Impacts of land use and landscape pattern on water quality at multiple spatial scales in a subtropical large river
Funding information: Ordinary University Characteristic Innovation Project of Guangdong Province, China, Grant/Award Number: 2020KTSCX140; Open Funding Project of the Key Laboratory of Aquatic Botany and Watershed Ecology, Chinese Academy of Science; National Natural Science Foundation of China, Grant/Award Numbers: 31922060, 31720103905, 32030069
Abstract
The coupling between land use/landscape pattern and water quality in river systems varies across different spatial and temporal scales. It is important to understand the association between water quality and land use/landscape pattern across different spatial and temporal scales for the protection of water resources. Here, we measured seasonal water quality at 12 sub-basins in the upper reaches of the Han River (UHR) between 2010 and 2018. We conducted factor analysis and redundancy analysis to determine the links between land use and water quality at multiple spatial scales and to identify the main factors influencing water quality. We found that the concentration of nutrients, including total nitrogen, total phosphorus, nitrate-N and ammonium-N, was higher during the wet season than the dry season. Total nitrogen was the main driver of nutrient pollution in the UHR, whereas total phosphorus was identified as another potential nutrient pollutant. We also found that water quality parameters were more strongly related to land use types during the wet season than the dry season. Croplands and urban lands increased phosphorus concentrations of river water, whereas forest and grasslands decreased the nitrogen concentrations of river water at the sub-basin scale. Land use at the riparian zone scales better explained variations in water quality than land use at sub-basin scales. The explained variations in landscape metrics were generally higher during the dry season than the wet season. The largest patch index and Shannon's diversity index were the main predictors of river water quality in the UHR.
1 INTRODUCTION
Water quality is important for habitat conservation, agriculture and industry (Zhang et al., 2019). Increased nutrient concentrations (nitrogen and phosphorus) can lead to serious degradation of water quality in aquatic ecosystems. The degradation of water quality has become an important challenge globally (de Mello et al., 2018; Zhang et al., 2020). The management of nitrogen and phosphorus inputs into water resources is particularly important to prevent eutrophication. The increasing nutrition concentrations in rivers are mainly due to numerous natural and anthropogenic activities, such as erosion, weathering, atmospheric deposition, and domestic and industrial effluents (Bu et al., 2014; Martin et al., 2017; Zhang et al., 2020). Previous studies have demonstrated that spatial and temporal variations also have strong effects on nutrition concentrations in river water (Li et al., 2012; Wijesiri et al., 2018; Yuan et al., 2020; Zhang et al., 2020).
The assessment of the spatial and temporal variation in water quality variables at the basin scale and the identification of the drivers of these variations have become particularly important because of the heterogeneity, regionality and seasonality of water quality in rivers. Spatial variations of water quality depend on land use patterns (Zhang et al., 2019). Past research has demonstrated that there can be a strong association between water quality and land use (de Mello et al., 2018; Zhang et al., 2020), including grassland, forest, bare lands, agricultural, industrial and urban land cover types (Ding et al., 2016; Zhang et al., 2020). The agricultural, industrial and urban land use strongly influences nutrient concentrations (N and P) and organic pollution in rivers (Campaneli et al., 2021). The excess application of inorganic fertilizers in agricultural areas along with soil erosion has resulted in the high loading of nutrients into water bodies (Bu et al., 2014; Li et al., 2009). Organic pollutants and heavy metals are associated with industrial and urban land uses (Bu et al., 2014; Yuan et al., 2017). However, vegetation plays a key role in filtering pollutants, sediments and nutrients and in reducing soil erosion. Therefore, forest and grasslands are important for preventing or reducing water quality degradation (Zhang et al., 2019). Further information on the links between land use and water quality in a watershed is crucial for effective management of water quality. Therefore, continuous long-term research on the relationship between land use and water quality in a water basin is necessary.
