Identifying the influence factors at multiple scales on river water chemistry in the Tiaoxi Basin, China

Ecological Indicators xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Identifying the influence factors at multiple scales on river water chemistry in the Tiaoxi Basin, China ⁎

Lijuan Cuia,b,1, , Wei Lia,b,1, Changjun Gaoa,b,c, Manyin Zhanga,b, Xinsheng Zhaoa,b, Zheng Yangd, Yinru Leia,b, Di Huanga,b, Wu Mae a

Institute of Wetland Research, Chinese Academy of Forestry, Beijing 100091, China Beijing Key Laboratory of Wetland Services and Restoration, Beijing 100091, China c Guangdong Academy of Forestry, Guangzhou 510520, China d School of Geographical Sciences, University of Bristol, United Kingdom e School of Natural Resources, West Virginia University, USA b

A R T I C L E I N F O

A B S T R A C T

Keywords: Redundancy analysis Multi-scale Physiography Human disturbance Tiaoxi river

The catchment environment and landscape is widely used as a predictor of stream-ecosystem condition, and the extent of its influence is closely linked to spatial scale. The aim of this study was to identify the influence factors on river water chemistry at multiple scales in a basin, namely the catchment, riparian corridor, and river reach scales. Information about the catchment and riparian corridor landscapes, reach-scale river properties, and catchment environments and river water chemistry data were collected monthly from 31 streams across the Tiaoxi Basin from July 2011 to June 2012. We used redundancy analysis to identify the relative influences of multi-scale variables on nine water quality indexes over both the whole study period and three sub-periods (before, during, and after rainy seasons). Results showed that all the selected factors helped to explain variations in water chemistry, although the relative effects of these factors changed considerably with variation in the spatial and temporal scales. Stream water chemistry across the entire study period was more sensitive to physiography and landscape variable at the catchment scale than at the reach and riparian corridor scales. From dry seasons to the rainy season, the influence of physiography and landscape variable at the catchment scale decreased slightly, while the effects of variables at the reach and riparian corridor scales increased noticeably. Besides, the influence of variables at the catchment scale was relatively strong and stable while the impacts of variables at the local scale were relatively weak and fluctuated widely with seasons. The findings from this study may improve our understanding of the main drivers of variations in stream water chemistry in different spatial and temporal scales, and will help managers protect and restore stream water environments using a basin-scale perspective.

1. Introduction

Sheldon et al., 2012; Allan, 2004). Also, the hydrological processes were dominantly affected by basin landscape pattern, which have scaledependent and seasonally variable influences on river ecosystems (Buck et al., 2004). Most early studies of the relationships between watershed landscapes and river ecosystems were limited in their spatial scale, such as the areas that were either directly connected with rivers and streams or within a few hundred meters of the river. Less consideration has been given to the importance of elements at larger spatial-scales (Allan, 2004; Marzin et al., 2013). With increasing human disturbance in catchments and riparian zone corridors, later studies related to the

Landscape patterns in catchments have important influences on the processes that control different forms of carbon (C), nitrogen (N), and phosphorus (P) discharged in river water (Ahearn et al., 2005; Allan, 2004). Basin landscape pattern was generally characterized by some geographical factors, such as climate, topography, surficial geology, and land use or land cover types (Frissell et al., 1986; Schoonover et al., 2005). In addition, hydrological processes within a basin can affect the water quality of rivers through changing the supply, transport, and transformations of C, N, and P in the river water column (Hynes, 1975;

⁎ Corresponding author at: Institute of Wetland Research, Chinese Academy of Forestry, Beijing 100091, China, and Beijing Key Laboratory of Wetland Services and Restoration, Beijing 100091, China. E-mail addresses: [email protected], [email protected] (L. Cui), [email protected] (W. Li). 1 These authors contributed equally to this work.

http://dx.doi.org/10.1016/j.ecolind.2017.08.053 Received 20 November 2016; Received in revised form 22 August 2017; Accepted 22 August 2017 1470-160X/ © 2017 Published by Elsevier Ltd.

Please cite this article as: Lijuan, C., Ecological Indicators (2017), http://dx.doi.org/10.1016/j.ecolind.2017.08.053

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

2.2. Sampling sites and water chemistry data

effects of landscape patterns on river ecosystems, such as biodiversity (Sandin and Johnson, 2004; Weijters et al., 2009), water quality (Johnson et al., 1997; Tong and Chen, 2002) and nutrients in rivers (Udy et al., 2006), have attracted increasing attention (Allan, 2004). Studies examined the importance of the influence of landscape patterns at different spatial scales on nutrient dynamics and variations in river water, however have produced inconsistent results (Buck et al., 2004; Dow et al., 2006; Johnson et al., 1997; Sliva and Dudley Williams, 2001). Some scholars believe that the variations in nutrients and river habitat conditions should be examined using catchment-scale landscape patterns due to their influence on driving the geomorphic processes of a catchment, e.g., shaping channel network, supplying water and sediments (Esselman and Allan, 2010; Frissell et al., 1986; Johnson et al., 1997; Roth et al., 1996), while the others consider that the relationships between river habitat and riparian zone landscape patterns on both sides of a river are much more significant. For examples, Peterson et al. (2011) indicated that upstream land use was more influential in larger streams, while local land use and other factors might be more important in smaller streams; and Sandin and Johnson (2004) concluded that local physical (24.4%) and local chemical (20.4%) variables explained the largest part of the among-site variability of river community assemblages. Moreover, most of these studies have not highlighted the effects of or seasonal variations in, the natural and anthropogenic factors at multi-scale on river habitat. Multi-scale studies of factors that affect river water chemistry variation are mostly focused on reach, local and catchment scale (Frissell et al., 1986; Esselman et al., 2010; Kings et al., 2005; Marzin et al., 2013). For the purposes of this study, reach-scale factors were selected to reflect local habitat conditions within ≤500 m sections of the river channel. Catchment-scale factors were defined as the integrated conditions in the landscape upstream of a given sampling location (e.g., percentage of different landscape types in a catchment), the geomorphology condition or the position relative to the sampling location (e.g., mean slope, distance from the river mouth etc.). Local-scale factors were restricted in the river riparian zone indicating the hydrological connectivity or the potential resource/sink for nutrient export from upland to downstream. The present study identify the relative importance of reach-, riparian- and catchment-scale landscape factors to variation in water chemistry in Tiaoxi River, one of the main tributaries to Taihu Lake. We then identified the variables that have most influence on the river habitat for individual spatial scales, and their combinations. By acknowledging the influence variables on nutrient dynamics at multiple scales and in different seasons (before, during, and after rainy seasons), we examined the hypothesis that the impacts of selected variables on nutrients varied seasonally. We anticipated that the findings from this study help basin water resource managers optimize the timing of special management measures, which will effectively control the supply, transport, and transformations of nutrients in catchments.

