Development and Application of a Novel Method for Regional Assessment of Groundwater Contamination Risk in the Songhua River Basin

Science of the Total Environment 605–606 (2017) 598–609

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Development and application of a novel method for regional assessment of groundwater contamination risk in the Songhua River Basin

Erik Nixdorf a,b,⁎, Yuanyuan Sun c, Mao Lin c, Olaf Kolditz a,b,d a Helmholtz Centre for Environmental Research, Department of Environmental Informatics, Permoserstr.15, 04318 Leipzig, Germany b Technical University Dresden, Faculty of Environmental Sciences, Helmholtzstr.10, 01069 Dresden, Germany c Chinese Research Academy of Environmental Sciences, Dayangfang, Beiyuan 8, 100012 Beijing, China d Sino-German Research Centre on Environmental Information Science (RCEIS), China

HIGHLIGHTS GRAPHICAL ABSTRACT

• Integrate numerical modelling, public datasets and web mapping services into a groundwater risk assessment framework • Evaluate various types of hazards from point and diffusive sources in a uniﬁed form • Assess groundwater risk by index- overlay of groundwater vulnerability and hazard • Generate maps for groundwater risk maps for the entire Songhua River Basin

article info abstract

Article history: The main objective of this study is to quantify the groundwater contamination risk of Songhua River Basin by ap- Received 16 March 2017 plying a novel approach of integrating public datasets, web services and numerical modelling techniques. To our Received in revised form 30 May 2017 knowledge, this study is the ﬁrst to establish groundwater risk maps for the entire Songhua River Basin, one of the Accepted 15 June 2017 largest and most contamination-endangered river basins in China. Available online xxxx Index-based groundwater risk maps were created with GIS tools at a spatial resolution of 30 arc sec by combining Editor: D. Barcelo the results of groundwater vulnerability and hazard assessment. Groundwater vulnerability was evaluated using the DRASTIC index method based on public datasets at the highest available resolution in combination with nu- Keywords: merical groundwater modelling. As a novel approach to overcome data scarcity at large scales, a web mapping Groundwater contamination risk service based data query was applied to obtain an inventory for potential hazardous sites within the basin. Hazard The groundwater risk assessment demonstrated that b1% of Songhua River Basin is at high or very high contam- Topic: ination risk. These areas were mainly located in the vast plain areas with hotspots particularly in the Changchun Intrinsic vulnerability metropolitan area. Moreover, groundwater levels and pollution point sources were found to play a signiﬁcantly DRASTIC larger impact in assessing these areas than originally assumed by the index scheme. Moderate contamination risk Model sensitivity was assigned to 27% of the aquifers, predominantly associated with less densely populated agricultural areas. Songhua River Basin However, the majority of aquifer area in the sparsely populated mountain ranges displayed low groundwater contamination risk. Sensitivity analysis demonstrated that this novel method is valid for regional assessments of groundwater contamination risk. Despite limitations in resolution and input data consistency, the obtained groundwater contamination

⁎ Corresponding author at: Helmholtz Centre for Environmental Research, Department of Environmental Informatics, Permoserstr.15, 04318 Leipzig, Germany. E-mail address: [email protected] (E. Nixdorf).

risk maps will be beneﬁcial for regional and local decision-making processes with regard to groundwater protection measures, particularly if other data availability is limited. © 2017 Elsevier B.V. All rights reserved.

1. Introduction For a detailed assessment of potential hazards, information about the pollution sources and the characteristic of the pollutants should be taken Since the 1950s, major advances in drilling technology and under investigation. The difficulties in obtaining this specificdataleadto hydrogeological knowledge had facilitated a massive expansion in the preference of an index based assessment system for hazards (Wang groundwater use across the developing world in order to satisfy the et al., 2012). An approach often applied for smaller catchments is to set needs of irrigation, domestic purposes and industrial demand (Foster up an inventory of potential hazard sources in combination with and Chilton, 2003). In China alone, N500 large cities rely on groundwa- weighting and rating schemes (e.g. Kuisi et al., 2014; Aliewi and ter resource for their drinking water supply (Van der Gun, 2012). These Al-Khatib, 2015). Alternatively, several previous studies used grid excessive exploitation and inappropriate activities at the land surface based data such as land cover or land use as a proxy parameter for de- lead to and foster degradation of groundwater resources in many scribing potential anthropogenic pollution hazard from diffusive and areas (Jiang, 2009). Moreover, remediation of contaminated aquifers is point sources (Bartzas et al., 2015). The obtained data is either included cost intensive and time consuming and may not reach remedial objec- in a modified groundwater vulnerability assessment (Gomezdelcampo tives due to large storage, physical inaccessibility and retardation of and Dickerson, 2008; Fritch et al., 2000) or in a separate index based haz- contaminants (Travis and Doty, 1990). In this context, groundwater ard assessment (Panagopoulos et al., 2006; Saidi et al., 2009). contamination risk assessment provides a useful tool to design and im- The present study is concerned with the assessment of groundwater plement groundwater protection measures to prevent or reduce vulnerability, groundwater hazard and the resulting groundwater risk for groundwater contamination (Zaporozec et al., 2002). Songhua River Basin, a large scale catchment providing water resources In a risk assessment, where risk is defined as hazard plus vulnerabil- for more than 62 million people in North-East China (Yuhong et al., ity, the combined rating of the potential harmfulness posed by a pollu- 2010). Although Songhua River Basin was one of the earliest urbanized tion source (hazard assessment) and the possibility of the spreading centers in China since the 1950s, a series of developmental strategies into and within the groundwater (vulnerability assessment) could be by the government calling for a “revitalization of the Northeast's old in- interpreted as the probability for groundwater contamination, both in dustrial base” brought in new economic development and promoted quantitative or qualitative terms depending on the used method rapid urbanization (Zhang, 2008). However, urbanization and industrial- (Johansson, 1999; Varnes, 1984). Regarding the groundwater vulnera- ization adversely affected the surrounding ecosystems and environment bility assessment, existing methods are classified into three categories: (Fu et al., 2016). One of the environmental issues of aquifers in the study Index methods, statistical methods (Erwin and Tesoriero, 1997; Li et area is an increased pollution of the aquifer system by Nitrate from fertil- al., 2015) and process-based methods (Sinkevich et al., 2005; Milnes, izer overuse (e.g. Zhang et al., 1996; Gu et al., 2013). As North-East China 2011). An overview about available methods is given by Zaporozec et is the old industrial heartland of China, discharges and leakages from in- al. (2002). Index methods, such as DRASTIC (Aller et al., 1987)orGOD dustrial sources and brownfields put an additional potential pressure on (Foster, 1987), which focus on the key factors controlling the solute groundwater resources (Nixdorf et al., 2015). Hence, assessing the transport processes, are the most commonly used procedures for vul- groundwater environmental risk in the region is very necessary. nerability mapping. They are relatively inexpensive, straightforward, In view of the aforementioned limitations, we chose a combined ap- adaptable on site-specific conditions (e.g. Guo et al., 2007; Lin et al., proach using public datasets and web services, remote sensing as well as 2016), have a minimum demand of data and produce an end product numerical groundwater flow modelling to generate input data for an embeddable into decision-making processes (Focazio et al., 2002). In index method based assessment of aquifer vulnerability, hazard and consequence, index based methods such as DRASTIC were applied to risk. Particularly the embedding of both a simplified 2D groundwater evaluate groundwater vulnerability for basins with very diverse condi- flowmodelandweb-querydataintheindex based assessment methods tions in many regions of the world such as Europe (e.g. Albuquerque aims to improve dealing with aforementioned problems of scarce field et al., 2013; Panagopoulos et al., 2006), the Middle East (e.g. Jamrah et data on larger scales. al., 2008; Awawdeh and Jaradat, 2010), Nepal (Shrestha et al., 2017) In this context the main objectives of this study are: Firstly, to devel- and China (e.g. Huan et al., 2012; Wang et al., 2012; Ye et al., 2015). op a methodology to combine web-query data from web mapping ser- However, the potential of standardized index methods based on a vices, numerical groundwater modelling and public datasets into a GIS- few key factors to predict groundwater vulnerability is questioned based groundwater risk assessment framework, secondly to apply the among researchers (Rupert, 2001). Gogu and Dassargues (2000) there- method to provide maps of groundwater vulnerability, groundwater fore emphasized the need to integrate process-based methods such as hazard potential and groundwater contamination in the entire Songhua numerical models into widely used index methods to take into account River Basin at the highest resolution available and thirdly, to identify the the physical processes of water movement and the associated fate and importance and interrelation of input parameters and obtained maps transport of contaminants in the environment. There have been a very using sensitivity analysis methods. limited number of studies to quantify intrinsic vulnerability by including numerical modelling (Yu et al., 2010; Sophocleous and Ma, 1998; 2. Study area description Connell and Van Den Daele, 2003). Process based methods need extensive field data, both for model setup and model calibration (Butscher The Songhua River Basin is located in the northeastern region of and Huggenberger, 2008). Especially for the assessment of larger catch- China (119°52′–132°31′ E and 41°42′–51°38′ N) and covers an area of ments, such data and information are scarce and unequally distributed N550,000 km2 (Fig. 1). Its water drains to Songhua River, which is with (Ireson et al., 2006; Wu et al., 2011). Furthermore, a considerable a length of N2300 km the largest tributary of Amur River. Songhua amount of computational power is required to simulate process-based River main tributaries are Nen River, which drains the Northern part of methods at larger scales, particularly if chemical and biological process- the basin and Second Songhua River coming from Jilin province in the es and unsaturated zone mechanisms are included in the simulation Southern part of the catchment. The basin covers vast areas of the (Anane et al., 2013). three Chinese provinces Heilongjiang, Jilin and Inner Mongolia as well 600 E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609

