Status assessment and causal factors diagnosis of river system health

A healthy river system means its structures can cooperatively and orderly work together and all functions can be well brought to exert when the external disturbance is under a limit extent. This paper had presented the concept framework, which had included status assessment (SA) and cause diagnosis (CD). On the base of the method of describing entropy, health index (HI) of a river system had been defined via order degrees of indicators, which had been calculated by distance far away from their criteria as well as the health grade had been classified and the main problems had been addressed. Using partial least square (PLS) regression, CD had been conducted to diagnose main factors inducing these problems and to establish the regression equation between external variables and internal variable. Meanwhile, the variable importance projection (VIP) had been quantified. In the case of Anxi River, results of status assessment indicated that this river system was healthy in general. However, abiotic indices were low in S2 and S3 reach. After CD had been conducted, it had been revealed that problems of this river system had been induced by the variable of upstream discharge. It had been suggested helpfully that upstream discharge should be controlled in management. Therefore, river system health diagnosis was very helpful and beneficial to river system management.


INTRODUCTION
As society had benefitted immeasurably from rivers (Karr, 1999), the rate and extent to which humans had altered rivers had been surpassing our ability to assimilate and understand the implications of our actions (Kay and Schneider, 1994;Kennard et al., 2006).It has resulted that many rivers have been becoming dramatically degraded (Karr, 1999;Pinto and Maheshwan, 2011).The progressive deterioration has aroused concerns about river health (Xu et al., 2004;Guan et al., 2011) and urgent restoration and maintenance measures of `healthy' river ecosystems have been becoming important objectives of river management (Norris and Thoms, 1999;Palmer and Allan, 2006;Vörösmarty et al., 2010).
Fortunately, increasing attention has been being paid on river health in many countries, such as Australia, the United States and the United Kingdom, South Africa, New Zealand, and China (Australian and New Zealand Environment and Conservation Council, 2000;Hohls, 1996;Norris and Thoms, 1999;Mekong River Commission, 2003;Li, 2004;Yang et al., 2005;Liu and Liu, 2008;Su et al., 2011).The diagnosis of stream and river health has been a vital issue in the field of stream restoration and management (Choi et al., 2011).Thus, bewildering varieties of river health concepts and assessment approaches, providing reliable information on the effect of long-term stressors and serve as the foundation for the restoration of disturbed systems, have been reported (Pinto and Maheshwan, 2011).For example, "DEPHI" forecasting and hierarchical model were used to select variables and assess river health (Pinto and Maheswan, 2011).All approaches can be classified into three categories: single perspective studies, ecological function based studies and composite studies (Pinto and Maheshwari, 2011).Composite approaches, using of comprehensive indicators, such as Index of Biotic Integrity (IBI) (Karr, 1981) and Rapid Bioassessment Protocols (RBPs) (Plafkin et al., 1989), Australian River Assessment System (AusRivAS) (Simpson et al., 1997), and River Invertebrate Prediction And Classification Scheme (RIVPACS) (Wright, 1995), etc., have been most common and widely accepted.Through these approaches, it has been easy to find symptoms and problems of river system.
Generally, a river system can be defined by their structures and functions.If its structures and functions are disturbed by external natural factors (e.g.climate, flood, etc.) or anthropogenic factors (e.g.pollution, land use, water resource utilization, etc.), a river system might become disorder or unhealthy.Thus, it is necessary to establish the relationships between the internal variables (indicators) of a river system and external factor.Most past works have paid to only the healthy status assessment of a river system.However, quantitative analyses of causal factors have not been done.The relationships between internal variables and external variables have not been effectively established till now.Therefore, an integrated diagnosing approach should be developed, through which not only symptoms and questions could be known, but also specific sources or mechanisms of causal factors could be quantified and revealed (Bunn et al., 1999;Vörösmarty et al., 2010).The objectives of this study was to: (a) construct the concept framework of river system health diagnosis; (b) figure out internal indicators and present method of status assessment of river system; (c) address the external causal factors and bridge the relationship between external causal factors and health status.

