation Satellite images (ALOS imagery) and aerial photography. All the images were registered to a 1:10,000 scale topographic image. We chose land use classification to represent different socioeconomic activities. For land use mapping, the classification system was set up considering both the ecological effects of different urban socio-economic activities and the existing national standard for land use classification (GB/T 21010-2007). Thus, five dominant land use types were identified in the study area, including new residential, old residential, villas, industrial and institutional land use (Table 1, Fig. 2). The visual interpretation was applied to map urban land uses by integrating the aerial photography taken in June 2010 with a spatial resolution of 0.3 m and the cadastre resources (Fig. 3). The visual interpretation ensured that land use patches were properly classified and represented the land use category. For land cover mapping, four major land cover types were identified to characterize the biophysical features of urban landscape, including impervious land, vegetated land, bare land and water body (Ridd, 1995). The ALOS imagery with a spatial resolution of 2.5 m was acquired on May 2010, and the object oriented classification method was applied to classify land covers. The overall accuracy of land cover classification was 85.3%. Specially, the user’s accuracy for vegetation, impervious, bare land and water body were 80.47, 88.37, 74.29 and 78.57%, respectively. The producer’s
W. Li et al. / Ecological Indicators 47 (2014) 171–178
173
Fig. 1. Location of the metropolis of Shanghai, China and the study site inner city.
accuracy for vegetation, impervious, bare land and water body were 83.78, 78.62, 92.86 and 89.19%. We applied the most important and broadly used landscape metrics to measure the land cover composition and configuration characteristics. The percentage of land cover (PLAND) was used to
measure the composition (McGarigal et al., 2002) (Table 2). The significant effects of land cover composition on LST are because the land surface characteristics such as albedo and evapotranspiration of different land cover components are discrepant and they can directly affect land surface temperature (Zhou et al., 2011; Li
Fig. 2. Aerial photography showing the characteristics of dominant urban land use types of Shanghai in June 2010.
174
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Table 2 Landscape metrics selected in this study. Variable
Description
Formulas
Landscape area (TLA)
Sum of areas of all patches of the same land use type polygon in the landscape (unit: ha)
A 104
Percent cover of class area (PLAND)
Percentage of the area of a particular patch type of total landscape area at class level (unit: %)
100 A
Patch density (PD)
Number of all patches at the given class divided by total landscape area at class level (unit: number per ha)
n A
Mean patch size (MPS)
Average patch size at class level (unit: ha)
Edge density (ED)
Amount of edge relative to the landscape area at class level (unit: m/ha)
×
n
× 104
n j=1
aij
ni
10000 A
×
Patch size standard deviation of all the patches at class level (unit: ha)
Area weighted mean shape index (AWMSI)
Average perimeter-to-area ratio for a landscape, weighted by the size of its patches at class level
1 10,000
n
eik
k=1
n
Patch size Standard deviation (PSSD)
ai
i=1
j=1
n
[aij −(
ni
j=1
2
aij /ni )]
n
0.25×p √
aij
ij
anij j=1
j=1
1 10000
aij
maxaij Largest patch index (LPI)
Percent of the total landscape that is made up by the largest patch at class level (unit: %)
j=1,n A
(100)
A: total landscape area (m2 ); aij : area (m2 ) of patch ij; eik : length of edge (m) of patch ik; pij : perimeter (m) of patch ij; n: number of patches.
The land use map was intersected with the land cover map to obtain the land cover types within each land use unit. Land cover composition and configuration of different land use types were calculated based on each land use patch (polygon), and all the landscape analysis was conducted in ArcGIS. 2.3. Land surface temperature
Fig. 3. Spatial distribution of the land use types within the study site in June 2010.
et al., 2011). For land cover configuration, we selected six landscape metrics to measure the size, shape and fragmentation of land cover patches. These land configuration metrics include patch density (PD), mean patch size (MPS), edge density (ED), standard deviation of patch size (PSSD), area weighted mean shape index (AWMSI) and largest patch index (LPI). MPS and PSSD were used to measure patch size of land cover. AWMSI was used to measure patch shape of land cover. LPI, PD and ED were used to measure the fragmentation characteristics of land cover. Table 2 lists the detailed definitions and formulas of these landscape metrics. These metrics have been well examined by previous studies to have significant impacts on LST, due to the fact that spatial configuration of land cover patches could affect the interactions of heat exchange among adjacent land cover patches (Zhou et al., 2011; Li et al., 2011, 2013; Maimaitiyiming et al., 2014).
In this study, the thermal infrared band 6 of the Landsat TM image (11.45–12.50 nm) with a spatial resolution of 120 m was used to estimate LST. Since Shanghai belongs to a subtropical monsoon climate and has a long summer with a lot of rain during the period of June to August, we selected the image acquired on September 9, 2009, a day with clear atmospheric conditions, when the urban heat island phenomenon was obvious. The widely used single-channel method was applied to calculate LST from the thermal band of Landsat data according to the thermal radiant transfer equation (Zhang et al., 2009). The main steps included: (1) correcting the radiometric and geometrical distortions; (2) converting calibrated DNs to absolute units of at-sensor spectral radiance; (3) converting atsensor spectral radiance to at-sensor brightness temperature (i.e., blackbody temperature); and (4) correcting for spectral emissivity of different land covers to generate the LST data. The mean LST value for each land use polygon was calculated by zonal statistical analysis in the ArcGIS platform. Then, the mean LST values and variance for different land use types were further summarized for analysis of the relations of LST and land use. 2.4. Statistical analysis Because the total number of land use polygons was approximately 8500 in the study, including 3400 of new residential, 1100 of old residential, 150 of villas, 2200 of industrial and 1700 of institutional land use, 10% of the land use polygons of each land use type were randomly selected as samples, except for villas, which were small in number. Considering the accuracy of land cover and LST derived for each land use polygon, the qualified samples for
W. Li et al. / Ecological Indicators 47 (2014) 171–178
175
Table 3 Description of LST for samples of the five land use types. LST
Mean (◦ C)
Std. dev. (◦ C)
New residential land use Old residential land use Villas Industrial land use Institutional land use
30.73 31.72 30.12 31.73 31.50
1.22 1.35 2.27 1.66 1.29
different land use types were 310 of new residential, 100 of old residential, 127 of villas, 199 of industrial and 152 of institutional land use. Relations between LST, land cover patterns and land use types were tested using Pearson’s and partial correlation with statistical significance at p < 0.05. Analysis of variance (ANOVA) was used to test the difference in mean LST among different land use types with statistical significance at p < 0.05. The main steps included: (1) A descriptive analysis was conducted to summarize the average features of land cover composition and configuration for different land use type samples. (2) The mean LST values of different land use types were calculated. By this step, the average land cover patterns distributed within different land use types and average LST of each land use type were known. (3) The spatial autocorrelation of LST among neighboring land use types was tested by using global Moran’s I and local indicators of spatial association (LISA) analysis (Moran, 1948; Anselin, 1995). Global Moran’s I values can range from −1 to 1, indicating the correlations with different degrees; while LISA measures the degree of local spatial autocorrelation at different locations by using a localized Moran’s I. In this study, we divided the study area into 148 polygons with 2 km × 2 km grid size, excluding the areas with low quality on LST inversion from satellite image due to cloud covering. For each divided grid polygons, we calculated the mean LST values of different land use types to test the spatial correlation among neighboring land use types. The neighboring land use patches were decided based on adjacency. We used ArcGIS software to carry out the spatial correlation analysis. (4) The relations between LST and land cover patterns for different land use types were analyzed by Pearson’s correlation analysis. Since many land cover configuration variables significantly relate to land cover composition, partial correlation analysis was further conducted by controlling the variable of land cover composition. (5) A one-way analysis of variance (ANOVA) was used to investigate whether the LST and land cover pattern varied among different land use types. 3. Results 3.1. Variation of LST among different land use categories The LST distribution of the inner city of Shanghai is shown in Fig. 4, which ranged from 16.7 to 44.3 ◦ C (Table 3). Among the five dominant urban land uses, industrial land use had the highest mean LST value of 31.73 ◦ C, followed by old residential (31.72 ◦ C), institutional (31.5 ◦ C), new residential land use (30.73 ◦ C), and villas (30.12 ◦ C). Interestingly, there was no statistical difference in LST between old residential, industrial and institutional land use, but mean LST values of other land uses were significantly different from each other (Table 4). Both global and local spatial autocorrelation indices were nonsignificant. The global Moran’s I score was 0.0 with Z score of 0.98, smaller than that of 1.96 for 95% confidence level. The local Moran’s
Fig. 4. LST distribution of the inner city of Shanghai, China in September 9, 2009.
