Exploring the Linkage between Land Use Type and Stream Water Quality of an Estuarine Island Applying GWR Model: A Case Study of Chongming, Shanghai

Lei Ding

Shanghai Municipal Engineering Design Institute (Group) CO., LTD., Shanghai, China.

DOI: 10.4236/gep.2022.107017                

Abstract

To understand the spatial correlations between land use type and water quality of an estuarine island is particularly essential to maintain its original ecological environment. Chongming Island is experiencing a rapid urbanization and agriculture land sprawl during the last decades, especially with the growth of agro-industry and the extension of old style residences. As a consequence, surface run-off from agricultural activities and domestic sewage discharge has a various linkage with stream water quality on the island. This study applied ageographical weight regression model approach to recognize the significance of the relationship between water quality and multiple land use. We also evaluated their spatial correlations which normally hidden from other traditional regression methods. The results reveal that the water quality of less-developed areas on Chongming Island was easily affected by land use types compared with other regions, TN, TP, BOD₅, COD were shown as the most significant responses among all the water quality indicators. Green land and water area had a reduced effect on nutrients, expansion of industrial land would continuously make a contribution of pollutants to the water environment. Suggestions should therefore be taken into consideration during the process of development planning, in order to prevent water contamination.

Keywords

Estuarine Island, Land Use, Stream Water Quality, Geographical Weight Regression

1. Introduction

As sustainable development is the main objective of management for small islands worldwide (Sandy A. Kerr, 2005; Tran, 2006; van der Velde et al., 2007; Romeela Mohee et al., 2015), maintaining the water environment under a healthy condition is an essential task to achieve the aim of sustainable development. However, issues like a plan-less urban sprawl, agricultural intensification, agro-industry, a lack of waste water control and treatment technique on small islands in China are threatening their own water environment and quality (Baorong Huang et al., 2008). Furthermore, stream water quality was easily affected by non-point source pollution generated by agricultural activity and domestic waste water, especially in rainy seasons, which is the period that represents a major risk against the water environment protection of the island. Under such circumstance, exploring the association between multiple land use types and stream water quality would provide important references for ecological planning and management for the islands.

Research concerning about stream water quality on an estuarine island has always been a key issue of the ecological environment protection on the islands. MARIE L. et al. examined a long-term water quality data set from the mid of 1980 to 1997 to study the trends of eutrophication on Western Long Island, trying to figure out the inducement. In 2010 Emily J. Shumchenia et al. assessed water quality in Rhode Island of Greenwich Bay in USA using evaluation of sediment profile imagery as the tool. Later on in 2011, Zhongyuan Yu et al. used a multiple-factor method to study spatial and industrial distribution traits and impact factors of water pollution on Hainan island. This study was aimed at discussing strategic ways to further building of eco-province on the island. Besides, many other studies on water quality of islands mainly focus on groundwater of volcanic islands, which discussed the effects of extra conditions such as climate change (Davood Mahmoodzadeh et al., 2014), surface soil (Syuntaro Hiradate et al., 2015) and sediments (Tatiana Cruz Fuentes et al., 2014) to ground water of the island. Researches concentrating on ground water of volcano island were relatively sufficient, however, there is a lack of studies focusing on the stream water quality and its corresponding impact factors of islands intervened by intense human activities. On a habitat island which is short of sewage treatment facilities, agriculture non-point source and urban point source pollution discharge have a significant effect on the stream water on the island and its downstream estuarine (Baker, 2003). Therefore, there is a missing of research on stream water quality of island associated with human settlements. Studying the influence of stream water quality of an island under anthropological development has important significance in protecting the ecological environment, as well as planning and implementing sustainable development of the small island (Ghina, 2003).

As spatial variations in land use type have affected the integrity and water quality of water resources, research on correlativity between land use type and stream water quality can provide valuable information for the development plan of the islands. For instance, many chemicals constitutes such as ammonia nitrogen concentration and chemical oxygen demand are affected both directly and indirectly by human land use activity (Bahar et al., 2008), which in turn cause water contamination problems, and potentially alter the ecological water environment. The relationship between land use type and water quality has been widely studied, discovering that stream water pollution is significantly related to agricultural non-point source pollution, discharge of industrial sewage and rural domestic waste (Ye et al., 2009). According to former researches (Dylan S. Ahearn et al., 2005; Roberts & Prince, 2010; María Laura Miserendino et al., 2011; Meneses et al., 2015; Songyan Yu et al., 2016), studying the influence of land use type on water quality is an effective and practical approach to predicting stream water quality trend in different scales of natural area. However, few researchers have explored the spatial relationships between land use type and stream water quality in a watershed on an island. Nevertheless, for some estuarine islands like Chongming which are close to a highly urbanized region, the land use type should have a major impact on its water environment and water resource. In addition, assessing the impact of land use type on islands is also an important subject faced by resource managers (Romeela Mohee et al., 2015). And this study will focus on figuring out the relationships between land use type and stream water quality of a habitat island.

Although many studies use conventional statistical methods such as ordinary least square (OLS) regression to address the relationships between land use type and alteration of water quality in different scales and places (Hejun Chang, 2008; Heidi Hertler et al., 2009; Guy Ragosta et al., 2010; Jinliang Huang et al., 2012; Meneses et al., 2015), this global regression lacks the ability to process relationships in local-specific area and spatial autocorrelation in model residuals (Songyan Yu et al., 2016). Besides, researchers found that different land use types have different impacts on water quality, and the relationship between a water quality indicator and a land use type may not constant over space (Jun Tu, 2011). In order to solve the inconsistency problem caused by spatial scale effects and watershed features, a geographically weighted regression (GWR) model, which has advantages for exploring the continuously spatial relationship variations over conventional global statistical methods was applied to examine the spatially varying relationships between land use type and water quality parameters (Fotheringham et al., 2003; Jun Tu et al., 2007; Jinliang Huang et al., 2015). An OLS model is also applied to compare the performance with GWR model.

This study aims at discovering spatial land use-water quality relationships of a human lived estuarine island in eastern coast of Shanghai (China) using 646 sampling data during the rainy seasons in 2011, as a deficiency of sampling sites in the whole study area has limitations on the effect of model simulation (Jinliang Huang et al., 2015). The GWR regression model was used for analyzing the impact of various land use types on different water quality parameters over space, and giving out the spatial relationship results. In order to examine which land use type can be applied as an important indicator to estimate and predict water quality, so as to improve the ecological management of land use of a scientific approach. Finally, the objectives of this study include: 1) discover the characteristic of spatial effect of different land use type of water quality of the island; 2) reveal the spatial relationship between stream water quality and multiple land use types; 3) discuss some general insights for island watershed management implications.