The correlations between water quality status and land use differ across different spatial and temporal scales. For instance, some studies have shown a link between land use types at the riparian zone scale and water quality to be stronger than land use at the basin scale (Shen et al., 2015); other studies have shown the opposite (de Mello et al., 2018; Ding et al., 2016; Zhang et al., 2019, 2020). The inconsistencies in these results are likely due to differences in locations of sampling sites. The reason may also be derived from differences in experimental design. Moreover, riverine water quality has been shown to be closely associated with both the land use and the spatial configurations of forest, agricultural and urban areas (Bu et al., 2014). Thus, the quantification of the impact of different land use configurations and scales on water quality is required for the improvement of water resource management. Technological advancements in geographic information systems and remote sensing have facilitated the use of land use metrics in the quantification of land use structures (Huang et al., 2016; Li et al., 2012; Zhang et al., 2019). Previous research has demonstrated that land use metrics such as patch density, aggregation and diversity, landscape shape and largest patch indices are strongly linked with water quality variation (Shen et al., 2015; Zhang et al., 2020).
The upper Han River (UHR) is an important water supply for North China, including to the cities of Beijing and Tianjin, via the Middle Route of the South-to-North Water Division Project, which transfers 14 billion m3 of water per year. Therefore, management of water quality of the UHR is crucial for socio-economic development. There have been many studies on the water quality challenges in the UHR (Li et al., 2009, 2012; Li & Zhang, 2010). Anthropogenic development has dramatically increased pollution loading into the UHR, causing the UHR to become a highly polluted tributary that feeds the Yangtze River (Li et al., 2009; Müller et al., 2012). However, there is a lack of knowledge of the long-term spatial and temporal water quality variations within the UHR and the relationship between water quality variations and patterns of land use at multiple spatial scales. Moreover, the extent to which pollution degrades water quality is also not very clear. Both the public and government sectors have been increasingly interested in the water quality of the basin. Therefore, our objectives for our study were to (1) identify the long-term temporal water quality variability of the UHR; (2) identify the main factors affecting water quality in the river and streams; and (3) quantify the correlations between patterns of land use and variation in water quality variables for different land use scales. Our results can assist in the development of water resource management strategies and conservation policies for the inter-basin water transfer project.
2 MATERIALS AND METHODS
2.1 Study area
The Han River is the biggest tributary feeding the Yangtze River in China. The upper reaches of the Han River (31°20′–34°10′ N, 106°–112° E; 210–3500 m above sea level [a.s.l.]) are situated in a mountainous region with a drainage area of 95,200 km2 and the river has a length of 925 km (Li et al., 2009, 2012). The UHR falls into the subtropical humid climate zone with an average annual temperature of 14°C and with lowest and highest temperatures of −13°C and 43°C, respectively. Annual precipitation ranges between 800 and 1250 mm with an uneven spatial distribution, and 70%–80% of rainfall is concentrated during the wet season from May to October (Li et al., 2012; Li & Zhang, 2008). The mean annual run-off of the UHR is 41.1 × 109 m3, which accounts for 70% of total basin run-off, and has a large seasonal variability (Jin & Guo, 1993; Yang et al., 1997). The Danjiangkou Reservoir marks the outlet of the UHR, with the dam regarded as the most important water infrastructure in the Han River basin as the South to North Water Transfer Project starts from this reservoir (Jin & Guo, 1993; Li et al., 2017).
Forests are the dominant land cover type in the UHR basin, accounting for 80% of the total basin area, and include deciduous, mixed deciduous and evergreen, and evergreen and subalpine meadow forests (Li et al., 2009; Li & Zhang, 2008; Yang et al., 1997). Agriculture and urban areas constitute approximately 16% and 1.8% of the total area of the UHR basin, respectively. As one moves from the headwaters to the Danjiangkou Reservoir along the river network, the river starts in the Hanzhong Plain, the Ankang Plain constitutes the middle section and the industrial centre of the UHR basin is in the region of the Danjiangkou Reservoir (Li et al., 2009; Shen & Liu, 1998). Large cities, including Danjiangkou, Ankang and Hanzhong, are located along the Han River corridor.
2.2 Water sampling and analysis
A total of 240 grab samples were collected from 24 sites across 12 sub-basins of the UHR during the years 2010 to 2018 (Figure 1). Water samples were collected in the months of January and July. The sampling was timed to capture seasonal variability, with January representing the dry season and July representing the wet season. High-density polyethylene sample bottles (3 L) were submerged in nitric acid for 48 h and then rinsed twice with distilled water before use. The sample bottles were used to collect water samples from a depth of 0.2 m. Upon returning to the laboratory, samples were immediately stored in the dark at 4°C for analysis the following day.