Thirty-one river sections (see Fig. 1) were selected as sampling sites in the main channels and tributaries of the Tiaoxi River. Sampling sites were distributed across the entire study area so that, as far as possible, all the characteristics of the natural geographical environment and the spatial layout of all land use types in the study area were included. We identified and sampled a reference site 1000 m upstream of the sampling area before formal sampling began where there were no impacts of industrial sewage outlets, livestock excrement, and a domestic waste disposal site. Sampling was completed monthly from July 2011 to June 2012. Two methods were used to collect water samples, depending on the nature of the river channel at the sampling site. A water collector with a volume of 2.5 L was used to collect water samples from nonwade able rivers in the middle and lower reaches where the water depth in the center of the river was greater than 1.5 m. At these sites, the samples were collected from a depth of 0.5 m and water surface in the center of the channel. The samples were mixed on collection and stored in acid-washed 1000 mL polyethylene bottles. For wadeable rivers in the upper reaches, water samples were collected using plastic cups from water at the surface on both sides of the river. Once collected, the samples were immediately placed in a portable fridge at 4 °C and transported to the laboratory, where they were processed and analyzed within 48 h of collection. The samples were analyzed for total nitrogen (TN), total phosphorus (TP), nitrate nitrogen (NO3eN), and soluble phosphate using the National Standard Methods (Editorial board of Water and wastewater monitoring analysis method, 2002). The dissolved organic carbon (DOC) concentrations were measured with a total organic carbon analyzer (Shimadzu TOC-V). Samples were filtered and the suspended solids on the filter paper were dried for 24 h at 60 °C, weighed, fumigated with concentrated hydrochloric acid for 24 h, and ground. Particulate organic carbon (POC) concentrations were determined after a series of treatments in an element analyzer (EA 3000 CHNS/O Analyzer). The sum of the POC and DOC concentrations was taken as the total organic carbon (TOC) concentration. The samples were analyzed in the State Key Laboratory of Lakes and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences. Based on the multi-year precipitation and runoff in the Tiaoxi Watershed (Huzhou Municipal Water Conservancy Bureau, 2004), the study period was divided into three periods: before the rainy season from March and May, during the rainy season from June to August, and after the rainy season in November, December and January. All the water quality data were normalized by log10 transformations. Results from normality analysis using the non-parametric K-S test showed all the indexes were normally distributed. 2.3. Selections of variable 2.3.1. Environmental variables Six natural environmental variables, including altitude, average slope, river discharge, annual precipitation, catchment area, and distance to the river source were used to represent the geomorphic processes of the sampling location and to determine their potential effects on nutrient dynamics in water (see Table 1). Elevation, mean slope, catchment area, and distance to the river source, were derived from spatial analysis of the 1:50, 000 DEM (see Fig. 1) of the study area and its derived data using ArcGIS 9.3 and ARCSWAT software. Annual precipitation data from 42 hydro-meteorological stations located in and around the Tiaoxi Basin were provided by the Hydrological Bureau of Huzhou City and were processed by the Kriging interpolation method in the geostatistical analysis module of ArcGIS 9.3. River discharge data for the middle and lower reaches of the main channel were supplied by the Huzhou Hydrological Bureau, while discharge data for the upstream source streams and other streams were collected using a portable flow meter (SonTek/YSI-Tracker). All environmental variable data were log10-transformed before data analysis to ensure that they met the

2. Material and methods 2.1. Study area The Tiaoxi Basin is located in the southwest of the Taihu Lake Basin (E:119°10′–120°11′; N:30°04′–120°02′) (see Fig. 1), and is divided into the east and west tributaries. Two tributaries converge at Bai quetang Bridge (hereinafter referred to as the Tiaoxi River) in Huzhou City and then pour into Taihu Lake (Huzhou City Water District, 2004). The main channel of Tiaoxi River is 157.4 km long, and the basin area covers 4576.4 km2, accounting for 12.54% of the total area of the Taihu Lake Basin (36500 km2). Tiaoxi River has a large annual runoff (14.93 × 109 m3) and is one of the main tributaries toTaihu Lake (Huzhou Water Resource Bureau, 2004).

2

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Fig. 1. The Tiaoxi River basin.

2.3.3. Landscape at the riparian corridor and catchment scales Riparian corridor refers to the riparian zone within 1000 m from both sides of the river channel and extends for a distance of up to 10,000 m along the sampling site. Because of its unique spatial structure and ecological functions, the riparian corridor is able to coordinate lateral (from land to water) and longitudinal (upstream to downstream) flows of material and energy by intercepting and filtering sediments, water, and nutrients (Meng et al., 2011) The term catchment-scale refers to the complete basin upstream, including the tributaries, of the stream sampling site. Changes in land use and landscape patterns in catchment areas are a concentrated expression of the integrated action of large-scale human activities (such as the expansion of urbanization, agricultural development, and forest harvesting) and natural factors, and they give rise to a variety of complex interactions between river hydrology, water quality, biological habitat, and aquatic organisms (King et al., 2005; Meng et al., 2011; Ou et al., 2012). Therefore, the catchment landscape pattern can be used as an important proxy of the effects of natural factors and human activities on the river water chemistry. In this study we used a Landsat TM image that covered the study area (Path/Row: 119/39; Date: 24 May 2010). The images were calibrated, cropped, and classified into six categories: forestland, urban land, cropland, bareland, water, and grassland (see Fig. 2). An accuracy assessment showed that the total accuracy was 88.46%, which met the requirements of this study. In this study, the percentage of each landscape type area was used as a variable to describe land use (see Table 1). The area of the landscape type was calculated with ArcGIS 9.3, and the data from the attribute table of each layer were exported into Excel 2010. The percentages of all the landscape types on both spatial scales were normalized with an inverse sine root-to-square transformation to ensure the data conformed to the normal distribution.

assumptions of the normal distribution.