I scores for the mean LST values of different land use types based on the divided grid polygons were ranged from 0.00 to 0.07, with all the corresponding Z-scores smaller than 1.96. 3.2. Relation of LST with land use and land cover The Pearson correlation analysis showed that the mean LST values were significantly related to both land cover composition and configuration variables (Table 5). However, after controlling the land cover composition percentage and the size of land use polygons, fewer land cover configuration variables were related to LST. For all the three residential land uses, the percent cover of vegetated land had significantly negative relations to LST, with the strongest relationship in old residential land use. By contrast, percent cover of impervious land both in new and old residential land use had significantly positive relation to LST, whereas no relation was found between impervious land cover of villas and LST. The percentage of impervious land cover in old residential land use had stronger relationship with LST than that in new residential land use. Furthermore, the relations between land cover configuration variables and LST were more complex than for land cover composition. In new residential land use, after controlling for the percentage of land cover compositions and the size of land use polygons, none of the vegetated land cover configuration metrics were related to LST, while the PD and ED of impervious land cover were significantly related to LST. Both the PD and ED of impervious land covers had positive relationship with LST. For old residential land use, the PD, ED, PSCOV and LPI of vegetated land cover were significantly related to LST. Among these vegetated cover configuration metrics, the PD, ED and PSCOV had negative effects on LST, while the LPI had a positive effect. By contrast, for the impervious land cover configuration metrics of old residential land use, only the AWMIS had a Table 4 Description of significance on mean difference in LST values of different land use types. “Values” in the table are the difference in LST between two different land used in the according rows and columns. “Values” with * indicate pairings where the mean difference exceeded critical value at ˛ = 0.01, and the others have no significant difference in pairings. LST
Old residential
Villas
Industrial
Institutional
New residential Old residential Villas Industrial
−.99*
0.61* 1.6*
−1.00* −.001 −1.61*
−.77* 0.21 −1.38* 0.23
176
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Table 5 Partial correlation between LST and land cover composition and configuration metrics of the five land use types. New residential Tla Vegetated land cover PLAND PD MPS ED PSSD AWMSI LPI Impervious land cover PLAND PD MPS ED PSSD AWMSI LPI * **
Old residential
Villas
−.241** −.295**
Industrial
Institutional
.278** −.511** −.313*
−.468**
−.322**
−.368** −.228** .179* −.244**
−.284** −.254**
.430**
.323**
−.217*
.357** .373** .209**
.452** −.595** .268** −.619**
.182** −.200*
.178* .198*
.150* −.164*
.193*
−.217**
Correlation is significant at the 0.05 level (two-tailed). Correlation is significant at the 0.01 level (two-tailed).
significantly negative relation to LST. In villas, none of the vegetated land cover configuration metrics was related to LST, whereas the PD, MPS, ED, AWMSI and LPI of the impervious land cover had significant relationship with LST. The PD and ED of the impervious land cover were negatively related to LST, but the MPS, AWMSI and LPI had positive relationships with LST. In industrial land use, both the mean size of the land use polygons and the percent impervious land cover were positively related to LST, while the percent cover of vegetated land was negatively related to LST. The PD and ED of vegetated land cover had significantly negative effects on LST, while the MPS of vegetated cover had positive effects. Similarly, the MPS of impervious land cover had significantly positive effects on LST, while the ED had negative effects. For institutional land use, both the percentage of vegetated and impervious land cover had significant relations with LST. The percentage of vegetated land cover was negatively related to LST, while the percent cover of impervious land was positively related to LST. The PD, ED and PSCOV of vegetated land cover had significantly negative relations with LST. The ED of impervious land cover was positively related to LST, while the LPI of impervious land cover had a negative effect on LST. 4. Discussion 4.1. Different impacts of land use and land cover on LST Table 4 shows that the mean values of LST vary among different land use types, suggesting that the factors affecting LST of each land use are different. Notably, for the land use types with similar land cover composition and configuration, such as new residential and industrial land uses, the mean LST values are significantly different. In contrast, for old residential, industrial and institutional land use, with significantly different land cover composition and configuration, the mean LST values are not statistically different. These results indicate that urban LST distribution is not consistent with land cover composition and configuration at the land use level, which might be the reason for contradictory results between LST and land cover patterns in previous studies (Zhou et al., 2007; Buyantuyev and Wu, 2010; Li et al., 2013; Connor et al., 2013). Table 5 also shows that the partial correlations between LST and land cover variables of each land use type are significant, which is consistent with the previous findings (Buyantuyev and Wu, 2010; Zhou et al., 2011; Li et al., 2011, 2013). Land cover composition has a stronger correlation with LST than land cover configuration,
which was also reported by the previous studies (Zhou et al., 2011; Li et al., 2011; Connor et al., 2013). However, the correlation coefficient of the land cover composition and configuration metrics among different land use types varies. For example, the percent cover of vegetated land has negative effects on LST for all the five land use types, but that for the old residential land use has the strongest negative relation with LST. This finding implies that the smaller green space composition a given land use polygon contains, the more sensitive to LST it is. Interestingly, in villas’ land use, no correlation was found between LST and percent cover of impervious land, as the percentage of impervious land cover is quite small compared with vegetated land cover, and the impact of impervious land on LST is lower compared with vegetated land cover. By contrast, the spatial arrangement of impervious land cover of villas is significantly related to LST, for example increasing PD and ED of impervious land might decrease LST, and decreasing MPS, AWMSI and LPI would reduce LST. These findings further suggest the complex mechanism of urban LST at very fine scales, which may derive not only from biophysical processes, but also from anthropogenic factors (Peng et al., 2012). At fine scale, the LST variation among different land use categories suggests that energy use and anthropogenic heat emission have important impacts on LST. Taha (1997), followed by Smith et al. (2009) and Sailor (2011), demonstrated that the anthropogenic heat flux sources such as human metabolism, buildings and traffic significantly contribute to the urban thermal environment. The heat intensity varies with climate, population density, and intensity of industrial and commercial activities. For example, on the LST distribution map (Fig. 4), locations with high LST values are mostly associated with the industrial and old residential land uses. The ANOVA analysis also shows that there is no significant difference in LST among old residential, industrial, and institutional land uses, although their land cover composition and configuration are significantly different. As Shanghai is the largest industrial city of China, it has enormous industries including steel, petrochemical, mechanical, textile, and food production, etc. (Yang and Yin, 2000). These industries consume a large amount of energy and release a great deal of sensible and latent heat (Sailor, 2011), which possibly leads to the industrial land use category having the highest LST, although the green space coverage is not the lowest among the studied land use types. Moreover, the green space percentage of the institutional category (38.81%) was significantly larger than that of the industrial one (27.64%), but there was no statistical difference in LST between them. This could be due to the higher anthropogenic heat emission from institutional than industrial land
W. Li et al. / Ecological Indicators 47 (2014) 171–178
use. Institutional land use is mainly composed of schools, colleges, universities and research laboratories that contain dense population and activities. The anthropogenic heat flux sources involve human metabolism, both indoor and outdoor, building energy use (such as lighting, heating, ventilation, air conditioning), and other specific activities. These all contribute to the relatively high LST. By contrast, the relatively high LST of old residential land use should be explained by its much lower percent cover of green space (16.79%) than the other land uses, suggesting the important role of vegetation in regulating LST under a given level of anthropogenic activities (Hu and Jia, 2010). The LST of new residential land use significantly differs from that of industrial, although their vegetated/impervious land cover compositions are similar. This indicates that the anthropogenic heat emission of new residential land use, mainly including human metabolism and residential building energy use, is much smaller than that of industrial land use (Sailor, 2011). The different anthropogenic activities explain the varied LST among different land use types and indicate that land cover can only explain part of LST variation. In other words, the relations between LST and land cover alone can be misleading. It is essential to identify the most important anthropogenic factors related to surface energy exchange.