Hydrochemical variables, including electrical conductivity (EC), pH, oxidation–reduction potential (ORP) and dissolved oxygen (DO), were measured in situ using a portable multimeter (YSI 6620, Yellow Springs, USA). Total phosphorus (TP), total nitrogen (TN), nitrate-nitrogen (NO3−-N) and ammonium-nitrogen (NH4+-N) were measured following standard methods (SEPA, 2002) for examining water and wastewater that are accepted worldwide (Bu et al., 2010, 2014).
2.3 Land use analysis
A method of supervised classification utilizing a support vector machine (SVM) was used to classify land uses in the ENVI 5.1 software. Land use in the UHR basin was mapped using imagery captured by the Landsat Thematic Mapper (30-m resolution) in 2018. The overall classification accuracy and kappa coefficient were 87.1% and 0.82, respectively. Six classes of land use were utilized, namely, croplands, forest land, grassland, water bodies, bare land and urban land. The proportions of different land use classes were represented as percentages of total land area.
We studied the effect of land use on river water quality in 12 rivers (R1–R12), which are distributed in different sub-basins (C1–C12) of the UHR basin (Figure 1). These 12 rivers are the main tributaries of the UHR, and their sub-basins had distinct land use and water quality. We investigated land use at three spatial scales: (1) within a 500-m buffer zone for the entire upstream length above the study site; (2) within a 1000-m buffer zone, which also extended the entire upstream length above the study site; and (3) at the sub-basin scale, which consisted of the entire area upstream of the study site (Figure 1). The 500- and 1000-m buffer zones were selected according to the resolution of the land classification data and previous studies (Ding et al., 2016). The buffer zones were delineated using ArcGIS9.3, and a digital elevation model (DEM) was used to calculate land use metrics.
Landscape pattern (land use structure) was quantified using the following metrics: patch number (NP), total area (TA), patch density (PD), edge density (ED), largest patch index (LPI) and Shannon's diversity index (SHDI). More information on these metrics is detailed in Huang et al. (2016) and Shi et al. (2017). We used FRAGSTATS V4.2.1 software to calculate these metrics.
2.4 Statistical analyses
We used variance analysis and other multivariate statistics. ANOVA (analysis of variance) was utilized to identify spatial and temporal differences in water quality variables at a level of significance of p < 0.05. Pollution factors affecting water quality were identified using factor analysis (FA) (Bu et al., 2014). The adequacy of the sampling data for FA was assessed using the Kaiser–Meyer–Olkin test. The correlations between land use/landscape pattern and water quality variables were identified using redundancy analysis (RDA) (de Mello et al., 2018; Zhang et al., 2019). In the RDA, land use/landscape pattern and water quality variables were utilized as environmental variables and species, respectively (de Mello et al., 2018). We used SPSS 18.0 (IBM Company, USA) to conduct statistical analyses whereas RDA was carried out using the CANOCO 5.0 software package (Microcomputer Power Company, USA). A complete workflow of our study is illustrated in Figure 2.
3 RESULTS
3.1 Spatial and temporal variation in water quality
Water quality variables had strong seasonal patterns (Figure 3a–h). DO levels were higher during the dry season, whereas the concentrations of ORP, TP, TN, NH4+-N and NO3−-N were elevated during the wet season. The sampling time (inter-annual) significantly affected water pH, EC, DO, ORP, TP, TN, NH4+-N and NO3−-N in both dry and wet seasons. The maximum and minimum observed levels of water quality variables over the study occurred in different years for different variables. EC in river water decreased from 2010 to 2018 (Figure 3a), whereas ORP and TN increased (Figure 3c). There was a significant positive correlation between time (year) and TN (R = 0.79, p < 0.05) during the wet season. The DO and NH4+-N content increased and then decreased in river water in both the wet and dry seasons (Figure 3d,g). In the wet season, the NO3−-N content increased and then decreased in the river water, and the NO3−-N content peaked in 2012 (Figure 3h). In general, the peaks in EC and TN occurred in 2010 and 2018, respectively, whereas ORP, NH4+-N and TP peaked in 2016, and DO and pH peaked in 2014.
