Environ Sci Pollut Res (2015) 22:16997–17011 DOI 10.1007/s11356-015-4795-x
RESEARCH ARTICLE
Water quality changes in response to urban expansion: spatially varying relations and determinants Wenjun Zhao 1 & Xiaodong Zhu 1 & Xiang Sun 1,2 & Yunqiao Shu 3 & Yangfan Li 1,4
Received: 22 January 2015 / Accepted: 27 May 2015 / Published online: 28 June 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract Urban expansion is an important stressor to water bodies, and the spatial variations of their relations are increasingly highlighted by recent studies. What remain unclear, however, are the underlying drivers to the spatial variability. The paper was not limited to modeling spatially varying linkages but also drew attention to the local anthropogenic influential factors that shape land-water relations. We employed geographically weighted regression to examine the relationships between urban expansion (measured by land use change intensity) and water quality changes (focusing on six water quality indicators) in a recently fast-growing Chinese city,
Responsible editor: Philippe Garrigues * Xiaodong Zhu [email protected]; [email protected] * Xiang Sun [email protected] Wenjun Zhao [email protected] Yunqiao Shu [email protected] Yangfan Li [email protected] 1
State Key Laboratory of Pollution Control and Resources Reuse, School of the Environment, Nanjing University, Nanjing 210023, People’s Republic of China
2
School of Environment, Guangxi University, Nanning 530004, People’s Republic of China
3
International Water Management Institute (IWMI) Southern Africa, Pretoria, South Africa
4
The Key Laboratory of the Coastal Zone Exploitation and Protection, Ministry of Land and Resource, Nanjing 210008, People’s Republic of China
Lianyungang. Specifically, we analyzed how the local characteristics including urbanization level, environmental management, industrial zone expansion, and land use composition, attributed to the varying responses of water quality changes. Results showed that urbanization level significantly affects land-water linkages. Remarkable water quality improvement was accompanied by urbanization in highly developed watersheds, primarily due to strong influence from extensive water management practices (particularly for COD, BOD, NH3-N, and TP). By contrast, water qualities of less-urbanized watersheds were more sensitive and negatively responsive to land use changes. Clustering industrial activities acted as distinct contributor to Hg contamination, while boosted organic pollution control in highly urbanized areas. The approach proposed in the study can locate and further zoom into the hotspots of human-water interactions, thereby contributing to better solutions for mitigating undesirable impacts of urbanization on water environment. Keywords River pollution . Land-water linkages . Geographically weighted regression . Anthropogenic factors . Urbanization level . China
Introduction Stream degradation is a key global issue in urbanizing areas (Defries and Eshleman 2004; Finkenbine et al. 2001; Hughes et al. 2014; Paul and Meyer 2001). Urban land expansion contributes to water pollution by increasing pollutant loads from both point and non-point sources (e.g., Atasoy et al. 2006; Dong et al. 2014; Halstead et al. 2014; Mehaffey et al. 2005; He et al. 2008). Exploring land-water relations in urban setting is essential to evaluate current and potential environmental consequences of human activities.
16998
Rivers flow through watersheds with varying natural and anthropogenic backgrounds, thus land-water linkages are inherently inconsistent. The spatial variations of land-water relations are demonstrated by water quality features (Carey et al. 2011; Li et al. 2009), watershed characteristics (e.g., land use composition, topography) (Huang et al. 2015; Su et al. 2013) and different analysis scales (e.g., sub-watersheds, riparian, or buffer zones) (Pratt and Chang 2012; Sliva and Williams 2001; Zhao et al., 2015). Conventional regression approach using global statistics may mask the spatial heterogeneity. Recent studies addressed the problem of generalization by employing a local regression statistics method named geographically weighted regression (GWR) (de Freitas et al. 2013; Tu 2011a; Yu et al. 2013). Most of the above-mentioned studies focused on exploring the most predictive models, but the anthropogenic driving forces behind the varying land-water linkages are less covered. Land use changes during urbanization did not always impact water quality directly but largely exerted the impacts through multiple influential factors (Teixeira et al. 2014). For example, urbanization pattern (Alberti et al. 2007; Carle et al. 2005; Hatt et al. 2004; McMahon and Cuffney 2000); landscape characteristics (Fu et al. 2005; Guo et al. 2010; Lee et al. 2009; Murray et al. 2010) and socio-economic factors (Huang et al. 2014; Juma et al. 2014; Pfeifer and Bennett 2011; Zhou et al. 2012) influence water pollutant features and runoff processes. These factors intervene and interact as part of the systematic processes, which may enhance, reduce, or even obscure the land use effects on water bodies (Jung et al. 2008; Mouri et al. 2012; Sun et al. 2014; Tu 2011b; Wang et al. 2013). As a result, it is critical to distinguish the influences from land use and possible stressors in understanding landwater relationships. The study therefore aims at tackling the above concern by including relevant influential factors in the analysis framework. Our research objective was not limited to investigating the spatially varying relations between urban expansion and water quality but further identifying the determinants to these variances. We conducted the study in Lianyungang, a rapidly urbanizing and expansive city during rapid urbanization period (2000– 2008). Specifically: & &
We firstly revealed the spatial variations of water quality changes in response to urban expansion, by modeling their relationships using GWR technique. Secondly, we analyzed how the anthropogenic factors contributed to the spatially varying responses of water quality changes. We analyzed whether urbanization level, industrial development, environmental management, and land use composition have significant impacts on the varying linkages using statistical tools.