4.2. Implications for urban planning and management The findings of the study imply that compared to land cover, land use has the potential to better explain the impacts of complex anthropogenic activities on urban ecosystems at fine scales. Land use category can reveal the linkage between the biophysical and socio-economic features of an urban ecosystem. The finding that land cover and land use have discrepant impacts on urban LST can not only provide an insight to better understand the mechanisms of urban heat islands, but also provide a practical guidance for urban land planning and management. On the coarse scale, city development needs a more rational and long-term master plan of land use to explicitly state the goal of urban land utilization according to local and regional development objectives. In recent years, the green space coverage of Shanghai has continuously increased, and many industries have been moved out of the downtown area under the guidance of the master plan for land utilization. This will help to reduce the detrimental impacts of high pollution and high emissions from industries to the environment. At present, new residential land use accounts for 22.44% of total area and is scattered in the inner city of Shanghai. Due to the negative relationship between the patch size of the new residential land use polygons and the LST, it is suggested that new residential land use patches developed in the future should be intensively built, which would be beneficial to mitigate the impacts of new residential land use on LST control on the whole city scale. On fine scale, it is suggested that a detailed land use development plan be established to direct surface landscape design by considering both the biophysical and socio-economic features within each land use polygon. LST may be mitigated by optimizing the land cover patterns for each land use type based on the relations of LST to land cover along with the specific human activity. For instance, in old residential area with a small amount of green space, the shape index of vegetated cover has strongly negative relation to LST; thus the increase in shape complexity of vegetated covers is recommended to reduce the LST effect. By contrast, for villas with high green space, LST is sensitive to the spatial arrangement of impervious land cover; therefore, it is better to design the buildings in a dispersed pattern. For institutional land use, increasing the shape complexity of vegetated cover and decreasing the shape complexity of impervious cover will benefit LST mitigation.
177
5. Conclusions This study comprehensively analyzes the impacts of land cover and land use category on LST by using the inner city of Shanghai as the case. New residential, old residential, villas, industrial and institutional land use types are the dominant land uses in this case, and each of them has two major land covers (impervious and vegetated land cover). The study reveals that land use and land cover have discrepant impacts on LST. Therefore, the use of land cover composition and configuration as sole indicators of LST might be misleading. Land use is more meaningful than land cover regarding the impact of urbanization on ecosystems at fine scales, as land use can link the biophysical processes to socio-economic activities. Further research needs to distinguish and quantify the contribution of biophysical and anthropogenic forces to urban heat islands, and then establish effective measures.
Acknowledgements The authors are grateful to the foundation support from the Remote Sensing Evaluation Project on Urbanized Region in China of Ministry of Environmental Protection of the People’s Republic of China (No. STSN-12-02). We thank Prof. Weiqi Zhou and Prof. Junxiang Li for their suggestions and helpful comments.
References Angel, S., Parent, J., Civco, D.L., et al., 2011. The dimensions of global urban expansion: estimates and projections for all countries, 2000–2050. Process Plan. 75, 53–107. Anselin, L., 1995. Local indicators of spatial association. Geogr. Anal. 27 (2), 93–115. Ball, S., Liu, M., Hou, L.J., et al., 2009. Pollutants in stormwater runoff in Shanghai China: implications for management of urban runoff pollution. Prog. Nat. Sci. 19, 873–880. Breuste, J., Niemela, J., Snep, R.P.H., 2008. Applying landscape ecological principles in urban environments. Landsc. Ecol. 23, 1139–1142. Buyantuyev, A., Wu, J.G., 2010. Urban heat islands and landscape heterogeneity: linking spatiotemporal variations in surface temperatures to land-cover and socioeconomic patterns. Landsc. Ecol. 25 (1), 17–33. Chen, B., Chen, G.Q., Yang, Z., 2006. Exergy-based resource accounting for China. Ecol. Model. 196, 313–328. Chen, B., Chen, G.Q., Yang, Z.F., Jiangc, M.M., 2007. Ecological footprint accounting for energy and resource in China. Energy Policy 35, 1599–1609. Chen, G.Q., Chen, Z.M., 2010. Carbon emissions and resources use by Chinese economy 2007: a 135-sector inventory and input–output embodiment. Commun. Nonlinear Sci. Numer. Simul. 15, 3647–3732. Chen, H., Chen, G.Q., Ji, X., 2010. Cosmic emergy based ecological systems modelling. Commun. Nonlinear Sci. Numer. Simul. 15 (9), 2672–2700. Cohen, B., 2004. Urban growth in developing countries: a review of current trends and a caution regarding existing forecasts. World Dev. 32 (1), 23–51. Connor, J.P., Galletti, C.S., Chow, W.T.L., 2013. Landscape configuration and urban heat island effects: assessing the relationship between landscape characteristics and land surface temperature in Phoenix, Arizona. Landsc. Ecol. 28 (2), 271–283. Diamond, M.L., Hodge, E., 2007. Urban contaminant dynamics: from source to effect. Environ. Sci. Technol. 41 (11), 3796–3805. Griffiths, P., Hostert, P., Gruebner, O., et al., 2010. Mapping megacity growth with multi-sensor data. Remote Sens. Environ. 114, 426–439. Grimm, N.B., Faeth, S.H., Golubiewski, N.E., et al., 2008. Global change and the ecology of cities. Science 319, 756–760. Grove, J.M., Burch Jr., W.R., 1997. A social ecology approach and applications of urban ecosystem and landscape analyses: a case study of Baltimore, Maryland. Urban Ecosyst. 1, 259–275. Han, M.Y., Shao, L., Li, J.S., Guo, S., Meng, J., Ahmad, B., Hayat, T., Alsaadi, F., Ji, X., Alsaedi, A., Chen, G.Q., 2014. Emergy-based hybrid evaluation for commercial construction engineering: a case study in BDA. Ecol. Indic., http://dx.doi.org/10.1016/j.ecolind.2014.02.020. Hu, Y.H., Jia, G.S., 2010. Influence of land use change on urban heat island derived from multi-sensor data. Int. J. Climatol. 30, 1382–1395. Hubacek, K., Sun, L., 2001. A scenario analysis of China’s land use and land cover change: incorporating biophysical information into input–output modelling. Struct. Change Econ. Dyn. 12, 367–397. Jones, P.D., Groisman, P.Y., Coughlan, M., et al., 1990. Assessment of urbanization effects in time series of surface air temperature over land. Nature 347, 169–172. Lasanta, T., Vicente-Serrano, S.M., 2012. Complex land cover change processes in semiarid Mediterranean regions: an approach using Landsat images in northeast Spain. Remote Sens. Environ. 124, 1–14.
178
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Lazzarini, M., Marpu, P.R., Ghedira, H., 2013. Temperature–land cover interactions: the inversion of urban heat island phenomenon in desert city areas. Remote Sens. Environ. 130, 136–152. Li, H.B., Yu, S., Li, G.L., et al., 2012. Lead contamination and source in Shanghai in the past century using dated sediment cores from urban park lakes. Chemosphere 88, 1161–1169. Li, J.J., Wang, X.R., Wang, X.J., et al., 2009. Remote sensing evaluation of urban heat island and its spatial pattern of the Shanghai metropolitan area, China. Ecol. Complex. 6, 413–420. Li, J.X., Song, C.H., Cao, L., et al., 2011. Impacts of landscape structure on surface urban heat islands: a case study of Shanghai, China. Remote Sens. Environ. 115 (12), 3249–3263. Li, X.M., Zhou, W.Q., Ouyang, Z.Y., 2013. Relationship between land surface temperature and spatial pattern of greenspace: what are the effects of spatial resolution? Landsc. Urban Plan. 114, 1–8. Maimaitiyiming, M., Ghulam, A., Tiyip, T., et al., 2014. Effects of green space spatial pattern on land surface temperature: implications for sustainable urban planning and climate change adaptation. ISPRS J. Photogramm. Remote Sens. 89, 59–66. McGarigal, K., Cushman, S.A., Neel, M.C., Ene, E., 2002. FRAGSTATS: Spatial Pattern Analysis Program for Categorical Maps. Computer Software Program Produced by the Authors at the University of Massachusetts, Amherst, Available from: http://www.umass.edu/landeco/research/fragstats/fragstats.html Moran, P.A.P., 1948. The interpretation of statistical maps. J. R. Stat. Soc. B 10 (2), 243–251. Peng, S.S., Piao, S.L., Ciais, P., et al., 2012. Surface Urban Heat Island Across 419 Global Big Cities. Environ. Sci. Technol. 46 (2), 696–703. Ramachandra, T.V., Aithal, B.H., Sanna, D.D., 2012. Insights to urban dynamics through landscape spatial pattern analysis. Int. J. Appl. Earth Obs. Geoinf. 18, 329–343. Redo, D., Millington, A.C., 2011. A hybrid approach to mapping land-use modification and land-cover transition from MODIS time-series data: a case study from the Bolivian seasonal tropics. Remote Sens. Environ. 115, 353–372. Ridd, M.K., 1995. Exploring a V-I-S (vegetation-impervious surface-soil) model for urban ecosystem analysis through remote sensing: comparative anatomy for cities. Int. J. Remote Sens. 16 (12), 2165–2185. Sailor, D., 2011. A review of methods for estimating anthropogenic heat and moisture emissions in the urban environment. Int. J. Climatol. 31, 189–199.
