Investigation and Determination of Land Use Effects on Surface Water Quality in Semi-Arid Areas: Case Study on Qarasu River in Iran
Subject Areas :
Jafar Mohammadi
1
,
Ebrahim Fataei
2
,
Akram Ojaghi
3
1 - Department of Environmental Science and Engineering, Ardabil Branch, Islamic Azad University, Ardabil, Iran
2 - Department of environmental Sciences, Ardabil Branch, Islamic Azad University, Ardabil, Iran
3 - Department of Environmental Sciences, Ardabil Branch, Islamic Azad University, Ardabil, Iran
Keywords:
Abstract :
Ahearn DS, Sheibley RW, Dahlgren RA, Anderson M, Johnson J, Tate KW: Land use/cover influence on water quality in the last free-flowing river draining the western Sierra Nevada, California. Journal of Hydrology.2005:313:234-247.
Amiri BJ, Nakane K: Entire catchment and buffer zone approaches to modeling linkage between river water quality and land cover-A case study of Yamaguchi Prefecture, Japan. Chinese Geographical Science. 2008:18: 85-92.
Amiri BJ, Nakane K: Modeling the linkage between river water quality and landscape metrics in the Chugoku District of Japan. Water resources management. 2009:23: 931-956.
Atkinson S, Johnson D, Venables B, Slye J, Kennedy J, Dyer S, Price B, Ciarlo M, Stanton K, Sanderson H : Use of watershed factors to predict consumer surfactant risk, water quality, and habitat quality in the upper Trinity River, Texas. Science of the Total Environment.2009: 407: 4028-4037.
Carey RO, Migliaccio KW, Li Y, Schaffer B, Kiker GA, Brown MT: Land use disturbance indicators and water quality variability in the Biscayne Bay Watershed, Florida. Ecological Indicators .2011:11: 1093-1104.
Davoudi Moghaddam D, Haghizadeh A, Tahmasebipour N, Zeinivand H. Spatial and Temporal Water Quality Analysis of a Semi-Arid River for Drinking and Irrigation Purposes Using Water Quality Indices and GIS. ECOPERSIA 2021; 9 (2) :79-93
Ding J, Jiang Y, Fu L, Liu Q, Peng Q, Kang M. Impacts of Land Use on Surface Water Quality in a Subtropical River Basin: A Case Study of the Dongjiang River Basin, Southeastern China. Water. 2015; 7(8):4427-4445.
Gleick PH, Cooley H, Katz D, Lee E, Morrison J, Palaniyappan M, et al. The World's Water 2006-2007: the Biennial Report on Freshwater Resources. Pacific institution for studies in development, environment and security; 2007.
Kaste,O. Henriksen,A. Hindar,A.Retention of atmospherically-derived nitrogen in subcatchments of the Bjerkreim river in southwestern Norway. Ambio. 1997: 26: 296–303.
Miller JD, Schoonover JE, Williard KW, Hwang CR. Whole catchment land cover effects on water quality in the Lower Kaskaskia River watershed. Water, Air, & Soil Pollution. 2011:221: 337-350.
Nasir MFM, Samsudin MS, Mohamad I, Awaluddin MRA, Mansor MA, Juahir H, Ramli N. River Water Quality Modeling Using Combined Principle Component Analysis (PCA) and Multiple Linear Regressions (MLR): A Case Study at Klang River, Malaysia. 2011.
Pirali Zefrehei A, Fallah M, Hedayati SA. Spatial‐temporal modeling of qualitative parameters and land use status in Anzali international wetland using GIS technique. Ecopersia. 2019;7(4):223-31.
Salajegheh, A., khalighi, S., Pourebrahim, S., & Beigi, H. (2022). Investigating the trend of temporal and spatial changes in river health using Flow Health (Case Study: Qarasu River). Iranian journal of Ecohydrology, 9(1), 35-46. doi: 10.22059/ije.2022.330043.1555
Seitz NE, Westbrook CJ, Noble BF. Bringing science into river systems cumulative effects assessment practice. Environmental Impact Assessment Review. 2011:31: 172-179.
Steele M, Aitkenhead-Peterson J. Long-term sodium and chloride surface water exports from the Dallas/Fort Worth region. Science of the Total Environment. 2011:409:3021-3032.