2.3.2. Reach-scale river properties A channel segment is the area between two defined cross-sections and is the smallest study scale in the study of a riparian zone. To achieve the objectives of this study, the river range was limited to the riparian zone that is within 30 m of both sides of the river channel and extends for a distance of up to 500 m along the bank. Bank slope, river vegetation type, type of traffic in the river, water conservation facilities were recorded with a tape measure and GPS instrument. The river types were identified using a combination of field observation and visual interpretation of high resolution images provided by the Google Earth platform (Quickbird and SPOT images, acquisition date: 4 July 2003 and 4 October 2010; resolution of 0.61 m and 5 m). After field investigation and expert judgement, six indicators were selected to indicate the hydrogeomorphic (e.g., instream habitat, channel form modifications) and hydro-connectivity (e.g., dam, lock) of modifications river characteristics (Table 1). Artificial embankment refers to the degree of artificial hardening of the riparian zone, and reflects the land-use intensity of the near shore. This indicator was classified into three types; namely natural (no hardening), semi-natural (partially hardened), and artificial (fully hardened). The degree of interference of vegetation on the bank slope reflects the vegetation coverage on the bank slope, and was divided into four classes, namely 3 (< 10%), 2 (10–45%), 1 (45–85%), and 0 (≥85%). The downstream sluice gate reflects the degree of connectivity of river water bodies on the vertical spatial scale, and it was categorized into three classes, depending on the type of sluice (Table 1). The river bed habitat disturbance reflects the degree of damage to the morphology and vegetation of the river bed. A good river bed habitat can, to a certain extent, decrease and impede the transport of nutrients in a water body. The main channels in the middle and lower reaches of the Tiaoxi River are navigable. Large cargo ships affect the transportation of nutrients in the water body by disturbing bottom sediment, changing the river bed habitat, and influencing the direction of flow.

2.4. Data analysis Spearman (ordinal variables) and Pearson correlations (quantitative variables) were calculated in SPSS 16.0 to determine the relationships 3

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

water quality indexes were close to the end of the arrows, the correlation was positive. If they were on the opposite site, then the correlation was negative (Sadyś et al., 2015).

Table 1 Summary statistics of influence factors at different scales. Variable

Unit

Data transformation

Variable statisticsa

3. Results Physiography Altitude (Alti) m Mean slope (Slope) % River discharge (RD) m3/s Annual precipitation mm (AP) Catchment area (CA) km2 Distance to the km source (DS) Reach-scale variable Artificial – embankment (AE) Reservoir upstream – (RU) Barrier downstream (BD) In-stream freighter – (ISF) Riparian vegetation – modified (RVM) Subfluvial habitat – modified (SHM) Riparian corridor zone landscape %forestland(FOR) % %urban land(URB) % %cropland (CROP) % %grassland (GRAS) % %water (WAT) % %bareland (BAR) % Catchment landscape %forestland % %urban land % %cropland % %grassland % %water % %bareland %

Log(x) Log(x) Log(x) Log(x) Log(x) Log(x)

157.01 (14.06–697.00) 25.05 (3.97–56.59) 5.16 (0.38–80.25) 1526.15 (1316.54–1872.21) 130.20 (23.59–355.18) 0.55 (0.19–1.91)

3.1. Variations in water quality variables in the Tiaoxi Basin Summary information on C, N, and P in the Tiaoxi Basin is presented in Table 2. During the monitoring period, the TN concentrations in the Tiaoxi River ranged from 0.94 to 9.23 mg/L, and the average concentration was 3.59 mg/L, 1.795 times greater than the water quality standard (2 mg/L). The average ammonium nitrogen (NH4eN) concentration was 0.38 mg/L, which was 10.58% of the TN concentration, and was lower than the Class II NH4eN threshold (0.5 mg/L). The average concentration of nitrate nitrogen (NO3eN) was 2.03 mg/L, the maximum concentration was less than 7.0 mg/L. For all components of TN, NO3eN represented 56.55%, and was the main form of TN. The mean concentration of TP was 0.11 mg/L. Concentrations ranged from 0.01 to 1.36 mg/L and were close to the limit (class II) for phosphorus (1 mg/L). The mean POC concentration was 200.63 mg/L, while that of DOC was only 2.39 mg/L. Therefore, POC was the main component of TOC. In contrast, the ecological stoichiometric ratios (N: P, C: N, and C: P) of C, N, and P in the Tiaoxi River were much higher than those in the sediments of Taihu Lake, which reflected the classic Redfield ratio (Liu et al., 2011; Qu et al., 2001; Redfield, 1958; Xu et al., 2010). This indicates that there is more C and N in the Tiaoxi River relative to P. As shown in Table 2, the ranges of the standard deviation and variability in the water quality variables of the Tiaoxi Basin were wide, indicating that there were large spatial differences in the concentrations of C, N, and P in river water. Fig. 3 shows the relationship between the water quality variables and quadrat samples in the Tiaoxi Basin. The first principal component mainly reflected the characteristics of DOC, TP, NH4eN, and POC, and the C:P and N:P ratios, while the second principal component mainly reflected NO3eN and the C:N ratio. There were strong correlations among some water quality variables, e.g., between DOC and TP or between POC and C:N ratio, indicating similar sources and input paths of these water quality variables to the water body. The figure also demonstrates that water quality variables varied within the different quadrats. When the rays of the water quality variables are extended, 31 squares are vertically projected onto the ray. Using the relative distance from the projection point to the solid arrow as the standard, the value increases in the direction of the arrow and decreases in the opposite direction. For example, the values for TN were largest in quadrats X1, X15, D5, D7, and D10. The squares of X4, X5, D3, and D4 were near the origin, which shows that the TN degree was approximately equal to the mean value and the TN degrees for X9, X10, X11 and X13 were the smallest. These samples roughly correspond to the spatial characteristics from downstream to upstream (Fig. 1), but these results also show that the water environment in the lower reaches of the Tiaoxi Basin has gradually deteriorated, the proportion of manmade landscapes has increased slightly, and human disturbance has gradually increased.