Fig. 1. River network and topography of Songhua River Basin. The upper-right map shows the location of the basin within the Peoples Republic of China. as a small part of Liaoning province. It contains mountainous areas in the contaminants, which are introduced into the ground surface. This is de- North-West (Greater Khingan Mountains and Smaller Khingan Moun- termined by two groups of indicators (Fig. 2): Firstly by the aquifer's in- tains) and South East (Changbai Mountains) with altitudes up to trinsic vulnerability, which describes the ease with which a contaminant N2500 m as well as extensive plain areas such as the Songnen and the introduced to the subsurface can reach and diffuse the groundwater Sanjiang plain in the middle and lower reaches of the Songhua River. Par- (Liggett and Talwar, 2009) and secondly by the hazard which is defined ticularly Songnen plain makes up one of the three largest black soil belts as a potential source of contamination resulting from human activities in the world and is an important location for commodity grain and (Andreo et al., 2006). lifestock husbandry in China (Wang et al., 2014). This region belongs to In this paper, aquifer vulnerability is assessed using the DRASTIC the northern temperate monsoon climate zone. Annual precipitation in method (Aller et al., 1987). The Acronym DRASTIC stands for the seven the basin is between 400 and 900 mm and is mainly concentrated on parameters, which are intended to use for groundwater vulnerability as- summer and fall (Li et al., 2016). Additionally, the amount of annual rain- sessment: Depth to groundwater, net Recharge, Aquifer media, Soil fall increases southeastward and with altitude. media, Topography, Impact of vadose zone and hydraulic Conductivity. From a hydrogeological perspective, Quaternary pore aquifers are Each parameter is classified into either ranges of physical quantity or a distributed in the plain areas of the Songhua river Basin and the most ex- qualitative description which correspond to a rating varying between 1 tensively exploited type of aquifer. They consist of unconsolidated mate- (low vulnerability) and 10 (large vulnerability). Finally the DRASTIC rial formed by alluvial, fluvial, lacustrine and glacial processes and reach index (DI) is computed by sum up the ratings which are each assigned a thickness of up to 300 m (Zaisheng et al., 2008). The Quaternary aqui- by specificweights: fers have a complex multilayered structure (Feng et al., 2014), including confined and unconfined aquifers consisting of sand, pebbly sand or DI ¼ Dw Dr þ Rw Rr þ Aw Ar þ Sw Sr þ Tw Tr þ Iw Ir þ Cw Cr ð1Þ gravel which are separated by aquitard layers made of clay. This aquifer system is underlined by a bedrock (aquiclude) made of argillaceous Where r is the variable rating and w is the weight assigned to that pa- sandstone (Wang et al., 2014). In the mountainous areas, local aquifers rameter. Standard weighting scheme with a parameter weighting from 1 consist either of pore water in unconsolidated fluvial sediments of the to 5 was applied as it covered a large variety of different subsurface con- river valleys or of pore-fissure water in clastic sedimentary rocks (e.g. ditions (Table 1).In general, a higher calculated DRASTIC index corre- sandstone, mudstone) of Tertian or Cretaceous origin (Jiang et al., sponds to a greater groundwater pollution potential of the assessed 2014). These aquifers typically contain only limited water reserves and aquifer and vice versa (Aller et al., 1987). are not used extensively (Zhang et al., 2015). The assessment of potential hazards based on two parameters, land cover used as a proxy to describe hazards from diffusive sources and 3. Materials and methods the web-query based point hazard inventory, which disclose potential point sources of hazards (Table 1). The hazard ratings for the available 3.1. Groundwater risk assessment index method land cover classes were adopted from Gomezdelcampo and Dickerson (2008). We used a German hazard class system for rating the point haz- The idea of an index based risk assessment is to find a quantitative ard inventory, which provides hazard classifications and ratings for in way of expressing the likelihood that an aquifer be polluted by total 284 industrial sectors (Sächsisches Landesamt für Umwelt und E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609 601

Fig. 2. Schematic ﬂowchart of the methodology adopted for this study. The brackets in the green boxes indicate the data source. For an explanation of the abbreviations, see the running text. (For interpretation of the references to color in this ﬁgure legend, the reader is referred to the web version of this article.)

Geologie, 1997). The hazardousness of these sites was rated between The basic groundwater contamination risk index (RI) was generated three (e.g. laundries) and nine (e.g. petrol stations). Map locations with by summing up the hazard and vulnerability index of each cell: no hazardous sites were rated with the lowest ranking of one. ¼ þ ð Þ The harmfulness of a hazard was quantiﬁed for each location by: RI DI HI 3

All resulting indices of hazard harmfulness, intrinsic vulnerability

HI ¼ LW LR þ PW PR ð2Þ and groundwater risk were classified into five equal sized categories de- scribing very low, low, moderate, high and very high vulnerability and harmfulness, respectively. where HI is the hazard index, LR and PR are the ratings for land cover and point hazard inventory and Lw and Pw the specificweights.Bothspecif- 3.2. Data collection and compilation methodology ic weights were set to 12 to ensure both, that each parameter has an equal impact on the HI calculation and that the maximum range of In this study, different data sources and collection techniques and values is similar with the vulnerability index calculation. procedures were applied to derive parameters and ratings required for

Table 1 Rating scheme applied for the aquifer vulnerability rating and the hazard rating. For the hazard inventory categories only a few examples are displayed due to their large total number.

