Journal of Hydrology 569 (2019) 218–229

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Water quality variability in the middle and down streams of Han River T
under the influence of the Middle Route of South-North Water diversion
project, China
⁎
Yi-Ming Kuoa,1, , Wen-wen Liua,1, Enmin Zhaoa, Ran Lia, Rafael Muñoz-Carpenab
a
School of Environmental Studies, China University of Geosciences, Wuhan 430074, China
b
Department of Agricultural and Biological Engineering-IFAS, University of Florida, Gainesville, FL, USA

A R T I C LE I N FO A B S T R A C T

This manuscript was handled by Huaming Guo, The middle and down streams of Han River are complex river systems influenced by hydrologic variations,
Editor-in-Chief, with the assistance of Chong- population distributions, and the engineering projects. The Middle Route of China’s South-to-North Water
Yu Xu, Associate Editor Transfer (MSNW) project planned to transfer 95 billion m3 annually from Han River to north China. The op-
Keywords: eration of the MSNW project may alter the flow rate and further influence the water quality of Han River. This
Min/max autocorrelation factor analysis study used min/max autocorrelation factor analysis (MAFA) and dynamic factor analysis (DFA) to analyze
Dynamic factor analysis spatio-temporal variations of the water quality variables in three typical tributary-mainstream intersection zones
Han River in Han River from June 2014 to April 2017. MAFA results showed that chlorophyll-a (Chl-a), chemical oxygen
Flow rate
demand (COD), suspend solid (SS) and phosphate (PO43−) (represented as trophic dynamics) are main con-
Water transfer
Water quality variation
cerned water quality variables in densely populated zones (Zones 1 and 3), and total nitrogen (TN), nitrate
nitrogen (NO3−), COD, and PO43− (regarded as nutrient formations dynamics) represent the underlying water
quality variations in agricultural cultivation zone (Zone 2). DFA results indicated that domestic and municipal
effluent pollutants influence the organic concentrations and nutrient formations in the mainstream in Zones 1
and 2. The non-point source nitrogen and phosphorus discharged from the tributaries Tangbai and Hanbei Rivers
elevate the nutrient concentrations and increase Chl-a concentration (i.e. promote the algal growth) in densely
populated zones. In addition, controlling the flow rates in low and middle flow rate conditions can avoid de-
grading water quality. The flow rate should be elevated to more than 700 cm (cubic meters per second) in the
middle stream and to more than 800 cm in the downstream of Han River for preventing water quality dete-
rioration from high loadings of organic pollutants and nutrients. The integrated MAFA and DFA method es-
tablishes an efficient analysis distinguishing spatio-temporal variation of water quality variables and provides
useful site-specific management to control water quality in various flow conditions.

1. Introduction
The ongoing urbanization and the augment of fertilizer application
substantially increase the loadings of nutrients (including nitrogen and
phosphorus) and organic matters into river streams, thereby deterior-
ating water quality. In big cities (more than 1 million population), the
discharge of industrial and domestic wastewaters containing high or-
ganic content may adversely affect the aquatic life, drinking water
quality, and agricultural activities (Kambole, 2003; Pekey et al., 2004).
Runoff volume carries agricultural pollutants (nitrogen and phos-
phorus) from farming lands into rivers in high precipitation period (Wei
et al., 2009). Municipal pollution and agricultural non-point source
pollution are two main sources providing organic pollutants and nu-
trients which may also influence the algal growth (Li et al., 2008).
Therefore, a better understanding of water quality variations and their
driving factors is critical for river water quality management.
Flow rate is considered as the key factor influencing the self-pur-
ification and interactions among water quality variables in river eco-
systems (Junk, 1989; Bunn and Arthington 2002). The flow rate can be
altered by precipitation and runoff volumes from various land uses for
agriculturalization or urbanization in a watershed (Vega et al., 1998;
Ahearn et al., 2005; Lin et al., 2013). In addition, the operation of
engineering projects may reduce the flow rate and consequently cause
eutrophic states in a river system (Huang et al., 2009; Poff et al., 2007;

⁎
Corresponding author.
E-mail address: airkuo@ntu.edu.tw (Y.-M. Kuo).
1
These authors contributed to the work equally.

https://doi.org/10.1016/j.jhydrol.2018.12.001
Received 28 May 2018; Received in revised form 30 November 2018; Accepted 1 December 2018
Available online 06 December 2018
0022-1694/ © 2018 Elsevier B.V. All rights reserved.
Y.-M. Kuo et al.
Wang et al., 2011a,b). However, the relation between water quality
variables and flow rate with strong spatio-temporal variation remains
unexplored because of the complex hydrochemical mechanism in river
system.
Dynamic process-based models such as SWAT (Arnold et al., 2014),
HSPF (Xie and Lian, 2013), and the AnnAGNPS model (Que et al.,
2015) have been applied to explain the relationship between variation
of environmental variables (e.g. temperature and precipitation) and
water quality dynamics. However, they need complex user-supplied
parameters (usually > 50 parameters) to perform physicochemical si-
mulations (Jackson-Blake et al., 2017). In addition, multivariate sta-
tistical analysis methods such as cluster analysis and factor analysis
(Momen et al., 1996; Lau and Lane, 2002; Arhonditsis et al., 2004)
mostly ignore the spatio-temporal variation of data instead of using the
average data of time or space (Yang et al., 2012). Therefore, a statistic
method considering the spatio-temporal variation may be an available
option for investigating the relationships among water quality vari-
ables.
Min/max autocorrelation factor analysis (MAFA) is used to extract
trends from multiple time series that represent the common behavior of
the original time series (Zuur et al., 2007). Dynamic factor analysis
(DFA), a dimension-reduction technique, is used to identify underlying
common trends among multivariate time series which also considers
the effects of explanatory variables on response variables (Kuo and Lin,
2010; Zuur and Pierce, 2004). MAFA has been widely applied for
spatio-temporal analysis, including the evolution of fish stocks in time
(Woillez et al., 2009), the water quality variation of river estuary (Kuo
et al., 2013), and the temporal variability of environmental variables in
German Bight (Shojaei et al., 2016). DFA has been used for studies of
aquatic ecological monitoring (Kuo and Lin, 2010; Ligas et al., 2010),
groundwater levels and quality (Muñoz-Carpena et al., 2005; Kuo and
Chang, 2010), atmospheric environment and pollution (Kuo et al.,
2011; Kuo et al., 2013), and the interactions between environmental
variables and algal abundance (Kuo and Wu, 2016). The combination of
MAFA and DFA can effectively deal with the regional water quality
characteristics and the relationship among the water quality variables
(Kuo et al., 2013).
The Han River basin where the population distributed unevenly has
more than 10 tributaries (Li et al., 2008). Several big cities located
along the mainstream of Han River with more than 1.5 million people
discharged in the range of 0.8–2.0 × 107 kg wastewaters into Han River
in 2010 (Zeng et al., 2014). Two seriously polluted feed-in tributaries
(Tangbai River and Hanbei River) discharged 2.1–3.8 × 106 kg/yr
phosphorus and 0.9–1.8 × 106 kg/yr ammonia nitrogen (NH4-N) into
Han River and finally deteriorated the water quality in the main stream
(Huang et al., 2009; Li et al., 2011). The Middle Route of China’s South-
to-North Water Transfer (MSNW) project planned to transfer 95 bil-
lion m3 annually from Han River to north China and the Yangtze-Han
Water Diversion (YHWD) project diverted water from the middle
stream (Jingzhou City) of Yangtze River to the downstream of Han
River (from Zhongxiang City to Wuhan City) to mitigate the effect of
the MSNW project on Han River since 2014 (Fig. 1). The operation of
the MSNW and YHWD projects may alter the flow rate of Han River
(Yang et al., 2012). These hydrological variations may influence water
self-purification and further influence the water quality of Han River
(Xie et al., 2004; Zhao et al., 2010).
This study statistically analyzes spatio-temporal variations of the
water quality variables in three typical zones (zones in the intersection
between the mainstream and two tributaries or an engineering project)
in Han River from June 2014 to April 2017. The objectives of the study
are to (i) determine the main concerned water quality variables in the
main stream of Han River of three zones, (ii) examine the environ-
mental factors influencing the temporal variations of main concerned
water quality variables in three zones, (iii) identify the relationships
between flow rate and environmental variables and provide useful
management to control water quality of Han River. With knowledge of
219
Journal of Hydrology 569 (2019) 218–229
these crucial factors, the results can be valuable for developing man-
agement strategies to improve water quality.
2. Study area and method
2.1. Study area
Han River located in the central China is the longest tributary of the
Yangtze River. The middle stream of Han River is defined from
Danjiangkou Reservoir to Jingzhou City (270 km) and the downstream
is defined from Jingzhou to Wuhan City (406 km) where it joins the
Yangtze River (Fig. 1). The total drainage area in the middle and down
streams is 64,000 km2, with an elevation from 500 m at Danjiangkou
Reservoir to 110 m at Wuhan City (Li et al., 2008). Annual precipita-
tions range from 700 to 1300 mm from the middle stream to the
downstream of Han River (Jiang et al., 2004). 80% of precipitation
concentrates in the period from May to October (Fig. 2). Jianghan plain,
with total drainage area of 46,000 km2, where 70% of land use is
agricultural land, is located in the downstream, with an elevation from
110 to 200 m (Zhang and Dong, 2009). Agricultural activities (such as
irrigation and fertilization) in Jianghan plain may affect the hydro-
logical conditions and water quality (Zhou, 2009; Li et al., 2012).
Undergoing rapid social and economic development has gradually
deteriorated water environment of the middle and down streams of Han
River watershed in the past 20 years (Chou et al., 2007; Zhong et al.,
2012). The average annual concentrations of organic substances, ni-
trogen, and phosphorus were increased by more than 20% from 1998 to
2009 (Luo et al., 2011). Tangbai River and Hanbei River are two
heavily polluted tributaries, which both flow through agricultural areas
and discharge nutrients (nitrogen and phosphorus) into the mainstream
of Han River (Liang et al., 2012). Xiangyang City and Zhongxiang City
(Fig. 1) located in the tributary-mainstream intersection zones with
more than 1 million population (Sun and Wu, 2012) discharged
2.1 × 107 kg/yr and 1.4 × 107 kg/yr domestic and municipal organic
pollutants into the mainstream respectively in 2007 (Wang et al.,
2011a,b). The high mass of organic pollutants and nutrients have re-
sulted in several large scale algal blooms in the middle and down
streams of Han River in recent 15 years (Chou et al., 2007; Zhang et al.,
2012; Zhang et al., 2016).
After the operation of the MSNW project, the flow rate discharged
from Danjiangkou Reservoir to Han River was decreased and the flow
rate in Tangbai River contributed more than 30% of that in the main-
stream of Han River (Liang et al., 2016). In addition, the YHWD project
diverts water from the middle stream (Jingzhou City) of Yangtze River
to Han River (Fig. 1) to mitigate the effect of the MSNW project on Han
River. The YHWD project-Han River intersection zone is located in
Jianghan plain where the agricultural land is the main land use pro-
ducing non-point pollutants (agricultural nitrogen and phosphorous)
into the mainstream (Zeng et al., 2014). Two tributaries (Tangbai and
Hanbei Rivers) and two engineering projects (MSNW and YHWD pro-
jects) play important roles for regulating water quality variation in the
mainstream of Han River. Therefore, the zones of tributary Tangbai-
Han River intersection, YHWD project-Han River intersection, and tri-
butary Hanbei-Han River intersection are defined as Zone 1, Zone 2,
and Zone 3 (Table 1 and Fig. 1), respectively, for further discussion in
the study.
The 9 water sampling sites were distributed from the middle and
down streams of Han River watershed. Among the 9 sampling sites,
sites Main1, Main2, and Main3 are located at the upstream from the
intersections of the tributary and the mainstream of Han River in Zones
1, 2, and 3, respectively, which represent the water quality in the
mainstream before the tributaries or YHWD project flows into Han
River. On behalf of the water quality in tributaries or YHWD project,
sites Tri1, Tri2, and Tri3 are located at tributary Tangbai River (Zone
1), YHWD project (Zone 2), and tributary Hanbei River (Zone 3), re-
spectively. Sites at the downstream from the intersections of the
Y.-M. Kuo et al. Journal of Hydrology 569 (2019) 218–229