Study area
Yunhe County is located in southwest of Zhejiang Province, Eastern China (27°53′ -28°9′ N, 119°21′ -119°44′ E) with approximately 10.9 million inhabitants and about land area of 984 km 2 .About 90% of lands are characterized by mountains and 5% is used as croplands and 5% is aquatic regions.In Yunhe, there are two main catchments both of which are the branches of Oujiang watershed.One is Wutongken River with the drainage areas of 14.84% of Yunhe's total territory.Another is Longquan River, flowing from southwest to northeast, whose drainage areas account for 85.16% of Yunhe's total territory.In the Longquan River catchment, there exist 13 tributaries such as Fuyun River, Mayangkeng River, Linhaikeng River and so on.Anxi River is the largest tributary of Fuyunxi River, which originates in Dongdai, 1178 m high above sea level.It flows through Shangcun, Chengzai and lastly feeds into Fuyunxi River at Gufang with length of 15.33 km, drainage area of 33.28 km 2 , average slope 6.8% (Figure 1).In this watershed, annual rainfall is about 1750 mm and flood events have frequently taken place.

Concept framework
'Health' is short hand for 'good condition' (Karr, 1999).A healthy river system implies good internal and external conditions.Good internal conditions are that all elements of structures can be organized orderly and supply good enough functions.Good external conditions imply that external disturbances do not surpass the self-adjustment capacity of a river system.Meanwhile, these structures and functions are easily disturbed by external factors of a river system.The mechanism is shown in Figure 2. Accordingly, health diagnosis of a river system includes two steps.The first step is assessment of internal condition which is called status assessment (SA).In this step, status of river system can be valued and graded so that some symptoms and problems can be known.Secondly, causal factor diagnosis (CD) is performed, in which the relationship between system status and internal factors is established and the degree to external factors acting on river system can be quantified.It is inferred that main external factors causing degradation can be investigated.The procedure framework is shown in Figure 2.

Status assessment
In order to exactly evaluate the status of a river system, health index should be calculated when comprehensive internal variables or indicators of structures and functions of a river system are selected.First of all, a set of internal indicators should be figured out, among of which there exists one indicator representing the most important function.This kind of indicator is named as controlling indicator while other indicators are grouped as cooperative indicators.If controlling indicator is satisfied with its criterion, a comprehensive assessment integrating controlling indicator into cooperative indicators would be conducted.If not, a comprehensive assessment, only considering cooperative indicators, would be performed.As results of the comprehensive assessment, health index could be quantified and health grade could be evaluated.Depending on health index, it would be inferred what the symptoms and problems of a river system identity would be.The procedure is shown in Figure 3.

Indicators of status assessment
In past few decades, some evaluating index systems such as the index of stream condition (ISC) (Landson and White, 1999;Kennard et al., 2006), Environmental Monitoring and Assessment Programme (EMAP) (Huhges et al., 2000), the Isle of Man a River Habitat Survey (RHS) (Raven et al., 2000) have been developed.In these index systems, all variables related to a river system have been considered as the state variables of a river system.However, status of a river system is denoted only by internal indicators, depending on its structures and functions (Bunn and Davies, 2000;Feio et al., 2010).Its structures include biotic elements and abiotic elements.Biotic elements consist of plants, animals and microorganisms, while abiotic elements include flowing water, sediments, nutrients, riparian, bed, etc.There exist comprehensive interactions and matter exchanges as well as energy flows among these elements.Due to these interactions and exchanges, a river system is dynamic and has self-organizing and self-adjusting capability.In functions, a river system has many essential goods and services as well as aesthetic and cultural values (Meyer, 1997;   Bunn , 2003;Bunn et al., 2010).Commonly, these functions can be classified two types: (a) natural functions which indicate a river system can provide clean drinking water, fisheries production, conservation and biodiversity values, flood management and ecological refuge (Bunn et al., 2010); (b) social functions which mean a river system accommodates communities by providing a medium for transport, recreation, tourism, worship, ecosystem services and a place to experience the serenity of nature (Pinto and Maheshwari, 2011).Therefore, the indicators of river system health include measures of structures and functions both of the biotic and of the physical components (Norris and Thoms, 1999), which are classified into three sub-indices: abiotic index, biotic index and service index.Abiotic index can be denoted by runoff variation ratio (C11), connectivity (C12), comprehensive stability (C13) and wetland conservation ratio (C14).Biotic index can be comprehensively reflected by biodiversity index (C21), vegetation covering ratio (C22), and ecological discharge insured ratio (C23).Service index can be measured by flood safety index (C31), landscape suitability index (C32), up-to standard rate of water quality (C33), water supply insured ratio (C34) and navigation insured ratio (C35).The hierarchy indicators of SA are shown in Table 1.

