Further understanding the mechanisms of landscape–water interactions is of great importance to water quality management in the Xitiaoxi catchment. Pearson's correlation analysis, stepwise multiple regression and redundancy analysis were adopted in this study to investigate the relation between water quality and landscape at the sub-catchment and 200 m riparian zone scales during dry and wet seasons. Landscape was characterized by natural environmental factors, land use patterns and four selected landscape configuration metrics. The obtained results indicated that land use categories of urban and forest were dominant landscape attributes, which influenced water quality. Natural environment and landscape configuration were overwhelmed due to land management activities and hydrologic conditions. In general, the landscape of the 200 m riparian zone appeared to have slightly greater influence on water than did the sub-catchment, and water quality was slightly better explained by all landscape attributes in the wet season than in the dry season. The results suggested that management efforts aimed at maintaining and restoring river water quality should currently focus on the protection of riparian zones and the development of an updated long-term continuous data set and higher resolution digital maps to discuss the minimum width of the riparian zone necessary to protect water quality.

INTRODUCTION

With the increase of rapid economic development, surface water has received large amounts of pollutants from a variety of sources such as industrial discharge, urban wastewater and agricultural activities (Zhao et al. 2011). Although point source pollution is controlled through regulations, nonpoint source pollution is harder to control and detect since it generally encompasses large areas in drainage basins and is influenced by a variety of watershed characteristics (Liu et al. 2012).

Human activities not only affect the land use, but also alter the landscape configuration (Chattopadhyay et al. 2005; Yin et al. 2005). In recent years, land use and landscape configuration have been used as predictors of river water quality in many areas; however, there are still many conclusions that are questionable or open to dispute. For example, Uuemaa et al. (2005) indicated that edge density (ED) was negatively correlated with total nitrogen (TN) concentrations in the Estonian river. However, Lee et al. (2009) reported positive relationship between ED and TN concentrations in South Korea. The reason for the conflicting conclusion was that landscape configuration metrics depend on pixel size, topographic scale and land use classification, and it is, therefore, useless to assess the effect of landscape configuration alone on water quality regardless of the influence of land use.

Recent studies attempt to explore the relationships between water quality and landscape at different spatio-temporal scales. For example, Sliva & Willams (2001) found that the catchment landscape appeared to have slightly greater influence on water quality than did the 100 m riparian zone. Pratt & Chang (2012) suggested that, while riparian land cover does have an effect on water quality, a wider contributing area needs to be included in order to account for distant sources of pollutants. Identifying a specific landscape within a catchment and its area of influence is important because each catchment has different characteristics of geography, different responses to seasonal changes and different land uses.

The Xitiaoxi catchment, which is located in the headwater upstream of the Taihu Basin, supplies 27.7% of the water volume for Taihu Lake. Due to human activities, the land use and land cover have undergone a dramatic change, and water pollution has an important impact on the human health and social economic development. Point sources of pollution of the Xitiaoxi catchment have been effectively controlled since 1999, while nonpoint source pollution is starting to create a greater concern than point sources. Therefore, the study on the relationship between landscape and water quality is of great significance to water quality management in the Xitiaoxi catchment and to prevention of Taihu Lake eutrophication.

The objectives of this study were: (1) to find the relationship between landscape attributes and surface water quality; (2) to determine the spatial scale at which landscape mostly influences water quality; and (3) to observe seasonal differences in the relationship.

MATERIAL AND METHODS

Study area

This study was conducted in the Xitiaoxi catchment of Huzhou city, northwest of Zhejiang Province, covering more than 2,200 km2. Xitiaoxi River is the main stream of the Xitiaoxi catchment, with a length of 145 km. This catchment was chosen because land use within its component drainages ranges from forest to urban area, and about 45% of the catchment area has been planted for bamboo. The main landform types of the Xitiaoxi catchment are mountain, hill, hillock and plain. High mountainous and hilly areas are distributed in the southwest with a maximum elevation of 1,587.4 m, whereas low alluvial plains lie in the northeast. Detailed information on land use is discussed in the section ‘Land use at different spatial scales’.