All water quality variables, except for OPR, had clear spatial variability (Figures 3 and S1). The lowest concentrations of TN, TP and NH4+-N occurred at site 19 during the dry season. Site 17 had the lowest concentrations of TN and NO3−-N, which occurred during the wet season. Sites 2, 3 and 4 had elevated values of TN, TP and NO3−-N. The Kaiser–Meyer–Olkin test yielded a result of 0.69 (p < 0.001), which indicated that FA could effectively reduce dimensionality. Factor analysis of the water quality variables identified four components that together accounted for >70% of total variance. Factor 1 reflected the nitrogen and phosphorus pollution level of river water due to the elevated loadings of TP, TN and NO3−-N. Factor 2 reflected the acidity or alkalinity of river water as high pH loadings (Table S1). The sampling sites 1, 2, 3, 4, 6, 13 and 24 were located on the right side of factor score 1, whereas the other sampling sites were on the left side of factor score 1. Sampling sites 1, 2, 3, 4 and 24 had significantly higher factor scores than the other sampling sites (Figure S1). These results indicated that sampling sites 1, 2, 3, 4 and 24 were seriously polluted by nitrogen and phosphorus.
According to national and international guidelines (National Environmental Protection Bureau [NEPB], 2002; Table S2), and factor analysis of water quality (Table S1), the TN concentrations of 4.17%, 41.67% and 54.17% of the sampling sites during the dry season were of grades III, IV and V, respectively. Most of the sampling sites (87.5%) had TP concentrations of grade II, whereas 12.5% of sampling sites were of grade III. All sampling sites had NH4+-N concentrations of grade II, whereas the DO concentrations of most sampling sites (95.83%) were of grade I, and 4.17% of the sampling sites were of grade II (Table S2; Figure S2; NEPB, 2002). During the wet season, the TP concentrations of 4.17%, 41.67%, 45.83%, 4.17% and 4.17% of the sampling sites were of grades I, II, III, IV and V, respectively (Table S2; Figure S3; NEPB, 2002).
3.2 Land use types of the UHR basin
The land use types (e.g., forest, urban and croplands) differed significantly among the buffer zones and sub-basins. Forest was the dominant land cover type of the sub-basins and constituted between 30.92% (in R6) and 93.15% (in R12) of the total land area (Table 1). The mean value of the percentage of forest area within each buffer zone was lower relative to the percentage of forest area within their sub-basins, with the lowest value being for the 1000-m buffer zone. Croplands were the second most important land use and comprised between 4.81% (R11, 500-m buffer) and 67.29% (R6, 500-m buffer) of land use area (Table 1). As shown in Table 1, R6 had the highest proportion of urban land cover across all spatial scales. Overall, croplands and urban land respectively accounted for 57.99% and 10.30% of the land cover at the sub-basin scale, 67.29% and 27.39% in the 500-m buffer zones, and 65.73% and 28.95% in the 1000-m buffer zones. Grassland was the smallest proportion of land cover in R6 and R7 (0%) but was the largest proportion of land cover (2.14%) in the R10 500-m buffer zone (Table 1). Water bodies showed extensive distributions from 0.12% (R3) to 16.31% (R11). The proportion of bare land was relatively low, reflecting the existence of a little amount of undeveloped land within this area.