Fig. 1. Monitoring sites of water quality in the study area (1: the MSNW project; 2: the YHWD project). Sites Main1, Tri1, and Int1 are located in Zone 1, Sites Main2,
Tri2, and Int2 are located in Zone 2, and Sites Main3, Tri3, and Int3 are located in Zone 3.

Fig. 2. The precipitation of the three zones (Xiangyang City in Zone 1, Zhongxiang City in Zone 2, and Wuhan City in Zone3) during the monitoring period.

Table 1
The regional watershed characteristics of the three zones. Yangtze River: water body was transferred and influenced by Yangtze River. Sites Int1, Int2, and Int3 are
influenced by both the mainstream sites (Main1, Main2, and Main3) and the tributary sites (Tri1, Tri2, and Tri3).
Zones Average Q (cms) Annual precipitation (mm) Sites City or Tributary Main land use Main pollutants

Zone 1 588 702 Main1 Xiangyang City Urban Municipal

wastewaters
Tri1 Tangbai River Agriculture Nutrients
Int1 None Agriculture From Main1 + Tri1
Zone 2 795 892 Main2 Zhongxiang City Urban + Agriculture Municipal wastewaters
+Nutrients
Tri2 YHWD project Yangtze River Nutrients
Int2 None Agriculture From Main2 + Tri2
Zone 3 1031 1245 Main3 None Urban + Agriculture Municipal
wastewaters
Tri3 Hanbei River Agriculture Nutrients
Int3 None Urban + Agriculture From Main3 + Tri3