Aquifer vulnerability rating Hazard rating

Name Depth to water Recharge Aquifer media Soil media Slope Vadose zone Conductivity Land cover Hazard [m] [mm/year] [−] [−] [%] [−] [m/day] [−] inventory

Weight 5 4 3 3 1 5 3 12 12

Rating Range/Category

1 N30.5 b51 Clay Nonshrinking N18 – b4.1 Water/wetland None clay 2 26.7–20.5 51–71.4 Loess, clastics Peat 17–18 Clay 4.1–12.2 Forests – 3 22.9–26.7 71.4–91.8 Metamorphic igneous rock Clay loam 15–17 Loess, clastics 12.2–20.3 – Turnery fine sand 4 15.2–22.9 91.8–117.2 Weathered igneous or Silty loam 13–15 Metamorphic/igneous 20.3–28.5 Grassland/shrubland Sawmill metamorphic rock fine sand 5 12.1–15.2 117.2–147.6 Medium sand Loam 11–13 – 28.5–34.6 Settlement Brewery 6 9.1–12.1 147.6–178 Sandstone/limestone Sandy loam 9–11 Sandstone/limestone 34.6–40.7 – Laundry medium sand 7 6.8–9.1 178–216 Weathered sand/limestone Shrinking 7–9 – 40.7–61.1 – Orchard clay 8 4.6–6.8 216–235 Coarse sand and gravel Peat 4–7 Coarse sand and gravel 61.1–71.5 Barren land Airport 9 1.5–4.6 235–254 – Sand 2–4 – 71.5–81.5 – Petrol station 10 1.5 N254 – Gravel/no b2 – N81.5 Cropland – soil 602 E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609 the GIS based computation and the subsequent creation of groundwater 13 rock classes by using literature values (Domenico and Schwartz, risk maps (Fig. 2). Three types of primary data sources were used for 1998; Heath, 1983). input data: public geospatial datasets, web service databases and point Land cover data was obtained from the GlobeLand30 Land Cover measurement data. Prior to all index calculations, specific preprocessing dataset (Jun et al., 2014). This dataset classifies land cover into 10 classes steps were needed for most datasets in order to bring them in an ade- based on processed images of Landsat and Chinese HJ-1 satellites obser- quate structure. All processed data grids were uploaded to the QGIS vations in the year 2010. GIS System (QGIS Development Team, 2009) and clipped to the extent of Songhua River Basin. Grids with a cell resolution other than 30″ 3.2.2. Hazard inventory establishment were modified to 30″ cell size, which represents a cell with an area of For the establishment of a hazard inventory place information on po- about 1 km2. The cell size was selected considering the spatial resolution tentially hazardous sites was queried from web service database. In gen- of available data and computational considerations. eral, places web service APIs allow searching for places on a variety of categories such as type and location and provide frequently updated information on businesses and points of interest. It can be applied on larger 3.2.1. Processing of public geospatial data scales and data scarce regions where data availability of catastral maps is The majority of parameters for the risk index assessment were calcu- limited. However, access to these web services has restrictions on related based on geospatial data obtained from web sites maintained by quest frequency and number of provided data per request. Additionally national agencies or international organizations. All datasets were freely for establishing of a hazard inventory, information on brownfields are available in the internet but differ in format, extent and resolution not obtainable from web service database. (Table 2). The topography of the study area was obtained from the For this study, we used the place web service API of Baidu Maps Digital elevation Model (DEM) dataset provided by the HydroSHEDS (Baidu Inc., 2016a), a web mapping service launched in 2005 by the Chi- project (Lehner et al., 2008). This dataset based on primary data obtained nese company Baidu Incorporation. The Baidu Place API was accessed during a Space Shuttle flight for NASA's Shuttle Radar Topography using a python client library (Song, 2015) and a valid API key. Each Mission (SRTM), which were reprocessed for hydrological purposes by place research request needed two input parameters, the search term e.g. “stream burning” procedures. Slopes for each cell were calculated (query) and the geographical boundaries which to retrieve place infor- from this DEM dataset using the Terrain Analysis tools in QGIS. Hereby mation (Fig. 3). Search terms were provided by German hazard class sys- the percentage maximum change in height of each cell was used as the tem (Sächsisches Landesamt für Umwelt und Geologie, 1997) excluding slope of the cell. terms related to military facilities. Recharge was directly computed from the difference between precip- The Baidu Place API allows up to 2000 requests per day and provides itation and evaporation (for an overview of recharge computation a maximum of 20 results per request. Songhua River Basin was delineat- methods see Shirazi et al., (2012)). Considering that this approach is ed into 94 cells with a cell size of 1° for the data requests. We automa- neglecting direct-runoff, derived values are reflecting an upper limit for tized the request procedure using a python script in a way that if 20 the DRASTIC rating procedure. Mean monthly precipitation data could results were received by a request, the affected geographical boundary be obtained from the WorldClim 1.4 – Global Climate Database which cell was quartered and four new requests were started. If necessary, based on weather station data from numerous sources (Hijmans et al., this procedure was repeated until the cell resolution of the resulting 2005). Actual Evaporation (i.e. evapotranspiration) data in a daily resolu- index map (30 arc sec) was reached. Data queries were submitted be- tion were provided by the GLEAM (Global Land Evaporation Amsterdam tween 29th of November and 19th of December 2016. Model) Version 3.a dataset which includes transpiration, bare-soil evap- Outputs of each request were returned as JSON arrays containing oration, interception loss, open-water evaporation and sublimation in its several fields such as name, type and location (Baidu Inc., 2016b). The calculation algorithm for evaporation (Miralles et al., 2011). Prior to cal- obtained arrays were used to evaluate the data and subsequently delete culation, both datasets were summed up to yearly averages in order to non-appropriate results. Baidu Place API provides the geographical co- derive a grid representing average spatial distribution of recharge. ordinates for the BD-09 coordinate system (Baidu Inc., 2016c), which The National Soil Database of China (SDC) provided the geospatial in- has an offset from the WGS-84 coordinate system of a few hundred me- formationonsoilcover(Shangguan et al., 2013). Required soil textures ters (Jianshuo, 2008). However, this offset was neglected, as it is smaller for DRASTIC method were identified by applying the triangle method than the cell resolution of the hazard index map. Finally, all results were of the Unified Soil Classification System on the vertically averaged parti- assigned with the corresponding hazard class ratings from the German cle size distribution of Sand, Silt and Clay (McMinn, 1960). hazard rating system and transferred into the raster file format. If more Information on hydrogeology were obtained from the high resolu- than one hazardous site was situated on the same single raster cell, the tion global lithological map (GliM) which was assembled from existing maximum value was assigned to the specific cell. regional geological maps (Hartmann and Moosdorf, 2012). GliM represents hydrogeology in Songhua River Basin with 3015 polygons of varying size and 88 different rock types. Hence, rock type classes were 3.2.3. Groundwater model merged to 13 classes consistent with the DRASTIC categories. The map For depth to groundwater calculation, the Chinese Ministry of Water was further converted into a geospatial grid and interpolated to the re- Resources (MWR) provides monthly drawdown measurements in a total quired resolution. Hydraulic conductivities were assigned to each of the 59 wells (Fig. 4) in Songhua River Basin (Ministry of Water Resources, 2015). Considering the large study area and the inhomogeneous spatial distribution of the wells, computed water tables using spatial interpola- Table 2 fi Characteristics of primary data sources used within this study. tion schemes signi cantly depend on the used interpolation scheme Abbreviations are explained in the running text. (Mitas and Mitasova, 1999). Furthermore, conventional spatial interpolation schemes are not able to include the impact of budget changes, Data type Data source Type Domain Resolution e.g. by groundwater recharge and pumping on the water table outside ″ DEM HydroSHEDS/SRTM Grid Global 3 the observation points. Subsequently, the depth to groundwater in Song- Precipitation WorldClim Grid Global 30″ Evaporation GLEAM Grid Global 0.25° hua River Basin was estimated by steady-state regional groundwater Soil SDC Grid China 30″ flow modelling using OpenGeoSys. OpenGeoSys is a scientific open Land cover GlobeLand30 Grid Global 1″ source project for the development of numerical methods to simulate 2 2 Hydrogeology GliM Polygons Global 150 m –23,000 km thermo-hydro-mechanical-chemical (THMC) processes in porous and GW-level MWR Points Local – fractured media based on the finite element method (Kolditz et al. E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609 603

Fig. 3. Pseudo-code for the schematic presentation of the Baidu Place API assessment using the python client.