Environ Sci Pollut Res (2015) 22:16997–17011
Study area Lianyungang is a recently fast-growing coastal city located on the western shore of the Yellow Sea (Fig. 1), which is facing multiple challenges from urbanization, industrial development, and environmental management (Table 1). The region is characterized by booming urbanization since 2000 with average annual growth of 9.8 % (while 0.5 % over the prior 20 years) and continuous growing industrial output value (Fig. 2). National and local development strategies accelerated urban spatial growth and stimulated the area as a rising center of clustering industrial zones. By contrast, the capacity of sewage collection and treatment grew slowly, and the first centralized sewage treatment plant was not installed until 2005. Lianyungang has implemented environmental management measures such as “total emission control policy” and “Eco-city program,” which are expected to improve environmental quality while facilitating urban and industrial use. However, rapidly degrading water quality is still a major concern as most of the rivers have been polluted. The study area (red boundary in Fig. 1) is the southeastern part of Lianyungang, covering three urban districts (Xinpu, Lianyun, Haizhou) and two counties (Guanyun, Guannan) with an area of 4053 km2. Urban districts are highly urbanized with mostly urban and suburban area while counties are predominantly agricultural. Despite of the availability of satellite images, this region is an adequate case study because it consists of urban centers as well as recently developing areas with gradients of anthropogenic influences.
Data and methods Land use data Landsat Thematic Mapper (TM) images from February 2000 and February 2008 were used to create land use classification (Fig. 3). We used two-step expert classification method (Sun et al. 2012) and supplementary material including Digital Elevation Model (DEM) data, vector coverages (e.g., industrial zones, salt marsh) to interpret land use information. The images were finally classified into six categories: (1) agricultural, (2) forest, (3) built-up, (4) industrial zone, (5) water bodies (other than salt marsh), and (6) salt marsh. Built-up land included urban/rural settlements and associated facilities. Industrial zones refer to stand-alone industrial areas named “industrial park” or “development zone.” Water bodies here refer to rivers, ponds, reservoirs, and sea water. Salt marsh is a typical coastal wetland form of Lianyungang, which has high development potential to be reclaimed or utilized for industrial use.