No (4)/Partial (13)/Yes (14) No(18)/Yes (13) No (18)/Partial (6)/Yes (7) No (15)/Slight (5)/ Intermediate (6)/High (5) No (4)/Slight (8)/ Intermediate (13)/High (6) No (3)/Partial (17)/Yes (11) √(arcsin(x)) √(arcsin(x)) √(arcsin(x)) √(arcsin(x)) √(arcsin(x)) √(arcsin(x))

34.67 (1.2–85.4) 16.28 (2.0–61.7) 40.58 (10.5–85.1) 11.00 (0.0–13.8) 6.53 (0.0–16.4) 0.56 (0.0–4.7)

√(arcsin(x)) √(arcsin(x)) √(arcsin(x)) √(arcsin(x)) √(arcsin(x)) √(arcsin(x))

57.53 (10.77–92.89) 9.08 (1.09–39.22) 28.09 (4.05–67.39) 0.99 (0.05–5.25) 3.44 (0.03–13.03) 0.81 (0.00–4.51)

a Modalities for qualitative variables (number of sites) and ranges for quantitative variables (median (min–max)).

among catchment physiography, landscape variables at three scales (reach, riparian and catchment), and water quality indices. PCA analysis was also used to describe the spatial distribution of the selected water quality indexes in CANOCO software (version 4.5, Plant Research International, the Netherlands). RDA was first carried out using CANOCO 4.5 to determine which multiple scale landscape variables (reach, riparian and catchment) were statistically suitable for explaining variations in the nine selected water quality indices. Statistically important correlations between explanatory (e.g., each possible factor) and response (e.g., variations of water chemistry) variables were determined using a Monte Carlo permutation test (with 499 permutations in the reduced model; Sadyś et al., 2015). In this case, RDA was applied to evaluate the relative influence of multi-scale influence factors on variations of each water quality index for the four periods (the whole study period, before the rainy season, during the rainy season, and after the rainy season). RDA results are presented graphically on biplots, where variations in each water quality index and landscape factor for each spatial scale are shown as arrows and their abbreviated names; the former are presented in gray, most of the latter are indicated in black, and the six physiography variables are indicated by black triangles. The significance of the correlation between the distribution of changes in the water quality indexes and landscape factors is shown by the length of the arrows. A longer arrow indicates a stronger correlation between them (Sadyś et al., 2015). A vector of relationships (directly proportional or inversely proportional) was interpreted from the position of the changed water quality index relative to the end of the arrow. If the

3.2. Correlations between water quality and variables Table 3 shows the correlation coefficients between natural environment characteristics, different landscape scales, and water quality variables. Almost all of the environmental variables were significantly correlated with TN, and with the N:P and C: P ratios. Correlations with TP, NO3eN, and the C:N ratio were weakest. Out of the variables, the catchment area (CA) was least able to explain the water quality variables, and was only significantly and positively correlated with the TN concentrations. The average annual precipitation (AP) was negatively correlated with TN, PO4-P, DOC, and POC, but was positively correlated with the N:P and C:P ratios. The distance to the source (DS) was 4

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Fig. 2. Distribution of land use types in the Tiaoxi River Basin.

Table 2 Summary information of water quality variables in the Tiaoxi Basin. Water quality index

Unit

Mean

Standard deviation

Min

Max

TN TP NO3eN NH4eN DOC POC N:P (TN:TP) C:N (TOC:TN) C:P (TOC:TP)

mg/L mg/L mg/L mg/L mg/L mg/L mg/mg mg/mg mg/mg

3.59 0.11 2.03 0.38 2.39 200.63 61.98 66.76 2817.55

1.72 0.14 1.30 0.13 2.01 137.53 64.21 48.04 1938.10

0.94 0.00 0.05 0.14 0.14 41.42 4.54 8.06 399.19

9.23 1.36 6.99 0.66 19.24 928.45 499.63 346.28 9908.09

Note: TOC equals to the sum of DOC and POC.

positively correlated with both TN and DOC concentrations, and was negatively correlated with the N:P and C:P ratios. The annual average discharge (RD) of the river was positively correlated with TP, NO3eN, PO4-P, and POC, and negatively correlated with the N:P and C:P ratios. The TN and DOC concentrations were negatively correlated, the N:P and C:P ratios were positively correlated, and the mean slope (slope) and altitude (Alti) were negatively correlated. The different correlations indicate that the POC concentrations were not correlated with Slope, but were negatively correlated with Alti. There were significant correlations between TP, PO4-P, and POC and the C:N ratios of all the river reach-scale variables. The artificial bank (AE) and cargo ships in the river channel (ISF) were strongly correlated with most of the water quality variables, and were positively correlated with TP, PO4-P, DOC, and COD and negatively correlated

Fig. 3. Biplots from relationships between water quality variables and sampling sites. Note: Arrows with italic letters indicate the nine selected water quality variables. Shallow black dots indicate the 31 sampling sites (D1–D15, X1–X16).