| Scale | River | Forest (%) | Grassland (%) | Cropland (%) | Urban (%) | Waters area (%) | Bare land (%) |
|---|---|---|---|---|---|---|---|
| Sub-basin | R1 | 73.30 | 0.61 | 23.35 | 1.79 | 0.83 | 0.12 |
| R2 | 74.21 | 0.99 | 20.41 | 3.80 | 0.49 | 0.11 | |
| R3 | 89.34 | 0.34 | 9.37 | 0.71 | 0.21 | 0.03 | |
| R4 | 75.68 | 0.00 | 21.37 | 2.60 | 0.20 | 0.05 | |
| R5 | 81.03 | 0.97 | 16.01 | 1.36 | 0.59 | 0.03 | |
| R6 | 30.92 | 0.00 | 57.99 | 10.30 | 0.78 | 0.00 | |
| R7 | 77.99 | 0.00 | 19.90 | 1.77 | 0.29 | 0.06 | |
| R8 | 86.46 | 0.33 | 11.56 | 1.24 | 0.31 | 0.09 | |
| R9 | 92.32 | 0.36 | 6.51 | 0.50 | 0.27 | 0.04 | |
| R10 | 85.60 | 0.36 | 10.97 | 1.99 | 0.91 | 0.17 | |
| R11 | 81.77 | 0.05 | 7.18 | 8.86 | 1.93 | 0.22 | |
| R12 | 93.15 | 0.14 | 5.71 | 0.17 | 0.79 | 0.04 | |
| 500-m buffer | R1 | 58.73 | 1.51 | 36.42 | 2.79 | 0.35 | 0.20 |
| R2 | 57.39 | 0.98 | 35.27 | 5.38 | 0.77 | 0.21 | |
| R3 | 85.66 | 0.00 | 10.01 | 4.21 | 0.13 | 0.00 | |
| R4 | 68.46 | 0.00 | 21.55 | 9.64 | 0.35 | 0.00 | |
| R5 | 64.75 | 1.37 | 22.64 | 5.08 | 6.02 | 0.13 | |
| R6 | 0.77 | 0.00 | 67.29 | 27.39 | 4.55 | 0.00 | |
| R7 | 56.37 | 0.00 | 34.16 | 7.65 | 1.81 | 0.00 | |
| R8 | 86.51 | 1.00 | 10.72 | 0.89 | 0.84 | 0.03 | |
| R9 | 83.19 | 0.12 | 12.17 | 1.68 | 2.80 | 0.04 | |
| R10 | 53.49 | 2.14 | 30.54 | 6.28 | 7.40 | 0.16 | |
| R11 | 21.61 | 0.00 | 4.81 | 56.94 | 16.31 | 0.32 | |
| R12 | 83.97 | 0.24 | 12.59 | 0.33 | 2.86 | 0.01 | |
| 1000-m buffer | R1 | 57.60 | 1.29 | 36.95 | 3.46 | 0.42 | 0.28 |
| R2 | 38.54 | 0.60 | 23.87 | 3.99 | 0.62 | 0.16 | |
| R3 | 87.43 | 0.01 | 9.12 | 3.32 | 0.12 | 0.00 | |
| R4 | 67.34 | 0.00 | 23.05 | 9.18 | 0.44 | 0.00 | |
| R5 | 64.83 | 1.62 | 21.96 | 5.03 | 6.47 | 0.08 | |
| R6 | 0.88 | 0.00 | 65.73 | 28.95 | 4.45 | 0.00 | |
| R7 | 55.94 | 0.00 | 34.73 | 7.49 | 1.84 | 0.00 | |
| R8 | 86.50 | 0.90 | 10.66 | 1.03 | 0.87 | 0.05 | |
| R9 | 83.59 | 0.09 | 12.16 | 1.75 | 2.36 | 0.04 | |
| R10 | 54.85 | 2.01 | 29.27 | 6.25 | 7.49 | 0.13 | |
| R11 | 28.86 | 0.00 | 6.44 | 51.64 | 12.35 | 0.71 | |
| R12 | 84.26 | 0.24 | 12.39 | 0.32 | 2.77 | 0.01 |
3.3 Landscape pattern of the UHR basin
The 12 sub-basins (C1–C12) also had large variations in landscape metrics (Figure 4). The levels of TA (9.56 × 105 km2) and NP (3.22 × 104) were the highest and lowest in C1 and C11, respectively. TA had a trend similar to NP from C1 to C12. The levels of PD (8.24 n/km2), ED (49.46 m/ha1) and SHDI (0.79) were the highest and lowest in C1 and C12, respectively. The landscape metric results indicated that PD, ED and SHDI tended to vary similarly from C1 to C12. LPI was highest in C12 (92.92%) and lowest in C1 (31.45%).
3.4 Correlations between land use and water quality
Our RDA results demonstrated that there were seasonal differences in the effect of land use types on water quality variables (Figure 5). The levels of explained variation were generally higher over the wet season compared to the dry season (Table 2). The predictors explained 76.2% of the variation for the 1000-m buffer zone during the wet season. However, the explanatory ability was lower at the sub-basins (63.6%). The variation in water quality was better explained by all axes for the 500-m buffer zone compared to the 1000-m buffer zone and the sub-basin scale over the dry season. Urban land was the most important explanatory variable (18.8%–20.6%) for the buffer zones and at the sub-basin scale during the wet season, whereas the most important variables explaining water quality during the dry season were grassland (22.6%), bare land (18.2%) and water bodies (11.5%) for the sub-basins, 500-m buffer zone and 1000-m buffer zone, respectively.
