1 Introduction

A river’s water quality is a comprehensive reflection of the multiscale landscape structure (Xiao et al., 2016). Landscape structures include landscape configuration (spatial distribution of landscape patches) (Wu & Lu, 2019a) and landscape pattern (land-use structure) (Shrestha et al., 2018). The interaction between these two factors affects the hydrological process and ecological environment of the river basin. Further, it affects the migration and transformation of pollutants at the surface, which is also a fundamental reason for the spatial difference in river water quality (Zhang, Liu, et al., 2018). Land-use patterns can reflect human activities and is the most crucial factor affecting nonpoint source pollution (Alvarez-Cabria et al., 2016; Sohail et al., 2019; Xiao & Ji, 2007). Soil erosion, urban expansion, excessive use of pesticide, and chemical fertilizer can be deemed responsible for non-point source pollution (dos Santos & Daniel, 2020; Wu & Lu, 2019b). However, many studies have shown that unreasonable use of land is the key to deteriorating water quality (Lu et al., 2021a; Ullah et al., 2018). Moreover, improper farmland management leads to soil erosion and excessive nitrogen and phosphorus loss with surface runoff, resulting in a large area of nonpoint source pollution of surface water. Nitrogen and phosphorus discharged from nonpoint pollution sources are the main causes of water eutrophication (Gushgari et al., 2017). Controlling nonpoint source pollution by changing land-use patterns can develop effective land management measures and improve water quality (Beck et al., 2016; Chen et al., 2020; Ullah et al., 2018). The landscape index (landscape patch) is a simple quantitative index that concentrates landscape pattern information and reflects some aspects of its structural composition and spatial configuration (Wang et al., 2020; Zhou & Li, 2015). As the landscape index can quantitatively express the degree of correlation between the landscape pattern and ecological processes, many scholars prefer to study the relationship of landscape pattern with water quality using the landscape index (Xu et al., 2019; Zhang et al., 2019; Zhang, Liu, et al., 2018).

In the past decade, scholars have studied the impact of landscape structure on water quality at different scales (riparian buffer zone, circular buffer zone, and sub-basin) (Khan et al., 2020; Li et al., 2009; Love & Nejadhashemi, 2011; Lv et al., 2015b; Wang & Zhang, 2018). However, it is still controversial what scale of landscape structure has a greater impact on water quality (Li et al., 2009). By comparing the relationship between landscape structure and water quality at different scales, identifying the best spatial scale of landscape structure affecting water quality in the basin is of great significance in adjusting the land-use structure, optimizing the landscape pattern, and improving water quality in the basin. The increasing concentration of pollutants, such as nutrients and heavy metals, is the primary reason for water-quality deterioration (Chen et al., 2020; Shrestha et al., 2018; Shu et al., 2020). Studies have shown that land use types has a significant impact on the concentration of nutrients in water (Chen et al., 2009; Wang et al., 2014). So, some scholars began studying the impact of combining the landscape type and configuration on water quality (Lv et al., 2015a; Uriarte et al., 2011; Zhang et al., 2021). However, this type of research regarding water quality as a whole without distinguishing indicators, such as nutrients and heavy metals. Thus, ignoring the differences in their source and responses to the landscape structure result in the poor interpretation of water quality by landscape structures and low result accuracy.

To improve the accuracy of the interpretation rate of water quality by landscape structure, the Xiangjiang River Basin with less human disturbance was selected as the research object, and water quality indexes were divided into two categories: heavy metals and conventional water quality indicators, from the perspective of the wet season and the dry season. The effects of landscape structure on nutrients and heavy metals at different spatial scales were analyzed, and the optimal spatial scale of landscape structures affecting the changes in nutrients and heavy metals in water was identified. The research conclusions provide practical suggestions or measures for watershed management.

2 Materials and Methods

The methodological framework of this research is shown in Fig. 1.

Fig. 1
figure 1

The methodological framework of the study

Full size image

The methodological framework describes the following:

Water quality monitoring: detection of heavy metals content and conventional water quality indicators content during wet season and dry season.

Land-use data: download and process land-use data in the Xiangjiang River Basin in 2020.