2. Study Area

Chongming Island is located in the Yangtze river delta, approximately 8 km off the coastal area of East China Sea at 121˚11'30'' to 121˚54'00''E longitude and 31˚27'00'' to 31˚51'15''N latitude (Figure 1). It has a major axis in the west-east direction. The island is the largest estuary alluvial island in the world which covers about 1200.68 km² land area with a history of 1300 years. In 2011, Chongming Island supported a population of about 703,722 including local and Outlanders. Chongming Island is situated in a subtropical zone with a monsoon climate, with a mean annual temperature of 15.3˚C. Its annual rainfall amounts to 1003.7 mm but 70.7% occurs in rainy seasons from April to September. It is characterized by a high density of river networks, fertile soil and relatively isolated environment with a long history of tillage. Besides, Chongming Island is ecologically sensitive as it is located in a boundary zone of ocean and estuary, which is close to a highly urbanized city area.

Figure 1. Study area and position of sampling points.

The island developed with agriculture based on the principle and supplemented by few manufacturers in a long term. Along with the construction of infrastructure and a local urbanization, industry gets a rapid speed of development. However, the development situation of agriculture will not change in recent years. Therefore, non-point source generated by agriculture and rural residence is assumed to be a long term problem. Beyond that, the island has to rely on the fresh water which is gained from the Yangtze river to fulfill the needs of lives. Thus, shortage of fresh water results in a high request for stream water quality inside the island.

3. Data and Methods

3.1. Data Sources

3.1.1. Water Quality Data

Water quality data of 2011 were obtained from Water Conservancy Management Office in Shanghai. The data were collected from the second water resource census of Shanghai. The whole process of surface water sampling and laboratory experimenting followed the Chinese National Standard (CSN) and also the International Organization for Standardization (ISO). Besides, the quality control procedures of laboratory work were also complied with CSN. Data from 646 water sampling sites were collected on the island, and the water quality parameters consisted of Biochemical Oxygen Demand (BOD₅), Chemical Oxygen Demand (COD), ammonia nitrogen (NH₃-N), total phosphorus (TP), total nitrogen (TN). We have calculated the average value of each water quality indicator at each spot to ensure the veracity of the regression results.

3.1.2. Land Use Data

Land use data for 2010 were obtained from the Key Laboratory of Geographic Information Science of Ministry of Education, East China Normal University as the original land use data sources. Land use types of the original data set were reclassified into six categories by integrating: industrial, built-up, transportation, agriculture, green land and water area. We summed public green land, functional green land and forest land as green land in this study. Built-up land is summed as a percentage of public architecture, residential, commercial and recreational land. Water area is summed as rivers, lakes and ponds.

In order to make land use indicators and water quality sampling sites better corresponded, the study area has been divided into 17 hydrology sub-catchments using Arcgis spatial hydrological module, based on digital elevation data provided by Geographical Information Monitoring Cloud Platform, and then calculate land use indicators.

All land use indicators (percentage of each land use area) were calculated in Arcgis by overlapping land use and water quality indicators in the sub-catchment drainage area of every sampling site. Thus, the linkage of land use indicators for each sampling site and sub-catchment were established, then used for further analysis.

3.2. Regression Method

3.2.1. Geographically Weighted Regression

The geographically weighted regression (GWR) model is an extension of the ordinary least squares (OLS) regression model that allows parameters to be estimated by a multiple regression model (Fotheringham et al., 2003). Instead of estimating global parameters which ignores the coordinates of the observations, geographically weighted regression can generate a continuous surface of parameter values at each local observation to denote the spatial variations of the surface. This article aims to discuss the spatial influence caused by land use type on Chongming Island, using the method of GWR perform more pertinently and making spatial analysis fit the real situation. For the operation of OLS, we need to compare the accuracy of the results generated by two different regression models so as to testify the advantage of GWR in a spatial analysis.

GWR is a model associating variables with localized regression coefficient, it is based on kernel regression which allows various relationships to exist at different points in space (Antonio Pasculli et al., 2014). A typical GWR model can be written as follows:

$y_i={\beta }_0\left({\mu }_i,{\nu }_i\right)+{\displaystyle \underset{i=1}{\overset{k}{\sum }}{\beta }_i\left({\mu }_i,{\nu }_i\right){\chi }_{ij}}+{\epsilon }_i$

where $y_i$ denotes the dependent variable, in this case the value of water quality parameters at location i , ${\mu }_i$ and ${\nu }_i$ denotes for spatial point coordinates for each location i , ${\beta }_0\left({\mu }_i,{\nu }_i\right)$ stands for the intercept coefficient at location i, ${\chi }_{ij}$ is the value of the jth explanatory variable at location i , ${\beta }_i\left({\mu }_i,{\nu }_i\right)$ is the local regression coefficient for the i th explanatory variable. Besides, k is the number of independent variables, ${\epsilon }_i$ is the random location specific error term. All observations around sampling sites were weights using a distance decay function in GWR, with an assumption that the observations closer to the sampling sites have higher effects on local estimates for the site (Tu & Xia, 2008). In this study the adaptive bi-square nearest neighbor formulation was adopted as the weighting function of the GWR model, as the data points were distributed freely over the study area, and the optimal bandwidths was generated through minimizing the AICc (Akaike Information Criterion) to the lowest, which on behalf of a better model performance (Hurvich et al., 1998). The kernel shape is defined by following equation:

${\omega }_{ij}=\{\begin{array}{l}{\left[1-{\left(d_{ji}/h_j\right)}^2\right]}^2{d}_{ji}<h_j\\ 0\text{otherwise}\end{array}$

where ${\omega }_{ij}$ is the weight of observation j for observation i , $d_{ji}$ is the distance between observation j and i, $h_j$ is the kernel bandwidth which stands for the nth nearest neighbor distance from j.

GWR model generates a separate regression equation for each observation and provides a method to measure the spatial relationships between dependent variables and independent variables, assessing degree of spatial non-stationarity. The regression results calculated by GWR model mainly include local parameter estimate value, the local R² value, t-test value of the local parameter estimates, and the local residual value. The local parameter estimate value was used to study the positive or negative relationships between variables, the local R² value reveals the situation of fitting optimization of the model, the t-test value was applied to examine the significance of the sampling points, and the local residual was used to calculate spatial autocorrelation of the model.