Uriarte M, Yackulic CB, Lim Y, Arce-Nazario JA. Influence of land use on water quality in a tropical landscape: a multi-scale analysis. Landscape ecology. 2011: 26: 1151-1164.
Verma RK, Verma S, Singh A, Tiwary R, Murthy S. Relationship between land use/ land cover patterns and surface water quality in the uper Damodar river basin, JHARKHAND.2010.
Yang X, Jin W. GIS-based spatial regression and prediction of water quality in river networks: A case study in Iowa. Journal of Environmental Management. 2010:91: 1943-1951.
Zampella RA, Procopio NA, Lathrop RG, Dow CL. Relationship of Land‐Use/Land‐Cover Patterns and Surface‐Water Quality in The Mullica River Basin. JAWRA Journal of the American Water Resources Association. 2007: 43: 594-604.
Zhou T, Wu J, Peng S. Assessing the effects of landscape pattern on river water quality at multiple scales: A case study of the Dongjiang River watershed, China. Ecological Indicators. 2012: 23: 166-175.
Investigation and Determination of Land Use Effects on Surface Water Quality in Semi-Arid Areas: Case Study on Qarasu River in Iran
Abstract:
Increasing population growth and civilization have caused great impacts on water quality. Among these, changes and excessive exploitation of land use have led to changes in the surface water quality to provide human needs. Therefore, the present study investigated the effects of land use on water quality using a spatial-statistical approach. To this end, the water quality data of 22 water quality-sampling stations were applied in the Qarasu watershed in Ardabil. Then, spatial analysis was conducted, including land use classification maps, watershed mapping and overlaying maps using GIS. Finally, correlation and multiple linear regression analysis were used to determine the relationship between water quality and land use variables. Pearson's correlation coefficient showed proper percentages of vegetation cover, agricultural lands, woodlands, residential users, barren land and forest as well as weak pastureland have significant correlation with water quality variables. Multiple linear regression analysis with stepwise approach also indicated that of dependent parameters as water quality variables, the parameters of pH, Cl, Ca, Mg, Na and SAR are associated with land use as independent parameters including irrigated agriculture, first-grade pasture, third-grade pasture, woodlands, moorland, forest and residential users. Furthermore, validation of the model based on two models of the survey of predicted and actual values as well as root-mean-square error (RMSE) demonstrated good accuracy of the resulting model.
Keywords: Water Resources Management, Land Use, Water Quality, Correlation
Human disturbances in watersheds besides increased water harvesting and changes in river ecosystems lead to adverse effects on the stability of freshwater resources around the world (Salajegheh et al., 2022; Gleick et al., 2007). Changes in the landscape caused by human activities can have undesirable impacts on water quality and quantity(Ding et al., 2015). Expansion and enhancement of the intensity of land use practices such as agriculture, forestry, urban growth and industrial development have harmful effects (various pollutions) on the health of the river flow(Davoudi Moghaddam et al., 2021;Fataei and Shiralipoor, 2011). The source of these pollutants can be as point or non-point in nature and occur in a large-scale spatio- temporal pattern (Seitz et al., 2011).
Atkinson et al. in 2009 using spatial analysis and multiple regression (stepwise) have developed a prediction model for surfactants, water quality and aquatic habitats based on the characteristics of the Trinity River sub-watershed basin in Texas ( Atkinson al et., 2009).
Amiri and Nakanh in 2009 investigated the correlation between the landscape patterns and water quality in Chugoku, Japan (Amiri and Nakane, 2009). Yang and Jin in 2010 examined the impact of watershed characteristics on nitrite and nitrate concentrations in Cedar River watershed (at Iowa Minnesota) (Yang and Jin, 2010). Verma and colleagues in 2010 conducted a study on the relationship between land use/cover and surface water quality in Damodar basin (Verma et al., 2010).
Nasir and colleagues (2011) determined the sources affecting water quality using five-year information of Klang River (Nasir et al., 2011). Miller and colleagues in 2011studied the effect of forest, agricultural and urban covers on water quality in the river basin of Kaskaskia, Illinois, America (Pirali Zefrehei et al., 2019;Miller et al., 2011). Uriarte et al. in 2011 using water quality data between 1977 and 2000, precipitation and land use maps investigated the effect of land use on river water quality in four spatial scales of watersheds, sub-watersheds, 60 m buffer along watersheds and 60 m buffer along sub-watersheds in Puerto Rico (Uriarte et al., 2011). Curry and colleagues in 2011 examined the relationship between land use disturbance indices and water quality changes in the Byskayn Bay watershed in Florida (Carey et al., 2011). Zhou and colleagues in 2012 studied the effects of landscape pattern on water quality in Dongjiang River watershed, China (Zhou et al., 2012).