Research scale division, landscape structure statistics at corresponding scales: 11 research scales of sub-basin, riparian buffer zones, and circular buffer zones. Calculate the land-use structure at the corresponding scale in Arcgis 10.3 software and calculate the landscape configuration at the corresponding scale in Fragstats 4.2 software.

Data analysis and screening: first step, the heavy metals and conventional indicators were condensed into a main component, respectively. Second step, water quality data and landscape structure data were analyzed using Pearson’s correlations, co-linearity diagnosis of landscape structure data.

Redundancy analysis: redundancy analysis (RDA) was conducted to calculate the contribution rate of landscape structures to nutrient and heavy metal indexes at corresponding scales.

2.1 Summary of the Study Area and Scale Division

The Xiangjiang River is a tributary of the Gongjiang River in the upper reaches of the Ganjiang River (a source of the Gongjiang River) (Xiong et al., 2019), emanating from Xunwu County, Ganzhou City, Jiangxi Province. It flows through Xunwu County and Huichang County, joining into the Gongjiang River in the northeast of Huichang County (Fig. 2). The total length of the Xiangjiang River is 103.4 km, and the basin area is 2049.3 km2. The basin is mainly in Huichang County, Ganzhou City, with many tributaries and abundant water, and is an important water source for Huichang County. The basin’s upper reaches belong to the middle–low mountainous areas with good vegetation, surrounded by green mountains, and the middle and lower reaches belong to the hilly region. There are two hydrological stations and 16 rainfall stations along the Xiangjiang River basin, and receives an annual average precipitation of 1560 mm, and an annual average water production of 1.57 billion cubic meters. The major stream of the Xiangjiang River is about 215 m wide in the rainy season and 35 m wide in the dry season. The water depth is about 3 m in the rainy season and 0.6 m in the dry season.

Fig. 2
figure 2

Spatial scale division of the study area and distribution of sampling sites

Full size image

The study area is divided into ten sub-watersheds through ArcSWAT (Fig. 2a). Six sampling points were set up in the major stream of the Xiangjiang River (X1–X6), and ten sampling points were set up on the tributary (S1–S10), and the basins upstream the sampling points were divided into 16 sub-basins. Based on the sampling points, the riparian (Fig. 2b) and circular buffer zones (Fig. 2c) were made at scales of 100 m, 200 m, 300 m, 500 m, and 1000 m (Table 1).

Table 1 The geographical coordinates of the sampling sites
Full size table

2.2 Water-Quality Data

Water samples were collected at a water depth of 0.5 m on the river’s surface for testing in January, June, July, and December 2020. Dissolved oxygen (DO) and electrical conductivity (EC) were measured on-site using a portable water-quality analyzer (SL1000, HACA, USA), and turbidity was measured using a portable turbidity meter (2100Q, HACA, USA). After the water samples were collected, they were filtered through a 0.45-μm filter membrane, poured into a polyethylene bottle soaked in acid for a long time, and immediately sent to a laboratory for analysis. The nutrient factors detected included CODMn (potassium permanganate method), NH3–N (salicylic acid spectrophotometry), TN (ammonium molybdate spectrophotometry), and TP (alkaline potassium persulfate digestion ultraviolet spectrophotometry). Heavy metals (Cu, Zn, Cd, and Pb) were determined using atomic absorption spectrophotometry (chelating extraction method), and the detection limits of each element were Cu 0.1 μg/L, Zn 0.1 μg/L, Cd 0.02 μg/L, and Pb 0.05 μg/L. The heavy metal Cr concentration was determined using diphenylcarbazide spectrophotometry, with a detection limit of 0.02 μg/L.