Schneider, A., 2012. Monitoring land cover change in urban and peri-urban areas using dense time stacks of Landsat satellite data and a data mining approach. Remote Sens. Environ. 124, 689–704. Sexton, J.O., Song, X.P., Huang, C.Q., et al., 2013. Urban growth of the Washington, D.C.–Baltimore, MD metropolitan region from 1984 to 2010 by annual, Landsatbased estimates of impervious cover. Remote Sens. Environ. 129, 42–53. Shanghai Municipal Statistics Bureau, 1982. Shanghai Statistical Year book 1982. China Statistics Press, Beijing. Shanghai Municipal Statistics Bureau, 2010. Shanghai Statistical Year book 2010. China Statistics Press, Beijing. Shanghai Municipal Statistics Bureau, 2011. Shanghai Statistical Year book 2011. China Statistics Press, Beijing. Smith, G., Lindley, S., Levermore, G., 2009. Estimating spatial and temporal patterns of urban anthropogenic heat fluxes for UK cities: the case of Manchester. Theor. Appl. Climatol. 98 (1-2), 19–35. Taha, H., 1997. Urban climates and heat islands: albedo, evapotranspiration, and anthropogenic heat. Energy Build. 25, 99–103. Taubenbock, H., Esch, T., Felbier, A., et al., 2012. Monitoring urbanization in mega cities from space. Remote Sens. Environ. 117, 162–176. Wang, Y.S., Ma, W.C., Wei, T., et al., 2013. A study on carbon emissions in Shanghai 2000–2008, China. Environ. Sci. Policy 27, 151–161. Wiedmann, T.O., Schandl, H., Lenzen, M., Moran, D., Suh, S., West, J., Kanemoto, K., 2013. The material footprint of nations. Proc. Natl. Acad. Sci. U. S. A., http://dx.doi.org/10.1073/pnas.1220362110. Wu, K.Y., Zhang, H., 2012. Land use dynamics, built-up land expansion patterns, and driving forces analysis of the fast-growing Hangzhou metropolitan area, eastern China (1978e2008). Appl. Geogr. 34, 137–145. Yang, W.Z., Yin, W.H., 2000. Spatial distribution of industries in Shanghai city. Res. Plan. 4, 2–5. Zhang, X.Y., Zhong, T.Y., Feng, X.Z., et al., 2009. Estimation of the relationship between vegetation patches and urban land surface temperature with remote sensing. Int. J. Remote Sens. 30 (8), 2105–2118. Zhou, W.Q., Huang, G.L., Cadenasso, M.L., 2011. Does spatial configuration matter? Understanding the effects of land cover pattern on land surface temperature in urban landscapes. Landsc. Urban Plan. 102 (1), 54–63. Zhou, Y.Y., Wen, Q.H., Gurney, K.R., et al., 2012. Estimation of the relationship between remotely sensed anthropogenic heat discharge and building energy use. ISPRS J. Photogramm. Remote Sens. 67, 65–72.
Contents lists available at ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
Discrepant impacts of land use and land cover on urban heat islands: A case study of Shanghai, China Weifeng Li a , Yang Bai b , Qiuwen Chen a,c,∗ , Kate He e , Xiaohua Ji a,d , Chunmeng Han a,d a
RCEES, Chinese Academy of Science, Shuangqing Road 18, Beijing 100085, China Shanghai Academy of Environmental Sciences, Shunghai 200233, China CEER, Nanjing Hydraulic Research Institute, Nanjing 210029, China d China University of Geosciences, Beijing 100083, China e Department of Biological Sciences, Murray State University, Kentucky 4207, USA b c
a r t i c l e
i n f o
Article history: Received 4 November 2013 Received in revised form 31 July 2014 Accepted 1 August 2014 Keywords: Land cover Landscape pattern Anthropogenic activities Land surface temperature Shanghai
a b s t r a c t The most visible aspect of urbanization is that more and more natural landscape is replaced by anthropogenic land cover/land use, which is the driving force of many ecological and environmental consequences such as urban heat islands. However, the difference between land use and land cover and their implications in ecology is often overlooked. The impacts of urban land cover composition and configuration on land surface temperature (LST) have been extensively investigated, but few studies have explored the relation between LST and land use category. This research takes the inner city of Shanghai as a case and comprehensively investigates the discrepant impacts of land use and land cover on LST. Land use and land cover data are derived respectively from aerial photography and high-resolution satellite imagery (ALOS), and the LST is estimated from Landsat TM images. There are five dominant land use types (new residential, old residential, villas, industrial, and institutional land use) and two major land cover types (vegetated and impervious land cover) in the study area. For most land use types, the land cover composition and configuration are varied. By contrast, no statistical difference is observed among old residential, industrial and institutional land uses for LST. The mean LST of new residential and industrial land use is significantly different, although their land cover compositions and configurations are quite similar. These results indicate that the key factors affecting urban LST are not only land cover patterns, but also other anthropogenic forces. Therefore, the explanation of urban LST by land cover alone is inadequate. Especially at fine spatial scales, information on land use is more meaningful than that of land cover to indicate the impacts of urbanization on ecosystems. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction It is well known that urbanization is one of the most powerful and visible anthropogenic forces on Earth (Cohen, 2004; Chen et al., 2010; Angel et al., 2011). The most evident aspect of urbanization is that more and more natural landscape is replaced by anthropogenic land cover/land use, which causes many ecological and environmental problems, such as the phenomenon of urban heat islands (Chen et al., 2007; Wiedmann et al., 2013). Therefore, better understanding of the driving forces of urban land cover/land use change and their impacts on ecosystems is critical for urban
∗ Corresponding author at: RCEES, Chinese Academy of Science, Shuangqing Road 18, Beijing 100085, China. Tel.: +86 10 62849326; fax: +86 10 62849326. E-mail address: [email protected] (Q. Chen). http://dx.doi.org/10.1016/j.ecolind.2014.08.015 1470-160X/© 2014 Elsevier Ltd. All rights reserved.
ecosystem research and sustainable urban planning (Grove, 1997; Breuste et al., 2008). Notably, the difference between land use and land cover and their implications in ecology is often overlooked. Land use generally refers to how people use the land in terms of social-economic functions, whereas land cover defines the physical pattern of land surfaces. Thus, it is important to distinguish between the two surface landscape indicators, which can reveal the linkage between biophysical features such as land cover and anthropogenic features such as social-economic activities (Hubacek and Sun, 2001; Chen et al., 2006; Chen and Chen, 2010; Han et al., 2014). Numerous studies have investigated urban land cover patterns using multi-resolution remote sensing imagery (Griffiths et al., 2010; Redo and Millington, 2011; Taubenbock et al., 2012; Schneider, 2012; Wu and Zhang, 2012; Ramachandra et al., 2012; Lasanta and Vicente-Serrano, 2012; Sexton et al., 2013), but few studies have quantified urban land use distribution at fine scales.