| Scale | Season | Forest (%) | Grass (%) | Cropland (%) | Urban (%) | Waters area (%) | Bare land (%) | Total explanation (%) |
|---|---|---|---|---|---|---|---|---|
| Sub-basin | Dry | 3.3 | 22.6 | 3.5 | 2.8 | 6.5 | 6.7 | 45.4 |
| Wet | 13.7 | 4.4 | 12.5 | 18.8 | 8.3 | 5.9 | 63.6 | |
| 500-m buffer | Dry | 3.1 | 11.4 | 5.7 | 3.6 | 9.6 | 18.2 | 51.6 |
| Wet | 11.6 | 10.6 | 14.0 | 20.9 | 10.7 | 5.6 | 73.4 | |
| 1000-m buffer | Dry | 6.9 | 8.0 | 2.4 | 3.6 | 11.5 | 5.9 | 38.3 |
| Wet | 13.5 | 9.8 | 10.7 | 20.6 | 7.8 | 13.8 | 76.2 |
Our RDA showed the relationships between the water quality variables and land use types (Figure 5). At the sub-basin scale, TN was positively related to the proportion of water bodies and to urban land during the dry season. NO3−-N and TP were positively related to the proportion of croplands during the wet season. NO3−-N and TN were positively and negatively related to the proportion of grassland, respectively. TP and TN were positively related to the proportion of bare land during the dry and wet seasons, respectively. In the 500-m buffer zone, TN was positively related to the proportion of grassland and croplands during the dry season. NO3−-N and TP were positively related to the proportion of croplands during the wet season. Croplands were positively associated with TN, EC and TP and had a negative association with ORP in the 1000-m buffer zone over the dry season. The proportion of croplands was positively associated with NO3−-N and TP and was negatively associated with TN in the 1000-m buffer zone over the wet season.
3.5 Correlations between landscape pattern and water quality
Figure 6 and Table S3 show ordination diagrams from RDA using sub-basin scale landscape metrics and water quality variables. The levels of explained variation were generally higher over the dry season compared to the wet season. LPI was the most important variable (22.5%) explaining water quality at the sub-basin scale over the dry season, whereas SHDI was the most important variable (15.6%) explaining water quality over the wet season (Table S3). For landscape metrics, TN and EC were positively related to ED, PD, NP and TA and negatively related to LPI at the sub-basins scale during the dry season. ED and SHDI were positively associated with TP, NO3-N and NH4+-N and negatively associated with TN at the sub-basin scale during the wet season. PD had a positive association with EC at the sub-basin scale during the wet season.
4 DISCUSSION
4.1 Spatial and temporal variations and assessment of water quality
The seasonal variability of nutrients concentrations in rivers is largely driven by natural processes, such as run-off and soil erosion, as well as by anthropogenic activity (Bu et al., 2010; Zhang et al., 2020). DO levels were elevated during the dry season (Figure 3), and colder water generally has higher DO (Manahan, 2017). During the wet season, the erosion of the soil or the input of leaf litter washed into nearby streams from forests patch may increase organic matter in rivers, leading to lower DO values (Yule & Gomez, 2009). We suspect that eutrophication during the wet season was responsible for the lower DO. Nutrients, such as TP, TN, NH4+-N and NO3−-N, had higher concentrations during the wet season compared to during the dry season (Figure 3), which was consistent with the results reported by some previous studies (Ai et al., 2015; Li et al., 2009; Shi et al., 2017; Tu, 2011). We found that diffuse inputs of nutrients from rainfall–run-off from urban areas and farmland contributed to elevated nutrient concentrations in the UHR basin. The inter-annual variations in water quality (Figure 3) were largely due to increases in point source pollution and human activities in the basin. The overall TN concentrations have increased since 2010, which can be attributed to population growth and economic development in the Han River basin. The UHR basin encompasses a large area containing a wide variety of land use types (e.g., farmland, forest, grassland and urban land). Thus, it was expected that there would be spatial variability in water quality in the UHR basin. The factor scores derived through factor analysis were used as indicators of pollution sources in the river system. In our study, the higher factor scores indicated that the pollution sources mainly come from the sampling sites 1, 2, 3, 4 and 24 (Figure S1; Table S1). The sampling sites had poor water quality, which demonstrated the influence of domestic sewage discharge from the cities of Shangluo, Nanyang and Shiyan. Sampling sites 17 to 19, however, had lower factor scores, which suggested the water was less polluted. Sites 17 to 19 could be regarded as non-impacted by anthropogenic influences as they were situated in the Jinshui River basin, which is characterized by extensive forest cover with limited human activity. Thus, the water quality at these sites was optimal.