2012). It has been widely used in previous studies for simulation of For calibration under steady-state conditions, we used the yearly av- groundwater flow (e.g. Sun et al., 2012; Wu et al., 2011). erage of hydraulic heads measured in the 59 monitoring wells. The cal- Songhua River Basin was transformed in a 2D triangular mesh ibration itself was conducted using the model-independent parameter consisting of 130,521 finite elements with an average size of about estimation code PEST (Doherty, 2015). In this context, hydraulic con- 4 km. Recharge and topography model input data were gathered from ductivity was used as the calibration parameter. Hydraulic conductivity the public geospatial databases (see Section 3.2.1). Average annual was defined as homogenous and isotropic for each cell of the model do- groundwater abstraction rates from 2010 to 2015 were taken from the main due to several reasons. Firstly, although information about the statistical yearbooks (SYB) of the provinces sharing Songhua River dominant hydrogeology structures was available (Table 2), a sufficient Basin (Heilongjiang Statistical Bureau, 2011–2016; Jilin Statistical number of bore log data was not available to set-up a structural model Bureau, 2011–2016; Inner Mongolia Statistical Bureau, 2011–2015) of the complex layered hydrostratigraphy in Songhua River Basin. Sec- and the Chinese National Resources Data Base (Zehui, 2002). Ground- ondly, the model simulates groundwater flow as saturated zone flow water recharge and abstraction rates were assigned as a surface bound- using the estimated recharge as a direct boundary condition of the ary condition to all model nodes except at nodes belonging to larger model. Hence, processes such as interflow in the unsaturated zone streams where a constant head condition was obtained from the topog- and preferential flow path along geological fractures are neglected sys- raphy. This approach followed the hydrogeological perspective that sur- tematically. This leads to an overestimation of hydraulic conductivity face water represents a surface manifestation of groundwater (Winter values by the model, particularly for the areas where non-sedimentary et al. 1998). Computed hydraulic heads by the model were interpolated aquifers are dominant. Additionally, the calibration process revealed to a resolution of 30 arc sec and finally subtracted from the elevation that using a heterogeneous (13 different hydraulic conductivity values dataset in order to calculate the depth to water of each grid cell. for each hydrogeological unit) set-up for the stationary groundwater

Fig. 4. Mesh structure and calibration results of the groundwater model. Left: 2D finite element mesh of the model domain. The mesh cells are colored by the different hydrogeological layers and the red dots show the location of the 59 monitoring wells used for the calibration. Right: Observed and measured groundwater wells for the stationary groundwater model with either homogeneous or heterogeneous hydraulic conductivity properties. A linear regression results in a correlation coefficient of 0.84 and 0.85, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 604 E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609

flow model did not significantly increase the linear correlation coeffi- (Fig. 5d). The second (about 18.6%) and third (3.6%) largest proportion cient between observed and simulated groundwater values (Fig. 4). were silt loam soils in the hilly parts of the catchment and clay loam soils mostly located in parts of the Songnen plain, both having a DRASTIC 4. Results and discussion rating of 4. The highest DRASTIC rating of 9 referred to sandy soils, which however, only covered a small area in the southern part of the catchment. 4.1. Vulnerability and hazard parameter maps According to the hydrogeological data, the groundwater in the plain areas of Songhua catchment was stored in sediments deposited since In Songhua River Basin, modelled stationary depth to water ranged the Pleistocene and Quaternary period. This type of aquifer covered between a few meters and N50 m covering the entire range of DRASTIC about 36% of the total aquifer area (Fig. 5e–f). The largest relative propor- indices (Fig. 5a). The groundwater model results showed that a shallow, tion of groundwater was, with 46% of the total subsurface area, stored in more vulnerable, groundwater system prevailed mainly in regions with igneous and magmatite lithological layers. They were located in the hilly low altitudes such as in the area of Songnen and Sanjiang plain as well and mountainous headwaters of the catchment, which contained differ- was along the main rivers draining the catchment. Low recharge rates ent types of sedimentary rock aquifers, too (18% of total area). The latter of b50 mm per year were calculated for vast parts of the Songnen plain one corresponded to DRASTIC values between two and seven reflecting as well as the Greater Khingan mountain range (Fig. 5b). High recharge the different type of sedimentary rock (e.g. sandstone, mudstone, lime- rates were calculated for the mountainrangesintheeasternpartofthe stone) and the impact of weathering processes. Groundwater in massive catchment with a peak of about 300 mm/a in the area of Changbai Moun- or thin-bedded carbonate rock aquifers was, with 0.3% in total, very rare tains which cover the highest parts of the catchment. in the catchment and limited to small areas in the Southern headwaters. A high topographical gradient with slopes up to 13.3% was found in The pattern of derived hydraulic conductivities reflected the subsur- the mountainous areas of the catchment (Fig. 5c). In contrast, calculated face properties (Fig. 5g). Higher hydraulic conductivities were associated slopes in the plain areas and the larger river valleys were usually smaller to parts of the Quaternary aquifers in the plain areas of Songhua catch- than 2% making potential pollutants more likely to infiltrate the subsur- ment, particularly to those, which consisted of sands with different face instead of being run offed. grain size. On the other hand, clay and silt rich parts of these aquifers re- Based on the USCS soil classification N72% of the soils in the catch- sulted in low DRASTIC ratings. The igneous, magmatite and sedimentary ment were loamy soils associated with a medium DRASTIC rating of 5 lithological layers forming large parts of the catchment headwaters