with the N:P and C:P ratios. The difference between the correlations for AE and ISF possibly means that the AE was also positively correlated with the NO3eN concentrations, but that the ISF were not correlated with the NO3eN concentrations. The correlations between the upstream reservoir and downstream dam were the weakest. 5

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Table 3 Relationships between water quality and selected variables at different scales. Variable Physiography CA AP DS RD Slope Altitude Reach-scale AE RU BD ISF RVM SHF Riparian-scale %forestland %urban land %cropland %grassland %bareland %water Catchment-scale %forestland %urban land %cropland %grassland %bareland %water

TN

TP

NO3

PO4

DOC

POC

N:P

C:N

C:P

0.44 −0.64 0.61 – −0.54 −0.53

– – – 0.33 – –

– – – 0.30 – –

– −0.35 – 0.32 – –

– −0.59 0.62 – −0.66 −0.60

– −0.40 – 0.32 – −0.31

– 0.62 −0.59 −0.37 0.50 0.53

– −0.33 0.32 – – –

– 0.62 −0.66 −0.44 0.65 0.62

0.53 – −0.30 0.60 – 0.58

0.70 – – 0.68 0.67 0.64

0.54 – – – 0.38 0.37

0.60 – – 0.37 0.49 0.36

0.37 – – 0.54 – 0.56

0.44 – – 0.48 0.38 0.55

−0.37 – – −0.40 – –

– – – – – –

−0.57 – – −0.69 – −0.47

0.61 – 0.32 −0.45 – 0.85

– – –

– – –

– 0.32 –

0.31 – –

– 0.31

– –

– 0.31

0.66 0.61 – −0.63 – 0.69

– –

−0.50 – −0.49 0.48 – −0.57

– – – – –

−0.61 – −0.50 0.64 – −0.64

0.51 0.59 0.64 −0.62 0.51 0.78

– – – – – 0.32

– – – – – –

– – – – – –

0.63 0.64 0.65 −0.64 0.32 0.66

– 0.31 – −0.34 – 0.32

– − 0.42 −0.60 0.64 −0.38 −0.60

– – – – – –

−0.42 −0.64 −0.69 0.68 −0.39 −0.64

Note: “–” means no significant correlations (P > 0.05). Listed numbers mean significant correlations (P < 0.05). Pearson correlations were computed for quantitative variables and Spearman correlations for ordinal variables.

between AP and DS were the highest, and were − 0.75 and 0.73 respectively. In addition, the elevation and slope variables were also linked to this axis, with a correlation coefficient of −0.66 between the two variables and the axis.

At the riparian corridor and the catchment scales, most of the landscape variables were positively correlated with TN, DOC, and the N:P and C:P ratios, but were not correlated with TP, NO3eN, PO4-P, POC, and C:N. The area occupied by woodland was negatively correlated with the concentrations of C, N, and P, while the other land use type variables were positively correlated with the concentrations of C, N, and P.

3.3.2. Before the rainy season As shown in Fig. 5, before the rainy season, the landscape variables were able to explain 48%, 28%, 40%, and 39% of the entire variation in C, N, and P at the natural environment, river reach, riparian corridor, and catchment area scales, respectively. Natural environment variables had the greatest influence on variability of nutrients in river water, followed by the riparian corridor scale and the catchment scale, the influences of which were similar. The river scale had least influence. Before the onset of the rainy season, the influence of the various landscape variables on nutrient variability in the water body was concentrated on the first ordination axis, such as the river scale, and the various landscape variable categories can explain 13.83% of the water quality variation on the first ordination axis (permutation F test, P = 0.002). The correlations between ISF and SHM and the first ordinal axis were the strongest, with correlation coefficients of 0.59 and 0.44, respectively, followed by AM with a correlation coefficient of 0.41. At the riparian corridor scale, the landscape variables in the initial ordinal axis were able to explain 27.15% of the variation in water quality (permutation F test, P = 0.002), with correlation coefficients of −0.68, 0.67, and 0.65, respectively. At the catchment scale, the landscape variables in the first ordination axis were together able to explain 45% of the variability in water quality (permutation F test, P = 0.002). The correlations were strongest between the landscape variables and the areas of cultivated land and water bodies, with correlation coefficients of 0.71 and 0.70, respectively. The natural environment variables (permutation F test, P = 0.002) are reflected on the first ordinal axis, and, out of all the environmental variables, the correlations were strongest with DS (r = −0.80), AP (r = 0.78), and Alti (r = 0.75).

3.3. Influences of multi-scale landscape variables on water chemistry in different seasons 3.3.1. Through the whole study period As shown in Fig. 4, at the natural environment, river reach, riparian corridor, and catchment scales, the landscape variables were able to explain 39%, 36%, 38% and 39% of the entire variation, respectively, in C, N, and P. The environment and catchment scale variables explained most of the variation, variables at the riparian corridor scale followed, and reach-scale variables explained the lowest percentage of the variation. The influence of the various landscape variables on the variability of nutrients in the water body is focused on the first two ordinal axes. At the river reach scale, the original ordinal axis explained 25.8% (permutation F test, P = 0.002) of the variation in river water quality caused by ISF and SHM. The correlation coefficients of ISF and SHM with the axis were 0.72 and 0.67, respectively. At the riparian corridor scale, the basic ordinal axis explained 30.21% of the variation in the water quality (permutation F test, P = 0.002). The water body and grassland, with correlation coefficients of − 0.97 and − 0.84, respectively, were most closely related to the axis. Forestland followed with a correlation coefficient of 0.73. At the catchment scale, the first ordinal axis explained 44.39% of the variability in water quality (permutation F test. P = 0.012) with correlation coefficients of − 0.75 and 0.79, respectively, between forestland, water and the axis. Out of the river basin environment variables, 36.98% of the variability in water quality was explained by the environmental variables in the initial ordinal axis (permutation F test, P = 0.012). The correlation coefficients 6