Calculation of health index
A river system constitute dissipative systems with many spatial freedom degrees which are reflected by entropy (Jackson, 1968;Rodríguez-Iturbe et al., 1992;Zhao et al., 2009).The more random and chaotic is a system, the higher is entropy.Commonly, entropy of a variable is quantified by its order degree which is determined by the relative distance far away from its ideal value.According to the responding trend of internal variables to order degree, there are three types of variables: increasing-orient variables, decreasingorient variables and middle-orient variables.The increasing-orient variables are beneficial to river system health with its value increasing.Inversely, those variables are the decreasing-orient variables.Additionally, there are middle-orient variables which have better effect on river system health when its value is more close to a middle fixed value.The order degree ε is defined as Express (1) where, ε is order degree of a variable; max R is the maximum value of a variable; min R is the minimum value of a variable; where, e is the entropy of variable.When the order degree and entropy of each indicator are obtained, the sub-indices are valued by Express (3).
Where, j SUBHI is the health index of the j-th subsystem (j=1,2,3); ij ω is the weight coefficient of the i-th indicator within the j-th subsystem; ij ε is the order degree of the i-th indicator in the j-th subsystem; n is the number of indicators in the j-th subsystem.Health index of a river system or reach is determined by subindices which depend on theirs indicators.It can be obtained via Express (4).
Where, HI is the health index of a river system or reach; wj is the weight coefficient of the j-th subsystem; j SUBHI is health index of the j-th subsystem.ij ω and wj can be obtained by use of analytic hierarchy process (AHP) (Saaty, 1990).

Causal factors diagnosis
The objective of causal factors diagnosis is to pick up the main causal factors that induce the symptoms and problems of river system health.Firstly, we would select some variables featuring external causal factors as exogenous variables and define the health index or sub-indices as endogenous variables, then we would establish the regression relationship between exogenous variables and endogenous variables by use of partial least square (PLS) regression (Haenlein and Kaplan, 2004).

Causal factors and endogenous variables
Ideally, the factors chosen should assist in the diagnosis of the probable cause of health degradation and inform management actions (Dole´dec et al., 2006;Bailey et al., 2007).A river system is usually affected by natural and human causes.Rainfall is the main natural cause which can result in flood, erosion and pollution.Particularly, runoff from roads and fields containing large amount of litter and other harmful pollutants such as petroleum hydrocarbons, zinc, lead, copper, chromium enters and pollute rivers.Today, storm water pollution is a big problem facing our rivers (http://www.melbournewater.com.au/).Commonly, the influence intensity of rainfall is interpreted by two variables of average rainfall and largest rainfall.
The short-term monetary gains of many past economy-driven management decisions have resulted in harmful long-term healthy consequences (Kay and Schneider, 1994).Many of rivers and streams have been changed dramatically due to human activities such as deforestation, farming, hydraulic engineering, pumping water, floodplain cultivation, navigation, urbanization and so on.These threats or causes can be grouped as seven categories: rainfall, pollutant, water resource development, illegal occupation activity, unnecessary hydraulic construction, urbanization, and management.Each type of cause can be quantified by its corresponding variables.Rainfall can be interpreted by average rainfall, and largest rainfall, and flood frequency and soil erosion.As there are four types of pollution sources: industry, farming, sewage, fishery, stockbreeding, pollution cause can be quantitatively described by variables of industrial pollutant, farming pollutant, sewage pollutant, fishery pollutant and stockbreeding pollutant.Water resource is the basic resource of industry, agriculture and living.The cause of water resource development depends on upstream discharge, industrial water consumption, agriculture water consumption and living water consumption.Sometimes, river systems may be illegally occupied for enlarging lands or sands from river bed.It results that aquatic area is reduced and river stability is declined.These changes can be measured by variables of water area variation ratio and digging sand ratio.In a river system, crossing constructions and bank protection may be built which have influence on regulating flood ratio and concreting riparian ratio.Urbanization is described via urbanization ratio and management cause can be illustrated by illegal event ratio.The cause variables and their denotations are shown in Table 2.