172
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Urban heat islands, as the most evident negative impact of urbanization, have been widely attributed to land cover/land use change (Jones et al., 1990; Grimm et al., 2008). Generally, urban heat islands can be characterized by air and land surface temperature (LST). Because of the limitation of in situ air temperature monitoring across large spatial-temporal scales, remote sensing provides a powerful way to quantify LST. Many studies have examined the relationship between LST and land cover composition and configuration on multiple scales (Zhang et al., 2009; Buyantuyev and Wu, 2010; Zhou et al., 2011; Li et al., 2011, 2013; Connor et al., 2013; Lazzarini et al., 2013). Most studies found the consistent result that urban vegetation could decrease LST, while impervious land cover increases LST. By contrast, the correlations between LST and land cover configuration were found inconsistent or contradictory in many studies (Zhou et al., 2011; Buyantuyev and Wu, 2010; Li et al., 2013; Connor et al., 2013), and were mostly explained by the scale effect (Zhou et al., 2007; Li et al., 2013). However, due to the notable effects of urban socio-economic activities on LST, the use of land cover pattern as the single driving factor for LST might be inadequate. In recent years, some studies have begun to explore the impacts of anthropogenic factors such as population density and night light on LST at the city or multi-city spatial scale over a region (Diamond and Hodge, 2007; Smith et al., 2009; Sailor, 2011; Peng et al., 2012; Zhou et al., 2012). Yet, a comprehensive study on the influence of anthropogenic factors on LST at fine scale is still lacking. This study takes the inner city of Shanghai, the largest metropolis in China, as a case to comprehensively examine the relations of urban land use and land cover with urban heat islands. The specific objectives are to: (1) quantify LST distribution among different land use categories; (2) investigate whether there are discrepant relations of LST with land use and land cover; (3) analyze the driving forces of anthropogenic factors on LST. This study expects to gain a better understanding of urban LST by analyzing the interactions of the biophysical and socio-economic drivers in term of land use and land cover. Results from this study can support urban land developers and managers in taking concrete measures to regulate urban social-economic activities and landscape patterns to mitigate urban heat islands.
2. Material and methods 2.1. Study area Shanghai is one of the largest and most important industrial centers of China. It is located in East China, between latitudes 30◦ 82 30 N and 31◦ 82 70 N, and longitudes 120◦ 85 20 E and 121◦ 84 50 E, surrounded by the Yangtze River estuary to the north, the East Sea to the east, and the Hangzhou Bay to the south. Shanghai belongs to a subtropical monsoon climate. The average total precipitation is 1067 mm per year, and average monthly temperature ranges from 2 to 27 ◦ C. The total area of Shanghai is about 6340.5 km2 and the total population of the city is about 23.0 million (Shanghai Municipal Statistics Bureau, 2011). Shanghai has experienced rapid urbanization since the implementation of the Reform and Open Policy in 1978. The urban area of Shanghai has increased from 149.85 km2 in 1982 to 998.8 km2 in 2010 (Shanghai Municipal Statistics Bureau, 1982, 2010). The study was conducted at the urban core of Shanghai, which is the origin and heart of this mega city (Fig. 1). The inner city is located within the out-ring road of Shanghai, covering 670 km2 , with very intensive and concentrated urban human activities. The landscape pattern is highly heterogeneous and the impacts of various human activities incur many ecological and environmental problems, such
Table 1 Land cover and land use category of the study area. Categories of urban landscape
Land cover
Impervious land cover
Vegetated land cover Bare soil Water New residential Land use
Old residential
Villas
Industrial
Institutional
Description Artificial lands such as roads, roofs and parking lots paved by impenetrable materials such as asphalt, concrete, bricks, etc. Natural and artificial lands paved by penetrable vegetation such as arbor, shrub and grasses Bare soil or sand not covered by vegetation or impervious lands Land covered by water body such as river, ponds and lakes Lands used by families for private residences or dwellings, with the age less than 15∼20 years old, high-rise apartment buildings and some open space Lands used by families for private residences or dwellings, with the age older than 15∼20 years, dense low-rise apartment buildings and less open space Lands used by single or multiple families for private residences or dwellings, with free-standing homes and greater open space Lands for industrial purposes, usually with multi-builds for different industrial activities, such as workspace, factories or warehouses and associated infrastructure Lands for schools, colleges, universities and research institutes, and associated infrastructure
as urban heat islands, water and air pollution (Li et al., 2009, 2012; Ball et al., 2009; Wang et al., 2013). 2.2. Land cover and land use pattern Land cover and land use have quite different meanings in urban ecosystems, as explained above. This study extracted the main land cover and land use types of the site by applying high resolution Advanced Land Observation Satellite images (ALOS imagery) and aerial photography. All the images were registered to a 1:10,000 scale topographic image. We chose land use classification to represent different socioeconomic activities. For land use mapping, the classification system was set up considering both the ecological effects of different urban socio-economic activities and the existing national standard for land use classification (GB/T 21010-2007). Thus, five dominant land use types were identified in the study area, including new residential, old residential, villas, industrial and institutional land use (Table 1, Fig. 2). The visual interpretation was applied to map urban land uses by integrating the aerial photography taken in June 2010 with a spatial resolution of 0.3 m and the cadastre resources (Fig. 3). The visual interpretation ensured that land use patches were properly classified and represented the land use category. For land cover mapping, four major land cover types were identified to characterize the biophysical features of urban landscape, including impervious land, vegetated land, bare land and water body (Ridd, 1995). The ALOS imagery with a spatial resolution of 2.5 m was acquired on May 2010, and the object oriented classification method was applied to classify land covers. The overall accuracy of land cover classification was 85.3%. Specially, the user’s accuracy for vegetation, impervious, bare land and water body were 80.47, 88.37, 74.29 and 78.57%, respectively. The producer’s
W. Li et al. / Ecological Indicators 47 (2014) 171–178
173
Fig. 1. Location of the metropolis of Shanghai, China and the study site inner city.
accuracy for vegetation, impervious, bare land and water body were 83.78, 78.62, 92.86 and 89.19%. We applied the most important and broadly used landscape metrics to measure the land cover composition and configuration characteristics. The percentage of land cover (PLAND) was used to
measure the composition (McGarigal et al., 2002) (Table 2). The significant effects of land cover composition on LST are because the land surface characteristics such as albedo and evapotranspiration of different land cover components are discrepant and they can directly affect land surface temperature (Zhou et al., 2011; Li
Fig. 2. Aerial photography showing the characteristics of dominant urban land use types of Shanghai in June 2010.
174
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Table 2 Landscape metrics selected in this study. Variable
Description
Formulas
Landscape area (TLA)
Sum of areas of all patches of the same land use type polygon in the landscape (unit: ha)
A 104
Percent cover of class area (PLAND)
Percentage of the area of a particular patch type of total landscape area at class level (unit: %)
100 A
Patch density (PD)
Number of all patches at the given class divided by total landscape area at class level (unit: number per ha)
n A
Mean patch size (MPS)
Average patch size at class level (unit: ha)
Edge density (ED)
Amount of edge relative to the landscape area at class level (unit: m/ha)
×
n
× 104
n j=1
aij
ni
10000 A
×
Patch size standard deviation of all the patches at class level (unit: ha)
Area weighted mean shape index (AWMSI)
Average perimeter-to-area ratio for a landscape, weighted by the size of its patches at class level
1 10,000
n
eik
k=1
n
Patch size Standard deviation (PSSD)
ai
i=1
j=1
n
[aij −(
ni
j=1
2
aij /ni )]
n
0.25×p √
aij
ij
anij j=1
j=1
1 10000
aij
maxaij Largest patch index (LPI)
Percent of the total landscape that is made up by the largest patch at class level (unit: %)
j=1,n A
(100)
A: total landscape area (m2 ); aij : area (m2 ) of patch ij; eik : length of edge (m) of patch ik; pij : perimeter (m) of patch ij; n: number of patches.
The land use map was intersected with the land cover map to obtain the land cover types within each land use unit. Land cover composition and configuration of different land use types were calculated based on each land use patch (polygon), and all the landscape analysis was conducted in ArcGIS. 2.3. Land surface temperature
Fig. 3. Spatial distribution of the land use types within the study site in June 2010.
et al., 2011). For land cover configuration, we selected six landscape metrics to measure the size, shape and fragmentation of land cover patches. These land configuration metrics include patch density (PD), mean patch size (MPS), edge density (ED), standard deviation of patch size (PSSD), area weighted mean shape index (AWMSI) and largest patch index (LPI). MPS and PSSD were used to measure patch size of land cover. AWMSI was used to measure patch shape of land cover. LPI, PD and ED were used to measure the fragmentation characteristics of land cover. Table 2 lists the detailed definitions and formulas of these landscape metrics. These metrics have been well examined by previous studies to have significant impacts on LST, due to the fact that spatial configuration of land cover patches could affect the interactions of heat exchange among adjacent land cover patches (Zhou et al., 2011; Li et al., 2011, 2013; Maimaitiyiming et al., 2014).