3.2.2. Spatial Autocorrelation Analysis

Spatial autocorrelation refers to the same variable in different spatial positions relevance, it can be divided into global and local indicators. Global Moran’s I is calculated to detect spatial types throughout the study area which uses a single value to reflect the degree of autocorrelation of the study area. If a significant spatial correlation exists in an OLS or GWR model, then the model is probably to be invalid as it violates the assumption of randomly distributed and independent residuals in regression models (Rodrigues et al., 2014).

Moran index of model residual is one of the important indicators to test spatial autocorrelation. Moran index ranges from −1 to 1, the closer it reaches to −1 or 1, the more significant spatial autocorrelation the model has. Besides, the index reaches to zero indicates a spatial randomness (Antonio Pasculli et al., 2014) of the sample. This paper has calculated the global and local Moran index of residuals of each GWR and OLS model to evaluate the model's ability to deal with autocorrelation. Local Moran’s I has been calculated for each spatial unit and the adjacent unit on the relevance of a particular property.

This study used spss 20.0 software to establish OLS model. GWR model was set up by GWR4.0 software, Global Moran’s I and Local Moran’s I statistics were carried out using Geoda software.

4. Results

4.1. Variation of Land Use Type and Water Quality among Sub-Catchments

According to variations of dominant land use type, the study area was divided into three categories, including urban sub-catchment, agriculture sub-catchment and less-developed sub-catchment so as to facilitate the research (Figure 2).

The land use varied among sub-catchment. In urban sub-catchment (U1, U2, U3) the total proportion of industrial land, built-up land, transportation land, has exceeded 22% of the sub-catchment, and in agriculture district (A1, A2, A3, A4) the percentage of agriculture land is more than 40% of the total area. Besides, the proportion of less-developed land (L1 - L10) which includes water quality and green land is beyond 63% of the corresponding sub-catchments.

The water quality of streams on Chongming Island generally met the requirements of landscape and recreation use set by surface water quality standard of Grade IV. However, spatial variations of water quality were found among different sub-catchments (Table 1).

The less-developed catchments had a lower concentration of DO and a higher COD, BOD, TP, TN, NH₃-N, than urban catchments and agriculture catchments. Compared to the standard concentrations of Grade IV, water quality parameters of DO, TP, TN, NH₃-N, had a higher exceeding rate than COD, BOD. Besides, the exceeding rate of DO, NH₃-N at the less-developed catchments (range: 65%, 63%) was higher than the urban catchments (range: 59%, 58%) and agricultural catchments (range: 59%, 59%). The results demonstrated that there was a potential risk of nutrient pollution in the less-developed catchments.

Normally, local parameter estimate (regression coefficients) of an independent variable of the GWR model is used to examine the level of dependent variable affected by independent variable. In this study, regression coefficients calculated by the GWR model were applied to analyze the relationships between land use type and water quality indicator at each regression point. Additionally, the

Figure 2. Land use proportion of Chongming Island.

Table 1. The variation of water quality parameters of different sub-catchment.

local R² generated by the model was used to demonstrate the ability of the corresponding models to explain the spatial variance of model regression results. At last, the significance of the relationships was examined by t-test in the GWR models during estimates of the independent variable (Dylan S. Ahearn et al., 2004).

4.2. Performance of OLS and GWR Model

Generally, the local R² and AICc (Hurvich et al., 1998) of GWR were calculated to evaluate model performance. The increase of local R² of the model indicates a strong explain ability from independent variable to dependent variable. And AICc value was calculated to evaluate the accuracy and complexity of the regression model. A lower value of AICc of the model accounts for a better ability to predict (Wang et al., 2013).

In this study, we found that the performance of GWR model had a significant promotion compared with OLS model. As shown in Table 2, all local R² value of GWR models displayed much higher than OLS models, revealing a great improvement of model ability to better explain spatial variance of the independent variables.

AICc is another important parameter to assess the accuracy and precision of regression model. A comparison of AICc values between different models of the

Table 2. Coefficient of determination local (R²) value of GWR models.

Abbreviations: industrial land (IN), transportation land (TR), built-up land (BU), agriculture land (AG), green land (GL), water area (WA). ${\text{R}}_{\text{G}}^2$ denotes coefficient of determination value of the geographical weight regression, ${\text{R}}_{\text{O}}^2$ denotes coefficient of determination value of the ordinary least square regression.

same dependent variables provides a simple way to detect the superiority of the model, in which lower AICc value denotes that the model has a closer prediction result than the measured value (Tu et al., 2007). As shown in Table 3, improvements for all AICc values were observed for GWR models over OLS models in this study. All the t-test values of GWR model are shown to be significant (p < 0.05) for different land use indicators with water quality parameters. Through the examination of AICc and local R², a significant advantage of the GWR model over OLS model was found, which was a guarantee for better accuracy of the GWR model.

The global Moran’s I was calculated to evaluate the autocorrelation of model residual. Statistically significant positive spatial autocorrelation was found in all the OLS models with Moran’s I value ranged from 0.098 to 0.695, which indicates the model violated the assumption of residual independence. In contrast, all GWR models had the global Moran’s I ranged from 0.005 to 0.024, revealing a much better ability to deal with spatial autocorrelation compared with OLS model. (Table 4)

4.3. Variations of Spatial Relationships between Land Use and Water Quality

Relationships between percentage of each land use type and water quality had shown an obvious spatial non-stationarity according to GWR model results, which suggests that the relationship between land use type and water quality is

Table 3. Statistical test results of AICc value.

Abbreviations: industrial land (IN), transportation land (TR), built-up land (BU), agriculture land (AG), green land (GL), water area (WA). AICc_G denotes the AICc value of the geographical weight regression, AICc_O denotes the AICc value of the ordinary least square regression.

Table 4. Moran’s I of the residuals from GWR model.

Abbreviations: industrial land (IN), transportation land (TR), built-up land (BU), agriculture land (AG), green land (GL), water area (WA). I_G denotes the value of Moran’s I for residuals from the geographical weight regression, I_O denotes the value of Moran’s I for residuals from the ordinary least square regression.

changing over space (Fotheringham et al., 2003). Land use indicators have both positive and negative relationships with all water quality parameters. But not all the relationships are significantly related.

For relationships between agricultural land and water quality, TN, TP, NH₃-N had higher significant levels than other water quality parameters. Nutrient of TN, TP, NH₃-N demonstrated a significant positive relationship (p < 0.05) with percentage of agriculture land in most of the less-developed sub-catchment. However, negative relationships were also appearing in agriculture sub-catchments, indicating the increase in the proportion of the agricultural land had little impact on the concentration of TN and TP (Figure 3).