220
Y.-M. Kuo et al.
tributary and the mainstream of Han River in Zones 1, 2, and 3, re-
spectively (Int1, Int2, and Int3, respectively), represent the water
quality in the mainstream influenced by both upstream Han River and
tributaries or engineering project.
2.2. Water quality analysis
Water samples were collected monthly from June 2014 and April
2017 (9 sites with 306 water samples in total) from 50 cm below a
water table by using a Kemmerer Water Sampler (Wildco, Yulee, FL,
USA). In each sample, a total of 13 environmental variables including
air temperature (TEMP), flow rate (Q), pH, chemical oxygen demand
(COD), dissolved oxygen (DO), electrical conductivity (EC), suspend
solid (SS), chlorophyll-a (Chl-a), NH4-N, nitrate-nitrogen (NO3-N), total
nitrogen (TN), total phosphorous (TP), and orthophosphate (PO43−)
were analyzed. TEMP, DO, EC, SS, and pH were measured immediately
in situ. All water samples were filtered (0.45 μm Millipore nitrocellulose
filter) in the field and the samples were acidified with ultra-pure HCl to
pH < 2 and stored at 4 °C in ice box for analyzing physiochemical
variables. TP was measured by a colorimetric method after digestion
(digested by the ammonium molybdenum blue method and analyzed by
the ascorbic acid method). NH4-N was measured by Nessler’s reagent
spectrophotometry (752 ltraviolet-visible spectrophotometer,
Biochrom, USA). NO3-N and TN were determined by the alkaline po-
tassium per sulfate oxidation–UV spectrophotometric method. COD was
measured by microwave digestion method (Sarkar et al., 1986), and
Chl-a was analyzed by the acetone frozen extraction spectro-
photometric method (Vollenweider, 1968).
2.3. Methods
2.3.1. Min/max autocorrelation factor analysis
MAFA is used to extract trends from multiple time series, to estimate
index functions from time series that represent the common behavior of
the original time series, and for smoothing (Switzer and Green 1984;
Shapiro and Switzer 1989; Solow, 1994). MAFA estimates the axes that
have decreasing auto-correlation at time lag 1 (Zuur et al., 2007).
MAFA divides the set of initial variables into a series of axes. The
number of important MAF axes was determined by obtaining p-values
for MAF axes through a randomization process (p > 0.80, Solow,
1994). The axis represents the underlying pattern, or the trend in the
data. The first MAFA axis (MAF1) is the main trend underlying the
entire time series that exhibits a maximal auto-correlation between
nearby observations, and the other axes represent less crucial trends.
Thus, the first MAFA axis has the highest auto-correlation with time lag
1 and the second axis has the second highest auto-correlation with time
lag 1. A similar trend occurs for all axes. The canonical correlation
coefficients (R) between the variables time series and the MAFA axes
were obtained to determine the relationship between each variable and
particular MAFA axes. The terms ‘high’, ‘moderate’, ‘weak’ and ‘minor’,
as applied to canonical correlation coefficients, refer to absolute values
of > 0.75, 0.75–0.5, 0.5–0.25 and < 0.25, respectively. In this study,
MAFA axes represent the water quality variation and the canonical
correlation coefficients were used to determine the significance of the
water quality variables, which reflect the regional characteristics.
Further details on MAFA are presented in Zuur et al. (2007) and Erzini
(2005).
2.3.2. Dynamic factor analysis
As a parameter optimization and dimension-reduction technique,
DFA is useful for determining interactions between time-series response
and possible explanatory variables (Zuur et al., 2003a). In a DFA model,
N time series are expressed as a function of a linear combination of M
common trends, one or more explanatory variables, a constant level
parameter, and noise (Zuur et al., 2003b; Zuur and Pierce, 2004). The
mathematical formulation of the DFA model is as follows:
221
Journal of Hydrology 569 (2019) 218–229
M K
Rn (t ) = ∑ γm,n αm (t ) + μn + ∑ βk,n z k (t ) + εn (t )
m=1 k=1 (1)
αm (t ) = α (t − 1) + ηm (t ) (2)
where Rn (t ) is the value of the nth response variable (RV) at time t;
αm (t ) is the mth common trend at time t, and γm, n is the factor loading or
weighting coefficient that indicates the importance of each αm (t ) trend
within the dynamic factor model. βk, n represents the regression coeffi-
cient for the kth explanatory variable (EV) z k (t ) . The t-test is used to
assess whether the EV is significantly related to the RV. γm, n and βk, n
allow for the identification of the most relevant common trends and EVs
for each RV. μn refers to the nth constant level parameter (intercept
term), which increases or decreases the linear combination of common
trends. The terms εn (t ) and ηm (t ) are independent, Gaussian-distributed
noise with zero mean and an unknown diagonal or symmetric/non-di-
agonal covariance matrix. The Nash-Sutcliffe Efficiency
(−∞ ≤ NSE ≤ 1; Nash and Sutcliffe 1970) and Akaike’s information
criterion (AIC; Akaike, 1974) were used to select the optimal-per-
forming DFA model. NSE = 1 indicated that the pairs of predicted and
observed data were consistent. Once a set of common trends was
identified, different combinations of common trends and explanatory
variables are incorporated in the analysis until the minimum AIC (best
DFA model) is identified without exceeding the variance inflation factor
(VIF) criterion (VIF = 5.0) (Zuur et al., 2007). DFA was performed by
using the Brodgar Version 2.6.6 statistical package (Highland Statistics
Ltd., Newburgh, UK), which uses the statistical software language “R”
version 2.9.1 (R Core Development Team, 2009). Further details re-
garding DFA are presented in Zuur et al. (2003a, 2007).
2.3.3. Analytical procedure
In this study, MAFA was first used to determine the main concerned
water quality variables in three zones and investigate the regional
watershed characteristics at the downstream sites from the intersections
of the tributary and the mainstream of Han River (i.e. sites Int1, Int2,
and Int3 shown as black points in Fig. 1). These main concerned water
quality variables were selected as RVs in DFA models. After that, DFA
was used to investigate the interactions among the water quality vari-
ables and what mechanisms of tributaries and engineering projects in-
fluence main concerned water quality variables in Han River. The water
quality variables at the upstream sites from the interaction between the
mainstream and tributaries (i.e. sites Main1, Main2, and Main3 shown
as red points in Fig. 1) and those in tributaries or in the YHWD project
(i.e. sites Tri1, Tri2, and Tri3 shown as red points in Fig. 1) were se-
lected as EVs. The VIF was computed to detect the multicolinearity of
each set of explanatory variables (Zuur et al., 2007). The VIF of ex-
planatory variables greater than five was removed in these analyses
(Ritter et al., 2009).
2.3.4. Model validation
The quality of the optimal DFA model was evaluated by the
FITEVAL code which uses a more general formulation of the coefficient
of efficiency (Ritter and Muñoz-Carpena, 2013). A combination of
graphical results, absolute value error statistics (e.g., root-mean-square
error, RMSE), and normalized goodness-of-fit statistics (e.g., NSE) was
used as quantitative statistics of the predictive accuracy of a model.
Hypothesis testing of the NSE exceeding threshold values is accom-
plished based on approximated probability distributions in the case of
time series data. In conjunction with graphs and indicator values, a
probability of the following model performance ratings was provided as
follows: “unsatisfactory” (NSE < 0.65), “acceptable” (0.65 ≤ NSE <
0.80), “good” (0.80 ≤ NSE < 0.90), and “very good” (NSE ≥ 0.90).
More details of application of FITEVAL code can be found in Wu et al.
(2014), Harmel et al. (2014), Hornero et al. (2016), and Mfumu
Kihumba et al. (2016).
Y.-M. Kuo et al. Journal of Hydrology 569 (2019) 218–229

Table 2
Water quality variables with mean value and coefficient of variation (CV) in 9 different sites of Han River. CV is shown in the parentheses. TEMP represents the air
temperature in each zone. –: no data observed.
Variables Zone1 Zone2 Zone3

Main1 Tri1 Int1 Main2 Tri2 Int2 Main3 Tri3 Int3

TEMP (°C) 14.5 (0.78) 15.9 (0.72) 16.5 (0.69)

Chl-a (μg/L) 4.44 6.04 5.14 4.76 6.51 5.60 5.91 7.93 6.81
(0.40) (0.47) (0.38) (0.34) (0.62) (0.43) (0.30) (0.50) (0.35)
pH 8.23 8.85 8.50 8.32 8.34 7.94 8.04 8.47 7.76
(0.09) (0.13) (0.10) (0.09) (0.08) (0.09) (0.10) (0.29) (0.16)
EC (μs/cm) 365 388 367 345 381 386 367 363 352
(0.09) (0.20) (0.19) (0.15) (0.13) (0.16) (0.12) (0.19) (0.16)
DO (mg/L) 7.31 6.92 7.03 7.57 7.45 6.18 6.19 4.28 6.06
(0.16) (0.15) (0.20) (0.18) (0.13) (0.18) (0.15) (0.11) (0.19)
COD (mg/L) 9.10 13.03 10.99 9.58 10.97 10.13 11.14 19.16 13.48
(0.50) (0.62) (0.53) (0.48) (0.54) (0.44) (0.44) (0.70) (0.44)
SS (mg/L) 26.10 36.34 30.33 27.13 35.23 29.38 33.05 38.25 37.18
(0.23) (0.41) (0.25) (0.24) (0.27) (0.17) (0.23) (0.22) (0.18)
NH4-N (mg/L) 0.59 0.73 0.68 0.55 0.64 0.61 0.67 0.68 0.71
(0.33) (0.31) (0.29) (0.41) (0.36) (0.40) (0.39) (0.43) (0.34)
NO3-N (mg/L) 0.53 0.93 0.33 0.34 0.33 0.41 0.46 0.51 0.35
(0.41) (0.37) (0.33) (0.45) (0.37) (0.33) (0.36) (0.41) (0.28)
TP (mg/L) 0.24 0.35 0.24 0.26 0.45 0.17 0.34 0.51 0.34
(0.19) (0.25) (0.29) (0.26) (0.41) (0.23) (0.25) (0.43) (0.28)
TN (mg/L) 1.67 2.37 1.91 1.26 2.71 1.72 1.01 2.72 1.02
(0.32) (0.32) (0.24) (0.20) (0.42) (0.19) (0.28) (0.28) (0.22)
PO43− (mg/L) 0.112 0.175 0.131 0.126 0.152 0.094 0.180 0.293 0.196
(0.43) (0.56) (0.47) (0.47) (0.55) (0.42) (0.52) (0.52) (0.47)
Q (cms) 445 481 824 770 110 881 998 – 1102
(0.59) (1.02) (0.58) (0.58) (1.81) (0.61) (0.69) – (0.74)