Establishing regression equation
The PLS was developed by Wold in the late 1960s (Carrascal et al., 2009).Its goal is to predict or analyze a set of dependent variables from a set of independent variables or predictors (Abdi, 2007) and to establish regression relationship between dependent variables and independent variables (Höskuldsson, 1988;Carrascal et al., 2009).Using of PLS, original multidimensionality is declined and intepretation of independent variables is maximized by extracted predictors.(Hubert and Branden, 2003;Maestre, 2004).We consider external variables of river system as independent variable set  Wold et al. (2001) and Li (2002) presented the procedure through which the regression equation could be established, as follows (Li et al., 2002).
Where, j e is a vector of errors and j b is estimated by: The latent variables are computed by , where, j w and j q unit length and are determined by maximizing the covariance between j t and j u .Then, , and The number of iteration steps is determined by use of cross-validation.Crossvalidation is a practical and reliable way to test this predictive significance (Wold et al., 2001).Wold's R criterion, which is based on cross-validation, has been the typical approach used to select the number of latent variables (Li et al., 2002).We suppose the i-th group data of Y is i y , and hi y ˆis the fitting value of the i-th group Y when h components have been extracted, and ) ( ˆi h y − is the fitting value of Y when the i-th group observation is omitted and h components have been extracted.Then, the sum of squares of these differences(SS) is computed and collected from all the parallel models to form the predictive residual sum squares (PRESS), which estimates the predictive ability of the model (Wold al., 2001).SSh and PESSh and predicted variation 2 h Q can be obtained as Express ( 6) to (8).
is larger than 0.095, the prediction and regression is satisfied.

Variable importance projection
The models are interpreted with the help of variable importance in the projection (VIPs).The VIPs represent the importance of the descriptors for the model, both with respect to correlation to Y and with respect to X (the projection).Each descriptor is uniquely and described by its VIP, which represents a measure of the contribution of the term to the decomposition of X, and the correlation with Y.Moreover, the VIPs are normalized so that they

Data
As Anxi River lies in mountain region, it has big variation in its geomorphology.Terrain slopes between different reaches vary from 0.75 to 10.69% and the mean slope is 6.8%.The steepest reach is from Dongdai to Shangcun, where the slope 10.69%.The flattest reach is from Zhifan to Gufang, where the slope is 0.75%.In terms of terrain slope, Anxi River can be divided four reaches as shown in Figure 1.The first reach(R1) is from Dongdai to Shangcun, 4.536 km long; the second reach (R2) is from Shangcun to Dongkeng, 2.864 km long, whose slope is 2.31%; the third reach (R3) is from Dongkeng to Zhifan, 1.689 km long, whose slope is 7.98%; the fourth reach is from Zhifan to Gufang, 6.241 km long, whose slope is 0.75%.Based on statistic data 2007, the values of these indices have obtained as shown in Table 3.To