2.3 Landscape Structure Data

Land-use data can represent the characteristics of a landscape structure (Yan et al., 2021). Land-use data in the Xiangjiang River Basin in 2020 was collected from the website of National Geomatics Center of China (http://www.globallandcover.com/). The data source is Landsat 8 with an accuracy of 30 m. The original data were subdivided into secondary land types. In this paper, the attribute fields were merged into the first-level land categories by ArcGIS, which were divided further into cropland (Cro), forest (For), grassland (Gra), water (Wat), settlement (Set), and others (Oth) (Fig. 2). Among them, the overall proportion of others in the cropped grids of each research scale in the Xiangjiang River Basin was less than 0.1%, so this category was eliminated in the follow-up analysis. The landscape configuration characteristics calculated in this study mainly reflect the impact on water quality by characterizing the connectivity, shape change, and fragmentation of land use in the basin. As the first-level classification of land use was utilized in this study, the landscape index at the same level reflects the similar ecological process of land use, and the results may cover up the real ecological and hydrological effects. Therefore, this paper calculates 12 landscape index quantitative landscape configuration features (Table 2); landscape index calculations were conducted using Fragstats v.4.2.

Table 2 Selected landscape index and meaning
Full size table

2.4 Data Analyses

DO, EC, turbidity, TN, NH3–N, CODMn, and TP were conventional water quality index, whereas Cu, Zn, Cd, Cr, and Pb were heavy metal indexes. According to the hydrological characteristics of Jiangxi Province, January, February, and December are dry seasons. June, July, August, and September are rainy seasons. Other months are normal seasons. The average value of 2 months in the same water period is taken as the water quality data of the water period.

The water quality data is standardized by the Z-score standardization method. One principal component was extracted from heavy metals and conventional water quality index in different seasonal periods through principal component analysis, respectively. The cumulative interpretation rate was above 70%. The new extracted factors were taken as dependent variables, whereas land-use data and landscape index were independent variables. The main steps of the analysis are as follows: ① the dependent and independent variables were analyzed using Pearson’s correlations, the independent variable data with correlation < 0.3 were eliminated; ② to avoid the high collinearity between remaining independent variables, multiple linear regression was performed between the dependent variable and the remaining independent variables, the independent variables with collinearity value VIF < 5 were eliminated (Liu & Yang, 2018), the remaining independent variable data were the best combination of landscape structures; ③ the nutrient and heavy metal indexes of different water periods were response variables, whereas the best combination of landscape structures was an explanatory variable. Redundancy analysis (RDA) was conducted to calculate the contribution rate of landscape structures to nutrient and heavy metal indexes at corresponding scales. The data were analyzed using SPSS v.21 and Canoco v.5.

3 Results and Discussion

3.1 Variation Characteristics of Water-Quality Indexes in Different Water Periods

The water quality of the Xiangjiang River Basin in different seasonal periods is shown in Fig. 3. Analyses showed that, overall, the water quality of the Xiangjiang River Basin met the Class III “Surface Water Environmental Quality Standards.” (Ministry of Ecology and Environment the People’s Republic of China 2002). In contrast, the concentrations of heavy metals in different seasonal periods met the Class II standards. Among the conventional water quality indicators, the average concentrations of DO, EC, NH3–N, and TN in the rainy season were lower than those in the dry season. The average concentrations of turbidity, CODMn, and TP in the dry season were higher than those in the rainy season; the average concentration of heavy metals in the rainy season was lower than that in the dry season. From the spatial scale perspective, there was little difference in the concentration of heavy metals at each point during the same period. The conventional indicators in the tributaries indicated good water quality, meeting the Class II standards. The closer to the major stream of the Xiangjiang River, the worse the water quality, especially NH3–N and TN, which only met the Class IV standards at X5 and X6, showing that nitrogen nutrients mainly pollute the Xiangjiang River.

Fig. 3
figure 3

Water environment in different water periods of the Xiangjiang River Basin. Grey represents the concentration of pollutant at each sampling point. The position of the red line represents the median level of the corresponding pollutant concentration. The blue lines represent 1/4 and 3/4 quantiles of the corresponding pollutant concentration, respectively

Full size image

3.2 Change Characteristics of Land-Use Structure and Landscape Index

The characteristics of land-use structure vary greatly at different spatial scales (Fig. 4). Forest, grasslands, and croplands dominate sub-watersheds. In the riparian and circular buffer zones of the Xiangjiang River, the proportions of the forest, grassland, and cropland are large. The major stream of Xiangjiang River is dominated by croplands in the circular buffer zone. Further downstream, the larger the proportion of croplands, the greater the increase in the circular buffer distance; also, the proportion of the settlement area was the largest at the 1000-m circular buffer scale. At the riparian buffer zone, the proportion of croplands in the major stream of the Xiangjiang River from the upper to the lower reaches decreases gradually, and the proportion of grasslands increases gradually. This is because the closer the Xiangjiang River is to the downstream, the closer it is to Huichang County (X6), and the greater the population density and the higher the urban greening rate. The land structure of the Xiangjiang River Basin is dominated by forest, grassland, and cropland, together accounting for more than 97% of its structure. The difference in land-use structure is the largest under the circular buffer zone.