Environ Sci Pollut Res (2015) 22:16997–17011
16999
Fig. 1 Location and watersheds delineation of the study area
Watersheds and water quality The study area has diverse topography with mountainous uplands with a maximum elevation of 607 m (above sea level) and flat alluvial plains with welldeveloped drainage network. Boundaries of watersheds are unambiguous in steep terrain, while difficult to delineate in flat areas. The accuracy of watershed determination in flat landscapes is challenged by the gentle slope differences and human activity influences. Adding input of ancillary drainage details on DEM was proved effective to better determine the river flows (Turcotte et al. 2001; Al-Muqdadi and Merkel 2011). We used DEM data, which was corrected with the river network provided by Lianyungang Water Authority, to produce the watershed boundaries based on the hydrological module of ArcGIS 9.3. Most of the rivers are natural so the watersheds extracted from the original 30×30 m DEM were consistent with the actual waterways. However, some rivers flow through Guanyun County are strongly impacted by land managers which were diverted as canals or flood-control channels. These rivers are mostly artificial so that the watersheds delineated were not completely matching the actual channels. We
did field checking and delineated the watersheds with the supervision of experts from Lianyungang Water Authority to improve the accuracy. Furthermore, in order to ensure the hydrological links between the runoff and the watersheds, only the monitoring sites suited in the river locks as well as in the outlet area of watersheds were selected. Thirty-three monitoring sites covering 19 major rivers were selected. Water quality data of the sites in 2000 and 2008 were collected from the Lianyungang Municipal Environmental Protection Bureau. Annual average data were gathered which allow us to avoid intervenes from temporal variability and focus on spatial variations. We defined the Pollution Load Index (PLI) as Eq. (1), which calculated the contribution of each pollutant, to screen the water quality indicators. PLIi ¼
n X
=Ci j0
Ci j
XX m
j¼1
n
ð1Þ
=Ci j0 100%
Ci j
i¼1 j¼1
Where PLIi is the pollution load index of pollutant i, m is the number of water quality indicators, Cij is the annual average value of indicator i at station j, n is the
17000
Environ Sci Pollut Res (2015) 22:16997–17011
Table 1
Anthropogenic context of Lianyungang during the study period
2000: 10th 5YP(5-year plan)a
,“Raise urbanization levels”
2005: 1 st Administrative Division Adjustment Landscape causes
Urban land expansion
2006: 2nd Administrative Division Adjustment 2008: 3rd Administrative Division Adjustment New Round of Urban Master Plan
Urbanization level
2000: Urban districts (69.1%), counties (16.4%) 2008: Urban districts (80.0%), counties (31.4%) 2000: Initiative of “National Environmental Protection model cities”
Relevant influential factors
Environmental management
2005: 11th 5YP “Restrictive target of COD emission reduction”; initiative of “Eco-city program” 1st centralized sewage treatment plant in Xinpu District 2007: 2nd centralized sewage treatment plant in Guanyun County 2000: 5 industrial zones, lead by food processing industry, mechanism manufacturing, and chemical industry
Industrial zone expansion
2008: 12 industrial zones, lead by food processing industry, chemical industry, and pharmaceutical industry Designated as “the steel and petrochemical industry base” in Jiangsu Coastal Development Strategy
Data source: Lianyungang Statistical Yearbook (2000-2008) a Aseries of social and economic development initiatives published by Chinese government every 5 years since 1953 (except for the period 1963 -1965) that set the national growth and reform targets
number of monitoring stations, and Cij0 is the water quality standard value1 of indicator i at station j. Six water quality parameters out of 11 standard monitoring indicators from the environmental protection bureau were employed. They are chemical oxygen demand (COD), biochemical oxygen demand (BOD5), ammonia nitrogen (NH3-N), oils, total phosphorous (TP), and Hg. These parameters experienced significant changes during the study period (PLI above 5 %, and their sum take up to 90 %) and demonstrated domestic, agricultural, and industrial discharges.
tially varying relationships, by producing a set of local regression results for each sampling sites (Brunsdon et al. 1998). GWR results including local coefficient (LC) estimates, R2 (coefficient of determination), t test values, and local residuals were calculated by SAM 4.0 and mapped with ArcGIS 9.3. The value of the local coefficient estimate indicates the relationship between the independent and the dependent variables. For example, negative LC indicates negative relationship and positive LC means positive linkage; bigger absolute value of LC indicates high sensitivity of water quality changes. GWR model is as follows:
Regression analysis
X β i u j ; v j xi j þ ε j Y j ¼ β0 u j ; v j þ
Water quality changes induced by land use practices tend to be spatial auto-correlated, since closer sites share similar natural background and human disturbances (de Freitas et al. 2013; Pratt and Chang 2012; Yu et al. 2013). GWR can effectively quantify the spa-
1
Water quality standard is designated by the surface water quality standard GB3838-2002 and Surface Water Environmental Functional Zoning of Jiangsu Province.