Data sets

Water quality data were obtained from the local government agency in the ANJI County Ministry of Environment, which takes monthly samples of surface water and analyzes it using standard methods (Institute of Hydrogeography and Engineering Geology, MGMR 1990). Twenty water quality stations (Figure 1) were chosen because they had complete water quality records for the period: 2003–2007. Five water quality parameters were chosen to represent the state of surface water quality. Dissolved oxygen (DO) is associated with habitat quality for fish and other aquatic animals. Chemical oxygen demand (CODMn), biochemical oxygen demand (BOD5), total phosphorus (TP) and ammonium (NH4+–N) are generally considered to increase as a function of effluents from point sources or nonpoint sources such as widespread agricultural practices of bamboo planting. In order to account for the seasonal variation in precipitation and the characteristic of average monthly discharge, the water quality data were split into dry season (November–March) and wet season (April–October). More than 75% of the annual precipitation falls in the wet season. The seasonal water quality data were averaged over a 5-year period (2003–2007) in order to reduce the effects of possible field sampling and laboratory analysis errors and any missing values.

Figure 1

Sample locations and the delineation of 200 m riparian zones and sub-catchments in the Xitiaoxi catchment of the Taihu Lake Basin, China.

Figure 1

Sample locations and the delineation of 200 m riparian zones and sub-catchments in the Xitiaoxi catchment of the Taihu Lake Basin, China.

Landscape was characterized by natural environmental factors, land use patterns and four selected landscape configuration metrics. The digital elevation model (DEM) of the catchment was derived from 1:10,000 topographic data and then resized to 30 × 30 m resolution for the model input. Land use data for the study area were obtained from Landsat TM in 2008. By supervised classification and manual interpretation, the overall accuracy of land use classification was about 90% and the overall Kappa coefficient was 0.86. The land use of the catchment was subdivided into four categories: forest land use (FOR), agriculture land use (AGR), urban land use (URB) and water body. Natural environmental factors included area, the averages of slope (slope) and precipitation (PCP). These acronyms are only used in the tables and figures. The averages of slope were derived from the DEM. Precipitations of 20 sub-catchments were generated by the spatial interpolation based on the Thiessen polygon method.

Methods

Geo-information system analysis

For each water sampling station, sub-catchment boundaries were delineated with ArcView's spatial analysis using the DEM data. Riparian zone analysis was used to extract landscape attributes for a region covering a 200 m riparian buffer zone on each side of the river of each sub-catchment. Four selected landscape configuration metrics including number of patches (NP), ED, contagion index (CONTAG) and Shannon's diversity index (SHDI) were calculated at the landscape level by FRAGSTATS, which is a spatial pattern analysis program for categorical maps, based on the land use of the Xitiaoxi catchment (Table 1).

Table 1

Landscape configuration metrics were chosen in this study (McGarigal et al. 2002)

Landscape configuration metrics Description 
Number of patches (NP) NP equals the number of patches in the landscape 
Edge density (ED) Total length of all edge segments per hectare in each catchment area. ED equals 0 when there is no edge in the landscape, and ED increases when the landscape is more complex 
Contagion index (CONTAG) CONTAG approaches 0 when land use types are maximally disaggregated and interspersed and approaches 100 when all land use types are maximally aggregated 
Shannon's diversity index (SHDI) SHDI increases as the number of different patch type increases 
Landscape configuration metrics Description 
Number of patches (NP) NP equals the number of patches in the landscape 
Edge density (ED) Total length of all edge segments per hectare in each catchment area. ED equals 0 when there is no edge in the landscape, and ED increases when the landscape is more complex 
Contagion index (CONTAG) CONTAG approaches 0 when land use types are maximally disaggregated and interspersed and approaches 100 when all land use types are maximally aggregated 
Shannon's diversity index (SHDI) SHDI increases as the number of different patch type increases 