In this study, the thermal infrared band 6 of the Landsat TM image (11.45–12.50 nm) with a spatial resolution of 120 m was used to estimate LST. Since Shanghai belongs to a subtropical monsoon climate and has a long summer with a lot of rain during the period of June to August, we selected the image acquired on September 9, 2009, a day with clear atmospheric conditions, when the urban heat island phenomenon was obvious. The widely used single-channel method was applied to calculate LST from the thermal band of Landsat data according to the thermal radiant transfer equation (Zhang et al., 2009). The main steps included: (1) correcting the radiometric and geometrical distortions; (2) converting calibrated DNs to absolute units of at-sensor spectral radiance; (3) converting atsensor spectral radiance to at-sensor brightness temperature (i.e., blackbody temperature); and (4) correcting for spectral emissivity of different land covers to generate the LST data. The mean LST value for each land use polygon was calculated by zonal statistical analysis in the ArcGIS platform. Then, the mean LST values and variance for different land use types were further summarized for analysis of the relations of LST and land use. 2.4. Statistical analysis Because the total number of land use polygons was approximately 8500 in the study, including 3400 of new residential, 1100 of old residential, 150 of villas, 2200 of industrial and 1700 of institutional land use, 10% of the land use polygons of each land use type were randomly selected as samples, except for villas, which were small in number. Considering the accuracy of land cover and LST derived for each land use polygon, the qualified samples for
W. Li et al. / Ecological Indicators 47 (2014) 171–178
175
Table 3 Description of LST for samples of the five land use types. LST
Mean (◦ C)
Std. dev. (◦ C)
New residential land use Old residential land use Villas Industrial land use Institutional land use
30.73 31.72 30.12 31.73 31.50
1.22 1.35 2.27 1.66 1.29
different land use types were 310 of new residential, 100 of old residential, 127 of villas, 199 of industrial and 152 of institutional land use. Relations between LST, land cover patterns and land use types were tested using Pearson’s and partial correlation with statistical significance at p < 0.05. Analysis of variance (ANOVA) was used to test the difference in mean LST among different land use types with statistical significance at p < 0.05. The main steps included: (1) A descriptive analysis was conducted to summarize the average features of land cover composition and configuration for different land use type samples. (2) The mean LST values of different land use types were calculated. By this step, the average land cover patterns distributed within different land use types and average LST of each land use type were known. (3) The spatial autocorrelation of LST among neighboring land use types was tested by using global Moran’s I and local indicators of spatial association (LISA) analysis (Moran, 1948; Anselin, 1995). Global Moran’s I values can range from −1 to 1, indicating the correlations with different degrees; while LISA measures the degree of local spatial autocorrelation at different locations by using a localized Moran’s I. In this study, we divided the study area into 148 polygons with 2 km × 2 km grid size, excluding the areas with low quality on LST inversion from satellite image due to cloud covering. For each divided grid polygons, we calculated the mean LST values of different land use types to test the spatial correlation among neighboring land use types. The neighboring land use patches were decided based on adjacency. We used ArcGIS software to carry out the spatial correlation analysis. (4) The relations between LST and land cover patterns for different land use types were analyzed by Pearson’s correlation analysis. Since many land cover configuration variables significantly relate to land cover composition, partial correlation analysis was further conducted by controlling the variable of land cover composition. (5) A one-way analysis of variance (ANOVA) was used to investigate whether the LST and land cover pattern varied among different land use types. 3. Results 3.1. Variation of LST among different land use categories The LST distribution of the inner city of Shanghai is shown in Fig. 4, which ranged from 16.7 to 44.3 ◦ C (Table 3). Among the five dominant urban land uses, industrial land use had the highest mean LST value of 31.73 ◦ C, followed by old residential (31.72 ◦ C), institutional (31.5 ◦ C), new residential land use (30.73 ◦ C), and villas (30.12 ◦ C). Interestingly, there was no statistical difference in LST between old residential, industrial and institutional land use, but mean LST values of other land uses were significantly different from each other (Table 4). Both global and local spatial autocorrelation indices were nonsignificant. The global Moran’s I score was 0.0 with Z score of 0.98, smaller than that of 1.96 for 95% confidence level. The local Moran’s
Fig. 4. LST distribution of the inner city of Shanghai, China in September 9, 2009.
I scores for the mean LST values of different land use types based on the divided grid polygons were ranged from 0.00 to 0.07, with all the corresponding Z-scores smaller than 1.96. 3.2. Relation of LST with land use and land cover The Pearson correlation analysis showed that the mean LST values were significantly related to both land cover composition and configuration variables (Table 5). However, after controlling the land cover composition percentage and the size of land use polygons, fewer land cover configuration variables were related to LST. For all the three residential land uses, the percent cover of vegetated land had significantly negative relations to LST, with the strongest relationship in old residential land use. By contrast, percent cover of impervious land both in new and old residential land use had significantly positive relation to LST, whereas no relation was found between impervious land cover of villas and LST. The percentage of impervious land cover in old residential land use had stronger relationship with LST than that in new residential land use. Furthermore, the relations between land cover configuration variables and LST were more complex than for land cover composition. In new residential land use, after controlling for the percentage of land cover compositions and the size of land use polygons, none of the vegetated land cover configuration metrics were related to LST, while the PD and ED of impervious land cover were significantly related to LST. Both the PD and ED of impervious land covers had positive relationship with LST. For old residential land use, the PD, ED, PSCOV and LPI of vegetated land cover were significantly related to LST. Among these vegetated cover configuration metrics, the PD, ED and PSCOV had negative effects on LST, while the LPI had a positive effect. By contrast, for the impervious land cover configuration metrics of old residential land use, only the AWMIS had a Table 4 Description of significance on mean difference in LST values of different land use types. “Values” in the table are the difference in LST between two different land used in the according rows and columns. “Values” with * indicate pairings where the mean difference exceeded critical value at ˛ = 0.01, and the others have no significant difference in pairings. LST
Old residential
Villas
Industrial
Institutional
New residential Old residential Villas Industrial
−.99*
0.61* 1.6*
−1.00* −.001 −1.61*
−.77* 0.21 −1.38* 0.23
176
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Table 5 Partial correlation between LST and land cover composition and configuration metrics of the five land use types. New residential Tla Vegetated land cover PLAND PD MPS ED PSSD AWMSI LPI Impervious land cover PLAND PD MPS ED PSSD AWMSI LPI * **
Old residential
Villas
−.241** −.295**
Industrial
Institutional
.278** −.511** −.313*
−.468**
−.322**
−.368** −.228** .179* −.244**
−.284** −.254**
.430**
.323**
−.217*
.357** .373** .209**
.452** −.595** .268** −.619**
.182** −.200*
.178* .198*
.150* −.164*
.193*
−.217**
Correlation is significant at the 0.05 level (two-tailed). Correlation is significant at the 0.01 level (two-tailed).