This study was conducted to examine the relationship between river water quality and land use/cover in Qarasu watershed in Ardabil province, northwestern Iran.
In terms of country division, Qarasu watershed is located in the center of Ardabil province in northwest of Iran.
Baliqlu River that originates from rain, snowmelt and springs on the slopes of Sabalan Mountain after crossing through the middle of Ardabil city in Anzob village, near the Samian village is connected to the Qarasu river that originates from the western slopes of Talesh mountains and after crossing through Ardebil plain flows into northwestern direction and is flowing down to Aras River within Aslanduz city. The Landsat 8 satellite images of 2014 were used in order to prepare a land use map. For this purpose, after atmospheric and geometric correcting carried out on image, existing land use in the region was extracted based on supervised classification.
Statistical correlation and regression analyses determined the relationship among variables in SPSS 16.0 software. In this section, since the condition of using linear regression is normality of dependent variables related to water quality, so firstly all data used in the study were tested for normal distribution through Shapiro-Wilk test. The results showed that the variables did not match with normal distribution. The outlier data were identified and removed and were tested for normality again. Since the data are quantitative, Pearson’s correlation test was used to analyze the relationship between water quality variables and predictor variables of land use. Finally, multiple linear regression analysis with stepwise approach was employed to examine the relationship between water quality and land use.
Results
Land use/cover map was prepared in 11 categories including agriculture, dry land areas, good pastureland, average pastureland, poor pastureland, third-grade forest, woodlands, urban areas, water bodies, land without coverage and barren land. Land use/cover information is marked in the study area of Qarasu River watershed in Figure. The classes of rain-fed agriculture, average rangeland and irrigated agriculture with 39.12, 25.86 and 22 coverage percentage respectively consist of dominant land cover.
Fig 1- Land use/cover map of study Qarasu watershed in Ardebil province
Water quality data
Data on river water quality collected from the sampling stations located on rivers are usually presented as tables with several water quality measurement parameters along with name, profile and geographic coordinates of sampling stations. The water quality data of Ardabil Regional Water Company were used in this study. The sampling stations located on Qarasu river watershed have been listed in Table 2. The water quality variables measured in these stations include temperature, total hardness, sum of anions and cations, TDS, EC, pH, HCo3, CL, So4, Ca, Mg, SAR, Na.
Correlation
The correlation analysis was used to evaluate the relationship between two variables. There are several methods for studying the correlation; appropriate coefficient can be chosen based on the type of analyzed data. Because of quantitative nature of variables in this stud, Pearson's correlation coefficient can be used for variables with normal distribution. After logarithmic data conversion, Shapiro-Wilk test results showed this condition in the data. Therefore, Pearson's correlation coefficient was used for the data, which the results are presented in the following.
Pearson’s correlation test results at the significance level of 0.05 between land use variables and river water quality showed that:
TDS variable were positively correlated with the variables of EC, HCo3, So4, Ca, Mg, Na, total hardness, SAR and sum of anions and cations. EC variable had positive correlation with the variables of TDS, HCo3, So4, Cl, Ca, Mg, Na, total hardness, SAR and sum of anions and cations. PH variable has a negative correlation with the variables of three-grade pastureland, woodlands and barren land. HCo3 variable were positively correlated with the variables of TDS, EC, So4, Cl, Ca, Mg, Na, total hardness, SAR and sum of anions and cations. So4 variable showed positive relationship with the variables of TDS, EC, Hco3, Cl, Ca, Mg, Na, total hardness, SAR and sum of anions and cations. Cl variable had positive association with the variables of TDS, EC, Hco3, So4, Mg, Na, total hardness, SAR and sum of anions and cations. Ca variable are positively correlated with the variables of TDS, EC, Hco3, So4, Cl, Mg, Na, total hardness, SAR and sum of anions and cations as well as is negatively correlated with third-grade forest land. Mg variable indicated positive correlation with the variables of TDS, EC, Hco3, So4, Cl, Na, total hardness, SAR and sum of anions and cations. Na variable has positive relationship with the variables of TDS, EC, Hco3, So4, Cl, Mg, Ca, total hardness, SAR and sum of anions and cations as well as is negatively correlated with residential use and first-grade pastureland. Total hardness variable were positively correlated with the variables of TDS, EC, Hco3, So4, Cl, Ca, Mg, Na, SAR and sum of anions and cations. SAR variable showed positive association with the variables of TDS, EC, Hco3, So4, Cl, Ca, Mg, Na, total hardness, sum of anions and cations as well as is negatively correlated with residential use and first-grade pastureland. Sum of anions variable were positively correlated with the variables of TDS, EC, Hco3, So4, Cl, Ca, Mg, Na, total hardness, SAR and sum of cations. Sum of cations variable had positive correlation with the variables of TDS, EC, Hco3, So4, Cl, Ca, Mg, Na, total hardness, SAR and sum of anion.