Statistical analysis

The Shapiro–Wilk test was used to test the normality of the water quality parameters and landscape attributes (Liu et al. 2012). Variables with a non-normal distribution were log-transformed before analysis. Descriptive statistics were calculated for the landscape and water quality databases. Seasonal differences were assessed using a t-test analysis (Johnson et al. 1997). Standardized coefficients of variation (CV = (SD/x) × 100%; SD is the standard deviation and x is the average of water quality parameter) for each of the seasonal water quality variables were calculated to indicate the spatial variability (Sliva & Willams 2001). Pearson's correlation analysis was used to examine the strength and significance of the relationships between single landscape attributes and separate water quality parameters (Buck et al. 2004). Stepwise multiple linear regression (R2 value) allowed us to find the landscape attributes with the strongest correlation with the water quality parameters. The above statistical analyses were performed using Statistical Product and Service Solutions. Redundancy analysis (RDA) is a multivariate direct gradient analysis method that describes variation between two multivariate data sets and is performed using the program CANOCO (canonical community ordination). The relationship between water quality, landscape attributes and sample sites were reflected on the RDA axis. RDA allows for the variance in response variables (water quality parameters) as a function of explanatory data set (landscape attributes) to be partitioned into different variable components; furthermore, the correlation of a landscape attribute with each axis indicates the strength of its relationship with the water quality. In order to avoid multicollinearity among landscape attributes, slope and NP were not included in the stepwise multiple regression and RDA analysis.

RESULTS AND DISCUSSION

Land use at different spatial scales

The dominant land use was forest, ranging from 37 to 97% (median 74.4%) in individual sub-catchments. The range of agriculture and urban area almost had the same proportion (median 12.36% and 12.37%, respectively). Land use in riparian zones was more variable than the sub-catchments. The average proportion of forest, agriculture and urban area in riparian zones were about 40%, 25% and 30%, respectively. These land use type proportions are consistent with the geography of the Xitiaoxi catchment, where high mountains in the upstream are mainly covered by forest, hilly areas distributed in the middle reaches are covered by forest and agriculture, while the flat area in the lower reaches is predominantly urban and agriculture.

Spatial and seasonal variations of water quality

Most spatial and seasonal variations in river water quality are driven by climatic and biotic factors and are, therefore, largely governed by the processes that are taking place in the terrestrial part of the watershed such as natural or human-induced vegetation cover changes (Allan 2004). Spatial and seasonal variations of water quality are illustrated in Table 2. According to the environmental quality standard for surface water of China, the maximum TP attained the fifth class standard while other water quality parameters basically kept third or fourth class water quality. DO varied significantly only with season (p < 0.001) and the concentrations of DO in a dry season were significantly higher than those of a wet season in all sub-catchments, due to the effect of temperature and biological activities. CODMn only had slightly significant variability among sub-catchments and during the seasons. The concentrations of BOD5, NH4+–N and TP were higher in the wet season than in the dry season, because there was a hydrologic connection between the catchment and the rivers during the wet season that was not present in the dry season.

Table 2

Seasonal averages, coefficients of variation (CV), and minimum (min) and maximum (max) values for water quality

  Dry season Wet season p-value 
Average (CV (%)) Min Max Average (CV (%)) Min Max 
DO (mg/L) 11.00 (7.43) 9.32 12.15 7.50 (15.13) 5.12 9.40 0.000 
CODMn (mg/L) 2.51 (35.33) 1.68 4.87 2.28 (32.38) 1.18 4.30 0.074 
BOD5 (mg/L) 2.35 (19.70) 1.63 3.10 2.46 (35.43) 1.05 4.38 0.568 
NH4+–N (mg/L) 0.22 (122.30) 0.09 1.33 0.26 (109.11) 0.07 1.41 0.019 
TP (mg/L) 0.056 (78.57) 0.02 0.20 0.080 (87.50) 0.03 0.32 0.079 
  Dry season Wet season p-value 
Average (CV (%)) Min Max Average (CV (%)) Min Max 
DO (mg/L) 11.00 (7.43) 9.32 12.15 7.50 (15.13) 5.12 9.40 0.000 
CODMn (mg/L) 2.51 (35.33) 1.68 4.87 2.28 (32.38) 1.18 4.30 0.074 
BOD5 (mg/L) 2.35 (19.70) 1.63 3.10 2.46 (35.43) 1.05 4.38 0.568 
NH4+–N (mg/L) 0.22 (122.30) 0.09 1.33 0.26 (109.11) 0.07 1.41 0.019 
TP (mg/L) 0.056 (78.57) 0.02 0.20 0.080 (87.50) 0.03 0.32 0.079 