significantly negative relation to LST. In villas, none of the vegetated land cover configuration metrics was related to LST, whereas the PD, MPS, ED, AWMSI and LPI of the impervious land cover had significant relationship with LST. The PD and ED of the impervious land cover were negatively related to LST, but the MPS, AWMSI and LPI had positive relationships with LST. In industrial land use, both the mean size of the land use polygons and the percent impervious land cover were positively related to LST, while the percent cover of vegetated land was negatively related to LST. The PD and ED of vegetated land cover had significantly negative effects on LST, while the MPS of vegetated cover had positive effects. Similarly, the MPS of impervious land cover had significantly positive effects on LST, while the ED had negative effects. For institutional land use, both the percentage of vegetated and impervious land cover had significant relations with LST. The percentage of vegetated land cover was negatively related to LST, while the percent cover of impervious land was positively related to LST. The PD, ED and PSCOV of vegetated land cover had significantly negative relations with LST. The ED of impervious land cover was positively related to LST, while the LPI of impervious land cover had a negative effect on LST. 4. Discussion 4.1. Different impacts of land use and land cover on LST Table 4 shows that the mean values of LST vary among different land use types, suggesting that the factors affecting LST of each land use are different. Notably, for the land use types with similar land cover composition and configuration, such as new residential and industrial land uses, the mean LST values are significantly different. In contrast, for old residential, industrial and institutional land use, with significantly different land cover composition and configuration, the mean LST values are not statistically different. These results indicate that urban LST distribution is not consistent with land cover composition and configuration at the land use level, which might be the reason for contradictory results between LST and land cover patterns in previous studies (Zhou et al., 2007; Buyantuyev and Wu, 2010; Li et al., 2013; Connor et al., 2013). Table 5 also shows that the partial correlations between LST and land cover variables of each land use type are significant, which is consistent with the previous findings (Buyantuyev and Wu, 2010; Zhou et al., 2011; Li et al., 2011, 2013). Land cover composition has a stronger correlation with LST than land cover configuration,
which was also reported by the previous studies (Zhou et al., 2011; Li et al., 2011; Connor et al., 2013). However, the correlation coefficient of the land cover composition and configuration metrics among different land use types varies. For example, the percent cover of vegetated land has negative effects on LST for all the five land use types, but that for the old residential land use has the strongest negative relation with LST. This finding implies that the smaller green space composition a given land use polygon contains, the more sensitive to LST it is. Interestingly, in villas’ land use, no correlation was found between LST and percent cover of impervious land, as the percentage of impervious land cover is quite small compared with vegetated land cover, and the impact of impervious land on LST is lower compared with vegetated land cover. By contrast, the spatial arrangement of impervious land cover of villas is significantly related to LST, for example increasing PD and ED of impervious land might decrease LST, and decreasing MPS, AWMSI and LPI would reduce LST. These findings further suggest the complex mechanism of urban LST at very fine scales, which may derive not only from biophysical processes, but also from anthropogenic factors (Peng et al., 2012). At fine scale, the LST variation among different land use categories suggests that energy use and anthropogenic heat emission have important impacts on LST. Taha (1997), followed by Smith et al. (2009) and Sailor (2011), demonstrated that the anthropogenic heat flux sources such as human metabolism, buildings and traffic significantly contribute to the urban thermal environment. The heat intensity varies with climate, population density, and intensity of industrial and commercial activities. For example, on the LST distribution map (Fig. 4), locations with high LST values are mostly associated with the industrial and old residential land uses. The ANOVA analysis also shows that there is no significant difference in LST among old residential, industrial, and institutional land uses, although their land cover composition and configuration are significantly different. As Shanghai is the largest industrial city of China, it has enormous industries including steel, petrochemical, mechanical, textile, and food production, etc. (Yang and Yin, 2000). These industries consume a large amount of energy and release a great deal of sensible and latent heat (Sailor, 2011), which possibly leads to the industrial land use category having the highest LST, although the green space coverage is not the lowest among the studied land use types. Moreover, the green space percentage of the institutional category (38.81%) was significantly larger than that of the industrial one (27.64%), but there was no statistical difference in LST between them. This could be due to the higher anthropogenic heat emission from institutional than industrial land
W. Li et al. / Ecological Indicators 47 (2014) 171–178
use. Institutional land use is mainly composed of schools, colleges, universities and research laboratories that contain dense population and activities. The anthropogenic heat flux sources involve human metabolism, both indoor and outdoor, building energy use (such as lighting, heating, ventilation, air conditioning), and other specific activities. These all contribute to the relatively high LST. By contrast, the relatively high LST of old residential land use should be explained by its much lower percent cover of green space (16.79%) than the other land uses, suggesting the important role of vegetation in regulating LST under a given level of anthropogenic activities (Hu and Jia, 2010). The LST of new residential land use significantly differs from that of industrial, although their vegetated/impervious land cover compositions are similar. This indicates that the anthropogenic heat emission of new residential land use, mainly including human metabolism and residential building energy use, is much smaller than that of industrial land use (Sailor, 2011). The different anthropogenic activities explain the varied LST among different land use types and indicate that land cover can only explain part of LST variation. In other words, the relations between LST and land cover alone can be misleading. It is essential to identify the most important anthropogenic factors related to surface energy exchange.
4.2. Implications for urban planning and management The findings of the study imply that compared to land cover, land use has the potential to better explain the impacts of complex anthropogenic activities on urban ecosystems at fine scales. Land use category can reveal the linkage between the biophysical and socio-economic features of an urban ecosystem. The finding that land cover and land use have discrepant impacts on urban LST can not only provide an insight to better understand the mechanisms of urban heat islands, but also provide a practical guidance for urban land planning and management. On the coarse scale, city development needs a more rational and long-term master plan of land use to explicitly state the goal of urban land utilization according to local and regional development objectives. In recent years, the green space coverage of Shanghai has continuously increased, and many industries have been moved out of the downtown area under the guidance of the master plan for land utilization. This will help to reduce the detrimental impacts of high pollution and high emissions from industries to the environment. At present, new residential land use accounts for 22.44% of total area and is scattered in the inner city of Shanghai. Due to the negative relationship between the patch size of the new residential land use polygons and the LST, it is suggested that new residential land use patches developed in the future should be intensively built, which would be beneficial to mitigate the impacts of new residential land use on LST control on the whole city scale. On fine scale, it is suggested that a detailed land use development plan be established to direct surface landscape design by considering both the biophysical and socio-economic features within each land use polygon. LST may be mitigated by optimizing the land cover patterns for each land use type based on the relations of LST to land cover along with the specific human activity. For instance, in old residential area with a small amount of green space, the shape index of vegetated cover has strongly negative relation to LST; thus the increase in shape complexity of vegetated covers is recommended to reduce the LST effect. By contrast, for villas with high green space, LST is sensitive to the spatial arrangement of impervious land cover; therefore, it is better to design the buildings in a dispersed pattern. For institutional land use, increasing the shape complexity of vegetated cover and decreasing the shape complexity of impervious cover will benefit LST mitigation.
177
5. Conclusions This study comprehensively analyzes the impacts of land cover and land use category on LST by using the inner city of Shanghai as the case. New residential, old residential, villas, industrial and institutional land use types are the dominant land uses in this case, and each of them has two major land covers (impervious and vegetated land cover). The study reveals that land use and land cover have discrepant impacts on LST. Therefore, the use of land cover composition and configuration as sole indicators of LST might be misleading. Land use is more meaningful than land cover regarding the impact of urbanization on ecosystems at fine scales, as land use can link the biophysical processes to socio-economic activities. Further research needs to distinguish and quantify the contribution of biophysical and anthropogenic forces to urban heat islands, and then establish effective measures.
Acknowledgements The authors are grateful to the foundation support from the Remote Sensing Evaluation Project on Urbanized Region in China of Ministry of Environmental Protection of the People’s Republic of China (No. STSN-12-02). We thank Prof. Weiqi Zhou and Prof. Junxiang Li for their suggestions and helpful comments.