3. Results and discussions
3.1. Descriptive statistical analysis
Table 2 summarizes the mean and coefficients of variation (CV) of
13 physicochemical parameters for the 9 sampling sites. High variations
(with CVs of > 50%) of environmental variables were observed for the
TEMP, Q, Chl-a, TP, PO43−, and COD variables at most sites, indicating
strong fluctuations during the study period from June 2014 to April
2017. Water diversion of the MSNW and YHWD projects result that CVs
of Q at Sites Main1 and Tri2 were greater than 100%.
Fig. 3 illustrates the temporal variations of water quality variables
in three zones. All water quality variables were relatively stable floating
around mean value of each variable. The concentrations of Chl-a among
the 3 zones were varied seasonally, which show high values in early
spring (February-April) and low values in summer (July-August) cor-
responding well with low and high flow rate conditions, respectively. In
addition, TN, TP, and PO43− concentrations showed similar trends
(Fig. 3). Q, Chl-a, TN, and TP concentrations decreased from June to
October and increased from January and April in all three zones. When
the abovementioned concentrations increased, Chl-a, TP and TN con-
centrations were in the range of 4.72–13.48 μg/L, 0.102–0.188 mg/L,
and 2.12–5.88 mg/L, respectively. The highest concentrations and error
estimates of water quality variables were distributed at Sites Tri1, Tri2,
and Tri3 among the three sites in each zone (Fig. 4). Therefore, con-
centrations of most water quality variables in the downstream section
were higher than those in the upstream section of Han River. In addi-
tion, mean values of most water quality variables in Zone 1 (except Site
Tri1) were lower than those in Zones 2 and 3 (Fig. 4), indicating that
water quality in middle stream was better than that in the downstream
Han River.
3.2. Min/max autocorrelation factor analysis (MAFA)
In both Zone 1 and Zone 3, only one MAF axis was selected for
representing the common behavior of the water quality time series in
these zones (Fig. 5 and Table 3) with an autocorrelation of 0.96
(p < 0.01) for Zone 1 and with that of 0.89 (p < 0.01) for Zone 3. Chl-
a, COD, and PO43− showed high positive canonical correlation coeffi-
cients with the MAF axis (R = 0.70, 0.84, and 0.88 for Chl-a, COD, and
PO43−, respectively), while SS moderately and negatively correlated
with the MAF axis (R = −0.71). Chl-a, which has been widely used as
an indicative indicator for estimating phytoplankton biomass (Felip and
Catalan, 2000), had the highest correlation coefficient with the MAF
axis in both Zone 1 (R = 0.61) and Zone 3 (R = 0.58). COD represents
the content of organic matter in water bodies (Chen et al., 2010; Liu
et al., 2018). The reach of Han River in Xiangyang City (Site Main1)
was heavily contaminated by domestic and municipal organic pollu-
tants (represented as COD) (Li et al., 2008). Tangbai River and Hanbei
River discharged more than 3.0 × 106 kg/yr of phosphorus to the
mainstream since 2009 (Li et al., 2011). The increases of COD and
PO43− concentrations can promote the growth of phytoplankton which
can increase Chl-a concentration and influence the trophic states
(Zheng et al., 2010). Large scale algal blooms have occurred five times
in the downstream Han River during 2002–2012, endangering the
safety of drinking and agricultural water (Luo et al., 2011; Zhang et al.,
2012). Therefore, the MAF axis can be interpreted as the trophic dy-
namics for water quality in these zones.
Two MAF axes were selected to represent the time series variation of
water quality in Zone 2 (Fig. 5 and Table 3), with an autocorrelation of
0.89 (p < 0.01) for the first MAF axis (MAF1) and that of 0.81
(p < 0.01) for the second MAF axis (MAF2), respectively. TN showed
moderate and positive correlation with the MAF1 (R = 0.63). COD was
positively correlated with the MAF2 (R = 0.61), while NO3− and
PO43− were negatively related to the MAF2 (R = −0.59 for NO3− and
R = −0.81 for PO43−). TN, NO3−, and PO43− represent the variations
of inorganic nutrient variables in water body (Hakanson, 1980; Elser
et al., 2000). Because agricultural cultivation land accounts for 70% of
Jianghan plain, nutrients such as agricultural fertilizers (organic pol-
lutants represented as COD) have been widely used and may finally
contaminate water quality. The organic agricultural fertilizers can be
oxidizing decomposed to inorganic nutrients (Dolman et al., 2012;
Paerl et al., 2011) and alter nutrient formations in water bodies.
Therefore, the MAF axis can be interpreted as the nutrient formation

222
Y.-M. Kuo et al.
Fig. 3. The averaged value of sampling sites at each zone for flow rate, Chl-a,
COD, SS, TN, TP, and PO43− monitored from July 2014 to April 2017.
dynamics for water quality in Zone 2.
3.3. Dynamic factor analysis
3.3.1. The optimal DFA model
Because high and positive correlation (r = 0.775–0.857, p < 0.05)
was found between Q and precipitation in all sampling sites and pre-
cipitation had a high VIF value (VIF = 12.78), precipitation was
223
Journal of Hydrology 569 (2019) 218–229
Fig. 4. Temporal mean (during the study period) concentrations of water
quality variables at 9 sites in Zones 1, 2, and 3, respectively. Error estimates are
expressed as one standard deviation of the mean values. The horizontal axis
shows the relative distance among the three zones.
removed from candidate EVs for determining the optimal DFA model
and should be considered together with Q, once Q was a significant EV
in the optimal model. The optimal DFA model (with the lowest AIC)
contained one common trend, various EVs, and a symmetric non-di-
agonal covariance matrix. In Zone 1, air temperature, COD at Site
Main1, and Chl-a, TP, TN, and Q at Site Tri1 were optimally fitted to the
time series of RVs. In Zone 2, Q, COD, TP, and PO43− at Site Main2, and
DO and TN concentrations at Site Tri2 were significant EVs in the op-
timal DFA model. Q, Chl-a, and TN at Site Main3, and COD, TP, and Q
at Site Tri3 were optimally fitted to the time series of the main con-
cerned water quality in Zone 3.
Table 4 shows the factor loadings (γ) of the optimal DFA models in
these three zones. The magnitude of factor loadings indicated the re-
lation of the common trend to the time series of the response variable
within the optimal DFA models. For all three zones, the magnitudes of
factor loadings for the common trend were less than 0.31, indicating
that the common trend is not important for influencing the variability
of response variables. Therefore, the explanatory variables are more
effective for the interpretation of what mechanisms of tributaries and
engineering projects influence water quality (response variables) in the
mainstream Han River.
Y.-M. Kuo et al. Journal of Hydrology 569 (2019) 218–229

(a)

(b) MAF1

(c) MAF2

(d)

Fig. 5. The MAFs and the canonical correlation coefficients of water quality variables of the MAFs of (a) Zone 1, (b) + (c) Zone 2, and (d) Zone 3, respectively.

Table 3 oxidizing decomposed to inorganic nutrients (e.g., PO43− and NO3−)

Main concerned water quality variables in each MAF axis at each zone. by chemoautotrophic bacteria (e.g., Nitrosomonas, Nitrobacter, etc.) in
Zone MAF axis Main concerned water quality Water quality dynamics
organic pollution states (high COD and low DO values) (McCauley
variables et al., 1989). Thus, a negative correlation between COD and NO3−
(r = −0.51) and that between COD and PO43− (r = −0.46) were
1 MAF1 Chl-a, COD, SS, PO43− Trophic dynamics found, and consequently COD and nutrients showed the opposite re-
2 MAF1 TN Nutrient formations
MAF2 NO3−, COD, PO43−
lationship in all three zones (Table 4). Precipitation can increase runoff
3 MAF1 Chl-a, COD, SS, PO43− Trophic dynamics volume carrying with sediment into a river, which increases SS in a
river (Qian et al., 2018). High and positive relationships between Q and
SS were found (r = 0.65, p < 0.05 for Zone 1 and r = 0.69, p < 0.05
3.3.2. Interactions among the water quality variables for Zone 3). In addition, increasing flow rate can dilute the con-
Table 4 illustrates the regression parameters (βk,n) which quantify centrations of Chl-a, COD, and nutrients (TN and TP) (Yu et al., 2016).
the relative contribution of each EV on each response time series with a Therefore, Q and SS concentrations were negatively related to Chl-a,
bold number indicating significant parameters by t test (t value > 2). A nutrients, and COD concentrations (Table 4).
total of nine water quality variables including Q, TEMP, COD, DO, TP, Chl-a concentrations at Sites Int1 and Int3 were positively related to
PO43−, TN, NO3−, and Chl-a were included in the optimal DFA models. both TP and COD in the tributaries in Zones 1 and 3 (Table 4). Some
Influencing mechanism among the water quality variables and the organic matters and high phosphorus concentration can promote the
source identification of pollutants are discussed in this section, and the activity of enzymes in the hydrophytes (Leboulanger et al., 2001),
relationship between flow rate and water quality variables is discussed which can be assimilated by aquatic plants, algae, and bacteria (Zheng
in Section 3.4. et al., 2010). In addition, Chl-a concentrations at Sites Int1 and Int3
In all three zones, PO43− (inorganic phosphorus) concentration were positively related with those at Sites Tri1 (Tangbai River) and
occupied more than 50% of TP (54%, 61%, and 59% in Zones 1, 2, and Main3 (upstream of zone 3 in the downstream Han River), respectively.
3, respectively), and the sum of NH4− and NO3− (inorganic nitrogen) In Zone 3, concentrations of nutrients and organic pollutants were
concentration occupied more than 69% of TN (77%, 75%, and 69% in much higher than those in Zones 1 and 2 (Fig. 4), especially during
Zones 1, 2, and 3, respectively) during the study period except July and early spring from February to April. During this period, the flow rate in
August in both 2015 and 2016 years. Therefore, most of TN and TP acts zone 3 is the lowest (Fig. 3) affected by the high backwater discharged
as inorganic compounds. The organic compounds in TN and TP can be from Yangtze River into Han River (Dong et al., 2010). Algae prefer to