According to RDA analysis results, the distribution of the sample sites (20 sub-catchments) was divided into three categories (A, B and C), which is shown in Figure 2. Samples from the A category were located in the places surrounded by the positive of the first axis and the positive of the second axis, especially for sub-catchment 19 with a high rate (44%) of urban area, which also showed the largest concentration of nutrients (NH4+–N, TP, CODMn, BOD5). Samples from the B category were located in places surrounded by the positive of the first axis and the negative of the second axis, which represented a high percentage of agriculture sub-catchments (e.g., sub-catchment 16, 36% agriculture), which exhibited the average values of NH4+–N and TP concentrations in both wet and dry season; samples from the C category were located in places surrounded by the negative of the first axis and the whole of the second axis, which represented a high percentage of forest sub-catchments, and water quality in these sub-catchments was good. The lowest concentrations of nutrients were in sub-catchment 2, which was nearly 100% forest.

Figure 2

Redundancy analysis results: (a) using water quality of dry season and 200 m riparian zone landscape attributes; (b) using water quality of dry season and sub-catchment landscape attributes; (c) using water quality of wet season and 200 m riparian zone landscape attributes; (d) using water quality of wet season and sub-catchment landscape attributes.

Figure 2

Redundancy analysis results: (a) using water quality of dry season and 200 m riparian zone landscape attributes; (b) using water quality of dry season and sub-catchment landscape attributes; (c) using water quality of wet season and 200 m riparian zone landscape attributes; (d) using water quality of wet season and sub-catchment landscape attributes.

Effects of landscape on water quality

Urban sources were identified as the pollution contributor, generating point and nonpoint source pollutants and increasing the imperviousness and thereby increasing runoff, while forests were proven to mitigate the effect of pollution (Xian et al. 2007). Agriculture degraded streams by increasing nonpoint inputs of pollutants, which affected riparian and stream channel habitat and altered flows (Allan 2004). The results agreed with previous findings. Tables 3 and 4 show that water quality was negatively affected by urban and agricultural activities, whereas in forested areas the water quality was better in both the 200 m riparian zones and sub-catchments. Table 5 shows that the main landscape attributes influencing NH4+–N and TP were forest in the 200 m riparian zone and urban at the sub-catchment scale.

Table 3

The relationship between water quality and landscape at 200 m riparian zone scale