n
ð2Þ
i¼1
Where Yj is the independent variable at location j(uj, vj), β0 (uj, vj) is the intercept, and βi is estimated local coefficient estimate for independent variable Xi at the location j and εj is the error term. GWR models assume closer independent units have stronger influence on the estimated value of a dependent variable. The influences decays by distance and are captured by a “weighting” procedure. The weight of each
Environ Sci Pollut Res (2015) 22:16997–17011
17001
Fig. 2 Profiles of urban districts and counties with a urbanization level and centralized sewage collection rate and b industrial output value and industrial waste water discharge
location is valued by decaying exponential distance: " # di j 2 W i j ¼ exp − ð3Þ b Where Wij is the weight of observation j that impacts observation i, dij is the distance between the two observation units, and b is the bandwidth which determines the scale of the units to be included in the regression. We adopt adaptive kernel bandwidth method (Fotheringham et al. 2002) to determine the optimal bandwidth. Usually, the land use and water quality data at different time period were pooled together in regression analysis, which
Fig. 3 Land use of the study area (2000, 2008)
provide more samples but lead to temporal-spatial mismatches. Also, seasonal differences and hydrological links may have influences on the regression results. We addressed the concerns by introducing temporal change intensity indexes, which allow us to focus on the sensitivity of water environment in responses to urban expansion. The indexes calculated the relative change value during the study period. The temporal change intensity of urban expansion was measured by land use change intensity (LCI) as Eq. (4), which has been applied to capture urbanization intensity in many studies (McMahon and Cuffney 2000; Tate et al. 2005; Xiao et al. 2006). Similarly, the water change intensity (WCI) of each water indicator was calculated as Eq. (5). Each GWR model
17002
Environ Sci Pollut Res (2015) 22:16997–17011
used the index of urban expansion as an independent variable to analyze its association with the index of every water quality indicator. LCIi ¼
LA j;tþn −LA j;t 100% n LAi
C i j;tþn −C i j;t 100% n C i j;t
ð5Þ
Where WCIij is the annual change intensity of water indicator i at monitoring station j (%/a); Cij,t+n and Cij,t is the concentration of water parameter i in the years of t+n and t. Statistical analysis In order to identify the determinant to the spatially varying relations, we employed non-parametric statistics to analyze the contributions of relevant influential factors to the variances in local models (Fig. 4). Four relevant influential factors are selected based on comparability with previous studies and actual anthropogenic background of the study area. They were (1) urbanization level, (2) environmental management, (3) industrial zones, and (4) land use composition. Watersheds were classified into two groups with our concerns about each influential factor (Table 2). Non-parametric tests are used to interpret whether there are significant differences of two groups by comparing their medians. The tests were firstly performed between GWR local coefficients (LC) and each influential factor in pairs. Statistical significant testing results indicated strong influences from a certain factor on the spatially varying land-water linkages. Otherwise, the tests were further applied between water quality change intensity (WCI) and influential factors to identify whether the factor had direct influence on water quality changes. Mann-Whitney U (M-W U) test, a non-parametric test when there are two groups involved, was performed by the PASW Statistics 18.0 (SPSS Inc.).
Results Spatial pattern of urban expansion and water quality changes Urban land expansion was recognized in all of the watersheds, but more intensive in counties with higher
Water quality changes GWR Regression
Independent variable (LCI, Land Change Intensity)
ð4Þ
Where LCIi is the annual land use change intensity of watershed i (%/a), LAj,t+n and LAj,t are the areas of built-up land in watershed i in the years of t+n and t, and LAi is the total area of watershed i. WCIi j ¼
Urban expansion
Dependent variable (WCI, Water Change Intensity)
Spatially varying relations Non-parametric Statistics
Urbanization
Environmental
Industrial
Land use
Level
management
zone
composition
Relevant influential factors
Fig. 4 Analysis framework of the study
average LCI at 1.11 %/a (Table 3). The most obvious urban expansion was observed in the central area of Guanyun County (Fig. 5). While urban growth in urban districts was relatively centered, urban land expansion in counties was dispersed as most of the watersheds had higher LCI of built-up land (over 1 %/a). Industrial zone exhibited dramatic growth particular in urban districts with a maximum LCI at 2.5 %/a. However, only three watersheds in counties had industrial zone settlements increase during the study period. The spatial pattern of water quality changes also varied between urban districts and counties. Except Hg, all the other indicators presented decreasing tendency in urban districts while increased in counties by comparing their means and medians (Table 3). Specifically, oils, NH3-N, and TP changed significantly over the study period with WCI ranges over ±10 %/a. Hg increased evidently in urban districts, while decreased slightly in counties.