224
Y.-M. Kuo et al. Journal of Hydrology 569 (2019) 218–229

Table 4
Common trends and regression coefficients of the explanatory variables for the optimal DFA models in Zones 1, 2, and 3. TEMP-Zone1 represents the variation of air
temperature in Zone 1, and COD-Main1 represents the COD concentration at Site Main1. The bold number of the explanatory variable is statistically significant (t
value > 2).
Zone Common trend No. (CT) Response variables (RV) Factor loadings (γ) Explanatory variables (EV) NSE

TEMP COD Q Chl-a TN TP

-Zone1 -Main1 -Tri1 -Tri1 -Tri1 -Tri1

1 1 Chl-a −0.212 −0.178 0.182 −0.126 0.605 0.395 0.248 0.81

COD 0.019 −0.250 0.364 −0.515 0.111 −0.308 −0.306 0.85
SS 0.115 −0.201 −0.973 0.305 −0.408 −0.831 −0.828 0.83
PO43− 0.283 −0.076 −0.057 0.169 −0.021 0.407 0.555 0.89
CT RV γ Q COD PO43− DO TN NSE
-Main2 -Main2 -Main2 -Tri2 -Tri2
2 1 NO3-N 0.273 0.076 −1.201 0.918 −0.013 0.270 0.87
COD −0.164 −0.042 1.212 −0.173 0.116 −0.015 0.84
TN 0.107 −0.279 −0.324 0.079 0.574 0.324 0.74
PO43− 0.233 −0.449 −0.161 0.534 −0.308 0.166 0.86
CT RV γ Q Chl-a TN COD SS TP NSE
-Main3 -Main3 -Main3 -Main3 -Tri3 -Tri3
3 1 Chl-a −0.175 −0.211 0.596 −0.092 0.430 0.103 0.316 0.90
COD 0.244 −0.116 0.917 −0.173 0.323 −0.410 −0.122 0.81
SS 0.252 0.129 −0.196 −0.013 −1.182 0.312 −0.049 0.84
PO43− 0.034 −0.081 0.892 −0.041 0.076 −0.051 0.851 0.89

grow abundantly in an approximate lentic environment with high or-
ganic pollutants and nutrients (Yang et al., 2016; Isabwe et al., 2018).
Therefore, large scale algal blooms have occurred five times in early
spring in the downstream Han River (Zone 3) rather than in Zones 1 and
2 during 2002–2012 (Chou et al., 2007; Zhang et al., 2012). More re-
lationships between various flow rate conditions and water quality
variables are further discussed in Section 3.4.
Organic pollutants and nutrients in the downstream of Han River
are influenced by multiple sources (i.e. tributaries and upstreams of
mainstream Han River) which exists high spatio-temporal variation. In
Zones 1 and 3, COD concentrations at Sites Main1 and Main3 sig-
nificantly influenced COD concentrations at Sites Int1 and Int3, re-
spectively, and TP concentrations at Sites Tri1 and Tri3 positively re-
lated to PO43− concentrations at Sites Int1 and Int3, respectively
(Table 4). In Zone 1, auto manufacturing is the pillar industry in
Xiangyang City which is also an important organic pollution source of
Han River (Sun et al., 2017). Based on our observations, Xiangyang City
(located near Site Main1) discharged 2.7 × 107 kg/yr domestic and
municipal organic pollutants in 2016 into the mainstream of Han River
and therefore influence the organic concentrations and nutrient cycles.
In Zone 3, high COD concentration in the mainstream may be derived
from the accumulation of pollutant emissions in Zones 1 and 2. The
tributaries Tangbai River and Hanbei River flowing through agri-
cultural regions discharged large amounts of nutrients into Han River
(Liang et al., 2012). Li et al. (2011) and Liang et al. (2012) reported that
the Tangbai River and Hanbei River discharged 6.2 × 106 kg/yr and
3.8 × 106 kg/yr phosphorus to the mainstream Han River, which oc-
cupied 32% and 15% of total amount of phosphorus discharged from
Han River to Yangtze River (at Wuhan City) in 2009, respectively.
In Zone 2, COD concentration and TP concentration at Site Main2
significantly influence the organic pollutants and nutrient concentra-
tions at Site Int2. Zhongxiang City, located at the upstream from the
intersection of YHWD project and the mainstream, discharged
1.7 × 107 kg/yr domestic and municipal organic pollutants in 2016
into Han River (Sun et al., 2017). The organic pollutants and phos-
phorus at Site Main2 may be also derived from the municipal pollutants
(discharged from Xiangyang City) and nutrients (discharged from
Tangbai River) in Zone 1.
3.3.3. Goodness-of-fit evaluation of the response variables
In each zone, time series data of the RVs in the optimal DFA model
were combined together and then were evaluated by the FITEVAL
model. The calibration process was repeated in an attempt to improve
the fit. Fig. 6 illustrates the main results of the evaluation. Data of RVs
in Zones 1 and 3 show “very good” fits (average NSE = 0.958, ranged
from 0.915 to 0.991 in Zone 1, and average NSE = 0.995, ranged from
0.991 to 0.998 in Zone 3; Fig. 6), and data of RVs in Zone 2 show a
“good” to “very good” fit (average NSE = 0.958, ranged from 0.862 to
0.989; Fig. 6). Among all the RVs, TN variable in Zone 2 (with the
lowest NSE, Table 4) fitted as “good” while the other variables were
fitted as “very good”, which resulted in the worse model fit in Zone 2
comparing with those in Zones 1 and 3. The spatio-temporal variability
of fertilizer application and the precipitation in the watershed lead to
the uncertainly of nitrogen fitting in Zone 2. In addition, the nitrogen
goodness-of-fit was difficult to evaluate because of the complex ni-
trogen cycles among various nitrogen formations (e.g., ammonium ni-
trogen, nitrate nitrogen, organic nitrogen) (Harmel and Smith, 2007;
Amiri, 2016).
3.4. Relationship between flow rate and environmental factors
In the optimal DFA model, the flow rates in Sites Tri1, Main2, and
Main3 are negatively correlated with most RVs (main concerned water
quality variables), indicating that increase in flow rate can effectively
improve the water quality. The flow rate can be regarded as an in-
dicative variable representing the regional characteristics. The levels of
flow rate were trisected as low, middle, and high flow rates conditions
of each zone to further discuss the relationships between flow rates and
the environmental variables (Fig. 7).
Low flow rate condition (defined Q less than 500, 700, and 800 cms
in Zones 1, 2, and 3, respectively) mostly occurred from November to
December and from January to March. In this condition, concentrations
of Chl-a, COD, and nutrient variables were high and varied strongly
(Table 5). In this period with low precipitation and low runoff volume
(Figs. 2 and 3), the self-purification ability of water quality variable in
low flow rate conditions can obviously alter water quality (You et al.,
2009). The low flow rate accompanied with low self-purification may
deteriorate water quality and increase concentrations of pollutants.
Therefore, the higher flow rate in Zone 3 possesses higher self-pur-
ification resulting in the lower CVs of Chl-a and COD compared with the
lower flow rate in Zones 1 and 2 (Fig. 7).
Middle flow rate conditions (defined Q between 500 and 1400,
700–1600, and 800–1800 cms in Zones 1, 2, and 3, respectively) mostly
occurred from April to May and from September to October (Figs. 2 and

225
Y.-M. Kuo et al. Journal of Hydrology 569 (2019) 218–229

Fig. 6. Goodness-of-fit evaluation of the response variables at (a) Zone 1, (b) Zone 2, and (c) Zone 3, respectively.