Landscape attribute DO CODMn BOD5 NH4+–N TP 
Dry Wet Dry Wet Dry Wet Dry Wet Dry Wet 
Slope 0.480* 0.412 −0.411 −0.484* −0.527* −0.259 −0.503* −0.252 −0.508* −0.274 
Area −0.375 −0.270 0.116 0.010 −0.196 −0.006 0.075 −0.152 0.013 −0.134 
PCP 0.347 0.264 0.035 0.077 −0.020 0.347 −0.147 0.205 0.347 0.115 
FOR 0.451* 0.372 −0.469* −0.462* −0.579* −0.324 −0.575* −0.309 −0.556* −0.349 
AGR −0.209 −0.303 0.362 0.063 0.267 −0.092 0.378 0.079 0.203 0.066 
URB −0.182 −0.022 0.627* 0.457* 0.673* 0.620* 0.545* 0.112 0.591* 0.561* 
NP −0.491* −0.340 0.211 0.111 −0.027 0.035 0.192 −0.055 0.052 −0.061 
ED −0.023 −0.186 0.281 0.129 0.015 −0.071 0.317 0.045 0.124 −0.034 
CONTAG −0.123 0.196 0.024 0.187 0.231 0.236 0.032 0.044 0.156 0.194 
SHDI −0.364 −0.347 0.219 0.383 0.050 0.183 0.270 0.236 0.394 0.123 
Landscape attribute DO CODMn BOD5 NH4+–N TP 
Dry Wet Dry Wet Dry Wet Dry Wet Dry Wet 
Slope 0.480* 0.412 −0.411 −0.484* −0.527* −0.259 −0.503* −0.252 −0.508* −0.274 
Area −0.375 −0.270 0.116 0.010 −0.196 −0.006 0.075 −0.152 0.013 −0.134 
PCP 0.347 0.264 0.035 0.077 −0.020 0.347 −0.147 0.205 0.347 0.115 
FOR 0.451* 0.372 −0.469* −0.462* −0.579* −0.324 −0.575* −0.309 −0.556* −0.349 
AGR −0.209 −0.303 0.362 0.063 0.267 −0.092 0.378 0.079 0.203 0.066 
URB −0.182 −0.022 0.627* 0.457* 0.673* 0.620* 0.545* 0.112 0.591* 0.561* 
NP −0.491* −0.340 0.211 0.111 −0.027 0.035 0.192 −0.055 0.052 −0.061 
ED −0.023 −0.186 0.281 0.129 0.015 −0.071 0.317 0.045 0.124 −0.034 
CONTAG −0.123 0.196 0.024 0.187 0.231 0.236 0.032 0.044 0.156 0.194 
SHDI −0.364 −0.347 0.219 0.383 0.050 0.183 0.270 0.236 0.394 0.123 

*Significant at the 0.05 level.

Table 4

The relationship between water quality and landscape at sub-catchment scale

Landscape attribute DO CODMn BOD5 NH4+–N TP 
Dry Wet Dry Wet Dry Wet Dry Wet Dry Wet 
Slope 0.455* 0.437 −0.508* −0.511* −0.466* −0.173 −0.592* −0.271 −0.483* −0.341 
Area −0.381 −0.223 0.072 −0.035 −0.140 −0.089 −0.009 −0.176 −0.148 −0.177 
PCP 0.347 0.264 0.035 0.077 −0.020 0.347 −0.147 0.205 0.347 0.115 
FOR 0.563** 0.446* −0.480* −0.434 −0.571** −0.323 −0.556* −0.351 −0.570* −0.319 
AGR −0.499* −0.490* 0.425 0.303 0.392 0.085 0.503* 0.179 0.295 0.157 
URB −0.349 −0.187 0.562* 0.421 0.714** 0.422 0.526* 0.283 0.731* 0.471* 
NP −0.429 −0.354 0.119 −0.001 −0.033 −0.113 0.089 −0.148 −0.065 −0.165 
ED −0.339 −0.299 0.415 0.263 0.351 0.042 0.389 0.075 0.283 0.127 
CONTAG 0.471* 0.401 −0.481* −0.393 −0.570** −0.223 −0.494* −0.326 −0.489* −0.328 
SHDI −0.566** −0.461* 0.429 0.461* 0.545* 0.316 0.526* 0.357 0.561* 0.303 
Landscape attribute DO CODMn BOD5 NH4+–N TP 
Dry Wet Dry Wet Dry Wet Dry Wet Dry Wet 
Slope 0.455* 0.437 −0.508* −0.511* −0.466* −0.173 −0.592* −0.271 −0.483* −0.341 
Area −0.381 −0.223 0.072 −0.035 −0.140 −0.089 −0.009 −0.176 −0.148 −0.177 
PCP 0.347 0.264 0.035 0.077 −0.020 0.347 −0.147 0.205 0.347 0.115 
FOR 0.563** 0.446* −0.480* −0.434 −0.571** −0.323 −0.556* −0.351 −0.570* −0.319 
AGR −0.499* −0.490* 0.425 0.303 0.392 0.085 0.503* 0.179 0.295 0.157 
URB −0.349 −0.187 0.562* 0.421 0.714** 0.422 0.526* 0.283 0.731* 0.471* 
NP −0.429 −0.354 0.119 −0.001 −0.033 −0.113 0.089 −0.148 −0.065 −0.165 
ED −0.339 −0.299 0.415 0.263 0.351 0.042 0.389 0.075 0.283 0.127 
CONTAG 0.471* 0.401 −0.481* −0.393 −0.570** −0.223 −0.494* −0.326 −0.489* −0.328 
SHDI −0.566** −0.461* 0.429 0.461* 0.545* 0.316 0.526* 0.357 0.561* 0.303 