Health index Anxi River
As Anxi River belongs to a typical mountain river, there are frequent flood disasters in this catchment.It would be harmful to river stability or ecological safety or resident safety.For instance, the most serious flood event happened in 2005.It results that some reaches, about 5 km long, were collapsed, and about 20 houses were broken out and approximate 10 km 2 crops were submerged.In terms of this, it is very important to keep Anxi River safety from flood disasters.Therefore, safety of flood control can be used as the controlling indicator.
Basically, controlling assessment can be conducted on the base of data in 2007.As results, it is inferred that controlling indicator satisfied its healthy criterion.Then the integrated assessment can be executed.Firstly, the order degree of each internal indicator can be obtained by use of Express (1) and weights of internal indicators can be valued through AHP.Secondly, sub-indices can be determined in terms of Express (3) and the weights of sub-systems can be calculated in AHP.When compounding sub-indices and their weights, we can get the health index (Table 5), in terms of which grade of Express ( 11) denotes the regression relationship between abiotic index and all these independent variables.However, different factors have different influence degree on dependent variable.

Conclusion
How can we know whether the outside effects overrun limitation of river system and what factors play main function on a river health?Quantitative diagnosis on a river system can help to solve these issues.In the case of Anxi River, its health status was assessed and its causal factors were diagnosed.In terms of value of health index (HI), we classified health status into three grades: Health, subhealth and disease.When HI is less than 0.50, a river system is disease; When HI lies in between 0.5 and 0.65, a river system is subhealty; When HI is larger than 0.65, river system is healthy (Suo et al., 2008;Liu and Liu, 2009).
As the general HI of Anxi River is 0.713, it is inferred that health status of this river belongs to healthy grade.But the sub-indices have large difference.For instance, the average value of biotic index is the largest, 0.885 which illustrates Anxi Stream has rich species components and fine ecological integrity.But in different reaches, health index of each reach varied greatly each other.Health index of S3, 0.720, is the highest.It means that S3 is best.Inversely, health index of S2, 0.699, is the least.It shows that S2 is not better than other reaches although S2 belongs to health grade.Moreover, the average abiotic index of Anxi River is only, 0.612, which implies that there exist some problems in abiotic structures of river system.According to the assessment of each reach, respectively, abiotic indices of S2 and S3 are 0.599 and 0.585, which is less than 0.6.It shows that structure statuses in S2 and S3 are the most serious.Thus, much more focus should be paid on this segment.
It is necessary to analyze what factors cause this problem.So, abiotic index has been considered as the dependent variable.The significance of each factor is illustrated by variable importance projection (VIP).Basically, the VIP can be obtained by the use of Equation ( 12) and ( 13).Respectively, the VIPs of X1, X2, X3 and X4 are 0.114, 3.851, 0.291 and 0.276 as shown in Figures 4 and 5.The VIP of X2 is the largest, which illustrates the upstream discharge plays the most significant role in abiotic index.In fact, because is a sudden turn and the terrain slope in the third reach varies dramatically, flow, flow regime has dramatic variation and carries large hydraulic energy when water flows through this reach.Intensive flow turbulence can result in potential risk of bank or bed damage.So as to keep healthy status of this river, it is importance to control the upstream and improve flow condition through suitable mending.It is helpful to manager to make good decisions to improve the health index.Therefore, the results of diagnosis can help decision makers to take ecological restoration treatments to reduce the intense variation of upstream discharge.Therefore, river system health diagnosis is a useful tool for river management.

Figure 3 .
Figure 3. Framework of river system health diagnosis.
real value of a variable. ε

Rainfall Land slide River System Treatm ent M an agem ent Land use Pollution Utilization of water resource Soil erosion Weather Figure
2. Mechanism of a river system

Table 1 .
Indicators of status assessment.
and entropy of a variable is defined as Express (2).

Table 2 .
External causal factors and endogenous variables.

Table 3 .
Values of indicators of Status Assessment (in 2007).

Table 4 .
Values of independent variables.

Table 5 .
Health Index and Sub-index(in 2007).