Percentage of green land had a significant negative relationship (p < 0.05) with the concentration of TP in agriculture and less-developed sub-catchments (A1, A3, L3, L6, L7, L8), and positive relationship at A2, A4, L4, L5 sub-catchments. Meanwhile, the concentration of TN was negatively influenced by proportion of green land on both ends of the island area (L1 - L3, L8, L10, A4), but positively influenced by proportion of green land in agriculture and urban sub-catchments (A2, A3, U2, L7) (Figure 4).

Impact from built-up land on water quality was alleviated. As shown in Figure 5, percentage of built-up land presented significant positive relationships with a concentration of BOD₅ and COD mainly in the north area of the island (L1, L2, L3, L4, L5, L6, A1, A2, A3), suggesting that a higher percentage of

Figure 3. Results of the GWR model between percentage of agriculture land and water quality, from top to down: TP, TN, NH₃-N. Abbreviations: (a) local coefficients for water parameters and percentage of agriculture land; (b) significance distribution for water parameters and percentage of agriculture land.

built-up land relates to higher concentration of water pollutants in the corresponding area. On the contrary, parameters of BOD₅ and COD were negatively influenced by built-up land in south and the central area which basically consist of urban and less-developed sub-catchments (U1, U2, L7, L8, L9, L10).

The percentage of industrial land was found to be closely related to all aerobic organic pollutant water quality parameters than nutrients with an obvious spatial non-stationarity existed. Proportion of industrial land had significant positive relationships with a concentration of BOD₅ and COD at all urban sub-catchments in the north region (U1, U3) for most of the sampling sites. Same positive relationships were found in agriculture sub-catchments. In the

Figure 4. Results of the GWR model for percentage of green land with water quality parameters, from top to down: COD, TP, TN. Abbreviations: (a) local coefficients for water parameters and percentage of green land; (b) significance distribution for water parameters and percentage of green land.

mean time, correlation between percentage of industrial land and concentration of DO had revealed a significant negative relationship concentrated in all urban sub-catchments (U1, U2, U3), part of agriculture sub-catchments, and a large proportion of less-developed sub-catchments (A1, A2, L1 - L10).

Significant negative relationships were found between percentage of the water area and concentration of TP in the less-developed sub-catchments areas located in the southern and central parts of the island (L6, L7, L8, L9, L10). Similarly, the concentration of NH₃-N had a significant negative relationship with percentage of water area, which was also distributed in the central and southern parts of the island (L5, L6, L7, L9, L10).

Figure 5. Results of the GWR model between percentage of built-up land and water quality, from top to down: BOD₅, COD, DO. Abbreviations: (a) local coefficients for water parameters and percentage of built-up land; (b) significance distribution for water parameters and percentage of built-up land.

5. Discussion

5.1. Explanatory Ability of Land Use Variables

Local R² is the fitting optimization between independent variables and dependent variables. Higher R² illustrates that an independent variable can explain more variance to a dependent variable. However, the meaning of the measurement equation is far more important than the statistical significance. As a relatively low fitting optimization of GWR model could also illustrate the relationships in a right way only if they have significance in environmental studies (Heidi Hertler et al., 2009).

The average local R² of all GWR models of this study is 0.27, which is relatively low compared with previous studies (Kibena et al., 2014), indicating that land use type could only account for 27% of the effect to stream water quality on Chongming Island (Qinghai Guo et al., 2010). Some researchers (Galbraith & Burns, 2007) also found that the explanatory ability of land use variables was less than 30%, with land use variables only contain composition of the land, meanwhile other studies covered configuration of land use type, geological and hydrologic indicators (Johnson et al., 1997; Sliva & Williams, 2001; Jun Zhao et al., 2015). The differences indicate that the explanatory ability on an estuary island is not only related to hydrologic factors and social economic level, but also to how many corresponding and water quality parameters are being considered.

5.2. Spatial Varying Effects from Land Use to Water Quality

Results of local parameter estimates, t-test, and local R² generated by GWR models allow the analyzation of the Spatial varying effects from land use to water quality parameters.

Agriculture land and Industrial land

Agriculture land was mainly associated with TN, TP, NH₃-N both positively and negatively over space. A positive significant relationship was found between the percentage of agriculture land and nutrients in less-developed areas. It means that an increasing of agricultural land proportion will lead to a high concentration of TN in stream water of the corresponding area. This is due to agricultural activities, including utilization of chemical fertilizer, livestock farming (Carpenter et al., 1998; Haidary et al., 2013; Lee et al., 2009; Sonoda et al., 2001; Sun et al., 2013; Tu, 2011; Wan et al., 2014) and excrement generated by aquaculture, were still the main source of total nitrogen and phosphorus (Carpenter et al., 1998; Jun Zhao et al., 2015) in the year of 2010.

However, the percentage of agriculture land also has a negative relationship with the concentration TN, TP, NH₃-N in agriculture sub-catchments (Figure 6), as in sub-catchments where the percentage of agricultural land has been already high, the change in proportion of agriculture land had little effect on water quality of nutrients.

In urban sub-catchments, the percentage of industrial land ranged from 0.1% - 3.6%, indicating that industrial development on Chongming Island was only in its primary stage, but industrial emission was an important source of water pollutants. According to the result of local parameter estimates, industrial land tends to have significant positive relationships with BOD₅, COD and a negative relationship with DO in urban sub-catchments, which agreed with some former studies (Evan J. Fedorko et al., 2004; Dylan S. Ahearn et al., 2005; Haidary et al., 2013; Lee et al., 2009; Pratt & Chang, 2012; Tu, 2013; Wan et al., 2014), as waste water discharged by factories from industrial park hardly meet the Chinese waste water treatment standard. Additionally, the local treatment plant has a

(a)(b)(c)

Figure 6. Prediction map: Kriging of all local parameter estimates of the GWR results between water quality parameters and agriculture land, from (a) to (c): TP, TN, NH₃-N.

deficiency on processing ability and capacity to treat a needed amount of waste water. Significant positive relationships were revealed between industrial land and concentration of BOD₅, COD in agriculture sub-catchments, which primarily attribute to expansion of agro-industry.

Green land and Water area

An increase of green land area could reduce the concentration of nutrients of water quality parameters, meanwhile both the concentration of COD, BOD, DO and TP, NH₃-N, TN will improve along with the increase of water area.