*Significant at the 0.05 level.

**Significant at the 0.01 level.

Table 5

Stepwise multiple regression models for water quality and landscape

  200 m riparian Regression model R2 p-value Sub-catchment Regression model R2 p-value 
Dry season DO 1.632 (FOR) + 8.428 0.246 0.026 DO 0.030 (FOR) + 8.778 0.333 0.008 
CODMn 3.573 (URB) + 0.614 0.336 0.007 CODMn 0.056 (URB) + 1.822 0.344 0.007 
BOD5 1.992 (URB) + 1.294 0.384 0.004 BOD5 0.033 (URB) + 1.944 0.443 0.001 
NH4+–N −0.783 (FOR) + 1.455 0.515 0.000 NH4+–N 0.025 (URB) − 0.089 0.747 0.000 
TP −0.102 (FOR) + 0.217 0.331 0.004 TP 0.003 (URB) + 0.020 0.389 0.003 
Wet season DOa – – – DO −0.064 (AGR) + 8.300 0.244 0.027 
CODMn −1.941 (FOR) + 5.335 0.428 0.001 CODMn 0.057 (URB) + 1.577 0.510 0.000 
BOD5 3.772 (URB) + 0.455 0.389 0.003 BOD5 0.053 (URB) + 1.802 0.324 0.009 
NH4+–N −0.752 (FOR) + 1.449 0.422 0.000 NH4+–N 0.025 (URB)–0.045 0.652 0.000 
TP −0.189 (FOR) + 0.378 0.479 0.000 TP 0.006 (URB) + 0.006 0.675 0.000 
  200 m riparian Regression model R2 p-value Sub-catchment Regression model R2 p-value 
Dry season DO 1.632 (FOR) + 8.428 0.246 0.026 DO 0.030 (FOR) + 8.778 0.333 0.008 
CODMn 3.573 (URB) + 0.614 0.336 0.007 CODMn 0.056 (URB) + 1.822 0.344 0.007 
BOD5 1.992 (URB) + 1.294 0.384 0.004 BOD5 0.033 (URB) + 1.944 0.443 0.001 
NH4+–N −0.783 (FOR) + 1.455 0.515 0.000 NH4+–N 0.025 (URB) − 0.089 0.747 0.000 
TP −0.102 (FOR) + 0.217 0.331 0.004 TP 0.003 (URB) + 0.020 0.389 0.003 
Wet season DOa – – – DO −0.064 (AGR) + 8.300 0.244 0.027 
CODMn −1.941 (FOR) + 5.335 0.428 0.001 CODMn 0.057 (URB) + 1.577 0.510 0.000 
BOD5 3.772 (URB) + 0.455 0.389 0.003 BOD5 0.053 (URB) + 1.802 0.324 0.009 
NH4+–N −0.752 (FOR) + 1.449 0.422 0.000 NH4+–N 0.025 (URB)–0.045 0.652 0.000 
TP −0.189 (FOR) + 0.378 0.479 0.000 TP 0.006 (URB) + 0.006 0.675 0.000 

aThe DO of wet season in 200 m riparian without regression models is not listed.