The relationship between land use and water quality
In the current study, the stepwise approach of multiple linear regressions was used for modeling the relationship between land use and river water quality. Table 4 shows the Pearson’s model implemented to study the relationship between land use and water quality.
Table 4- Pearson’s model output for relationship between land use and water quality
| IR | DF | IF1 | IF2 | IF3 | Fr | Th | Ci | Wa | Na | Ba |
|---|---|---|---|---|---|---|---|---|---|---|---|
TDS | .068 | .132 | .148 | .249 | .107 | .081 | .202 | .095 | .377 | .421 | .107 |
Ec | .054 | .136 | .207 | .255 | .104 | .068 | .203 | .096 | .362 | .431 | .104 |
PH | .182 | .477 | .183 | .142 | .013 | .153 | .010 | .162 | .460 | .211 | .013 |
Hco3 | .084 | .117 | .300 | .456 | .087 | .141 | .199 | .106 | .306 | .348 | .087 |
So4 | .089 | .140 | .160 | .362 | .103 | .208 | .247 | .064 | .287 | .337 | .103 |
Cl | .110 | .119 | .116 | .204 | .232 | .041 | .334 | .202 | .331 | .420 | .232 |
Ca | .074 | .211 | .286 | .203 | .097 | .053 | .165 | .213 | .364 | .493 | .097 |
mg | .054 | .080 | .291 | .352 | .130 | .088 | .261 | .225 | .239 | .469 | .130 |
na | .075 | .139 | .033 | .175 | .104 | .156 | .185 | .048 | .477 | .434 | .104 |
Sakhti-k | .064 | .145 | .284 | .257 | .111 | .064 | .200 | .219 | .314 | .484 | .111 |
sar | .106 | .166 | .011 | .155 | .127 | .250 | .204 | .033 | .453 | .415 | .127 |
Sum.A | .067 | .120 | .140 | .263 | .110 | .091 | .207 | .097 | .371 | .414 | .110 |
Sum.K | .068 | .131 | .144 | .251 | .105 | .083 | .198 | .092 | .383 | .441 | .105 |
According to ANOVA regression equations and correlation shown in Table 4, it is revealed that:
In the predictive model of pH, among independent variables, variables of third-grade pastureland, woodlands and barren land have been significant. The coefficient of determination (R = 0.719) showed that the mentioned variables predict 71% of pH concentration changes in sub-watersheds of study area. To check the validity of model, the Shapiro-Wilk test was used to examine the normality of residuals; the results indicated that residuals were normal in 95% confidence interval.
H = 0.002 IF3 + (-0.549) Th + 0.517 Ba +7.842
In the predictive model of cl, among independent variables, forest variable has been significant. The coefficient of determination (R = 0.718) indicated that the mentioned variable can predict 71% of cl concentration changes in sub-watersheds of study area. To confirm the validity of model, the Shapiro-Wilk test was used to examine the normality of residuals; the results indicated that residuals were normal in 95% confidence interval.
Cl =-0.079 Fr +4.799
In the predictive model of Ca, among independent variables, the forest variable has been significant. The coefficient of determination (R=0.631) demonstrated that the mentioned variable can predict 63% of Ca concentration changes in sub-watersheds of study area. To approve the validity of model, the Shapiro-Wilk test was used to examine the normality of residuals; the results indicated that residuals were normal in 95% confidence interval.