Fig. 5. Selected rating maps used for DRASTIC (Fig.3 a-g) and hazard index (Fig. 3h–i) computation in Songhua River Basin. The ratings of the point hazardous sites are shown as vector points. The categorization from 1 (dark green color) to 10 (red color) corresponds with the rating scheme explained in Table 1 whereas higher values indicate a higher contribution to groundwater vulnerability or hazard potential. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609 605 aquifer system usually had, except of strong weathered or fractured simultaneously, the areas with the highest contamination risk class parts, hydraulic conductivities orders of magnitude smaller than 1 m/d, were located in the plain areas near large agglomerations. An area of par- giving these aquifers the lowest possible DRASTIC rating. ticularly high groundwater risk could be identified in the Southern parts The assessment of the land cover distribution, revealed the impor- of the catchment near Changchun with computed risk index values up to tance of Songhua River Basin for crop production. Cultivated land associ- 425. In contrast, areas with high groundwater risk were much less in the ated with a high hazard rating of 9 covered about 42% of the land surface, large metropolitan area of Harbin city, the second provincial capital in particularly in the area of the large plains (Fig. 5h). Forests were the the catchment due to different subsurface properties and climatic condi- dominating land cover in the mountainous areas with 33% of total land tions. Additional smaller scattered areas associated with high and cover followed by grassland with about 19%. Urban settlements covered highest risk to be polluted were mainly located in the urban plain areas about 2.5% of the land surface either as large settlement agglomerations of the catchment such as around the cities of Baicheng and Songyuan such as Changchun, Harbin or Jilin or as dispersed spots at other parts of in the Southeastern part of the catchment. Low groundwater risks the catchment. Minor parts of the catchments surface of b4.5% consisted were obtained for the mountainous headwaters of Songhua River mainly of open water, bare land and wetlands. duetolesspotentialhazardsandfavorable vulnerability conditions such The data query on Baidu Place API returned an inventory of in total as steep slopes and low hydraulic permeability. This was particularly the 20,234 sites classified into 118 different categories according to the case for the North-Western Mountain ranges, where relatively low re- used hazardous sites catalogue (Fig. 5i). The most widespread hazardous charge rates further decreased the load of potential contaminants to- sites were laundries with 19.6% percent of all results followed by petrol wards the groundwater. stations (18.1%), repair shops (11.4%) and building companies (9.0%). Al- though smaller in individual numbers, the data query revealed that at 4.3. Parameter sensitivity analysis least 440 chemical producers and 334 companies employing metal working technologies were situated at Songhua River Basin. The spatial Although subjectivity is practically unavoidable when developing distribution of potentially hazardous sites was related to the population risk assessment maps on a regional-scale, sensitivity analysis helps to distribution in the basin. Most sites were found at much higher densely identify how the composition of factors and weights works and is min- populated areas in the central region of the catchment in the proximity imizing doubts on the accuracy of the assessment results (Pacheco et al., to the large cities as well as along the main roads connecting the urban 2015; Wang et al., 2012). It is widely applied in groundwater risk assess- areas. In contrast, apart from river valley, pollutant point sources were ment to assess the relative sensitivity of the model output with respect rare at the mountainous headwater areas of the catchment. to changes of model parameters (e.g. Bartzas et al., 2015; Babiker et al., 2005). 4.2. Groundwater index maps For the study area, a sensitivity analysis was conducted by performing the single-parameter sensitivity analysis (Napolitano and After the construction of all parameters maps GIS raster calculator Fabbri, 1996). This method evaluates the impact of each of the parame- was used to combine the different attributes to compute maps showing ters on the index generation by comparing the effective weight (Wp) the distribution of the DRASTIC vulnerability index, the hazard index and that each parameter had in each cell with the theoretical weights the risk index in Songhua catchment (Fig. 6a–c). All indices were assigned by the index assessment. The effective weight for each cell reclassified into five equal sized classes between the minimum and max- can be calculated by: imum possible index values. Derived DRASTIC Indices (DI) were in a ðÞ ðÞ range between 37 and 211. In Songhua River Basin, 75.1% of the aquifers PR i PW i WpðÞ¼i ð4Þ had a low or very low vulnerability, 20.3% a medium vulnerability and IndexðÞ i 4.6% a high or very high vulnerability. Zones of high vulnerability were mainly located in the eastern part of Songnen plain, the northwestern Where PR and PW are the ratings and weightings of the parameter P part of Sanjiang plain and close to the riverbanks of the major streams. assigned to each raster cell i. Thus, the effective weight of each parameter Particularly a few clusters in the Southeast of the basin received DRASTIC of each cell expresses the value of each parameter in the context of the values of N200. The majority of the aquifers in the basin lowlands were values of the other parameters (Napolitano and Fabbri, 1996). For our rated by medium or low vulnerability whereas very low groundwater analysis, the weighting for each parameter was evaluated for both, the vulnerability could be estimated predominantly in the Northwestern DRASTIC and the hazard assessment. The statistical analysis of the results mountainous areas of the catchment. The high proportion of these was independently conducted for all cells belonging to each of the 5 areas revealed that the Songhua River Basin is relatively protected from index classes. contamination on the groundwater surface. The effective weights of the DRASTIC parameter showed some signif- Potential hazards were evaluated by assigning hazard indices to each icant derivations from their theoretical weights (Fig. 7). Regarding the raster cell according to Eq. (2). The calculations showed that more than entire domain of Songhua River Basin, the vadose zone tends to be the half of the area in Songhua River Basin was associated with a very low most effective parameter with an average effective weight of 25.0%, or low potential harmfulness (Fig. 6b). The grouping of hazard point which was also larger in comparison to the theoretical weight defined sources and the large forest covered areas, particularly in the headwater by the DRASTIC method (21.7%). Additionally the parameters topogra- regions of the catchment resulted that area with very low potential phy (slope) and soil exceeded their theoretical weight due to the flat to- harmfulness were about 15 times more frequently than areas with low pography in many areas of the basin and the large occurrence of silty and potential harmfulness. Medium potential harmfulness was derived for loamy soils. In contrast, for parts of the catchment with higher vulnera- N40% of the basin and was, to a large part, assigned to areas of Songnen bility, the depth to groundwater parameter inflicted the larges impact plain and Sanjiang plain used for agricultural purposes. Areas of high and on the vulnerability of the specific aquifers whereas the impact of most even very high potential harmfulness were rare with in sum b1% of the other parameters decreased. For the areas with medium, high and very total area, found mostly in more densely populated areas in the Songnen high vulnerability, the average effective weight of this parameter on and Sanjiang plain. Hot spots of highest hazard class were located in the the resulting vulnerability was between 2.9% and 5.0% higher than the metropolitan area of large cities such as Harbin, Jilin and Changchun and theoretical weight. Considering, that areas with higher total vulnerability along the main traffic routes connecting these mega-cities. are priority areas for groundwater management, the significance of the The groundwater risk index (Fig. 6c) was obtained by spatial overlay depth to water and the geological parameters highlights the importance of the hazard map and the vulnerability map according to Eq. (3). Taking of continuously improving the accurateness, resolution and reliability of into account the hazard potential and groundwater vulnerability obtained information about these factors. 606 E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609

Fig. 6. Groundwater index maps of Songhua River Basin. DRASTIC vulnerability index (a), hazard index (b) and risk index (c) were classified in five categories from very low (dark green color) to very high (red color) groundwater vulnerability, potential harmfulness and pollution risk. The percentage frequency distribution of all index maps categories is given by (d). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The theoretical weight of the two hazard assessment parameters is hazard inventory with the lowest possible rating. On the other hand, ef- 50% according to the weights distribution in Eq. (2). The effective weight fective weights of both parameters were almost equal for grid cells analysis showed that the land cover parameter was the dominant pa- assigned with a high and very high hazard potential. The particular rele- rameter for the entire domain, particularly for areas with low to medium vance of those areas clearly showed that beside land cover data (e.g. hazard potential. Main reason is the point-like distribution of the hazard- Fritch et al., 2000; Gomezdelcampo and Dickerson, 2008) the additional ous sites parameter, which assigned grid cells without an entry in the set-up and integration of a point hazard inventory into the groundwater risk assessment is of great signiﬁcance for the delineation of hot-spots of potential groundwater pollution.

4.4. Independence of parameter and indices

It is evident that parameters of the applied index methods are interacting dependent variables. On the one hand a strong correlation between parameters is favorable as it decreases the effect of individual misjudgement and data errors on the derived vulnerability index but can also amplify errors if the parameters are not classed separately (Rosen, 1994). Furthermore some effects of pollutant transport towards the groundwater are only implicitly included in the index based method. For instance, although hydraulic conductivity is the parameter directly related to groundwater velocity only the interrelation with other parameters such as net recharge, aquifer media and topography finally defines the velocity conditions (Aller et al., 1987). Similarly, interrelations between parameters of the vulnerability and the hazard potential assessment may exist, too. As an example, extensive land use over long periods can change result the degree of percolation through the soil ma- Fig. 7. Statistics of the single parameter sensitivity analysis. Up: Effective mean weights of trix by altering its colloidal nature, which is one of the parameters incor- all DRASTIC parameters for all grid cells belonging to a certain vulnerability class. The poratedintheDRASTICmethod(Merchant, 1994). errorbars show the standard deviation for each dataset. The dashed lines indicate the theoretical weight for each parameter. Low: Effective mean weights and standard The statistical summary of correlation between all parameters, evalu- deviations for the two categories of the hazard assessment. ated by the correlation coefficients of each parameter pair for all cells in E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609 607

Fig. 8. Correlation analysis of risk assessment parameters and indices in Songhua River Basin. Left: Table of correlation coefﬁcients between all parameters used for vulnerability and hazard assessment. Right: Correlation between unique DRASTIC vulnerability index values and associated mean hazard index computed over all cells of Songhua River Basin. A linear regression resulted in a correlation coefﬁcient of 0.84.