Slope is a fundamental parameter for predicting the rates of water flow across surfaces. Sliva & Willams (2001) indicated that increased slope was generally associated with greater pollution. Unlike previous studies, our research found that the water quality was good in the high slope sub-catchments. The reason was that the high slope sub-catchments are usually covered by forest and the influence of forest on water quality most likely masked the interaction between the slope and pollution. The area of catchment is related to the distance of pollution transportation. The farther the distance, the greater the amount of residual pollutant degradation. However, the area did not show significant effect on water quality parameters, because the main topography of this study area is mountain and the rates of water flow are relatively fast, which caused less retention and degradation of pollution.

Disentangling the landscape configuration from land use is difficult. In this study, sub-catchments which showed the landscape configuration characteristics of high SHDI and ED and low CONTAG had the poor water quality (Table 4). The reasons could be that landscape metrics are often correlated due to the nature of their calculations as there is evidence that CONTAG is inversely correlated with ED and SHDI (Table 4 and Figure 2), and the dominant land use in most sub-catchments was forest. The increase of SHDI means the number of agricultural and urban land use areas was increased and may lead to more pollution. The ED is an effective indicator of landscape heterogeneity because the division of contiguous landscape directly results in an increased edge length of patch, which caused the degradation of filter function in highly fragmented forest landscape. The increase of CONTAG means the forest has formed a good connectivity, and large or clustered forest patches are retaining more organic matter.

However, from stepwise multiple regression method results and RDA analysis, the natural environment and landscape configuration showed little influence on water quality at various temporal and spatial scales. They both were not included in the regression models and were mainly represented by the second axis of RDA diagrams, which illustrate fewer water quality parameters. The reason could be that the combination of land management activities and hydrologic conditions, especially during the wet season, may have overwhelmed the natural environment and landscape configuration.

Riparian vs sub-catchment landscape and seasonal variations

Many researchers have addressed the issue of whether the landscape of a riparian buffer zone is a better predictor of water quality than the entire catchment (Johnson et al. 1997; Sliva & Willams 2001). The results of these studies may have been influenced by how closely land use in the riparian mirrors land use throughout the catchment by data resolution, by the interplay of anthropogenic and natural gradients and by specifics of study design. These uncertainties also remain partly because each catchment and riparian zone has a unique combination of characteristics that influence water quality, and partly because thorough investigations at the catchment and riparian zone scale are extremely time and resource consuming. As shown in Table 6, total variation explained by all axes (%) showed 0.712 water quality in the 200 m riparian zone during the dry season and 0.731 in the wet season, slightly better than the sub-catchment (0.651 and 0.718, respectively). The results indicated that the 200 m riparian zone landscape characteristics appeared to have a slightly greater influence on water quality than did the sub-catchment. The reason could be that the poor-quality water reaches are where the catchment may be highly mixed in land use but there are poor-quality inflows from urban areas along the river and give a large percentage of urban land use within the 200 m riparian zone.

Table 6

RDA results showing the proportion of total variance in water quality

Season Spatial scale Total variation explained by all axes (%) Cumulative percentage of canonical variance accounted for by axes 1–4 
Axis 1 Axis 2 Axis 3 Axis 4 
Dry 200 m riparian zone 0.712 77.1 89.9 97.3 99.9 
Sub-catchment 0.651 80.3 89.7 95.8 100.0 
Wet 200 m riparian zone 0.731 88.3 97.1 99.1 99.8 
Sub-catchment 0.718 88.0 96.8 99.1 99.8 
Season Spatial scale Total variation explained by all axes (%) Cumulative percentage of canonical variance accounted for by axes 1–4 
Axis 1 Axis 2 Axis 3 Axis 4 
Dry 200 m riparian zone 0.712 77.1 89.9 97.3 99.9 
Sub-catchment 0.651 80.3 89.7 95.8 100.0 
Wet 200 m riparian zone 0.731 88.3 97.1 99.1 99.8 
Sub-catchment 0.718 88.0 96.8 99.1 99.8 