Spatially varying relations explored by GWR models As shown in Fig. 6, urban expansion had consistent positive impacts on oils’ contamination and consistent negative influences on COD. Other indicators showed both positive and negative responses. In particular, BOD, NH3-N, and TP had similar response patterns. They were negatively associated with urban expansion in urban districts, whereas positively related to urban land increase at most of the watersheds in counties. Results also indicated that built-up land expansion significantly increased Hg concentration, mostly found in urban districts (p
RESEARCH ARTICLE
Water quality changes in response to urban expansion: spatially varying relations and determinants Wenjun Zhao 1 & Xiaodong Zhu 1 & Xiang Sun 1,2 & Yunqiao Shu 3 & Yangfan Li 1,4
Received: 22 January 2015 / Accepted: 27 May 2015 / Published online: 28 June 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract Urban expansion is an important stressor to water bodies, and the spatial variations of their relations are increasingly highlighted by recent studies. What remain unclear, however, are the underlying drivers to the spatial variability. The paper was not limited to modeling spatially varying linkages but also drew attention to the local anthropogenic influential factors that shape land-water relations. We employed geographically weighted regression to examine the relationships between urban expansion (measured by land use change intensity) and water quality changes (focusing on six water quality indicators) in a recently fast-growing Chinese city,
Responsible editor: Philippe Garrigues * Xiaodong Zhu [email protected]; [email protected] * Xiang Sun [email protected] Wenjun Zhao [email protected] Yunqiao Shu [email protected] Yangfan Li [email protected] 1
State Key Laboratory of Pollution Control and Resources Reuse, School of the Environment, Nanjing University, Nanjing 210023, People’s Republic of China
2
School of Environment, Guangxi University, Nanning 530004, People’s Republic of China
3
International Water Management Institute (IWMI) Southern Africa, Pretoria, South Africa
4
The Key Laboratory of the Coastal Zone Exploitation and Protection, Ministry of Land and Resource, Nanjing 210008, People’s Republic of China
Lianyungang. Specifically, we analyzed how the local characteristics including urbanization level, environmental management, industrial zone expansion, and land use composition, attributed to the varying responses of water quality changes. Results showed that urbanization level significantly affects land-water linkages. Remarkable water quality improvement was accompanied by urbanization in highly developed watersheds, primarily due to strong influence from extensive water management practices (particularly for COD, BOD, NH3-N, and TP). By contrast, water qualities of less-urbanized watersheds were more sensitive and negatively responsive to land use changes. Clustering industrial activities acted as distinct contributor to Hg contamination, while boosted organic pollution control in highly urbanized areas. The approach proposed in the study can locate and further zoom into the hotspots of human-water interactions, thereby contributing to better solutions for mitigating undesirable impacts of urbanization on water environment. Keywords River pollution . Land-water linkages . Geographically weighted regression . Anthropogenic factors . Urbanization level . China
Introduction Stream degradation is a key global issue in urbanizing areas (Defries and Eshleman 2004; Finkenbine et al. 2001; Hughes et al. 2014; Paul and Meyer 2001). Urban land expansion contributes to water pollution by increasing pollutant loads from both point and non-point sources (e.g., Atasoy et al. 2006; Dong et al. 2014; Halstead et al. 2014; Mehaffey et al. 2005; He et al. 2008). Exploring land-water relations in urban setting is essential to evaluate current and potential environmental consequences of human activities.
16998
Rivers flow through watersheds with varying natural and anthropogenic backgrounds, thus land-water linkages are inherently inconsistent. The spatial variations of land-water relations are demonstrated by water quality features (Carey et al. 2011; Li et al. 2009), watershed characteristics (e.g., land use composition, topography) (Huang et al. 2015; Su et al. 2013) and different analysis scales (e.g., sub-watersheds, riparian, or buffer zones) (Pratt and Chang 2012; Sliva and Williams 2001; Zhao et al., 2015). Conventional regression approach using global statistics may mask the spatial heterogeneity. Recent studies addressed the problem of generalization by employing a local regression statistics method named geographically weighted regression (GWR) (de Freitas et al. 2013; Tu 2011a; Yu et al. 2013). Most of the above-mentioned studies focused on exploring the most predictive models, but the anthropogenic driving forces behind the varying land-water linkages are less covered. Land use changes during urbanization did not always impact water quality directly but largely exerted the impacts through multiple influential factors (Teixeira et al. 2014). For example, urbanization pattern (Alberti et al. 2007; Carle et al. 2005; Hatt et al. 2004; McMahon and Cuffney 2000); landscape characteristics (Fu et al. 2005; Guo et al. 2010; Lee et al. 2009; Murray et al. 2010) and socio-economic factors (Huang et al. 2014; Juma et al. 2014; Pfeifer and Bennett 2011; Zhou et al. 2012) influence water pollutant features and runoff processes. These factors intervene and interact as part of the systematic processes, which may enhance, reduce, or even obscure the land use effects on water bodies (Jung et al. 2008; Mouri et al. 2012; Sun et al. 2014; Tu 2011b; Wang et al. 2013). As a result, it is critical to distinguish the influences from land use and possible stressors in understanding landwater relationships. The study therefore aims at tackling the above concern by including relevant influential factors in the analysis framework. Our research objective was not limited to investigating the spatially varying relations between urban expansion and water quality but further identifying the determinants to these variances. We conducted the study in Lianyungang, a rapidly urbanizing and expansive city during rapid urbanization period (2000– 2008). Specifically: & &
We firstly revealed the spatial variations of water quality changes in response to urban expansion, by modeling their relationships using GWR technique. Secondly, we analyzed how the anthropogenic factors contributed to the spatially varying responses of water quality changes. We analyzed whether urbanization level, industrial development, environmental management, and land use composition have significant impacts on the varying linkages using statistical tools.