Fig. 4
figure 4

Characteristics of land-use structure at different spatial scales

Full size image

Patch density (PD), edge density (ED), patch richness density (PRD), interspersion and juxtaposition index, and Shannon’s evenness index (SHEI) at the sub-basin are smaller than those in the buffer zone; moreover, the landscape shape index (LSI), contagion index (CONTAG), percentage of landscape (PLADJ), and patch cohesion index (COHESION) are higher than those in the buffer zone (Fig. 5). This shows that landscape connectivity is better at the sub-basin scale, and most of them are distributed evenly and in regular-shaped patches, the degree of landscape fragmentation is lower (Martin et al., 2008; Zhang, Chen, et al., 2018). The Xiangjiang River is a tributary of the Gongjiang River in the upper reaches of the Ganjiang River (a source of the Gongjiang River), with good vegetation, surrounded by green mountains, less human disturbance to the degree of land use in watershed, which also shows that the landscape is more stable at the sub-basin scale in terms of natural lands, and the ability to resist external disturbance is stronger (Ewane, 2020). The largest patch index (LPI) index in the circular buffer zone and sub-basin scale is significantly higher than that in the riparian buffer zone, and the LPI index is higher when the circular buffer distance is 100 m, 200 m, and 300 m. This shows that the circular buffer zone is dominated by single homogeneous large patches, showing the fragmentation distribution of single land use (Ou et al., 2016). Furthermore, the Shannon’s diversity index (SHDI) and Euclidean nearest neighbor index-mean (ENN_MN) were the highest at the riparian buffer scale, indicating that the land use was abundant and landscape fragmentation was severe at this scale, showing a multi-landscape interactive distribution structure (Yue et al., 2008).

Fig. 5
figure 5

Variation characteristics of the landscape index in Xiangjiang River Basin at different spatial scales

Full size image

3.3 Landscape Structure Index Selection and Core Index Determination

According to the correlation and collinearity diagnosis above (Table 3 and Table 4), PLADJ, ED, LSI, and SHDI were eliminated, and the remaining 13 indicators were used for follow-up analyses (five land-use type indicators and eight landscape index indicators). For redundant analysis of the remaining data and water-quality indicators, indicators with higher scores on the first three axes were selected as the main landscape variables. The results showed that settlement (Set), cropland (Cro), PRD, grassland (Gra), and COHESION were the main landscape structure indexes affecting the conventional indicators content in water (Table 5). As the main variable, the frequency reached 65.15%. ENN_MN, grassland (Gra), PD, LPI, and SHEI were the main landscape structure indicators affecting the heavy metal content in water; as the main variable, the frequency reached 66.67%, showing that the impact of landscape structure indicators on different components of water quality is different.

Table 3 Pearson correlation between water quality indicators and landscape indicators in different seasonal periods
Full size table
Table 4 The collinear coefficient of landscape structure at different scales
Full size table
Table 5 Main landscape indicators selected based on redundancy analysis (RDA)
Full size table