Songhua River Basin, is shown in Fig. 8. Large positive correlations are measures must be taken for densely settled plain areas of the catchment, assigned with a correlation coefficient close to 1 and values close to e.g. in the vicinity of Changchun metropolitan area. Furthermore, inten- −1 refer to a strong negative correlation. Strongest correlation was cal- sive agriculture takes place in the chernozem soil belts of the catchment culated between the parameters “impact of vadose zone” and “aquifer over aquifers with medium to groundwater contamination risk posing a media” which refers to their construction from a common database. Ad- threat to groundwater contamination by fertilizer residuals or pesticides ditionally, high correlations were evident between these two parameters from diffuse sources. In contrast, the elevated mountainous parts of the and the other hydrogeological parameter hydraulic conductivity. On the catchment displayed low to very low contamination risk. other hand, weakest correlations appeared where recharge was in- Although subjectivity was unavoidable in the study, some measures, volved. Both hazard index parameters had a weak correlation between such as the aggregation and classification of different parameters in a each other mainly attributed to the large areas of the basin for which unified form and the computation of risk-class-dependent parameter no point pollution sources could be obtained by the web-based hazard sensitivity, were able to overcome this disadvantage to some extent. inventory assessment. The significant correlation between derived vulnerability and harmful- Although the correlation table showed no correlation between ness of an area highlighted the necessity to integrate both within the DRASTIC vulnerability and hazard index parameters, we found a strong contamination risk assessment. Hence, the derived groundwater vulner- tendency that a higher vulnerability index value was associated with a ability, potential hazard and groundwater contamination risk maps may higher hazard index value (Fig. 8). One reason for this is the fact that support decision-making in the field of future land planning and ground- human settlements and related anthropogenic activities are clustered water management. Moreover, provided maps and assessment method- in areas with patches of good soil, access to water bodies, absence of ology could be combined with a chemical specific exposure assessment steep slopes and dense forests. This landscape description fits well to to assess human health risks from groundwater pollution in the Songhua the large plain areas of Songhua River Basin which, however, have aqui- River Basin. fers more vulnerable to the potential pollution associated with human Further efforts should be conducted to overcome present resolution activities. Subsequently, this interrelation stresses the fact that both in- limitations in order to make the maps more useful on a local scale and to trinsic vulnerability and the hazard potential needed to be analysed to- improve validation concepts. Moreover, further studies are needed to gether in order to evaluate the potential groundwater pollution risk of a investigate temporal changes of groundwater contamination risk in certain area. the area, e.g. due to the man-made lowering of the groundwater table or climate change impacts, in order to predict short-term effects of 5. Conclusions land-planning decisions.

This study provides a new method for assessing groundwater contamination risk on a large scale for the (upper) aquifers of Songhua Conflict of interest River Basin. This was achieved by the integration of intrinsic vulnerability assessment using the DRASTIC model and an assessment of potential The authors have no conflicts of interest. point and diffusive hazards in order to calculate the contamination risk indices. A new approach was proposed which combines public datasets, web service databases and groundwater flow modelling to derive the Acknowledgements input datasets for the index computation. A GIS platform was used to scale and overlay input datasets from different sources and to compute We would like to thank Dr. Qiang Wang from the Heilongjiang Re- and visualize evaluation results in a grid resolution of 30 arc sec, which search Institute of Environmental Science for data support. This work is reflected the resolution limit for most of the currently available datasets. part of the H2O SUSTAIN project and financially supported by the EU- The sole requirement of public available datasets and open source soft- China Environmental Sustainability Programme of the European Union ware for our assessment approach ensured its reproducibility and (EuropeAid/133582/L/ACT/CN). allowed it to be transferable to assess the groundwater risk in other catchments in China and the world with limited access to field data. References The application of this methodology to the entire Songhua River Basin demonstrated that vulnerability and harmfulness of a significant Albuquerque, M.T.D., Sanz, G., Oliveira, S.F., Martinez-Alegria, R., Antunes, I.M.H.R., 2013. Spatio-temporal groundwater vulnerability assessment - a coupled remote sensing part of the catchment's plain areas was emphasized. According to the and GIS approach for historical land cover reconstruction. Water Resour. Manag. study results, the most urgent need for groundwater protection 27:4509–4526. http://dx.doi.org/10.1007/s11269-013-0422-0. 608 E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609