Environ Sci Pollut Res (2015) 22:16997–17011
Study area Lianyungang is a recently fast-growing coastal city located on the western shore of the Yellow Sea (Fig. 1), which is facing multiple challenges from urbanization, industrial development, and environmental management (Table 1). The region is characterized by booming urbanization since 2000 with average annual growth of 9.8 % (while 0.5 % over the prior 20 years) and continuous growing industrial output value (Fig. 2). National and local development strategies accelerated urban spatial growth and stimulated the area as a rising center of clustering industrial zones. By contrast, the capacity of sewage collection and treatment grew slowly, and the first centralized sewage treatment plant was not installed until 2005. Lianyungang has implemented environmental management measures such as “total emission control policy” and “Eco-city program,” which are expected to improve environmental quality while facilitating urban and industrial use. However, rapidly degrading water quality is still a major concern as most of the rivers have been polluted. The study area (red boundary in Fig. 1) is the southeastern part of Lianyungang, covering three urban districts (Xinpu, Lianyun, Haizhou) and two counties (Guanyun, Guannan) with an area of 4053 km2. Urban districts are highly urbanized with mostly urban and suburban area while counties are predominantly agricultural. Despite of the availability of satellite images, this region is an adequate case study because it consists of urban centers as well as recently developing areas with gradients of anthropogenic influences.
Data and methods Land use data Landsat Thematic Mapper (TM) images from February 2000 and February 2008 were used to create land use classification (Fig. 3). We used two-step expert classification method (Sun et al. 2012) and supplementary material including Digital Elevation Model (DEM) data, vector coverages (e.g., industrial zones, salt marsh) to interpret land use information. The images were finally classified into six categories: (1) agricultural, (2) forest, (3) built-up, (4) industrial zone, (5) water bodies (other than salt marsh), and (6) salt marsh. Built-up land included urban/rural settlements and associated facilities. Industrial zones refer to stand-alone industrial areas named “industrial park” or “development zone.” Water bodies here refer to rivers, ponds, reservoirs, and sea water. Salt marsh is a typical coastal wetland form of Lianyungang, which has high development potential to be reclaimed or utilized for industrial use.