Studies have shown that cropland is the main pollution source of nutrients in water (Kibena et al., 2014); nutrients that are not absorbed in croplands are mainly adsorbed on the surface layer of the soil (Wu & Lu, 2019a), resulting in the accumulation of nutrients in water during drainage. The settlement includes pollution sources, such as urban land, rural land, industrial and mining land, and transportation land, that contribute to the output of pollutants (Carter & Dzialowski, 2012). They affect water quality directly through point source pollution and increase the proportion of impervious surfaces. The interception and absorption of pollutants and the natural infiltration of soil are affected by the lack of natural vegetation during the rainstorm period (Amin et al., 2013). This results in many pollutants passing through the impermeable ground into the river in a short time. Affected by the large-scale development of cities, settlement is concentrated and connected; domestic industrial wastewater is often concentrated and discharged, resulting in an increased water pollution load. This study found that PRD and COHESION are also the main landscape variables affecting nutrients in water. PRD represents the abundance of patches per unit area (Bateni et al., 2013), that is, the diversity of patch types. The more ample the patches, the more stable the landscape. COHESION represents the degree of landscape aggregation, which reflects the degree of connectivity of the natural landscape (Wu & Lu, 2019b). Under the influence of low COHESION, land use types such as grassland which could purify water show dispersed and inconsistent patch distribution, which cannot effectively purify water quality; as water-quality degradation usually occurs in highly fragmented and low-dispersion landscapes (Amiri & Nakane, 2008), this type of landscape structure facilitates the transport of pollutants in the soil surface, leading to long-distance diffusion and the concentrated transport of nutrients. As the main “sink” landscape type, grasslands purify water by absorbing and fixing the pollutants passed by runoff (Martinez et al., 2009). In conclusion, the risk of water-quality deterioration can be reduced effectively by accelerating the scale cultivation of farmland in the Xiangjiang River Basin, enhancing the degree of farmland aggregation in the basin, improving the utilization rate of fertilizer, controlling urban expansion, and improving the urban greening rate.

In this study, ENN_MN, grassland (Gra), PD, LPI, and SHEI were the main landscape structure indexes affecting the content of heavy metals in water. The ENN_MN index reflects the probability of passing through a variety of landscape types in the confluence process (Wu & Lu, 2019b), and its value determines the ecological characteristics of the dominant species in the landscape, the abundance of internal species richness, etc. LPI reflects the proportion of the entire landscape area occupied by the largest patch in a certain patch type(Li et al., 2015). The change in its value depicts the intensity and frequency of human interference, and reflects the direction and strength of human activities. SHEI reflects that the landscape is dominated by one or a few dominant patch types. The closer the SHEI value is to 1, the lower the superiority degree, indicating that there is no obvious advantage type in the landscape and each patch type is evenly distributed in the landscape. The study area is in a mountainous area, and industrial development is backward. The heavy metal content in the middle and upper reaches of the Xiangjiang River is extremely low. Human activities have little effect on the degree of land use. The heavy metal content in the lower reaches of the Xiangjiang River rose slightly, mainly from the concentrated discharge of domestic sewage and industrial wastewater. Downstream belongs to population concentrated areas, indicating that human activities are the main source of heavy metals in water. The changes of ENN_MN, LPI, and SHEI values can reflect the degree of change to natural land caused by human activities, and therefore serve as the main landscape structure indexes affecting the heavy metal content in water. It is hard to analyze why grassland is a landscape index affecting heavy metal content in this study, which may not consider the difference of soil properties.

*Significantly correlated at the 0.05 level (bilateral).

**Significantly correlated at the 0.01 level (bilateral).

3.4 Spatial Scale of Landscape Structure Affecting Water Quality

Table 6 shows the explanation rate of landscape structure on conventional indicators and heavy metal changes in water at different spatial scales. The results showed that the average interpretation rate of landscape structures on the conventional indicators content in water was the highest (93.6%) in the 300-m riparian buffer zone, this is similar to the conclusions of existing studies (Sadeghi et al., 2020). This is because riparian zones, as a transitional landform connecting water and highlands, can influence the nutrient content and chemical composition transported to the river. Pollutants can be transported for a long distance and are affected by the conditions along the riverbank. Studies have shown that there is a good correlation between land use type and nutrients in the 100-m buffer zone of riparian zone (Li et al., 2009). However, they did not conduct multi scale research and used only land-use type as the main landscape indicator, and only analyzed the correlation between land use type and water quality in the 100-m buffer zone of riparian zone at nine sampling sites, which is also the primary reason for the gap between the two conclusions. Cropland in the Xiangjiang River Basin is concentrated mainly in the riparian zone, and the 200–300-m scale buffer zone is affected most strongly by human activities, a high degree of patch fragmentation, and poor aggregation intensity (Fig. 5). This extends the retention time of runoff on the surface of farmland (Shehab et al., 2021), enriches the nutrients in the runoff, forms the input of nutrients to the river, and accelerates the deterioration of water quality.