Green land is often associated with good water quality (Schoonover et al., 2005) . The percentage of green land had shown a significant negative relationship with TP and TN in most of the less-developed sub-catchments, indicating a rise of green land proportion could reduce the concentration of COD and TP in stream water. Non-point source pollutants from agriculture activities had been sent into rivers and streams loaded by land run-off. During the period that run-off flow through grassland or forest land, nutrient molecules have been filtrated when they were affected by ecological processes, including soil adsorption, biological fixation of nitrogen from plants in green land and forest land (Sang-Woo Lee et al., 2009).

However, concentration of TN had shown a positive relationship with percentage of green land in some sub-catchments, reflecting a spatial different form GWR model. These results probably attributed to the contribution of fertilizer plants nearby (Water Conservation Bureau of Chongming, 2009).

The impact of water area on water quality was seldom reported in former researches. The analysis of GWR model results discovered that almost all water quality parameters had significant spatial relationships with the percentage of water area. TP and NH₃-N had negative relationships with water area in the south less-developed sub-catchments, while high percentage of water area were associated with lower concentration of COD, BOD₅ and DO. (Figure 7 and Figure 8) Both of the corresponding sub-catchments had a low percentage of water area ranging from 2.7% to 9.2%. The local water environment assimilative capacity will be improved with the increase of water area, which might be the main reason why concentration of TP, NH₃-N, TN were negatively associated with the percentage of water area in the less-developed sub-catchments.

Built-up land

A clear spatial variance was found between the percentage of built-up land and water quality using GWR model. In contrast with the findings of previous research, high percentage of built-up land was not associated with bad water quality in urban sub-catchments and less-developed sub-catchments on Chongming Island.

Strong significant negative relationships between BOD₅ and COD and percentage of built-up land were found in south and north central area which basically consisted of urban and less-developed sub-catchments, and the results may have two primary cause. On one hand, Chongming Island had widely adopted the technology of rural domestic sewage treatment facilities before 2010, and the

(a)(b)(c)

Figure 7. Prediction map: Kriging of all local parameter estimates of the GWR results between water quality parameters and green land, from (a) to (c): COD, TP, TN.

(a)(b)(c)

Figure 8. Prediction map: Kriging of all local parameter estimates of the GWR results between water quality parameters and built-up land, from (a) to (c): BOD₅, COD, DO.

treatment process works well at reducing the input nutrients generated by domestic sewage and rural livestock farming (Sheng Xie, 2013). On the other hand, in a sub-catchment where percentage of built-up land is low, it normally contains higher percentage of agriculture land and green land, which generates less water pollutants. As a consequence, high percentage of built-up land might not always associate with adverse water quality.

The GWR model results denoted that a lower percentage of built-up land was always associate with higher concentrations of BOD₅ and COD of stream water. In a sub-catchment where percentage of built-up is lower, it normally contains higher percentage of agriculture land and green land which generates less water pollutants. Therefore, high percentage of residential land might not always associated with adverse water quality.

However, in agriculture sub-catchments and less developed sub-catchments at the south of the island, the model results of built-up land and water quality had some significant positive relationships, which indicates built-up land still made certain contributions to non-point source pollution in those area.

5.3. Various Relationship between Land Use and Water Quality of Different Scales

This paper has simulated the spatial relationship between land use type and stream water quality of an estuary island on catchment scale, the results have many difference with most previous research (Dylan S. Ahearn et al., 2005; Haidary et al., 2013; Johnson et al., 1997; Lee et al., 2009; Pratt & Chang, 2012; Sonoda et al., 2001; Tu, 2013; Wan et al., 2014). The proximity at which land-use influences water quality has been studied in several different ways with varying results.

In some naturally upstream watershed scales, water quality of nutrients was strongly and positively related with built-up land (Sun et al., 2013). Meanwhile in coastal watershed scale, agriculture land was negatively correlated with nutrients of NH₃-N, TN (Jinliang Huang et al., 2013). Dylan studied the relationship between water quality and land use at a rural watershed scale showed that the proportions of built-up land including rural habitation and commercial land use had positive correlations to TN, TP, NH₃-N, COD concentrations (Dylan S. Ahearn et al., 2005). Besides, percentage of industrial land can’t predict the loading of NH₃-N and TN. However, a combination of grassland and agricultural land plays an important role in driving the nutrient loading, which increased with increasing agricultural coverage. This relationship was also identified by our study.

In some highly developed watersheds, Johnson, Zhao had summarized that agriculture is no longer a pollution source but a container for pollution under the circumstances of rapid urbanization (Zhao, 2008) . Christopher argues that urbanization had the strongest negative influence over water quality in different buffer scales of land use (Christopher P. Tran et al., 2010). Furthermore, Zhao found that urban land (built-up land, commercial land) had a positive relationship with most water quality indicators (TN, TP, COD_Mn, and NH₃-N) in metropolitan area, which is consistent with most previous research (Dylan S. Ahearn et al., 2005; Haidary et al., 2013; Johnson et al., 1997; Lee et al., 2009; Pratt & Chang, 2012; Sonoda et al., 2001; Tu, 2013; Wan et al., 2014).

Chongming Island wasn’t deeply affected by human disturbance, and nor the upstream water. It was self-closed using sluice on stream connected to the sea. The urbanization and industrialization were still in its primary age. Although agriculture land use had the most important impact on water quality in this study, green land and water area had a negative impact on water quality. Particularly, green land and built-up land had a more complex relationship with water quality. Green land had a positive relationship with COD in water while built-up land was positively related with DO. This result had been confirmed by previous study (Dylan S. Ahearn et al., 2005). However, this relationship was only significant within the sub-catchment scale. When the scale size increased to the whole island, the correlation between industrial land use and water quality decreased.

5.4. Island Water Environment Conservation and Land Use Development Plan

This research has studied the spatial correlations between various land use types and stream water quality on an estuarine island, in which agriculture is the dominated land use type.

The application of GWR model results have provided specific references on protecting the stream water quality of the island: 1) a rapid development of agricultural activity and agro-industry in the past few years were the main sources of water pollution, 2) negative impact on water quality from built-up land had been alleviated after the popularization of domestic sewage treatment service, but it was still a pollution source in the southern less-developed area, 3) industrial emission still had important effects on water quality in urban regions and some agriculture sub-catchments, 4) water quality in less-developed sub-catchment was more vulnerable compared with urban sub-catchment and agriculture sub-catchment, 5) on the whole scale of the island, transportation land was not significantly related with water quality.