Environ Sci Pollut Res (2015) 22:16997–17011
16999
Fig. 1 Location and watersheds delineation of the study area
Watersheds and water quality The study area has diverse topography with mountainous uplands with a maximum elevation of 607 m (above sea level) and flat alluvial plains with welldeveloped drainage network. Boundaries of watersheds are unambiguous in steep terrain, while difficult to delineate in flat areas. The accuracy of watershed determination in flat landscapes is challenged by the gentle slope differences and human activity influences. Adding input of ancillary drainage details on DEM was proved effective to better determine the river flows (Turcotte et al. 2001; Al-Muqdadi and Merkel 2011). We used DEM data, which was corrected with the river network provided by Lianyungang Water Authority, to produce the watershed boundaries based on the hydrological module of ArcGIS 9.3. Most of the rivers are natural so the watersheds extracted from the original 30×30 m DEM were consistent with the actual waterways. However, some rivers flow through Guanyun County are strongly impacted by land managers which were diverted as canals or flood-control channels. These rivers are mostly artificial so that the watersheds delineated were not completely matching the actual channels. We
did field checking and delineated the watersheds with the supervision of experts from Lianyungang Water Authority to improve the accuracy. Furthermore, in order to ensure the hydrological links between the runoff and the watersheds, only the monitoring sites suited in the river locks as well as in the outlet area of watersheds were selected. Thirty-three monitoring sites covering 19 major rivers were selected. Water quality data of the sites in 2000 and 2008 were collected from the Lianyungang Municipal Environmental Protection Bureau. Annual average data were gathered which allow us to avoid intervenes from temporal variability and focus on spatial variations. We defined the Pollution Load Index (PLI) as Eq. (1), which calculated the contribution of each pollutant, to screen the water quality indicators. PLIi ¼
n X
=Ci j0
Ci j
XX m
j¼1
n
ð1Þ
=Ci j0 100%
Ci j
i¼1 j¼1
Where PLIi is the pollution load index of pollutant i, m is the number of water quality indicators, Cij is the annual average value of indicator i at station j, n is the
17000
Environ Sci Pollut Res (2015) 22:16997–17011
Table 1
Anthropogenic context of Lianyungang during the study period
2000: 10th 5YP(5-year plan)a
,“Raise urbanization levels”
2005: 1 st Administrative Division Adjustment Landscape causes
Urban land expansion
2006: 2nd Administrative Division Adjustment 2008: 3rd Administrative Division Adjustment New Round of Urban Master Plan
Urbanization level
2000: Urban districts (69.1%), counties (16.4%) 2008: Urban districts (80.0%), counties (31.4%) 2000: Initiative of “National Environmental Protection model cities”
Relevant influential factors
Environmental management
2005: 11th 5YP “Restrictive target of COD emission reduction”; initiative of “Eco-city program” 1st centralized sewage treatment plant in Xinpu District 2007: 2nd centralized sewage treatment plant in Guanyun County 2000: 5 industrial zones, lead by food processing industry, mechanism manufacturing, and chemical industry
Industrial zone expansion
2008: 12 industrial zones, lead by food processing industry, chemical industry, and pharmaceutical industry Designated as “the steel and petrochemical industry base” in Jiangsu Coastal Development Strategy
Data source: Lianyungang Statistical Yearbook (2000-2008) a Aseries of social and economic development initiatives published by Chinese government every 5 years since 1953 (except for the period 1963 -1965) that set the national growth and reform targets
number of monitoring stations, and Cij0 is the water quality standard value1 of indicator i at station j. Six water quality parameters out of 11 standard monitoring indicators from the environmental protection bureau were employed. They are chemical oxygen demand (COD), biochemical oxygen demand (BOD5), ammonia nitrogen (NH3-N), oils, total phosphorous (TP), and Hg. These parameters experienced significant changes during the study period (PLI above 5 %, and their sum take up to 90 %) and demonstrated domestic, agricultural, and industrial discharges.
tially varying relationships, by producing a set of local regression results for each sampling sites (Brunsdon et al. 1998). GWR results including local coefficient (LC) estimates, R2 (coefficient of determination), t test values, and local residuals were calculated by SAM 4.0 and mapped with ArcGIS 9.3. The value of the local coefficient estimate indicates the relationship between the independent and the dependent variables. For example, negative LC indicates negative relationship and positive LC means positive linkage; bigger absolute value of LC indicates high sensitivity of water quality changes. GWR model is as follows:
Regression analysis
X β i u j ; v j xi j þ ε j Y j ¼ β0 u j ; v j þ
Water quality changes induced by land use practices tend to be spatial auto-correlated, since closer sites share similar natural background and human disturbances (de Freitas et al. 2013; Pratt and Chang 2012; Yu et al. 2013). GWR can effectively quantify the spa-
1
Water quality standard is designated by the surface water quality standard GB3838-2002 and Surface Water Environmental Functional Zoning of Jiangsu Province.