According to our factor analysis of water quality (Table S1), the water quality variables that failed to meet the Chinese State Standard (CSS) were TN and TP, with TN being the most important indicator. Thus, we identified TN as the key factor driving water quality impairment in the UHR basin, and we identified TP as a potential nutrient pollutant. Our results indicated that controlling TN could effectively improve the water quality in the UHR basin and were consistent with the results reported by previous studies (Nobre et al., 2020).
4.2 Impact of land use on water quality
The degradation of water quality is related to the proportion of croplands and urban land (Bu et al., 2014; Ding et al., 2016; Zhang et al., 2020). We found that urban and cropland are detrimental to water quality at the sub-basin scale (Figure 5). Our results agreed with some previous studies (Shi et al., 2017; Zhang et al., 2019). The effects of terrestrial N and P on water quality have been broadly reported, and the effects of agricultural sources of N and P on water quality have been documented at river (David & Gentry, 2000; Mao et al., 2017), watershed (Yuan et al., 2013) and basin scales (Turner & Rabalais, 1991). The poor water quality might be a result of the excess use of pesticides and fertilizers on croplands in the UHR basin, and nutrients in croplands run-off can accelerate eutrophication in surface water bodies (Wan et al., 2014). The water quality variables trait of urban land at the sub-basin scale (Figure 5) were supported by previous studies that found positive correlations between river water nitrogen and urban land use patterns (Ding et al., 2016; Tu, 2011). Pollutants and effluent in urban areas are transported to river systems through run-off over impervious surfaces, thereby adversely impacting riverine water quality during rainfall events (Johnson et al., 2013; Wan et al., 2014). However, our RDA results demonstrated that the TP concentration was negatively related to the proportion of urban land, which was in contrast to some previous studies that found positive relationships between TP and some water quality variables in the 500- and 1000-m buffer zones during the dry season. Our contrasting result could possibly be explained by the cropland types and the small amount of rainfall during our sampling time.
The good water quality had positive correlations with forest and grassland proportions (Ding et al., 2016; Huang et al., 2016; Zhang et al., 2020). Vegetation, such as forest and grassland, has been identified as a major factor in mitigating degradation of water quality (Shi et al., 2017; Zhang et al., 2019). We found that the grassland was negatively associated with TN at the sub-basin scale during the wet season (Figure 5). Our results could be attributed to a reduction in soil erosion, nutrient assimilation and absorption. Therefore, the locations and proportions of vegetation-type land cover categories should be considered for sustainable management (Liu et al., 2018). However, the proportion of grassland had a positive association with TN in the 500-m buffer zone and at the sub-basin scale during the dry season, which was likely due to the relatively small amount of run-off and a lot of detritus being deposited directly into the river during the dry season. Some water quality variables, such as TN, were positively correlated with the proportion of water bodies. Water bodies have large surface–volume ratios and can therefore facilitate the removal of some nutrients through nitrification and absorption to sediment (Zhang et al., 2020). TN and TP were positively related to bare land (Figure 5), which indicated the natural contribution of nutrients through rock weathering.