Table 6 Interpretation rate of landscape structure on water quality at different spatial scales based on redundancy analysis (RDA)
Full size table

Table 6 shows in the 200-m and 300-m riparian buffer zones, the average interpretation rates of heavy metal content were relatively strong (93.8% and 92.7%, respectively). Generally speaking, the discharge of industrial wastewater from chemical and mining industries is the primary source of heavy metal pollution (Adeola Fashae et al., 2019); therefore, heavy metal pollution occurs mainly from point sources. Furthermore, the landscape structure of the circular buffer zone should have a higher interpretation rate for heavy metals. The main reason for the difference between the two conclusions is that ① Xiangjiang River is located at the source of the Ganjiang River, Huichang County, is in a mountainous area, and industrial development is backward, the heavy metal content is low; ② the main industry in Huichang County is the cultivation of navel oranges, which requires a large amount of phosphate fertilizer. Generally speaking, phosphate fertilizers contain heavy metals such as Hg, Cd, As, Zn, and Pb, heavy metal pollution mainly comes from non-point source pollution. Therefore, the landscape structure at riparian buffer zones has a higher interpretation rate for heavy metal indicators than circular buffer zones.

At the sub-basin scale, the interpretation rate of landscape structures on conventional water quality indicators and heavy metals were relatively low (70.9% and 86.1%, respectively). The reason is that compared to the sub-basin scale, ① the landscape structure at riparian buffer zone and circular buffer zone can better regulate the nutrient content and chemical composition transported to the river, thus can more effectively reduce the pollutant content in water; ② landscape structure at buffer zones can better reflect the migration of pollutants around the sampling points, which makes the correlation between water quality and landscape structure more remarkable at the buffer zone scale.

Overall, under the two spatial scales of the circular and riparian buffer zones, the interpretation rate of landscape structures on conventional indicators was higher in the rainy season than in the dry season, and there was no significant difference in the interpretation rate of landscape structures on heavy metals in the rainy and dry seasons. The rainy season is the rice heading period, which is the peak period of demand for fertilizer. During crop fertilization, the risk of nutrient loss, such as nitrogen and phosphorus, is increased. Additionally, in the rainy season, the intensity of surface runoff is high, and the loss of nonpoint sources is enhanced (Li et al., 2009), which may lead to a high concentration of nutrients, such as nitrogen and phosphorus. Farmland is fallow during the dry season (Li et al., 2009); nutrients in water come mainly from the discharge of domestic sewage; the total amount of nutrients decreases; the interference ability of the landscape structure is weakened (Li et al., 2018). With an increase in the buffer scale, the more complex the landscape structure is, the more distant the riparian zone is from the river, and the weaker the regulation ability is.

The sampling points in this paper are divided into two categories. One is located upstream of the settlements and most houses uniform landscapes like forest or grasslands (such as sampling points like S1, S2, S3, S6, S7, S9, S10, and X4), and the other is located upstream and downstream of settlements along the tributaries(such as sampling points like S4, S5, S8, X1, X2, X3, X5, and X6). The changes of landscape indicators under different buffer zone scales are analyzed in two groups below: for sampling points located upstream of the settlements, PD reflects the degree of landscape fragmentation. The greater the value, the higher the degree of fragmentation (Li et al., 2015). The PD value decreases with the increase of the buffer scale, indicating that the disturbance of human activities is mainly concentrated near the river bank. This is because human activities have little effect on the degree of land use in the Xiangjiang River Basin, and the riparian zone is the concentrated area of human activities, which is consistent with the above conclusions. With the increase of buffer distance, the CONTAG value increases gradually, indicating that the degree of landscape fragmentation decreases gradually. The greater the buffer distance, the better the connectivity of some dominant patch type. As the buffer scale increased, both SHDI and SHEI values decreased, indicating that the dominance degree of a certain landscape increased gradually. The reason is that the larger the research scale, the greater the proportion of certain types of land use (forest, cropland), and thus the more obvious the dominance of the landscape; For sampling points located upstream and downstream of settlements along the tributaries, the trends of PD values are similar to the sampling points upstream of the settlements. With the increase of riparian buffer distance, the trends of CONTAG, SHDI, and SHEI values are similar to the sampling points upstream of the settlements. With the increase of the circular buffer distance, the variation trends of CONTAG, SHDI, and SHEI values are not obvious, which remains to be further studied. At each buffer scale, the closer to the lower reaches of the Xiangjiang River, the greater the IJJJ value (IJJJ value at point S1 is between 10 and 30, and greater than 90 at point S6. It shows that the upstream block type is only adjacent to a few other types, and the downstream development intensity is relatively large, which is consistent with the development status of the study area.