The protection of stream water quality of an estuarine island is a priority for island ecological conservation, as sustainable development is the main objective of management for Chongming Island. Few suggestions have been proposed for conservation of water quality on the island. Firstly, land use development plan should pay more attention to controlling the total amounts of agricultural activity and intensification of agro-industry, improving the efficiency of domestic waste water treatment, so as to reduce non-point source of contaminants. Secondly, the administrative department should restrict industrial growth in the conservation area, and ensure waste water from factories is strictly treated to meet the emission standard before it is discharged into the river. Finally, local development programs should enlarge the percentage of green land in urban and agricultural sub-catchments, and put river network protection job as the focus, especially in highly developed areas.

5.5. Limitations of This Study

This study attempted on applying a relatively large sample size in the GWR model application compared with former studies, (Christopher P. Tran et al., 2009; Huang et al., 2015) trying to figure out the spatial association between water quality and various land use types on Chongming Island respectively. Nevertheless, there were several limitations. On one hand, there is a lack of comparison between high and low run-off condition of the land surface, this study only concern about water quality in the wet season of the year 2011. On the other hand, the classification of sub-catchments in this article is an approach which mainly suits when illustrating effects of surface water quality from non-point pollution source. In sub-catchments where point source discharge was the dominated pollution source, the results of the model may be affected. Besides, this study haven’t verified the temporary effect of land use change on stream water quality of the island, as only one year of the water quality and land use data were analyzed. Both positive and negative relationships appeared in some other sub-catchments, which might be affected by the sampling frequency and the way the sub-catchments had been divided.

6. Conclusion

In order to assess the spatial relationship between island stream water quality and land use type, we applied a geographically weighted regression methodology to analyze data collected from a river network of an estuarine island.

The characteristic of spatial effect from each land use type to water quality was evaluated by analyzing the results calculated by GWR model. We conclude that the water quality of less-developed sub-catchments at northern and southern parts of Chongming Island was regions which can be easily affected by land use types. Of all the water quality parameters are used to evaluate the impact from different land uses, water parameters of TN, TP, BOD₅, COD have shown the most obvious and significant responses.

The spatial relationship between land use type and water quality was evaluated through the local parameter estimate statistic analysis, which discovered that nutrients, mainly revealed positive and negative relationships with agriculture land and green land respectively. Meanwhile BOD₅, COD were positively related to industrial land and built-up land. Green land and water area had a reduced effect on nutrients in agriculture and less-developed area. Expansion of industrial land will continuously be a pollution source of the river simultaneously. Unexpectedly, built-up land denoted a negative impact to water quality in parts of the study area.

By the application of GWR model the character of spatial non-stationarity relationships and potential spatial differences between water quality and land use type was revealed, which could provide practical reference to island managers. In order to explore the effects from land-use to water quality, so as to protect the water environment of the island from been degenerated, further studies should be conducted in the future.

Acknowledgements

The research leading to these results has received funds from Research of Chongming Water Asset Accounting Technology (15dz1208102), also the authors would like to thank the all the support from the project of Sediment Quality Evaluation and Analysis of Water Resource Census (48903270), and for cooperation with Water Conservancy Management Office in Shanghai, Hydrological Station of Shanghai.

Funding

Funding Name: Study on Comprehensive Benefit Evaluation of Sewage Quality Improvement and Efficiency Increase in Yangtze River Delta Region; Funding No.: K2020J028A; Affiliation: Shanghai Municipal Engineering Design Institute (Group) CO., LTD.

Conflicts of Interest

The authors declare no conflicts of interest regarding the publication of this paper.