Hydrology is a strong regulator of the linkages between landscape attributes and water quality, and is subject to seasonal variations (Johnson et al. 1997; Allan 2004). Runoff from impervious surfaces and agricultural areas can more efficiently transport nutrients, sediments and contaminants, thus further degrading in-stream habitat. According to Table 6, the water quality was a little better explained by landscape attributes in the wet season than in the dry season. In the Xitiaoxi catchment, flood magnitude and frequency were increased and the changes of surface runoff and streamflow were more pronounced during the wet season (Xu et al. 2010; Zhou et al. 2013).

CONCLUSION

In general, water quality was associated with land use, natural environment and landscape configuration. The dominant landscape attributes were urban and forest land uses. The 200 m riparian zone landscape attributes appeared to have slightly greater influence on water quality than did the sub-catchment area, and the water quality was slightly better explained by landscape attributes in the wet season than in the dry season.

Management efforts aimed at maintaining and restoring stream water quality should currently focus on the protection of riparian buffer zones. However, the difficulty of achieving this goal lies in determining the minimum width of the riparian buffer zone. The width of this buffer zone is affected by the spatial variations in physical, ecological and land use conditions within the streamside areas of the catchment. Knowledge of the impact of landscape on river water quality is useful and should be taken into account in planning and managing landscape in catchments. An updated long-term continuous data set and higher resolution digital maps are recommended to gain a better understanding of the mechanisms of land–water processes in riparian buffer zones.

ACKNOWLEDGEMENTS

This research work is financially supported by National Natural Science Foundation of China (Grant Nos 41371046, 41401035), Commonwealth and Specialized Program for Scientific Research, Ministry of Water Resources of China (Grant Nos 201201072, 201301075), Natural Science Fund for Colleges and Universities in Jiangsu Province (Grant No. 14KJB170021), Open Project of Key Laboratory for Ecology and Pollution Control of Coastal Wetlands (Environmental Protection Department of Jiangsu Province) (Grant No. KLCW1205) and Jiangsu Province Ordinary University Graduate Student Research Innovation Project (Grant No. CXLX11_0022). The authors would also like to express appreciations to colleagues in our laboratory for their valuable comments and other help.

REFERENCES

REFERENCES
Allan
J. D.
2004
Landscapes and riverscapes: the influence of land use on stream ecosystems
.
Annual Review of Ecology, Evolution, and Systematics
35
,
257
284
.
Chattopadhyay
S.
Rani
L. A.
Sangeetha
P. V.
2005
Water quality variations as linked to land use pattern: A case study in Chalakudy river basin, Kerala
.
Current Science
89
,
2163
2169
.
Institute of Hydrogeography and Engineering Geology, MGMR
1990
Water Quality Analysis
.
Geology Press
,
Beijing
(
in Chinese)
.
Johnson
L. B.
Richards
C.
Host
G. E.
1997
Landscape influences on water chemistry in Midwestern stream ecosystems
.
Freshwater Biology
37
,
193
208
.
McGarigal
K.
Cushman
S. A.
Neel
M. C.
Ene
E.
2002
FRAGSTATS, Spatial Pattern Analysis Program for Categorical Maps. http://www.umass.edu/landeco/research/fragstats/fragstats.html
.
Xu
Y. P.
Xu
J. T.
Ding
J. J.
Chen
Y.
Yin
Y. X.
Zhang
X. Q.
2010
Impacts of urbanization on hydrology in the Yangtze River Delta, China
.
Water Science and Technology
62
,
1221
1229
.
Zhao
G. J.
Gao
J. F.
Tian
P.
Tian
K.
Ni
G. H.
2011
Spatial-temporal characteristics of surface water quality in the Taihu Basin, China
.
Environmental Earth Sciences
64
,
809
819
.