n
ð2Þ
i¼1
Where Yj is the independent variable at location j(uj, vj), β0 (uj, vj) is the intercept, and βi is estimated local coefficient estimate for independent variable Xi at the location j and εj is the error term. GWR models assume closer independent units have stronger influence on the estimated value of a dependent variable. The influences decays by distance and are captured by a “weighting” procedure. The weight of each
Environ Sci Pollut Res (2015) 22:16997–17011
17001
Fig. 2 Profiles of urban districts and counties with a urbanization level and centralized sewage collection rate and b industrial output value and industrial waste water discharge
location is valued by decaying exponential distance: " # di j 2 W i j ¼ exp − ð3Þ b Where Wij is the weight of observation j that impacts observation i, dij is the distance between the two observation units, and b is the bandwidth which determines the scale of the units to be included in the regression. We adopt adaptive kernel bandwidth method (Fotheringham et al. 2002) to determine the optimal bandwidth. Usually, the land use and water quality data at different time period were pooled together in regression analysis, which
Fig. 3 Land use of the study area (2000, 2008)
provide more samples but lead to temporal-spatial mismatches. Also, seasonal differences and hydrological links may have influences on the regression results. We addressed the concerns by introducing temporal change intensity indexes, which allow us to focus on the sensitivity of water environment in responses to urban expansion. The indexes calculated the relative change value during the study period. The temporal change intensity of urban expansion was measured by land use change intensity (LCI) as Eq. (4), which has been applied to capture urbanization intensity in many studies (McMahon and Cuffney 2000; Tate et al. 2005; Xiao et al. 2006). Similarly, the water change intensity (WCI) of each water indicator was calculated as Eq. (5). Each GWR model
17002
Environ Sci Pollut Res (2015) 22:16997–17011
used the index of urban expansion as an independent variable to analyze its association with the index of every water quality indicator. LCIi ¼
LA j;tþn −LA j;t 100% n LAi
C i j;tþn −C i j;t 100% n C i j;t
ð5Þ
Where WCIij is the annual change intensity of water indicator i at monitoring station j (%/a); Cij,t+n and Cij,t is the concentration of water parameter i in the years of t+n and t. Statistical analysis In order to identify the determinant to the spatially varying relations, we employed non-parametric statistics to analyze the contributions of relevant influential factors to the variances in local models (Fig. 4). Four relevant influential factors are selected based on comparability with previous studies and actual anthropogenic background of the study area. They were (1) urbanization level, (2) environmental management, (3) industrial zones, and (4) land use composition. Watersheds were classified into two groups with our concerns about each influential factor (Table 2). Non-parametric tests are used to interpret whether there are significant differences of two groups by comparing their medians. The tests were firstly performed between GWR local coefficients (LC) and each influential factor in pairs. Statistical significant testing results indicated strong influences from a certain factor on the spatially varying land-water linkages. Otherwise, the tests were further applied between water quality change intensity (WCI) and influential factors to identify whether the factor had direct influence on water quality changes. Mann-Whitney U (M-W U) test, a non-parametric test when there are two groups involved, was performed by the PASW Statistics 18.0 (SPSS Inc.).
Results Spatial pattern of urban expansion and water quality changes Urban land expansion was recognized in all of the watersheds, but more intensive in counties with higher
Water quality changes GWR Regression
Independent variable (LCI, Land Change Intensity)
ð4Þ
Where LCIi is the annual land use change intensity of watershed i (%/a), LAj,t+n and LAj,t are the areas of built-up land in watershed i in the years of t+n and t, and LAi is the total area of watershed i. WCIi j ¼
Urban expansion
Dependent variable (WCI, Water Change Intensity)
Spatially varying relations Non-parametric Statistics
Urbanization
Environmental
Industrial
Land use
Level
management
zone
composition
Relevant influential factors
Fig. 4 Analysis framework of the study
average LCI at 1.11 %/a (Table 3). The most obvious urban expansion was observed in the central area of Guanyun County (Fig. 5). While urban growth in urban districts was relatively centered, urban land expansion in counties was dispersed as most of the watersheds had higher LCI of built-up land (over 1 %/a). Industrial zone exhibited dramatic growth particular in urban districts with a maximum LCI at 2.5 %/a. However, only three watersheds in counties had industrial zone settlements increase during the study period. The spatial pattern of water quality changes also varied between urban districts and counties. Except Hg, all the other indicators presented decreasing tendency in urban districts while increased in counties by comparing their means and medians (Table 3). Specifically, oils, NH3-N, and TP changed significantly over the study period with WCI ranges over ±10 %/a. Hg increased evidently in urban districts, while decreased slightly in counties.
Spatially varying relations explored by GWR models As shown in Fig. 6, urban expansion had consistent positive impacts on oils’ contamination and consistent negative influences on COD. Other indicators showed both positive and negative responses. In particular, BOD, NH3-N, and TP had similar response patterns. They were negatively associated with urban expansion in urban districts, whereas positively related to urban land increase at most of the watersheds in counties. Results also indicated that built-up land expansion significantly increased Hg concentration, mostly found in urban districts (p