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Fig. 4. Results of redundancy analysis of water chemistry for the whole study period and landscape variables at three scales. Note: a – reach-scale; b – riparian scale; c – catchment scale; d – physiography

period, the natural environment of the study area still had the most influence on variation in nutrients in the water, followed by the catchment scale, the riparian corridor scale, and the river scale. After the rainy season, the various landscape variables on the first ordinal axis explained 19.12% of the variation in water quality (permutation F test, P < 0.01) The correlation coefficient between ISF and the first ordinal axis was 0.64. The correlations between the axis and both SHM and AM were the most significant (permutation F test, P < 0.01), with correlation coefficients of 0.47 and 0.51, respectively. At the riparian corridor scale, the first ordinal axis explained17.28% of the variation in water quality (permutation F test, P < 0.01), and the correlation with the water body variable was strongest (r = −0.62), followed by the percentages of grass, forest, and arable land, which had correlation coefficients of −0.53, 0.44, and 0.36, respectively, with the first ordinal axes. At the catchment scale, the landscape variables on the first ordinal axis explained 32.8% of the variation in the water quality (permutation F test, P < 0.01), and forestland was the sole variable with negative correlation coefficients of −0.63. For other variables, the correlation coefficients of the water body, bare land, urban, cropland, and grassland, were 0.60, 0.58, 0.57, 0.54, and 0.50, respectively. The natural environment variable explained 30.32% of the variation in water quality (permutation F test, P < 0.01). The correlations with AP (r = 0.73), DS (r = −0.67), Alti (r = 0.61), and Slope (r = 0.52) variables were the most significant.

3.3.3. During the rainy season As shown in Fig. 6, during the rainy season, landscape variables at the natural environment, river reach, riparian corridor, and catchment area scales explained 39%, 32%, 44%, and 37%, respectively, of the entire variation in C, N, and P. Variables at the riparian corridor-scale had most influence on the variability of nutrients in the water, followed by the those at the natural environment and catchment scales. Variables at the river reach scale had least influence. During the rainy season, the impacts of various landscape variables on the variability of nutrients in the water body were also distributed on the first ordinal axis. The four types of variables, i.e. river reach scale, riparian corridor, catchment scale, and natural environment, can explain 20.57%, 24.57%, 35.45%, and 28.57% of the variation in water quality on the first ordinal axis, respectively (permutation F test, P < 0.01). At the river reach scale, the most significant correlation was between ISF and the first ordinal axis (r = 0.68), followed by SHM and AM, the correlation coefficients of which were 0.57 and 0.54, respectively. At the riparian corridor scale, the strongest correlation was between the areas of water body and grassland, with correlation coefficients of −0.84 and − 0.61, respectively. At the catchment scale, the correlation coefficients between landscape variables represented by water, cultivated land, and forestland were 0.74, − 0.69, and − 0.63, respectively. Out of the environmental variables, the DS (r = −0.63) and AP (r = 0.68) were most strongly correlated with the first ordinal axis.

3.3.4. After the rainy season The period after the rainy season is presented in Fig. 7. In period, the natural environment, river section, riparian corridor, catchment area landscape variables explained 47%, 31%, 36%, 43%, respectively, of the total variation in C, N, and P. During

this and and this 7

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Fig. 5. Results of redundancy analysis of water chemistry before the rainy season and landscape variables at three scales. Note: a – reach-scale; b – riparian scale; c – catchment scale; d – physiography

4. Discussions

human disturbance, which is consistent with the actual situation in the Tiaoxi Basin (see in Fig. 2), where forestland dominates (60%), and cultivated land accounts for 28%. In the riparian zone, the proportion of forestland has decreased (30%), and the proportion of cultivated land has increased (38%). At the same time, the explanatory variables at all scales explained less than 50% of the variation in water quality, indicating that variables, such as the size distribution of sediment, river velocity or river order, the integrity of aquatic organisms, and the physical distance or hydrological distance between a land-use type and a river at the catchment or riparian corridor scale, could better reflect the water quality parameters of the study area (Wang et al., 2006; Peterson et al., 2011; Wang et al., 2014; Novikmec et al., 2016) It is generally accepted that forests are nutrient sinks, while anthropogenically-influenced landscapes such as croplands and towns are nutrient sources (Allan, 2004; Kim et al., 2013). The difference is that there were no correlations between the percentage areas of forestland, cultivated land, and urban area and TP concentrations, but the N:P and C:P ratios were correlated, which perhaps indicates that the nutrients (P, C, and N) that migrate from land to water somehow interact rather than being individually controlled by different land use and landscape types within the catchment (Frost et al., 2002). Some studies have shown that increases in the area of grassland in a basin are closely related to nutrient enrichment in water (Ahearn et al., 2005; Xiao and Ji, 2007), while others have shown that grassland causes nutrient enrichment in water bodies (Huang et al., 2011; Ou et al., 2012). Our results indicate that grassland promoted nutrient enrichment in the Tiaoxi Basin. Grassland in the Tiaoxi drainage area mainly evolved from abandoned farmland, and is distributed in the near the river banks. It is

4.1. Relationships between selected landscape variables and water chemistry In the whole study period, the water quality parameters selected in this study were significantly correlated with land use types, river properties on reach-scale, and hydrological landforms at the catchment scale. The correlations between the land use types and the selected water quality parameters were slightly larger at the catchment scale than at the local scales (riparian corridor and river reach scale). Consistent with other earlier studies, this implies that, when solving water quality problems at the local scale, we should focus on the effects of catchment-scale variables, such as land use or hydrogeomorphic features, on river water quality (Roth et al., 1996; Sliva and Dudley Williams, 2001). Wang et al. (2006) indicated that the ecological quality of river water bodies with less human activities was mainly controlled by local environmental factors in watersheds. As human activities increase in the river basin and in the riverbank area, the influence of variables at the river basin scale on the river habitat gradually increases, while the influence of the local-scale variables gradually decreases. Therefore, future studies of how human activities impact on river habitat quality should consider interactions between the effects from pressures at different scales. Results from RDA for the whole study period show that landscape variables had slightly more influence on variations in water quality at the river catchment scale than at the riparian zone scale. This shows that the landscape in the study area is subject to some degree of 8