Ca=-0.068 Fr +1.916
In the predictive model of Mg, among independent variables, irrigated agriculture variable has been significant. The coefficient of determination (R = 0.627) revealed that the mentioned variable can predict 62% of Mg concentration changes in sub-watersheds of study area. To check the validity of model, the Shapiro-Wilk test was used to examine the normality of residuals; the results indicated that residuals were normal in 95% confidence interval.
Mg =-0.02 IR +1.089
In the predictive model of Na, among independent variables, first-grade pastureland and residential use variables have been significant. The coefficient of determination (R = 0.718) showed that the mentioned variables predict 71% of pH concentration changes in sub-watersheds of study area. To prove the validity of model, the Shapiro-Wilk test was used to examine the normality of residuals; the results indicated that residuals were normal in 95% confidence interval.
Na =-0.058 IF1+0.124 Ci + 2.915
In the predictive model of SAR, among independent variables, first-grade pastureland and residential use variables have been significant. The coefficient of determination (R = 0.741) found that the mentioned variables predict 74% of SAR concentration changes in sub-watersheds of study area. To determine the validity of model, the Shapiro-Wilk test was used to examine the normality of residuals; the results indicated that residuals were normal in 95% confidence interval.
SAR=-0.052 IF1+0.092 Ci + 2.294
Discussion
The results of Pearson’s correlation test at the significance level of 5% between land use and water quality variables in the Qarasu river watershed demonstrated that, among the 11 categories of land use, there are significant correlations among percentage of variables including good pastural cover, agricultural land, woodland, residential use, barren land, forest and poor pastural cover with water quality variables. More specifically, the variable of good pastural cover percentage has a direct relationship with the average concentration of water quality parameters including Na and SAR; in other words, the values of Na and SAR parameters have elevated with increasing the percentage of good pasture in the region, reducing the values of sodium and sodium adsorption ratio in downstream. The findings of the present study confirms the results of Steele and Peterson's works completed in 2011; they found that the sodium concentration has a direct relationship with the percentage of urban land use (Steele and Peterson, 2011).
The results showed a direct and negative relationship between agricultural land use and average concentration of water quality factors including Mg in areas with high percentage of agricultural land uses. The percentage of woodlands variable showed direct and reversed relationship with the average concentration of water quality parameters including pH; the woodland use in the study area has acidic pH. The percentage of residential use variable had direct relationship with the average concentration of water quality parameters including Na and SAR; this means that the values of Na and SARA variables have also enhanced with increasing the percentage of residential use. This can be due to human activities and pollutions and in some cases because of changes in soil hydrological groups in these areas. Zampella and colleagues in 2007 found that the percentage of land use in urban areas is directly related to the changes in the values of Ca, Cl, EC, pH, and Mg (Zampella et al, 2007).
Barren land variable has a direct and positive relationship with an average concentration of water quality parameters including pH; probably no certain activities such as grazing and human activities have led to alkaline pH in these areas due to calcareous combination of river bedrock. The percentage of forestland use variable has a direct subtractive relationship with the average concentration of water quality parameters including Cl and Ca; in other words, the concentrations of mentioned parameters have been reduced in downstream by increasing the percentage of forest cover in the watersheds of study area. The results in this regard are in line with the findings of Li and Zhang in 2008. They have observed the lowest concentration of anions and cations in water in areas with dominant forest cover (Li and Zhang, 2008).
Conclusion
The results obtained from the present study demonstrated that the values of water quality parameters have changed with increasing the percentage of different land uses in the Qarasu watershed. This means that the river water quality has been largely influenced toward downstream due to increased pollution load within the river and entering multiple sources of contaminants caused by tributaries join and the establishment of various residential, agricultural and industrial land uses in the region and connecting to each other along the river. The values of water quality parameters have been changed in downstream in this area of study region, indicating the effects of land use changes on surface water quality of the Qarasu River.
Ahearn DS, Sheibley RW, Dahlgren RA, Anderson M, Johnson J, Tate KW: Land use/cover influence on water quality in the last free-flowing river draining the western Sierra Nevada, California. Journal of Hydrology.2005:313:234-247.
Amiri BJ, Nakane K: Entire catchment and buffer zone approaches to modeling linkage between river water quality and land cover-A case study of Yamaguchi Prefecture, Japan. Chinese Geographical Science. 2008:18: 85-92.
Amiri BJ, Nakane K: Modeling the linkage between river water quality and landscape metrics in the Chugoku District of Japan. Water resources management. 2009:23: 931-956.