226
Y.-M. Kuo et al. Journal of Hydrology 569 (2019) 218–229

3). During this period with high precipitation (Fig. 3), increasing runoff
volume carries high amounts of organic fertilizers, pesticides, and nu-
trients into Han River, and consequently triggers the algal growth.
Therefore, the flow rate positively relates to the nutrients, COD and Chl-
a concentrations in middle flow rates (Fig. 7). The flow rate at Site Tri2
is negatively correlated with COD (r = −0.65) and Chl-a (r = −0.59)
at Site Main2, indicating that increase in the flow rate at the YHWD
project may be an effective way for improving the water quality in the
mainstream.
High flow rate conditions (more than 1400, 1600, and 1800 cms in
Zones 1, 2, and 3, respectively) have high self-purification ability and
high self-rehabilitation capacity (Vagnetti et al., 2003; Vandermarken
et al., 2018). Therefore, the correlations between flow rate and the
environmental factors including nutrients, organic matters, and sus-
pend solids are not obviously. The CVs of these environmental factors
are also the lowest among all flow rate conditions (Table 5). Because
the concentrations of the water quality variables are low and stable, the
variations of these variables are not drastic in high flow rate conditions.
Flow rate, as a key variable, can be regulated by the MNSW and
YHWD projects for controlling water quality in the middle and down
streams of Han River. During low flow rate conditions, the flow rate
discharged from the MNSW project should be reduced and that from the
YHWD project to Han River should be increased. The flow rate should
be controlled to more than 700 cm in the middle stream and to more
than 800 cm in the downstream of Han River for preventing water
quality deterioration from high loadings of organic pollutants and nu-
trients. In middle flow rate conditions, decreasing the flow rate trans-
ferred from the MNSW project and increasing the flow rate discharged
from the YHWD project may be effective ways for improving the water
quality in the mainstream. The relationship between water quality and
flow rate can provide useful site-specific management to control water
quality in various flow rate conditions.

4. Conclusion

Fig. 7. The nutrients, organic pollutants, and Chl-a concentrations in various

The MSNW and YHWD projects have been operated since 2014,
flow rate conditions in three zones. The rectangle, triangle, and roundness re-
which may influence the hydrologic conditions and deteriorate water
present Zones 1, 2, and 3, respectively, and the low, middle, and high flow rate
conditions in these zones are shown in red, black, and blue, respectively. The quality in Han River. MAFA was used to statistically analyze regional
liner regression represents relation between middle flow rates and water quality characteristics of water quality in three tributary-mainstream inter-
variables of three zones. (For interpretation of the references to colour in this section zones of Han River. In densely populated zones (Zones 1 and 3),
figure legend, the reader is referred to the web version of this article.) Chl-a, COD, SS and PO43− (represented as trophic dynamics) represent
the main variations of water quality variables. In the agricultural cul-
tivation zone (Zone 2), TN, NO3−, COD, and PO43− (regarded as nu-
Table 5
Water quality variables with mean value and coefficient of variation (CV) in
trient formations) are main concerned water quality variables. From
various flow rates in three zones of Han River. CV is shown in the parentheses. DFA results, we identified pollutants source and infer possible me-
chanisms of tributaries and engineering projects influence main con-
Zone Flow Q Chl-a COD SS TP TN PO43−
cerned water quality in the mainstream Han River. Domestic and mu-
conditions (cms) (μg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L)
nicipal organic pollutants are mainly discharged from Xiangyang City
1 Low 466 4.99 13.45 33.23 0.30 1.29 0.112 and Zhongxiang City into Han River, which influence the concentra-
(0.65) (0.65) (0.56) (0.63) (0.69) (0.41) (0.63) tions of organic pollutants and nutrients in Zones 1 and 2. The non-
Middle 978 4.87 16.28 24.57 0.26 1.21 0.125
point source nitrogen and phosphorus containing high fraction of or-
(0.45) (0.41) (0.33) (0.25) (0.52) (0.58) (0.43)
High 1390 4.55 14.8 25.47 0.19 1.01 0.106 ganic compounds emitted from the tributaries Tangbai and Hanbei
(0.12) (0.33) (0.31) (0.31) (0.45) (0.37) (0.36) Rivers elevate the nutrient concentrations and increase Chl-a con-
2 Low 463 4.99 18.16 32.45 0.25 1.42 0.115 centration (i.e. promote the algal growth) in densely populated zones.
(0.54) (0.40) (0.39) (0.37) (0.33) (0.29) (0.45) In low flow rate conditions, the flow rate should be elevated to more
Middle 1023 6.49 15.25 30.33 0.16 1.70 0.109
than 700 cm in the middle stream of Han River and to more than
(0.37) (0.39) (0.35) (0.33) (0.45) (0.43) (0.49)
High 1596 3.86 7.72 18.2 0.11 0.65 0.081 800 cm in the downstream of Han River. Controlling the flow rate in the
(0.08) (0.21) (0.22) (0.19) (0.22) (0.29) (0.26) MSNW and YHWD projects can avoid degrading water quality in low
3 Low 715 7.4 25.47 39.84 0.46 1.66 0.151 flow rate conditions. The study can establish an efficient analysis dis-
(0.81) (0.39) (0.47) (0.60) (0.46) (0.48) (0.56)
tinguishing spatio-temporal variation of water quality variables and
Middle 1262 6.14 18.14 42.21 0.36 1.56 0.103
(0.42) (0.38) (0.34) (0.33) (0.46) (0.44) (0.54)
provide useful site-specific management to control water quality in
High 1727 3.97 10.44 20.06 0.19 0.61 0.071 various flow conditions. With knowledge of relationships between the
(0.09) (0.21) (0.24) (0.19) (0.32) (0.31) (0.31) flow rate and the water quality variables, we can effectively improve
water quality of Han River.