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Fig. 6. Results of redundancy analysis of water chemistry during the rainy season and landscape variables at three scales. Note: a – reach-scale; b – riparian scale; c – catchment scale; d – Physiography most significant.

continued to decline. During this period, the variations in water quality at the river reach scale became the most important. The land use in the riparian corridor category is dominated by cultivated land, forests, and urban areas. Planted bamboo and orchards dominate the forests (see in Fig. 2) (Liang et al., 2008). During the rainy season, persistent rainfall is the strongest driver of soil erosion in the region, enhancing the linkages between land use in the landscape and nutrients in the river basin. The increases in the soil erosion rate and surface runoff during the rainy season mean that large quantities of suspended matter with adsorbed particulate nutrients are delivered to the water body (Li and Zhang, 2008; Ou et al., 2012). Comparison of the three time periods, namely before, during, and after the rainy season, indicates that changes in the influence of landscape variables on water quality variation are similar in two of these periods. The explanatory power of landscape variables on water quality variability is low in the periods before and after the rainy season, perhaps because there was little surface runoff before and after the rainy season, when vegetation was growing. Nutrient absorption was greater and more surface runoff was intercepted in these periods, resulting in a decrease in the influence of landscape variables on variations in stream water quality.

low-lying and steeply sloped, with grazing livestock, poultry, and resting and foraging birds like egrets scattered throughout. Nutrients migrate easily from the grassland into the water body in surface runoff, resulting in enrichment of C, N, and P in the water body. The increase in the proportions of water bodies relative to the catchment at the basin and riparian corridor scale means that the enrichment of C, N, and P will persist throughout the water body. Previous studies have also shown that the ratio of the area occupied by reservoirs and lakes in the upstream areas can lead to hydrological degradation and an increase in particulate C concentrations in water bodies (Frost et al., 2009; Marzin et al., 2013). An increase in the percentage of the area of water may indicate, to some extent, an increase in the transport of nutrients in a region.

4.2. Landscape influences on water chemistry at different seasons and spatial scales The effects of different scales and different landscape types on C, N, and P in river water varied seasonally (Buck et al., 2004; Peterson et al., 2011). Compared to the whole study period, the influence of landscape variables on water quality in the river basin before the rainy season increase significantly in both the natural environment and at the riparian corridor scale, and that the influence and explanatory power of the catchment scale and the river reach scale variables decline visibly. From before∼ to during the rainy season, the influence of landscape variables on changes in water quality continued to increase, while the influence of natural environment and catchment scale variables

5. Conclusions All of the selected variables at three spatial-scales had an effect on the variation of water quality in the Tiaoxi Basin, and the relative influence of variables varies at different temporal and spatial scales. From 9

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Fig. 7. Results of redundancy analysis of water chemistry after the rainy season and landscape variables at three scales. Note: a – reach-scale; b – riparian scale; c – catchment scale; d – physiography

References

dry seasons (e.g., before and after rainy season) to the rainy season, the influence of physiography and landscape variables at the catchment scale decreased slightly, while the effects of variables at the reach and riparian corridor scales increased noticeably. Generally, the influence of variables at the catchment scale was relatively strong and stable while the impacts of variables at the local scale were relatively weak and fluctuated widely with seasons. So, RDA method adopted in this study was proven to be a powerful tool in identifying the optimal time-space scales, variables or their combination on variations of water chemistry within a basin. Considering their different responses, the use of multiple pace-time scale landscape groups may be appropriate to monitor river water quality. Because the selected variables could explain less than 50% variations of water chemistry, revised methods (e.g., the more fine-scale division, the introduction of new landscape variables) should be introduced in exploring the scale effects of landscape on river water chemistry in the future.

Ahearn, D.S., Sheibley, R.W., Dahlgren, R.A., Anderson, M., Johnson, J., Tate, K.W., 2005. Land use and land cover influence on water quality in the last free-flowing river draining the western Sierra Nevada, California. J. Hydrol. 313, 234–247. Allan, J.D., 2004. Annu. Rev. Ecol. Evol. Syst. 35, 257–284. Buck, O., Niyogi, D.K., Townsend, C.R., 2004. Scale-dependence of land use effects on water quality of streams in agricultural catchments. Environ. Pollut. 130, 287–299. Dow, C.L., Arscott, D.B., Newbold, J.D., 2006. Relating major ions and nutrients to watershed conditions across a mixed-use: water-supply watershed. J. North Am. Benthol. Soc. 25, 887–911. Editorial board of Water and wastewater monitoring analysis method, 2002. Water and Wastewater Monitoring Analysis Method(fourth). China Environmental Science Press, Beijing. Esselman, P., Allan, J., 2010. Relative influences of catchment- and reach-scale abiotic factors on freshwater fish communities in rivers of northeastern Mesoamerica. Ecol. Freshw. Fish 19, 439–454. Frissell, C., Liss, W., Warren, C., Hurley, M., 1986. A hierarchical framework for stream habitat classification: viewing streams in a watershed context. Environ. Manage. 10, 199–214. Frost, P.C., Stelzer, R.S., Lamberti, G.A., Elser, J.J., 2002. Ecological stoichiometry of trophic interactions in the benthos: understanding the role of C:N:P ratios in lentic and lotic habitats. J. North Am. Benthol. Soc. 21, 515–528. Frost, P.C., Kinsman, L.E., Johnston, C.A., Larson, J.H., 2009. Watershed discharge modulates relationships between landscape components and nutrient ratios in stream seston. Ecology 90, 1631–1640. Huang, J.L., Li, Q.S., Hong, H.S., Lin, J., Qu, M.C., 2011. Preliminary study on linking land use & landscape patternand water quality in the Jiulong River watershed. Environ. Sci. 32 (1), 64–72. Huzhou Water Resource Bureau, 2004. http://www.hzwr.gov.cn/newsmb.asp?f= 2004102509140.htm. Hynes, H., 1975. Edgardo Baldi, memorial lecture, The stream and its valley. Verhandlungen Der Internationalen Vereinigung Fur Theoretische Und Angewandte Limnologie 19, 1–15. Johnson, L., Richards, C., Host, G., Arthur, J., 1997. Landscape influences on water chemistry in Midwestern stream ecosystems. Freshw. Biol. 37, 193–208.