References

[1]	Ahearn, D. S., Sheibley, R. W., Dahlgren, R. A., & Keller, K. E. (2004). Temporal Dynamics of Stream Water Chemistry in the Last Free-Flowing River Draining the Western Sierra Nevada, California. Journal of Hydrology, 295, 47-63. https://doi.org/10.1016/j.jhydrol.2004.02.016
[2]	Ahearn, D. S., Sheibleya, R. W., Dahlgrena, R. A., Andersonb, M., Johnsonc, J., & Tate, K. W. (2005). Land Use and Land Cover Influence on Water Quality in the Last Free-Flowing River Draining the Western Sierra Nevada, California. Journal of Hydrology, 313, 234-247. https://doi.org/10.1016/j.jhydrol.2005.02.038
[3]	Bahar, M. M., Ohmori, H., & Yamamuro, M. (2008). Relationship between River Water Quality and Land Use in a Small River Basin Running through the Urbanizing Area of Central Japan. Limnology, 9, 19-26. https://doi.org/10.1007/s10201-007-0227-z
[4]	Baker, A. (2003). Land Use and Water Quality. Hydrological Processes, 17, 2499-2501. https://doi.org/10.1002/hyp.5140
[5]	Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., & Smith, V. H. (1998). Nonpoint Pollution of Surface Water with Phosphorus and Nitrogen. Ecological Applications, 8, 559-568. https://doi.org/10.1890/1051-0761(1998)008[0559:NPOSWW]2.0.CO;2
[6]	Chang, H. (2008). Spatial Analysis of Water Quality Trends in the Han River Basin, South Korea. Water Research, 42, 3285-3304. https://doi.org/10.1016/j.watres.2008.04.006
[7]	Fedorko, E. J., Pontius Jr., R. G., Aldrich, S. P., Claessens, L., Hopkinson Jr., C., & Wollheim, W. M. (2004). Spatial Distribution of Land Type in Regression Models of Pollutant Loading. Ecology and Population Biology, 207, 173. https://doi.org/10.1086/BBLv207n2p173
[8]	Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2003). Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. John Wiley & Sons.
[9]	Fuentes, C. T., del Carmen Cabrera, M., Heredia, J., & Custodio, E. (2014). Groundwater Salinity and Hydrochemical Processes in the Volcano-Sedimentary Aquifer of La Aldea, Gran Canaria, Canary Islands, Spain. Science of the Total Environment, 484, 154-166. https://doi.org/10.1016/j.scitotenv.2014.03.041
[10]	Galbraith, L. M., & Burns, C. W. (2007). Linking Land-Use Water Body Type and Water Quality in Southern New Zealand. Landscape Ecology, 22, 231-241. https://doi.org/10.1007/s10980-006-9018-x
[11]	Ghina, F. (2003). Sustainable Development in Small Island Developing States: The Case of the Maldvies. Environment, Development and Sustainability, 5, 139-165. https://doi.org/10.1023/A:1025300804112
[12]	Guo, Q. H., Ma, K. M., Yang, L., & He, K. (2010). Testing a Dynamic Complex Hypothesis in the Analysis of Land Use Impact on Lake Water Quality. Water Resources Management, 24, 1313-1332. https://doi.org/10.1007/s11269-009-9498-y
[13]	Haidary, A., Amiri, B. J., Adamowski, J., Fohrer, N., & Nakane, K. (2013). Assessing the Impacts of Four Land Use Types on the Water Quality of Wetlands in Japan. Water Resources Management, 27, 2217-2229. https://doi.org/10.1007/s11269-013-0284-5
[14]	Hertler, H., Boettner, A. R., Ramírez-Toro, G. I., Minnigh, H., Spotila, J., & Kreeger, D. (2009). Spatial Variability Associated with Shifting Land Use: Water Quality and Sediment Metals in La Parguera, Southwest Puerto Rico. Marine Pollution Bulletin, 58, 672-678. https://doi.org/10.1016/j.marpolbul.2009.01.018
[15]	Hiradate, S., Morita, S., Morita, S., Osawa, T., Sugai, K., & Kachi, N. (2015). Effects of Soil Erosion and Seabird Activities on Chemical Properties of Surface Soils on an Oceanic Island in Ogasawara Islands, Japan. Catena, 133, 495-502. https://doi.org/10.1016/j.catena.2015.04.013
[16]	Huang, B. R., Ouyang, Z. Y., Zheng, H., Zhang, H. Z., & Wang, X. K. (2008). Construction of an Eco-Island: A Case Study of Chongming Island, China. Ocean & Coastal Management, 51, 575-588. https://doi.org/10.1016/j.ocecoaman.2008.06.007
[17]	Huang, J. L., Huang, Y. L., Pontius Jr., R. G., & Zhang, Z. Y. (2015). Geographically Weighted Regression to Measure Spatial Variations in Correlations between Water Pollution versus Land Use in a Coastal Watershed. Ocean & Coastal Management, 103, 14-24. https://doi.org/10.1016/j.ocecoaman.2014.10.007
[18]	Huang, J. L., Li, Q. S., Pontius Jr., R. G, Klemas, V., & Hong, H. S. (2013). Detecting the Dynamic Linkage between Landscape Characteristics and Water Quality in a Subtropical Coastal Watershed, Southeast China. Environmental Management, 51, 32-44. https://doi.org/10.1007/s00267-011-9793-2
[19]	Huang, J. L., Pontius Jr., R. G., Li, Q. S., & Zhang, Y. J. (2012). Use of Intensity Analysis to Link Patterns with Processes of Land Change from 1986 to 2007 in a Coastal Watershed of Southeast China. Applied Geography, 34, 371-384. https://doi.org/10.1016/j.apgeog.2012.01.001
[20]	Hurvich, C. M., Simonoff, J. S., & Tsai, C.-L. (1998). Smoothing Parameter Selection in Nonparametric Regression Using an Improved Akaike Information Criterion. Journal of the Royal Statistical Society Series B, 60, 271-293. https://doi.org/10.1111/1467-9868.00125
[21]	Johnson, L., Richards, C., Host, G., & Arthur, J. (1997). Landscape Influences on Water Chemistry in Midwestern Stream Ecosystems. Freshwater Biology, 37, 193-208. https://doi.org/10.1046/j.1365-2427.1997.d01-539.x
[22]	Kerr, S. A. (2005). What Is Small Island Sustainable Development about? Ocean & Coastal Management, 48, 503-524. https://doi.org/10.1016/j.ocecoaman.2005.03.010
[23]	Kibena, J., Nhapi, I., & Gumindoga, W. (2014). Assessing the Relationship between Water Quality Parameters and Changes in Landuse Patterns in the Upper Manyame River, Zimbabwe. Physics and Chemistry of the Earth, Parts A/B/C, 67-69, 153-163. https://doi.org/10.1016/j.pce.2013.09.017
[24]	Lee, S.-W., Hwang, S.-J., Lee, S.-B., Hwang, H.-S., & Sung, H.-C. (2009). Landscape Ecological Approach to the Relationships of Land Use Patterns in Watersheds to Water Quality Characteristics. Landscape and Urban Planning, 92, 80-89. https://doi.org/10.1016/j.landurbplan.2009.02.008
[25]	Mahmoodzadeh, D., Ketabchi, H., Ataie-Ashtiani, B., & Simmons, C. T. (2014). Conceptualization of a Fresh Groundwater Lens Influenced by Climate Change: A Modeling Study of an Arid-Region Island in the Persian Gulf, Iran. Journal of Hydrology, 519, 399-413. https://doi.org/10.1016/j.jhydrol.2014.07.010
[26]	Meneses, B. M., Reis, R., Vale, M. J., & Saraiva, R. (2015). Land Use and Land Cover Changes in Zêzere Watershed (Portugal)—Water Quality Implications. Science of the Total Environment, 527-528, 439-447. https://doi.org/10.1016/j.scitotenv.2015.04.092