Aliewi, A., Al-Khatib, I.A., 2015. Hzard and risk assessment of pollution on the groundwa- Jamrah, A., Al-Futaisi, A., Rajmohan, N., Al-Yaroubi, S., 2008. Assessment of groundwater ter resources and residents' health of Salfit District, Palestine. J. Hydrol. Reg. Stud. 4: vulnerability in the coastal region of Oman using DRASTIC index method in GIS envi- 472–486. http://dx.doi.org/10.1016/j.ejrh.2015.07.006. ronment. Environ. Monit. Assess. 147:125–138. http://dx.doi.org/10.1007/s10661- Aller, L., Lehr, J.H., Petty, R., 1987. A Standardized System to Evaluate Ground Water Pol- 007-0104-6. lution Potential Using Hydrogeologic Settings. Natl. Water Well Assoc. Jiang, H., Shan, W., Hu, Z., 2014. Bedding landslide formation mechanism and traits in Anane, M., Abidi, B., Lachaal, F., Limam, A., Jellali, S., 2013. GIS-based DRASTIC, Pesticide Lesser Khingan Mountain. In: Shan, W., Guo, Y., Wang, F., Marui, H., Strom, A. DRASTIC and the Susceptibility Index (SI): comparative study for evaluation of pollu- (Eds.), Landslides in Cold Regions in the Context of Climate Change. Springer Interna- tion potential in the Nabeul-Hammamet shallow aquifer, Tunisia. Hydrogeol. J. 21: tional Publishing, Cham:pp. 71–81 http://dx.doi.org/10.1007/978-3-319-00867-7_6. 715–731. http://dx.doi.org/10.1007/s10040-013-0952-9. Jiang, Y., 2009. China's water scarcity. J. Environ. Manag. 90:3185–3196. http://dx.doi.org/ Andreo, B., Goldscheider, N., Vadillo, I., Vias, J.M., Neukum, C., Sinreich, M., Jimenez, P., 10.1016/j.jenvman.2009.04.016. Brechenmacher, J., Carrasco, F., Hötzl, H., Perles, M.J., Zwahlen, F., 2006. Karst ground- Jianshuo, W., 2008. All Maps in China are Transformed ([WWW Document]. URL http:// water protection: first application of a Pan-European Approach to vulnerability, haz- home.wangjianshuo.com/20081109_all_maps_in_china_are_transformed.htm ard and risk mapping in the Sierra de Libar (Southern Spain). Sci. Total Environ. 357: (accessed 15.12.16)). 54–73. http://dx.doi.org/10.1016/j.scitotenv.2005.05.019. Jilin Statistical Bureau, 2011–2016. Jilin Statistical Yearbook. Beijing. Awawdeh, M.M., Jaradat, R.A., 2010. Evaluation of aquifers vulnerability to contamination Johansson, P.-O., 1999. Framework for ground water protection-the Managua ground in the Yarmouk River basin, Jordan, based on DRASTIC method. Arab. J. Geosci. 3: water system as an example. Ground Water 37, 204–213. 273–282. http://dx.doi.org/10.1007/s12517-009-0074-9. Jun, C., Ban, Y., Li, S., 2014. China: open access to Earth land-cover map. Nature 514:434. Babiker, I.S., Mohamed, M.A.A., Hiyama, T., Kato, K., 2005. A GIS-based DRASTIC model for http://dx.doi.org/10.1038/514434c. assessing aquifer vulnerability in Kakamigahara Heights, Gifu Prefecture, central Japan. Kolditz, O., Bauer, S., Bilke, L., Boettcher, N., Delfs, J.O., Fischer, T., Goerke, U.J., Kalbacher, T., Sci. Total Environ. 345:127–140. http://dx.doi.org/10.1016/j.scitotenv.2004.11.005. Kosakowski, G., McDermott, C.I., Park, C.H., Radu, F., Rink, K., Shao, H., Shao, H.B., Sun, Baidu Inc., 2016a. 概述 Web服务API [WWW Document]. URL http://lbsyun.baidu.com/ F., Sun, Y.Y., Singh, A.K., Taron, J., Walther, M., Wang, W., Watanabe, N., Wu, Y., Xie, M., index.php?title=webapi (accessed 15.12.16). Xu, W., Zehner, B., 2012. OpenGeoSys: An open-source initiative for numerical simu- Baidu Inc., 2016b. Place API Web服务API [WWW Document]. URL http://lbsyun.baidu.com/ lation of thermo-hydro-mechanical/chemical (THM/C) processes in porous media. index.php?title=webapi/guide/webservice-placeapi (accessed 15.12.16). Environ. Earth Sci. 67:589–599. http://dx.doi.org/10.1007/s12665-012-1546-x. Baidu Inc., 2016c. Baidu Web API 常见问题 [WWW Document]. URL http:// Kuisi, M. Al, Mashal, K., Al-qinna, M., Hamad, A.A., Margana, A., 2014. Groundwater vul- developer.baidu.com/map/question.htm (accessed 15.12.16). nerability and hazard mapping in an arid region: case study, Amman-Zarqa Basin Bartzas, G., Tinivella, F., Medini, L., Zaharaki, D., Komnitsas, K., 2015. Assessment of (AZB) - Jordan. J. Water Resour. Prot. 6:297–318. http://dx.doi.org/10.4236/jwarp. groundwater contamination risk in an agricultural area in north Italy. Inf. Process. 2014.64033. Agric. 2:109–129. http://dx.doi.org/10.1016/j.inpa.2015.06.004. Lehner, B., Verdin, K., Jarvis, A., 2008. New global hydrography derived from spaceborne Butscher, C., Huggenberger, P., 2008. Intrinsic vulnerability assessment in karst areas: a elevation data. EOS Trans. Am. Geophys. Union 89:93–104. http://dx.doi.org/10. numerical modeling approach. Water Resour. Res. 44. http://dx.doi.org/10.1029/ 1029/2008EO100001. 2007WR006277. Li, F., Zhang, G., Xu, Y., 2016. Assessing climate change impacts on water resources in the Connell, L.D., Van Den Daele, G., 2003. A quantitative approach to aquifer vulnerability Songhua River Basin. WaterSA 8:420. http://dx.doi.org/10.3390/w8100420. mapping. J. Hydrol. 276:71–88. http://dx.doi.org/10.1016/S0022-1694(03)00038-6. Li, X., Zuo, R., Teng, Y., Wang, J., Wang, B., 2015. Development of relative risk model for Doherty, J., 2015. Calibration and Uncertainty Analysis for Complex Environmental regional groundwater risk assessment: a case study in the lower Liaohe River Plain, Models. Watermark Numerical Computing, Brisbane, Australia. China. PLoS One 10:1–22. http://dx.doi.org/10.1371/journal.pone.0128249. Domenico, P.A., Schwartz, F.W., 1998. Physical and Chemical Hydrogeology. Wiley. Liggett, J.E., Talwar, S., 2009. Groundwater vulnerability assessments and integrated water Erwin, M.L., Tesoriero, A.J., 1997. Predicting ground-water vulnerability to nitrate in the resource management. Watershed Manag. Bull. 13, 18–29. Puget Sound Basin. USGS Factsheet 61, 1–4. Lin, M., Ji, D., Cui, C., Sun, Y., Su, J., Xi, B., Nixdorf, E., 2016. Groundwater Vulnerability Par- Feng, B., Xiao, C., Zhou, Y., 2014. Groundwater level forecast and prediction of Songnen tition in Ashi River Basin. Res. Environ. Sci. 29, 1773–1781. Plain in Jilin Province. Phys. Numer. Simul. Geotech. Eng. 55, 993–997. McMinn, J., 1960. Identifying soils by a triangle based on unified soil classification system. Focazio, M., Reilly, T.E., Rupert, M.G., Helsel, D.R., 2002. Assessing ground-water vulnera- Papers on Soils 1959 Meetings. American Society for Testing and Materials, Baltimore. bility to contamination: providing scientifically defensible information for decision Merchant, J.W., 1994. GIS-based groundwater pollution hazard assessment: a critical re- makers. USGS Circ. 1224. view of the DRASTIC model. Photogramm. Eng. Remote. Sens. 60, 1117–1128. Foster, S.S.D., 1987. Fundamental concepts in aquifer vulnerability, pollution risk and pro- Milnes, E., 2011. Process-based groundwater salinisation risk assessment methodology: tection strategy. Vulnerability soil Groundw. to Pollut. 38, 69–86. Application to the Akrotiri aquifer (Southern Cyprus). J. Hydrol. 399:29–47. http:// Foster, S.S.D., Chilton, P.J., 2003. Groundwater: the processes and global significance of dx.doi.org/10.1016/j.jhydrol.2010.12.032. aquifer degradation. Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 358:1957–1972. Ministry of Water Resources, 2015. 地下水动态月报.Beijing. http://dx.doi.org/10.1098/rstb.2003.1380. Miralles, D.G., Holmes, T.R.H., De Jeu, R.A.M., Gash, J.H., Meesters, A.G.C.A., Dolman, A.J., Fritch, T.G., McKnight, C.L., Yelderman, J.C., Dworkin, S.I., Arnold, J.G., 2000. A predictive 2011. Global land-surface evaporation estimated from satellite-based observations. modeling approach to assessing the groundwater pollution susceptibility of the Hydrol. Earth Syst. Sci. 15:453–469. http://dx.doi.org/10.5194/hess-15-453-2011. paluxy aquifer, central texas using a geographic information system. Environ. Geol. Mitas, L., Mitasova, H., 1999. Spatial interpolation. In: Longley, P.A., Goodchild, M.F., 39:1063–1069. http://dx.doi.org/10.1007/s002549900078. Maguire, D.J., Rhind, D.W. (Eds.), Geographical Information Systems: Principles, Tech- Fu, X., Shen, Y., Dong, R., Deng, H., Wu, G., 2016. Analysis of urbanization based on center- niques, Management and Applications. John Wiley & Sons, pp. 481–492. of-gravity movement and characteristics in Songhua River basin of China and its Napolitano, P., Fabbri, A.G., 1996. Single-parameter sensitivity analysis for aquifer vulner- southern source sub-basin between 1990 and 2010. Chin. Geogr. Sci. 26:117–128. ability assessment using DRASTIC and SINTACS. HydroGIS 96. http://dx.doi.org/10.1007/s11769-015-0757-y. Nixdorf,E.,Chen,C.,Sun,Y.,Kolditz,O.,2015.PersistentorganicpollutantscontaminateChi- Gogu, R., Dassargues, A., 2000. Current trends and future challenges n groundwater vul- nese water resources: overview of the current status, challenges and European strate- nerability assessment using overlay and index methods. Environ. Geol. gies. Environ. Earth Sci. 74:1837–1843. http://dx.doi.org/10.1007/s12665-015-4448-x. Gomezdelcampo, E., Dickerson, J.R., 2008. A modified DRASTIC model for siting Confined Pacheco, F.A.L., Pires, L.M.G.R., Santos, R.M.B., Sanches Fernandes, L.F., 2015. Factor Animal Feeding Operations in Williams County, Ohio, USA. Environ. Geol. 55: weighting in DRASTIC modeling. Sci. Total Environ. 505:474–486. http://dx.doi.org/ 1821–1832. http://dx.doi.org/10.1007/s00254-007-1133-8. 10.1016/j.scitotenv.2014.09.092. Gu, B., Ge, Y., Chang, S.X., Luo, W., Chang, J., 2013. Nitrate in groundwater of China: Panagopoulos, G.P., Antonakos, A.K., Lambrakis, N.J., 2006. Optimization of the DRASTIC sources and driving forces. Glob. Environ. Chang. 23:1112–1121. http://dx.doi.org/ method for groundwater vulnerability assessment via the use of simple statistical 10.1016/j.gloenvcha.2013.05.004. methods and GIS. Hydrogeol. J. 14:894–911. http://dx.doi.org/10.1007/s10040-005- Guo, Q., Wang, Y., Gao, X., Ma, T., 2007. A new model (DRARCH) for assessing groundwater 0008-x. vulnerability to arsenic contamination at basin scale: a case study in Taiyuan basin, QGIS Development Team, 2009. QGIS Geographic Information System. northern China. Environ. Geol. 52:923–932. http://dx.doi.org/10.1007/s00254-006- Rosen, L., 1994. A study of the DRASTIC methodology with emphasis on Swedish condi- 0534-4. tions. Ground Water 32, 278. Hartmann, J., Moosdorf, N., 2012. The new global lithological map database GLiM: a rep- Rupert, M.G., 2001. Calibration of the DRASTIC ground water vulnerability mapping resentation of rock properties at the Earth surface. Geochem. Geophys. Geosyst. 13: method. Ground Water http://dx.doi.org/10.1111/j.1745-6584.2001.tb02350.x. 1–37. http://dx.doi.org/10.1029/2012GC004370. Sächsisches Landesamt für Umwelt und Geologie, 1997. Handbuch zur Heath, R.C., 1983. Basic ground-water hydrology. US Geological Survey Water-Supply Altlastenbehandlung Teil 2: Verdachtsfallerfassung und formale Erstbewertung. Paper 2220. Saidi, S., Bouri, S., Dhia, H. Ben, 2009. Groundwater vulnerability and risk mapping of the Heilongjiang Statistical Bureau, 2011–2016. Heilongjiang Statistical Yearbook. Beijing. hajeb-jelma aquifer (central tunisia) using a gis-based drastic model. Environ. Earth Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high resolution in- Sci. 59:1579–1588. http://dx.doi.org/10.1007/s12665-009-0143-0. terpolated climate surfaces for global land areas. Int. J. Climatol. 25:1965–1978. Shangguan, W., Dai, Y., Liu, B., Zhu, A., Duan, Q., Wu, L., Ji, D., Ye, A., Yuan, H., Zhang, Q., http://dx.doi.org/10.1002/joc.1276. Chen, D., Chen, M., Chu, J., Dou, Y., Guo, J., Li, H., Li, J., Liang, L., Liang, X., Liu, H., Liu, Huan, H., Wang, J., Teng, Y., 2012. Assessment and validation of groundwater vulner- S., Miao, C., Zhang, Y., 2013. A China data set of soil properties for land surface model- ability to nitrate based on a modified DRASTIC model: a case study in Jilin City of ing. J. Adv. Model. Earth Syst. 5:212–224. http://dx.doi.org/10.1002/jame.20026. northeast China. Sci. Total Environ. 440:14–23. http://dx.doi.org/10.1016/j. Shirazi, S.M.M., Imran, H.M.M., Akib, S., 2012. GIS-based DRASTIC method for groundwa- scitotenv.2012.08.037. ter vulnerability assessment: a review. J. Risk Res. 15:991–1011. http://dx.doi.org/10. Inner Mongolia Statistical Bureau, 2011–2015. Inner Mongolia Statistical Yearbook. Beijing. 1080/13669877.2012.686053. Ireson, A., Makropoulos, C., Maksimovic, C., 2006. Water resources modelling under data Shrestha, S., Kafle, R., Pandey, V.P., 2017. Evaluation of index-overlay methods for ground- scarcity: coupling MIKE BASIN and ASM groundwater model. Water Resour. Manag. water vulnerability and risk assessment in Kathmandu Valley, Nepal. Sci. Total Envi- 20:567–590. http://dx.doi.org/10.1007/s11269-006-3085-2. ron. 575:779–790. http://dx.doi.org/10.1016/j.scitotenv.2016.09.141. E. Nixdorf et al. / Science of the Total Environment 605–606 (2017) 598–609 609