Acknowledgments This research was funded by the Special Forestry Project of Public Interests (grant numbers 201404305 and 200904001) and “The Lecture and Study Program for Outstanding Scholars from Home and Abroad” (grant number CAFYBB2011007). We also thank Edanz Editing China for linguistic assistance.

10

Ecological Indicators xxx (xxxx) xxx–xxx

L. Cui et al.

Sci. 46, 205–221. Roth, N., Allan, J.D., Erickson, D., 1996. Landscape influences on stream biotic integrity assessed at multiple spatial scales. Landsc. Ecol. 11, 141–156. Sadys, M., Strzelczak, A., Grinn-Gofron, A., Kennedy, R., 2015. Application of redundancy analysis for aerobiological data. Int. J. Biometeorol. 59 (1), 25–36. Sandin, L., Johnson, R.K., 2004. Local landscape and regional factors structuring benthic macroinvertebrate assemblages in Swedish streams. Landsc. Ecol. 19, 501–514. Schoonover, J.E., Lockaby, B.G., Pan, S., 2005. Changes in chemical and physical propertiesof stream water across an urban-rural gradient in western Georgia. Urban Ecosyst. 8 (1), 107–124. Sheldon, F., Peterson, E.E., Boone, E.L., Sippel, S., Bunn, S.E., Harch, B.D., 2012. Identifying the spatial scale of land use that most strongly influences overall river ecosystem health score. Ecol. Appl. 22, 2188–2203. Sliva, L., Dudley Williams, D., 2001. Buffer zone versus whole catchment approaches to studying land use impact on river water quality. Water Res. 35, 3462–3472. Tong, S.T., Chen, W., 2002. Modeling the relationship between land use and surface water quality. J. Environ. Manage. 66, 377–393. Udy, J.W., Fellows, C.S., Bartkow, M.E., Bunn, S.E., Clapcott, J.E., Harch, B.D., 2006. Measures of nutrient processes as indicators of stream ecosystem health. Hydrobiologia 572, 89–102. Wang, L., Seelbach, P.W., Hughes, R.M., 2006. Introduction to landscape influences on stream habitats and biological assemblages. In: American Fisheries Society Symposium 48. American Fisheries Society. pp. 1. Wang, G., Yinglan, A., Xu, Z., Zhang, S., 2014. The influence of land use patterns on water quality at multiple spatial scales in a river system. Hydrol. Processes 28 (20), 5259–5272. Weijters, M., Janse, J., Alkemade, R., Verhoeven, J., 2009. Quantifying the effect of catchment land use and water nutrient concentrations on freshwater river and stream biodiversity. Aquat. Conserv. Mar. Freshw. Ecosyst. 19, 104–112. Xiao, H., Ji, W., 2007. Relating landscape characteristics to non-point source pollution in mine waste-located watersheds using geospatial techniques. J. Environ. Manage. 82, 111–119.

Kim, J., Choi, J., Choi, C., Park, S., 2013. Impacts of changes in climate and land use/land cover under IPCC RCP scenarios on streamflow in the Hoeya River Basin Korea. Sci. Total Environ. 452–453, 181–195. King, R.S., Baker, M.E., Whigham, D.F., Weller, D.E., Jordan, T.E., Kazyak, P.F., Hurd, M.K., 2005. Spatial considerations for linking watershed land cover to ecological indicators in streams. Ecol. Appl. 15 (1), 137–153. Li, L.J., Zhang, Q., 2008. Application of a surface runoff and groundwater coupled model to Xitiaoxi catchment. J. Soil Water Conserv. 22 (4), 56–61. Liang, T., Wang, H.N., Cao, C.Y., Zhang, C.S., Li, H.T., Li, H.P., Song, W.C., Chong, Z.Y., 2008. Estimation of ammonia nitrogen load from nonpoint sources in the Xitiao River catchment, China. J. Environ. Sci. 20 (10), 1195–1201. Liu, X., Lu, X., Chen, Y., 2011. The effects of temperature and nutrient ratios on Microcystis blooms in Lake Taihu, China: an 11-year investigation. Harmful Algae 10, 337–343. Marzin, A., Verdonschot, P.M., Pont, D., 2013. The relative influence of catchment, riparian corridor, and reach-scale anthropogenic pressures on fish and macroinvertebrate assemblages in French rivers. Hydrobiologia 704, 375–388. Meng, W., Zhang, Y., Qu, X.D., 2011. River Ecology Investigation Technique. China SciencePress, Beijing. Novikmec, M., Hamerlík, L., Kočický, D., Hrivnák, R., Kochjarová, J., Oťaheľová, H., Paľove-Balang, P., Svitok, M., 2016. Ponds and their catchments: size relationships and influence of land use across multiple spatial scales. Hydrobiologia 774 (1), 155–166. Ou, Y., Wang, X.Y., Geng, R.Z., 2012. The influences of different landscape characteristics on water quality in the upper watershed of Miyun Reservoir. Acta Sci. Circumst. 32 (5), 1219–1226. Peterson, E.E., Sheldon, F., Darnell, R., Bunn, S.E., Harch, B.D., 2011. A comparison of spatially explicit landscape representation methods and their relationship to stream condition. Freshw. Biol. 56 (3), 590–610. Qu, W., Mike, D., Wang, S., 2001. Multivariate analysis of heavy metal and nutrient concentrations in sediments of Taihu Lake, China. Hydrobiologia 450, 83–89. Redfield, A.C., 1958. The biological control of chemical factors in the environment. Am.

11