In conclusion, there are significant differences in the interpretation ability of riparian and circular buffer zones to water-quality changes in the different seasons. For different water pollution types, two division methods help analyze the impact of landscape structure on water quality.

3.5 Impact of Landscape Structure on Water Quality

RDA was conducted on the landscape structure and water-quality indicators, with high interpretation rates of the two pollutant types mentioned above. In the 300-m riparian buffer zone, the first characteristic axes in the rainy and dry seasons were 0.613 and 0.8194, respectively. The correlation coefficients between landscape structure and conventional indicators ordination axes were 0.9995 and 0.9547, respectively. Under a scale of the 200-m riparian buffer zone, the first characteristic axes in the rainy and dry seasons were 0.6994 and 0.7187, respectively, and the correlation coefficients between landscape structure and heavy metal ordination axes were 0.981 and 0.9828, respectively (Table 7). This shows that the two-scale ordination axes can better reflect the correlation between landscape structure and water-quality indicators.

Table 7 Redundancy analysis of the variance interpretation rate of each sort of axis
Full size table

A sequencing graph of the impact of landscape structure on water quality at the corresponding scale was drawn according to the RDA results (Fig. 6). NH3–N, CODMn, TP, and EC were correlated positively with cropland, settlement, and COHESION, whereas they were correlated negatively with grassland in the rainy season. Settlement has a great impact on NH3–N, CODMn, TP, indicating that settlement is the main output source of these pollutants. COHESION has little effect on NH3–N, CODMn, TP. This is due to the fact that the natural land in the study area remains better and the landscape connectivity well. DO correlated positively with grassland and correlated negatively with settlement, water, and PRD, indicating that the stronger the human interference, the heavier the water pollution (Griffith, 2002). All conventional indicators correlated negatively with ENN_MN. In addition, the length of the ENN_MN arrow is relatively short, indicating that ENN_MN has little effect on nutrient index.

Fig. 6
figure 6

Redundancy analysis between landscape structure and water quality. The arrow length of each landscape structure indicates the degree of influence on the water-quality index. The angle between landscape structure and water-quality indicators indicates the correlation. If the included angle is less than 90°, it indicates a positive correlation, greater than 90° = negative correlation, and close to or equal to 90° = no correlation

Full size image

In the dry season, DO correlated negatively with PD, SHDI, grassland, and settlement. The negative correlation between SHDI and DO is strong, and the positive correlation between SHDI and TN and NH3–N is strong. This is due to the high degree of land fragmentation under the buffer scale of 300 m riparian zone and the strong interference of human activities on land use, which is similar to the existing research conclusions (de Mello et al., 2020; Tahiru et al., 2020). The correlation between SHDI and TP, CODMn is extremely weak, and the specific reasons need to be further demonstrated. Other conventional indicators correlated positively with settlement. The positive correlation between CODMN and settlement in the dry season is not as strong as that in the rainy season, because the main source of pollution in the study area comes from domestic sewage discharge, while the residential water consumption is larger in the rainy season. TN and EC were correlated positively with PD, SHDI, SHEI, and grassland. Among them, TN and PD are slightly correlated positively with PD, indicating that the greater the heterogeneity of the landscape per unit area, the greater the landscape fragmentation, and the heavier the TN pollution. TN slightly correlated negatively with grassland. This is because the vegetation in the study area is good, water and soil erosion is not easy to occur. Pollutants are easily absorbed and intercepted after passing through the grassland, which is consistent with existing research conclusions (Lu et al., 2021b). Turbidity correlated positively with cropland and settlement, but correlated weakly with most other landscape structure indicators. Cropland and settlement are the main output sources of nutrient pollutants (Banadda et al., 2009); large amounts of chemical fertilizers and pesticides are needed in croplands, which may cause water pollution. Grasslands play a positive role in protecting water quality, probably because grasslands can effectively retain and absorb pollution elements, indirectly purifying water quality (Singh et al., 2017).