227
Y.-M. Kuo et al.
Declaration of interests
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
Acknowledgement
Great thanks to the Editor, Prof. Huaming Guo, Associate Editor,
Prof. Chong-Yu Xu, and all reviewers for insightful and constructive
comments on early versions of the manuscript. This research was fi-
nancially supported by the CRSRI Open Research Program
(CKWV2018491/KY), National Natural Science Foundation of China
(41877534), the Fundamental Research Funds for the Central
Universities, China University of Geosciences (Wuhan) (CUG130616
and CUGCJ1704), and the China Geological Survey Projects
(121201001000150121).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.jhydrol.2018.12.001.
References
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 (3), 234–247.
Akaike, H., 1974. A new look at the statistical model identification. IEEE T. Automat.
Contr. 19 (6), 716–723.
Amiri, E., 2016. Calibration and testing of the Aquacrop model for rice under water and
nitrogen management. Commun. Soil Sci. Plan. 80 (3), 396–406.
Arhonditsis, G.B., Winder, M., Brett, M.T., Schindler, D.E., 2004. Patterns and mechan-
isms of phytoplankton variability in lake Washington (USA). Water Res. 38 (18),
4013–4027.
Arnold, J.G., Moriasi, D.N., Gassman, P.W., Abbaspour, K.C., White, M.J., Srinivasan, R.,
Santhi, C., Harmel, R.D., Griensven, A.V., Liew, M.W.V., Kannan, N., Jha, M.K., 2014.
SWAT: model use, calibration, and validation. Trans. ASABE 55 (4), 1491–1508.
Bunn, S.E., Arthington, A.H., 2002. Basic principles and ecological consequences of al-
tered flow regimes for aquatic biodiversity. Environ. Manage. 30, 492–507.
Chen, C.Y., Chen, L.K., Yu, F.C., Lin, S.C., Lin, Y.C., Lee, C.L., Wang, Y.T., 2010.
Landslides affecting sedimentary characteristics of reservoir basin. Environ. Ear. Sci.
59, 1693–1702.
Chou, W.S., Lee, T.C., Lin, J.Y., Yu, S.L., 2007. Phosphorus load reduction goals for feitsui
reservoir watershed, Taiwan. Environ. Monit. Assess. 131 (1–3), 395–408.
Dolman, A.M., Rücker, J., Pick, F.R., Fastner, J., Rohrlack, T., Mischke, U., et al., 2012.
Cyanobacteria and Cyanotoxins: the influence of nitrogen versus phosphorus. PLoS
One 7 (6), e38757.
Dong, X.R., Yang, X.M., Lu, Y., Ma, K.P., Bao, J., Zhang, H.T., 2010. Comparison of the
biotoxicity of organic compounds in source and tap water from the Yangtze and Han
river, China. Environ. Sci. 30 (2), 263–268.
Elser, J.J., Fagan, W.F., Denno, R.F., Dobberfuhl, D.R., Folarin, A., Huberty, A., et al.,
2000. Nutritional constraints in terrestrial and freshwater food webs. Nature 408
(6812), 578–580.
Erzini, K., 2005. Trends in NE Atlantic landings environmental landings. Fish. Oceanogr.
14 (3), 195–209.
Felip, M., Catalan, J., 2000. The relationship between phytoplankton biovolume and
chlorophyll in a deep oligotrophic lake: decoupling in their spatial and temporal
maxima. J. Plankton Res. 22 (1), 1987–1997.
Hakanson, L., 1980. An ecological risk index for aquatic pollution control-a sedimento-
logical approach. Water Res. 14 (8), 975–1001.
Harmel, R.D., Smith, P.K., Migliaccio, K.W., Chaubey, I., Douglas-Mankin, K.R., Benham,
B., et al., 2014. Evaluating, interpreting, and communicating performance of hy-
drologic/water quality models considering intended use: a review and re-
commendations. Environ. Modell. Softw. 57, 40–51.
Harmel, R.D., Smith, P.K., 2007. Consideration of measurement uncertainty in the eva-
luation of goodness-of-fit in hydrologic and water quality modeling. J. Hydrol. 337
(3), 326–336.
Hornero, J., Manzano, M., Ortega, L., Custodio, E., 2016. Integrating soil water and tracer
balances, numerical modelling and gis tools to estimate regional groundwater re-
charge: application to the Alcadozo aquifer system (SE Spain). Sci. Total Environ.
568, 415–432.
Huang, F., Wang, X.Q., Lou, L.P., Zhou, Z.Q., Wu, J.P., 2009. Spatial variation and source
apportionment of water pollution in Qiantang River (China) using statistical tech-
niques. Water Res. 44, 1562–1572.
Isabwe, A., Yang, J.R., Wang, Y., Liu, L., Chen, H., Yang, J., 2018. Community assembly
processes underlying phytoplankton and bacterioplankton across a hydrologic
change in a human-impacted river. Sci. Total Environ. 630 (2), 658–667.
228
Journal of Hydrology 569 (2019) 218–229
Jiang, X., Huang, W., Liu, C., Huang, Q., 2004. water supply need analysis for the lower
Yellow River. Water Int. 29, 415–422.
Junk, W.J., 1989. The flood pulse concept in river-floodplain systems. Can. J. Fish. Aquat.
Sci. 106, 110–127.
Kambole, M.S., 2003. Managing the water quality of the Kafue river. Phys. Chem. Earth.
28 (20–27), 1105–1109.
Kuo, Y.M., Chang, F.J., 2010. Dynamic factor analysis for estimating ground water arsenic
trends. J. Environ. Qual. 39 (1), 176–184.
Kuo, Y.M., Jang, C.S., Yu, H.L., Chen, S.C., Chiu, C.H., 2013. Identifying nearshore
groundwater and river hydrochemical variables influencing water quality of Kaoping
River Eestuary using dynamic factor analysis. J. Hydrol. 486, 39–47.
Kuo, Y.M., Lin, H.J., 2010. Dynamic factor analysis of long-term growth trends of the
intertidal seagrass Thalassia hemprichii in southern Taiwan. Estuar. Coast. Shelf S. 86
(2), 225–236.
Kuo, Y.M., Wang, S.W., Jang, C.S., Yu, H.L., Yeh, N., 2011. Identifying the factors in-
fluencing PM2.5 in southern Taiwan using dynamic factor analysis. Atmos. Environ.
45 (39), 7276–7285.
Kuo, Y.M., Wu, J.T., 2016. Phytoplankton dynamics of a subtropical reservoir controlled
by the complex interplay among hydrological, abiotic, and biotic variables. Environ.
Monit. Assess. 188, 689–701.
Lau, S.S., Lane, S.N., 2002. Biological and chemical factors influencing shallow lake eu-
trophication: a long-term study. Sci. Total Environ. 288 (3), 167–181.
Leboulanger, C., Rimet, F., Lacotte, M.H.D., Bérard, A., 2001. Effects of atrazine and
nicosulfuron on freshwater microalgae. Environ. Int. 26 (3), 131–135.
Li, J.X., Wang, Y.X., Xie, X.J., Su, C.L., 2012. Hierarchical cluster analysis of arsenic and
flX.–ride enrichments in groundwater from the Datong basin, Northern China. J.
Geochem. Explor. 118, 77–89.
Li, S., Li, J., Zhang, Q., 2011. Water quality assessment in the rivers along the water
conveyance system of the Middle Route of the South-to-North Water Transfer project
(China) using multivariate statistical techniques and receptor modeling. J. Hazard.
Mater. 195 (1), 306–317.
Li, S., Xu, Z., Cheng, X., Zhang, Q., 2008. Dissolved trace elements and heavy metals in
the Danjiangkou Reservoir, China. Environ. Geol. 55, 977–983.
Liang, K., Wang, X., Zhang, D., Zhou, Y., 2012. Ecological conditions of diatom water
bloom formulation in the middle and lower reach of the Hanjiang River and strategy
for water bloom control. Environ. Sci. Technol. 35 (12), 113–116 (in Chinese).
Ligas, A., De Ranieri, S., Micheli, D., Reale, B., Sartor, P., Sbrana, M., Belcari, P., 2010.
Analysis of the landings and trawl survey time-series from the Tyrrhenian Sea (NW
Mediterranean). Fish. Res. 105, 46–56.
Lin, Y.C., Chang-Chien, G.P., Chiang, P.C., Chen, W.H., Lin, Y.C., 2013. Multivariate
analysis of heavy metal contaminations in seawater and sediments from a heavily
industrialized harbor in Southern Taiwan. Mar. Pollut. Bull. 76, 266–275.
Liu, Y., Cao, S.L., Zhang, X., Li, F.Z., Li, X.T., 2018. Joint improvement of river water
quality indicators based on a multivariate joint probability distribution of the dis-
charge and water quality. Hydrol. Res. https://doi.org/10.2166/nh.2018.196.
Luo, P., He, B., Takara, K., Razafindrabe, B.H.N., Nover, D., Yamashiki, Y., 2011.
Spatiotemporal trend analysis of recent river water quality conditions in Japan. J.
Environ. Monit. 13, 2819–2829.
McCauley, E., Downing, J.A., Watson, S., 1989. Sigmoid relationships between nutrients
and chlorophyll among lakes. Can. J. Fish. Aquat. Sci. 46 (7), 1171–1175.
Mfumu Kihumba, A., Ndembo Longo, J., Vanclooster, M., 2016. Modelling nitrate pol-
lution pressure using a multivariate statistical approach: the case of Kinshasa
groundwater body, democratic republic of congo. Hydrogeol. J. 24 (2), 1–13.
Momen, B., Eichler, L.W., Boylen, C.W., Zehr, J.P., 1996. Application of multivariate