References Angel, S., Parent, J., Civco, D.L., et al., 2011. The dimensions of global urban expansion: estimates and projections for all countries, 2000–2050. Process Plan. 75, 53–107. Anselin, L., 1995. Local indicators of spatial association. Geogr. Anal. 27 (2), 93–115. Ball, S., Liu, M., Hou, L.J., et al., 2009. Pollutants in stormwater runoff in Shanghai China: implications for management of urban runoff pollution. Prog. Nat. Sci. 19, 873–880. Breuste, J., Niemela, J., Snep, R.P.H., 2008. Applying landscape ecological principles in urban environments. Landsc. Ecol. 23, 1139–1142. Buyantuyev, A., Wu, J.G., 2010. Urban heat islands and landscape heterogeneity: linking spatiotemporal variations in surface temperatures to land-cover and socioeconomic patterns. Landsc. Ecol. 25 (1), 17–33. Chen, B., Chen, G.Q., Yang, Z., 2006. Exergy-based resource accounting for China. Ecol. Model. 196, 313–328. Chen, B., Chen, G.Q., Yang, Z.F., Jiangc, M.M., 2007. Ecological footprint accounting for energy and resource in China. Energy Policy 35, 1599–1609. Chen, G.Q., Chen, Z.M., 2010. Carbon emissions and resources use by Chinese economy 2007: a 135-sector inventory and input–output embodiment. Commun. Nonlinear Sci. Numer. Simul. 15, 3647–3732. Chen, H., Chen, G.Q., Ji, X., 2010. Cosmic emergy based ecological systems modelling. Commun. Nonlinear Sci. Numer. Simul. 15 (9), 2672–2700. Cohen, B., 2004. Urban growth in developing countries: a review of current trends and a caution regarding existing forecasts. World Dev. 32 (1), 23–51. Connor, J.P., Galletti, C.S., Chow, W.T.L., 2013. Landscape configuration and urban heat island effects: assessing the relationship between landscape characteristics and land surface temperature in Phoenix, Arizona. Landsc. Ecol. 28 (2), 271–283. Diamond, M.L., Hodge, E., 2007. Urban contaminant dynamics: from source to effect. Environ. Sci. Technol. 41 (11), 3796–3805. Griffiths, P., Hostert, P., Gruebner, O., et al., 2010. Mapping megacity growth with multi-sensor data. Remote Sens. Environ. 114, 426–439. Grimm, N.B., Faeth, S.H., Golubiewski, N.E., et al., 2008. Global change and the ecology of cities. Science 319, 756–760. Grove, J.M., Burch Jr., W.R., 1997. A social ecology approach and applications of urban ecosystem and landscape analyses: a case study of Baltimore, Maryland. Urban Ecosyst. 1, 259–275. Han, M.Y., Shao, L., Li, J.S., Guo, S., Meng, J., Ahmad, B., Hayat, T., Alsaadi, F., Ji, X., Alsaedi, A., Chen, G.Q., 2014. Emergy-based hybrid evaluation for commercial construction engineering: a case study in BDA. Ecol. Indic., http://dx.doi.org/10.1016/j.ecolind.2014.02.020. Hu, Y.H., Jia, G.S., 2010. Influence of land use change on urban heat island derived from multi-sensor data. Int. J. Climatol. 30, 1382–1395. Hubacek, K., Sun, L., 2001. A scenario analysis of China’s land use and land cover change: incorporating biophysical information into input–output modelling. Struct. Change Econ. Dyn. 12, 367–397. Jones, P.D., Groisman, P.Y., Coughlan, M., et al., 1990. Assessment of urbanization effects in time series of surface air temperature over land. Nature 347, 169–172. Lasanta, T., Vicente-Serrano, S.M., 2012. Complex land cover change processes in semiarid Mediterranean regions: an approach using Landsat images in northeast Spain. Remote Sens. Environ. 124, 1–14.
178
W. Li et al. / Ecological Indicators 47 (2014) 171–178
Lazzarini, M., Marpu, P.R., Ghedira, H., 2013. Temperature–land cover interactions: the inversion of urban heat island phenomenon in desert city areas. Remote Sens. Environ. 130, 136–152. Li, H.B., Yu, S., Li, G.L., et al., 2012. Lead contamination and source in Shanghai in the past century using dated sediment cores from urban park lakes. Chemosphere 88, 1161–1169. Li, J.J., Wang, X.R., Wang, X.J., et al., 2009. Remote sensing evaluation of urban heat island and its spatial pattern of the Shanghai metropolitan area, China. Ecol. Complex. 6, 413–420. Li, J.X., Song, C.H., Cao, L., et al., 2011. Impacts of landscape structure on surface urban heat islands: a case study of Shanghai, China. Remote Sens. Environ. 115 (12), 3249–3263. Li, X.M., Zhou, W.Q., Ouyang, Z.Y., 2013. Relationship between land surface temperature and spatial pattern of greenspace: what are the effects of spatial resolution? Landsc. Urban Plan. 114, 1–8. Maimaitiyiming, M., Ghulam, A., Tiyip, T., et al., 2014. Effects of green space spatial pattern on land surface temperature: implications for sustainable urban planning and climate change adaptation. ISPRS J. Photogramm. Remote Sens. 89, 59–66. McGarigal, K., Cushman, S.A., Neel, M.C., Ene, E., 2002. FRAGSTATS: Spatial Pattern Analysis Program for Categorical Maps. Computer Software Program Produced by the Authors at the University of Massachusetts, Amherst, Available from: http://www.umass.edu/landeco/research/fragstats/fragstats.html Moran, P.A.P., 1948. The interpretation of statistical maps. J. R. Stat. Soc. B 10 (2), 243–251. Peng, S.S., Piao, S.L., Ciais, P., et al., 2012. Surface Urban Heat Island Across 419 Global Big Cities. Environ. Sci. Technol. 46 (2), 696–703. Ramachandra, T.V., Aithal, B.H., Sanna, D.D., 2012. Insights to urban dynamics through landscape spatial pattern analysis. Int. J. Appl. Earth Obs. Geoinf. 18, 329–343. Redo, D., Millington, A.C., 2011. A hybrid approach to mapping land-use modification and land-cover transition from MODIS time-series data: a case study from the Bolivian seasonal tropics. Remote Sens. Environ. 115, 353–372. Ridd, M.K., 1995. Exploring a V-I-S (vegetation-impervious surface-soil) model for urban ecosystem analysis through remote sensing: comparative anatomy for cities. Int. J. Remote Sens. 16 (12), 2165–2185. Sailor, D., 2011. A review of methods for estimating anthropogenic heat and moisture emissions in the urban environment. Int. J. Climatol. 31, 189–199.
Schneider, A., 2012. Monitoring land cover change in urban and peri-urban areas using dense time stacks of Landsat satellite data and a data mining approach. Remote Sens. Environ. 124, 689–704. Sexton, J.O., Song, X.P., Huang, C.Q., et al., 2013. Urban growth of the Washington, D.C.–Baltimore, MD metropolitan region from 1984 to 2010 by annual, Landsatbased estimates of impervious cover. Remote Sens. Environ. 129, 42–53. Shanghai Municipal Statistics Bureau, 1982. Shanghai Statistical Year book 1982. China Statistics Press, Beijing. Shanghai Municipal Statistics Bureau, 2010. Shanghai Statistical Year book 2010. China Statistics Press, Beijing. Shanghai Municipal Statistics Bureau, 2011. Shanghai Statistical Year book 2011. China Statistics Press, Beijing. Smith, G., Lindley, S., Levermore, G., 2009. Estimating spatial and temporal patterns of urban anthropogenic heat fluxes for UK cities: the case of Manchester. Theor. Appl. Climatol. 98 (1-2), 19–35. Taha, H., 1997. Urban climates and heat islands: albedo, evapotranspiration, and anthropogenic heat. Energy Build. 25, 99–103. Taubenbock, H., Esch, T., Felbier, A., et al., 2012. Monitoring urbanization in mega cities from space. Remote Sens. Environ. 117, 162–176. Wang, Y.S., Ma, W.C., Wei, T., et al., 2013. A study on carbon emissions in Shanghai 2000–2008, China. Environ. Sci. Policy 27, 151–161. Wiedmann, T.O., Schandl, H., Lenzen, M., Moran, D., Suh, S., West, J., Kanemoto, K., 2013. The material footprint of nations. Proc. Natl. Acad. Sci. U. S. A., http://dx.doi.org/10.1073/pnas.1220362110. Wu, K.Y., Zhang, H., 2012. Land use dynamics, built-up land expansion patterns, and driving forces analysis of the fast-growing Hangzhou metropolitan area, eastern China (1978e2008). Appl. Geogr. 34, 137–145. Yang, W.Z., Yin, W.H., 2000. Spatial distribution of industries in Shanghai city. Res. Plan. 4, 2–5. Zhang, X.Y., Zhong, T.Y., Feng, X.Z., et al., 2009. Estimation of the relationship between vegetation patches and urban land surface temperature with remote sensing. Int. J. Remote Sens. 30 (8), 2105–2118. Zhou, W.Q., Huang, G.L., Cadenasso, M.L., 2011. Does spatial configuration matter? Understanding the effects of land cover pattern on land surface temperature in urban landscapes. Landsc. Urban Plan. 102 (1), 54–63. Zhou, Y.Y., Wen, Q.H., Gurney, K.R., et al., 2012. Estimation of the relationship between remotely sensed anthropogenic heat discharge and building energy use. ISPRS J. Photogramm. Remote Sens. 67, 65–72.