In the rainy season, Cu, Cd, Zn, and Cr correlated positively with settlement, PD, SHEI, and IJI, and correlated negatively with water, grassland, cropland, COHESION, PRD, CONTAG, and LPI. This shows that heavy metal pollution sources in the rainy season are mainly from the settlement. The more the landscape patches per unit area, the lower the connectivity between landscapes, the weaker the ability of landscapes to resist external interference, and the more the heavy metal pollution (Li et al., 2018; Lv et al., 2015b; Yue et al., 2008). Cu, Cd, Zn, and Cr correlated positively with IJI, SHEI during the rainy season, and the correlation is strong. The reason is that IJI reflects the degree of adjacency between a certain patch type and other patch types (Zhang, Liu, et al., 2018), and the larger the value, the more patch types, and the greater human interference. The lower SHEI value indicates that there is no obvious dominant type in the landscape (Huang et al., 2016), and each patch type is evenly distributed in the landscape. The original land types in Xiangjiang River Basin are relatively single (mainly cropland, grassland, and forest). The diversity of patch types shows that human beings have changed the degree of land use greatly, which is consistent with the above conclusions.

In the dry season, Cu, Cd, Zn, and Cr correlated positively with cropland, water, PD, and SHEI. This shows that the heavy metal pollutants in the dry season come mainly from croplands. The larger the PD, the more landscape patches per unit area (Zhang, Liu, et al., 2018), and the more changes human activities have made to the land use. The larger the SHEI, indicates that there are no evident dominant types in the landscape, and the patch types are evenly distributed in the landscape, that is, the lower the landscape connection, the heavier heavy metal pollution. This is consistent with the above research conclusions on the relationship between heavy metals and landscape indexes during the rainy season. Pb correlated negatively with IJJ and grassland, and the proportion of grassland and IJJJ value have great influence on Pb. Generally speaking, although heavy metals cannot be degraded, grassland should have intercept effect on heavy metals (Lu et al., 2021b). However, the conclusion that Pb is negatively correlated with grassland in this study is still difficult to explain and needs further demonstration. The correlation between the area of water and water-quality indicators showed different directions in the rainy and dry seasons. Further, the correlation between landscape structure and heavy metals was different in the wet and dry seasons, further explaining that the two classification methods should be used to determine the impact of landscape structures on water quality.

4 Conclusion

  • (1) The water quality of the Xiangjiang River tributary is better than that of the major stream, and the closer the Xiangjiang River major stream is to the downstream, the worse the water quality is. However, the water quality of the Xiangjiang River Basin met the Class III standards.

  • (2) The land structure of the Xiangjiang River Basin is dominated by forests, grasslands, and croplands, accounting for more than 97%. The difference in land-use structure is the largest in the circular buffer zone.

  • (3) Settlement, croplands, PRD, grasslands, and COHESION were the main landscape structure indexes affecting the nutrient content in the water. Alternatively, ENN_MN, grassland, PD, LPI, and SHEI are the main landscape structure indicators affecting the heavy metal content in water.

  • (4) At the sub-basin scale, the interpretation rate of landscape structures on conventional water quality indicators heavy metals are 70.9% and 86.1%, respectively.

  • (5) The average interpretation rate of landscape structure on the nutrient content in water was the highest (93.6%) in the 300-m riparian buffer zone. In the 200-m and 300-m riparian buffer zones, the average interpretation rates of heavy metal content were relatively strong (93.8% and 92.7%, respectively).

  • (6) Cropland and settlement are the main output sources of nutrient pollutants; also, grasslands protect water quality. Heavy metals in the Xiangjiang River come mainly from croplands in the dry season and settlements in the rainy season.

5 Conflict of Interest

The authors declare that they have no conflict of interest.