[27]	Miserendino, M. L., Casaux, R., Archangelsky, M., Di Prinzio, C. Y., Brand, C., & Kutschker, A. M. (2011). Assessing Land-Use Effects on Water Quality, In-Stream Habitat, Riparian Ecosystems and Biodiversity in Patagonian Northwest Streams. Science of the Total Environment, 409, 612-624. https://doi.org/10.1016/j.scitotenv.2010.10.034
[28]	Mohee, R., Mauthoor, S., Bundhoo, Z. M. A., Somaroo, G., Soobhany, N., & Gunasee, S. (2015). Current Status of Solid Waste Management in Small Island Developing States: A Review. Waste Management, 43, 539-549. https://doi.org/10.1016/j.wasman.2015.06.012
[29]	Pasculli, A., Palermi, S., Sarra, A., Piacentini, T., & Miccadei, E. (2014). A Modelling Methodology for the Analysis of Radon Potential Based on Environmental Geology and Geographically Weighted Regression. Environmental Modelling & Software, 54, 165-181. https://doi.org/10.1016/j.envsoft.2014.01.006
[30]	Pratt, B., & Chang, H. (2012). Effects of Land Cover, Topography, and Built Structure on Seasonal Water Quality at Multiple Spatial Scales. Journal of Hazardous Materials, 209-210, 48-58. https://doi.org/10.1016/j.jhazmat.2011.12.068
[31]	Ragosta, G., Evensen, C., Atwill, E. R., Walker, M., Ticktin, T., Asquith, A., & Tate, K. W. (2010). Causal Connections between Water Quality and Land Use in a Rural Tropical Island Watershed. EcoHealth, 7, 105-113. https://doi.org/10.1007/s10393-010-0299-9
[32]	Roberts, A. D., & Prince, S. D. (2010). Effects of Urban and Non-Urban Land Cover on Nitrogen and Phosphorus Runoff to Chesapeake Bay. Ecological Indicators, 10, 459-474. https://doi.org/10.1016/j.ecolind.2009.07.017
[33]	Rodrigues, M., de la Riva, J., & Fotheringham, S. (2014). Modeling the Spatial Variation of the Explanatory Factors of Human-Caused Wildfires in Spain Using Geographically Weighted? Logistic Regression. Applied Geography, 48, 52-63. https://doi.org/10.1016/j.apgeog.2014.01.011
[34]	Schoonover, J. E., Lockaby, B. G., & Pan, S. F. (2005). Changes in Chemical and Physical Properties of Stream Water across an Urban-Rural Gradient in Western Georgia. Urban Ecosystems, 8, 107-124. https://doi.org/10.1007/s11252-005-1422-5
[35]	Sheng, X. (2013). Discussion on Application of Technological Processes for Rural Sewage Treatment Engineering. Water Purification Technology, 32, 79-83.
[36]	Sliva, L., & Williams, D. D. (2001). Buffer Zone versus Whole Catchment Approaches to Studying Land Use Impact on River Water Quality. Water Research, 35, 3462-3472. https://doi.org/10.1016/S0043-1354(01)00062-8
[37]	Sonoda, K., Yeakley, J. A., & Walker, C. E. (2001). Near-Stream Land Use Effects on Stream Water Nutrient Distribution in an Urbanizing Watershed. Journal of the American Water Resources Association, 37, 1517-1532. https://doi.org/10.1111/j.1752-1688.2001.tb03657.x
[38]	Sun, R. H., Chen, L.D., Chen, W. L., & Ji, Y. H. (2013). Effect of Land-Use Patterns on Total Nitrogen Concentration in the Upstream Regions of the Haihe River Basin, China. Environmental Management, 51, 45-58. https://doi.org/10.1007/s00267-011-9764-7
[39]	Tran, C. P., Bodeb, R. W., Smithb, A. J., & Kleppela, G. S. (2010). Land-Use Proximity as a Basis for Assessing Stream Water Quality in New York State (USA). Ecological Indicators, 10, 727-733. https://doi.org/10.1016/j.ecolind.2009.12.002
[40]	Tran, K. C. (2006). Public Perception of Development Issues: Public Awareness Can Contribute to Sustainable Development of a Small Island. Ocean & Coastal Management, 49, 367-383. https://doi.org/10.1016/j.ocecoaman.2006.02.005
[41]	Tu, J. (2011). Spatially Varying Relationships between Land Use and Water Quality across an Urbanization Gradient Explored by Geographically Weighted Regression. Applied Geography, 31, 376-392. https://doi.org/10.1016/j.apgeog.2010.08.001
[42]	Tu, J. (2013). Spatial Variations in the Relationships between Land Use and Water Quality across an Urbanization Gradient in the Watersheds of Northern Georgia, USA. Environmental Management, 51, 1-17. https://doi.org/10.1007/s00267-011-9738-9
[43]	Tu, J., & Xia, Z.-G. (2008). Examining Spatially Varying Relationships between Land Use and Water Quality Using Geographically Weighted Regression I: Model Design and Evaluation. Science of the Total Environment, 407, 358-378. https://doi.org/10.1016/j.scitotenv.2008.09.031
[44]	Tu, J., Xia, Z.-G., Clarke, K. C., & Frei, A. (2007). Impact of Urban Sprawl on Water Quality in Eastern Massachusetts, USA. Environmental Management, 40, 183-200. https://doi.org/10.1007/s00267-006-0097-x
[45]	Van der Velde, M., Green, S. R., & Vanclooster, M. (2007). Sustainable Development in Small Island Developing States: Agricultural Intensification, Economic Development, and Freshwater Resources Management on the Coral Atoll of Tongatapu. Ecological Economics, 61, 456-468. https://doi.org/10.1016/j.ecolecon.2006.03.017
[46]	Wan, R. R., Cai, S. S., Li, H. P., Yang, G. S., Li, Z. F., & Nie, X. F. (2014). Inferring Land Use and Land Cover Impact on Stream Water Quality Using a Bayesian Hierarchical Modeling Approach in the Xitiaoxi River Watershed, China. Journal of Environmental Management, 133, 1-11. https://doi.org/10.1016/j.jenvman.2013.11.035
[47]	Wang, R. Z., Xu, T. l., Yu, L. Z., Zhu, J. J., & Li, X. Y. (2013). Effects of Land Use Types on Surface Water Quality across an Anthropogenic Disturbance Gradient in the Upper Reach of the Hun River, Northeast China. Environmental Monitoring and Assessment, 185, 4141-4151. https://doi.org/10.1007/s10661-012-2856-x
[48]	Water Conservation Bureau of Chongming (2009). Water Conservation Annals of Chongming. (Chapter 4, pp. 212-215). Chongming Water Conservation Bureau.
[49]	Ye, L., Cai, Q.-H., Liu, R.-Q., & Cao, M. (2009). The Influence of Topography and Land Use on Water Quality of Xiangxi River in Three Gorges Reservoir Region. Environmental Geology, 58, 937-942. https://doi.org/10.1007/s00254-008-1573-9
[50]	Yu, S. Y., Xu, Z. X., Wu, W., & Zuo, D. P. (2016). Effect of Land Use Types on Stream Water Quality under Seasonal Variation and Topographic Characteristics in the Wei River Basin, China. Ecological Indicators, 60, 202-212. https://doi.org/10.1016/j.ecolind.2015.06.029
[51]	Zhao, J. (2008). Landscape Pattern Change and Its Environmental Response across Multiple Spatial Scales in Tidal Plain (Chapter 5). East China Normal University.
[52]	Zhao, J., Lin, L. Q., Yang, K., Liu, Q. X., & Qian, G. R. (2015). Influences of Land Use on Water Quality in a Reticular River Network Area: A Case Study in Shanghai, China. Landscape and Urban Planning, 137, 20-29. https://doi.org/10.1016/j.landurbplan.2014.12.010