Atkinson S, Johnson D, Venables B, Slye J, Kennedy J, Dyer S, Price B, Ciarlo M, Stanton K, Sanderson H : Use of watershed factors to predict consumer surfactant risk, water quality, and habitat quality in the upper Trinity River, Texas. Science of the Total Environment.2009: 407: 4028-4037.
Carey RO, Migliaccio KW, Li Y, Schaffer B, Kiker GA, Brown MT: Land use disturbance indicators and water quality variability in the Biscayne Bay Watershed, Florida. Ecological Indicators .2011:11: 1093-1104.
Davoudi Moghaddam D, Haghizadeh A, Tahmasebipour N, Zeinivand H. Spatial and Temporal Water Quality Analysis of a Semi-Arid River for Drinking and Irrigation Purposes Using Water Quality Indices and GIS. ECOPERSIA 2021; 9 (2) :79-93
Ding J, Jiang Y, Fu L, Liu Q, Peng Q, Kang M. Impacts of Land Use on Surface Water Quality in a Subtropical River Basin: A Case Study of the Dongjiang River Basin, Southeastern China. Water. 2015; 7(8):4427-4445.
E. Fataei and S. Shiralipoor, “Evaluation of Surface Water Quality Using Cluster Analysis: A Case Study,” World Journal of Fish and Marine Sciences, Vol. 3, 2011, pp. 366-370
Gleick PH, Cooley H, Katz D, Lee E, Morrison J, Palaniyappan M, et al. The World's Water 2006-2007: the Biennial Report on Freshwater Resources. Pacific institution for studies in development, environment and security; 2007.
Kaste,O. Henriksen,A. Hindar,A.Retention of atmospherically-derived nitrogen in subcatchments of the Bjerkreim river in southwestern Norway. Ambio. 1997: 26: 296–303.
Miller JD, Schoonover JE, Williard KW, Hwang CR. Whole catchment land cover effects on water quality in the Lower Kaskaskia River watershed. Water, Air, & Soil Pollution. 2011:221: 337-350.
Nasir MFM, Samsudin MS, Mohamad I, Awaluddin MRA, Mansor MA, Juahir H, Ramli N. River Water Quality Modeling Using Combined Principle Component Analysis (PCA) and Multiple Linear Regressions (MLR): A Case Study at Klang River, Malaysia. 2011.
Pirali Zefrehei A, Fallah M, Hedayati SA. Spatial‐temporal modeling of qualitative parameters and land use status in Anzali international wetland using GIS technique. Ecopersia. 2019;7(4):223-31.
Salajegheh, A., khalighi, S., Pourebrahim, S., & Beigi, H. (2022). Investigating the trend of temporal and spatial changes in river health using Flow Health (Case Study: Qarasu River). Iranian journal of Ecohydrology, 9(1), 35-46. doi: 10.22059/ije.2022.330043.1555
Seitz NE, Westbrook CJ, Noble BF. Bringing science into river systems cumulative effects assessment practice. Environmental Impact Assessment Review. 2011:31: 172-179.
Steele M, Aitkenhead-Peterson J. Long-term sodium and chloride surface water exports from the Dallas/Fort Worth region. Science of the Total Environment. 2011:409:3021-3032.
Uriarte M, Yackulic CB, Lim Y, Arce-Nazario JA. Influence of land use on water quality in a tropical landscape: a multi-scale analysis. Landscape ecology. 2011: 26: 1151-1164.
Verma RK, Verma S, Singh A, Tiwary R, Murthy S. Relationship between land use/ land cover patterns and surface water quality in the uper Damodar river basin, JHARKHAND.2010.
Yang X, Jin W. GIS-based spatial regression and prediction of water quality in river networks: A case study in Iowa. Journal of Environmental Management. 2010:91: 1943-1951.
Zampella RA, Procopio NA, Lathrop RG, Dow CL. Relationship of Land‐Use/Land‐Cover Patterns and Surface‐Water Quality in The Mullica River Basin. JAWRA Journal of the American Water Resources Association. 2007: 43: 594-604.
Zhou T, Wu J, Peng S. Assessing the effects of landscape pattern on river water quality at multiple scales: A case study of the Dongjiang River watershed, China. Ecological Indicators. 2012: 23: 166-175.