statistics in detecting temporal and spatial patterns of water chemistry in lake
George, New York. Ecol. Model. 91 (1), 183–192.
Muñoz-Carpena, R., Ritter, A., Li, Y.C., 2005. Dynamic factor analysis of groundwater
quality trends in an agricultural area adjacent to Everglades National Park. J.
Contam. Hydrol. 80 (1–2), 49–70.
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models. Part 1-
A discussion of principles. J. Hydrol. 10, 282–290.
Paerl, H.W., Xu, H., Mccarthy, M.J., Zhu, G., Qin, B., Li, Y., et al., 2011. Controlling
harmful Cyanobacterial blooms in a hyper-eutrophic lake (lake Taihu, china): the
need for a dual nutrient (n p) management strategy. Water Res. 45 (5), 1973–1983.
Pekey, H., Karakaş, D., Bakoğlu, M., 2004. Source apportionment of trace metals in
surface waters of a polluted stream using multivariate statistical analyses. Mar.
Pollut. Bull. 49 (9), 809–818.
Poff, N.L., Olden, J.D., Merritt, D.M., Pepin, D.M., 2007. Homogenization of regional river
dynamics by dams and global biodiversity implications. P. Natl. Acad. Sci. USA 104
(14), 5732–5737.
Qian, F., Huang, J.S., Liu, J.J., Liu, H.H., 2018. Effects of flow hydraulics on total nitrogen
loss on steep slopes under simulated rainfall conditions. Hydrol. Res. 49 (4),
1088–1100.
Que, Z., Seidou, O., Droste, R.L., Wilkes, G., Sunohara, M., Topp, E., Lapen, D.R., 2015.
Using AnnAGNPS to predict the effects of tile drainage control on nutrient and se-
diment loads for a river basin. J. Environ. Qual. 44, 629–641.
R Development Core Team, 2009. R: A language and environment for statistical com-
puting. R Foundation for Statistical Computing, Vienna.
Ritter, A., Muñoz-Carpena, R., 2013. Performance evaluation of hydrological models:
statistical significance for reducing subjectivity in goodness-of-fit assessments. J.
Hydrol. 480 (4), 33–45.
Ritter, A., Regalado, C.M., Muñoz-Carpena, R., 2009. Temporal common trends of topsoil
water dynamics in a humid subtropical forest watershed. Vadose Zone J. 8 (2),
437–449.
Sarkar, J.M., Hennebert, G.L., Mayaudon, J., 1986. Optimization and characterization of
an extracellular polysaccharide produced bymoniliella, pollinis. Biotechnol. Lett. 8
Y.-M. Kuo et al.
(5), 319–322.
Shapiro, D.E., Switzer, P., 1989. Extracting Time Trends from Multiple Monitoring Sites.
Department of Statistics. Stanford University. Tech. Rep, pp. 132.
Shojaei, A., Moradi, S., Alaeddini, F., Khodadoost, M., Abdizadeh, A., Khademi, A., 2016.
Evaluating the temporal trend of completed suicides referred to the Iranian forensic
medicine organization during 2006–2010. J. Forensic Leg. Med. 39, 104–108.
Solow, A.R., 1994. Detecting changes in the composition of a multispecies community.
Biometrics. 50, 556–565.
Switzer, P., Green, A.A., 1984. Min/max Autocorrelation Factors for Multivariate Spatial
Imagery. Tech. rep. 6, Department of Statistics, Stanford University.
Sun, H., Giesy, J.P., Jin, X., Wang, J., 2017. Tiered probabilistic assessment of organo-
halogen compounds in the Han river and danjiangkou reservoir, central China. Sci.
Total Environ. 586, 163–173.
Sun, C., Wu, H., 2012. Pollution from animal husbandry in China: a case study of the Han
river basin. Water Sci. Technol. 66 (4), 872.
Vagnetti, R., Miana, P., Fabris, M., Pavoni, B., 2003. Self-purification ability of a re-
surgence stream. Chemosphere 52 (10), 1781–1795.
Vandermarken, T., Croes, K., Van, L.K., Boonen, I., Servais, P., Garcia-Armisen, T., 2018.
Endocrine activity in an urban river system and the biodegradation of estrogen-like
endocrine disrupting chemicals through a bio-analytical approach using dre- and ere-
calux bioassays. Chemosphere 201, 540–549.
Vega, M., Pardo, R., Barrado, E., Debán, L., 1998. Assessment of seasonal and polluting
effects on the quality of river water by exploratory data analysis. Water Res. 32 (12),
3581–3592.
Vollenweider, R.A., 1968. Scientific fundamentals of the eutrophication of lakes and
flowing waters, with particular reference to nitrogen and phosphorus as factors in
eutrophication. Organisation for Economic Co-operation and Development.
Wang, S.W., Kuo, Y.M., Kao, Y.H., Jang, C.S., Maji, S.K., Chang, F.J., Liu, C.W., 2011a.
Influence of hydrological and hydrogeochemical parameters on arsenic variation in
shallow groundwater of southwestern Taiwan. J. Hydrol. 408, 286–295.
Wang, X., Hao, F., Cheng, H., Yang, S., Zhang, X., Bu, Q., 2011b. Estimating non-point
source pollutant loads for the large-scale basin of the Yangtze river in China. Environ.
Earth Sci. 63 (5), 1079–1092.
Wei, W., Chen, L., Fu, B., Lü, Y., Gong, J., 2009. Responses of water erosion to rainfall
extremes and vegetation types in a loess semiarid hilly area, NW China. Hydrol.
Process. 23 (12), 1780–1791.
Woillez, M., Rivoirard, J., Petitgas, P., 2009. Using min/max autocorrelation factors of
survey-based indicators to follow the evolution of fish stocks in time. Aquat. Living
Resour. 22 (2), 193–200.
Wu, L., Muñozcarpena, R., Gao, B., Yang, W., Pachepsky, Y.A., 2014. Colloid filtration in
surface dense vegetation: experimental results and theoretical predictions. Environ.
Sci. Technol. 48 (7), 3883.
Xie, H., Lian, Y.Q., 2013. Uncertainty-based evaluation and comparison of SWAT and
229
Journal of Hydrology 569 (2019) 218–229
HSPF applications to the Illinois River Basin. J. Hydrol. 481, 119–131.
Xie, H., Zafiriou, O.C., Cai, W., Zepp, R.G., Wang, Y., 2004. Photooxidation and its effects
on the carboxyl content of dissolved organic matter in two coastal rivers in the
Southeastern United States. Environ. Sci. Technol. 38 (15), 4113–4119.
Yang, Q., Xie, P., Shen, H., Xu, J., Wang, P., Zhang, B., 2012. A novel flushing strategy for
diatom bloom prevention in the lower-middle Hanjiang river. Water Res. 46 (8),
2525–2534.
Yang, J., Lv, H., Yang, J., Liu, L., Yu, X., Chen, H., 2016. Decline in water level boosts
cyanobacteria dominance in subtropical reservoirs. Sci. Total Environ. 557 (558),
445–452.
Yu, S., He, L., Lu, H.W., 2016. An environmental fairness based optimisation model for
the decision support of joint control over the water quantity and quality of a river
basin. J. Hydrol. 472 (535), 366–376.
Zeng, Z.X., Zhang, H., Shan, B.Q., Yang, H.G., 2014. Analysis of industrial pollution
sources of the middle and lower reaches of hanjiang river basin. Resour. Environ.
Yangtze Basin 23 (2), 252–259 (in Chinese).
Zhao, H., Li, X., Wang, X., Tian, D., 2010. Grain size distribution of road–deposited se-
diment and its contribution to heavy metal pollution in urban runoff in Beijing,
China. J. Hazard Mater. 183, 203–210.
Zhang, C., Dong, S.H., 2009. A new water quality assessment model based on projection
pursuit technique. J. Environ. Sci. S1, 154–157.
Zhang, W., Wei, X., Zheng, J., Zhu, Y., Zhang, Y., 2012. Estimating suspended sediment
loads in the Pearl River Delta region using sediment rating curves. Cont. Shelf Res. 38
(15), 35–46.
Zhang, D., Qian, B., Zhang, Y., 2016. Bloom characteristic parameters of water quality
analysis in Hanjiang River. J. Water Resour. Res. 5 (4), 402–408.
Zheng, W., Zhao, H., Mancera, E., Steinmetz, L.M., Snyder, M., 2010. Genetic analysis of
variation in transcription factor binding in yeast. Nature 464 (7292), 1187–1191.
Zhong, H., Kraemer, L., Evans, D., 2012. Effects of aging on the digestive solubilization of
Cu from sediments. Environ. Pollut. 164, 195–203.
Zhou, Y., 2009. Pesticide Transport and Enrichment in Water–Soil Environment: A Case
Study of Jianghan Plain. A Dissertation Submitted to China University of Geosciences
for the Doctor Degree of Philosophy, 77–89.
Zuur, A.F., Fryer, R.J., Jolliffe, I.T., Dekker, R., Beukema, J.J., 2003a. Estimating common
trends in multivariate time series using dynamic factor analysis. Environmetrics 14,
665–685.
Zuur, A.F., Leno, E.N., Smith, G.M., 2007. Analysing Ecological Data. Springer, pp. 1317-
1317.
Zuur, A.F., Pierce, G.J., 2004. Common trends in Northeast Atlantic Squid time series. J.
Sea Res. 52, 57–72.
Zuur, A.F., Tuck, I.D., Bailey, N., 2003b. Dynamic factor analysis to estimate common
trends in fisheries time series. Can. J. Fish. Aquat. Sci. 60, 542–552.