4.3 Influence of landscape pattern on water quality
The landscape pattern also played an important role in the mediation of ecological processes, such as energy flows and nutrient cycles, and can adjust their effects on water quality (Bu et al., 2014; Ding et al., 2016; Zhang et al., 2020). A ‘patch’ is defined as an area with relatively homogeneous land cover, such as patches of forest or grassland (Lee et al., 2009). TA was found to be related to basin area and NP was correlated with the degree of fragmentation of a basin (Zhang et al., 2019). We found that TN and EC were positively associated with NP and TA at the sub-basin scale during the dry season (Figure 6). There was a positive correlation between TA and areas of land use categories associated with human impacts. Thus, elevated TA levels contributed to nutrient loading. PD represents the number of corresponding patches divided by total area (Ding et al., 2016). ED is the total length of all edge segments per unit area for the landscape (Huang et al., 2016). The high values of these metrics indicated abundant land use and elevated fragmentation, which might result in increased surface run-off and soil erosion. Therefore, higher PD represents degraded water quality and is unfavourable for managing pollutant migration (Shi et al., 2017; Zhang et al., 2019). PD and ED were positively associated with TN and EC at the sub-basin scale in the dry season (Figure 5). Huang et al. (2016) similarly identified positive correlations between certain water quality variables, including EC and TN, with PD. SHDI represents the diversity of patches within a basin (Zhang et al., 2019). Degraded water quality was positively correlated with SHDI. Increased TP, NH4+-N and NO3−-N were related to higher SHDI values during the wet season (Figure 6). In contrast, LPI was negatively associated with TN in the dry season and TP in the wet season (Figure 6). LPI represents the largest area of a particular patch type divided by total area. Our results indicated that forest was the dominant land cover type and large forest patches played a role in ameliorating water pollution by filtering pollutants and nutrients. Thus, the dominance of forest was associated with high-quality water.
4.4 Impact of spatial and temporal scales on the water quality variation
RDA showed that the impact of land use types and landscape metrics on the water quality variables had significant seasonal variability. Similar findings have been reported in previous research (Ding et al., 2016; Pak et al., 2021). The water quality had a stronger link to land use types during the wet season compared to over the dry season (Table 2), which was consistent with previous reports (Shi et al., 2017). This linkage can be explained by the run-off of fertilizer and domestic sewage from croplands or urban land into rivers during the wet season, which results in water quality degradation. Therefore, the effects of land use types on water quality were more distinct over the wet season. However, other studies have found the opposite (Pak et al., 2021; Zhang et al., 2019). These different results are likely to be attributed to differences in basin topography, farming practices, human activities and land use. In contrast, the explained variations of landscape metrics were generally higher in the dry season than in the wet season (Table S3). The most important factors explaining water quality over the dry and wet seasons were LPI and SHDI, respectively.
Scale dependency was evident in the influence of land use types on water quality. We found that land cover within the 500- and 1000-m buffer zones better explained variations in water quality than land cover at the sub-basin scale (Table 2). Thus, more attention should be placed on land management within riparian buffer zones. Shi et al. (2017) found that riparian land use had a larger effect on water quality than land use at larger spatial scales, but the opposite conclusion has been made in other studies (Ding et al., 2016; Zhang et al., 2019). These conflicting reports may be attributed to differences in regional characteristics. However, different land use types had different scale effects. In our study, water quality was strongly related to forest and grassland at larger scales (Table 2). Croplands, water bodies, urban land and bare land within the buffer zones were strongly associated with water quality (Table 2).
5 CONCLUSIONS
Riverine water quality in the UHR has significant spatial and temporal variations. We found that nitrogen was the key factor impairing water quality of the UHR, and phosphorus was a potential pollutant. Our results indicated that controlling TN could effectively improve the water quality in the UHR. We found water quality to be poor at sites 1, 2, 3, 4 and 24, likely due to domestic sewage discharge from the cities of Shangluo, Nanyang and Shiyan. RDA showed that that croplands and urban land were detrimental to water quality at the sub-basin scale. The forest and grassland were negatively associated with TN at the sub-basin scale during the wet season. Water quality was strongly linked to land use types over the wet season. In contrast, the explained variations of landscape metrics (e.g., LPI and SHDI) were in general higher during the dry season than the wet season. RDA showed that land cover type in the 500- and 1000-m buffer zones better explained variations in water quality than land cover at the sub-basin scale. However, different land use metrics had different impacts at different spatial scales. Forest and grassland were more strongly associated with water quality at larger scales. Croplands, water bodies, urban land and bare land within the buffer zones were highly associated with water quality. Our results provide key information for water quality conservation and land use planning at multiple scales.
ACKNOWLEDGEMENTS
This work was financially supported by the National Natural Science Foundation of China (32030069, 31720103905, 31922060); the Open Funding Project of the Key Laboratory of Aquatic Botany and Watershed Ecology, Chinese Academy of Science; and the Ordinary University Characteristic Innovation Project of Guangdong Province, China (2020KTSCX140).
CONFLICT OF INTEREST
The authors declare no conflict of interests.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author.