Sinkevich, M.G., Walter, M.T., Lembo, A.J., Richards, B.K., Peranginangin, N., Aburime, S.A., Ye, X., You, C., Du, X., Wang, L., 2015. Groundwater vulnerability to contamination near Steenhuis, T.S., 2005. A GIS-based ground water contamination risk assessment tool the river: a case study of the second Songhua River. Sci. Technol. Rev. 33:78–83. for pesticides. Gr. Water Monit. Remediat. 25:82–91. http://dx.doi.org/10.1111/j. http://dx.doi.org/10.3981/j.issn.1000-7857.2015.17.008. 1745-6592.2005.00055.x. Yu, C., Yao, Y., Hayes, G., Zhang, B., Zheng, C., 2010. Quantitative assessment of groundwater Song, E., 2015. An Unofficial Python Client Library for Baidu Maps Web Services APIs. vulnerability using index system and transport simulation, Huangshuihe catchment, GitHub Repos. China. Sci. Total Environ. 408:6108–6116. http://dx.doi.org/10.1016/j.scitotenv.2010. Sophocleous, M., Ma, T., 1998. A decision support model to assess vulnerability to salt 09.002. water intrusion in the great bend prairie aquifer of Kansas. Ground Water 36, 476–483. Yuhong, Y., Baixing, Y., Wanbin, S., 2010. Assessment of point and nonpoint sources pol- Sun, F., Shao, H., Wang, W., Watanabe, N., Bilke, L., Yang, Z., Huang, Z., Kolditz, O., 2012. lution in Songhua River Basin, Northeast China by using revised water quality model. Groundwater deterioration in Nankou—a suburban area of Beijing: data assessment Chin. Geogr. Sci. 20:030–036. http://dx.doi.org/10.1007/s11769-010-0030-3. and remediation scenarios. Environ. Earth Sci. 67:1573–1586. http://dx.doi.org/10. Zaisheng, H., Jayakumar, R., Ke, L., Hao, W., Rui, C., 2008. Review on transboundary aqui- 1007/s12665-012-1600-8. fers in People's Republic of China with case study of Heilongjiang-Amur River Basin. Travis, C.C., Doty, C.B., 1990. Can contaminated aquifers at Superfund sites be remediated? Environ. Geol. 54:1411–1422. http://dx.doi.org/10.1007/s00254-007-0922-4. Environ. Sci. Technol. 24, 1464–1466. Zaporozec, A., Aureli, A., Conrad, J.E., Hirata, R., Johansson, P.-O., Nonner, J.C., Romijn, E., Van der Gun, J., 2012. Groundwater and global change: trends, opportunities and chal- Weaver, J.M.C., 2002. Groundwater Contamination Inventory: a methodological lenges. United Nations World Water Assessment Programme. guide. Unesco Ihp-Vi, Ser. Groundw. 2, 17–. 21 Varnes, D.J., 1984. The principles and practice of landslide hazard zonation. Bull. Int. Assoc. Zehui, L., 2002. China's Natural Resources Database (CNRD). Data Sci. J. 1, 238–246. Eng. Geol. http://dx.doi.org/10.1007/BF02594720. Zhang, B., Song, X., Zhang, Y., Han, D., Tang, C., Yang, L., Wang, Z.L., 2015. The relationship be- Wang, J., He, J., Chen, H., 2012. Assessment of groundwater contamination risk using hazard tween and evolution of surface water and groundwater in Songnen Plain, Northeast quantification, a modified DRASTIC model and groundwater value, Beijing Plain, China. China. Environ. Earth Sci. 73:8333–8343. http://dx.doi.org/10.1007/s12665-014-3995-x. Sci. Total Environ. 432:216–226. http://dx.doi.org/10.1016/j.scitotenv.2012.06.005. Zhang, P., 2008. Revitalizing old industrial base of Northeast China: process, policy and chal- Wang, R., Bian, J.-M.M., Gao, Y., 2014. Research on hydrochemical spatio-temporal charac- lenge. Chin. Geogr. Sci. 18:109–118. http://dx.doi.org/10.1007/s11769-008-0109-2. teristics of groundwater quality of different aquifer systems in Songhua River Basin, Zhang, W.L., Tian, Z.X., Zhang, N., Li, X.Q., 1996. Nitrate pollution of groundwater in northeastern Songnen Plain, Northeast China. Arab. J. Geosci. 7:5081–5092. http://dx.doi. ern China. Agric. Ecosyst. Environ. 59:223–231. http://dx.doi.org/10.1016/0167- org/10.1007/s12517-014-1324-z. 8809(96)01052-3. Winter, T.C., Harvey, J.W., Franke, O.L., Alley, W.M., 1998. Ground water and surface water, A single resource. U.S. Geol. Surv. Circ. 1139, 79. Wu, Y., Wang, W., Toll, M., Alkhoury, W., Sauter, M., Kolditz, O., 2011. Development of a 3D groundwater model based on scarce data: the Wadi Kafrein catchment/Jordan. Environ. Earth Sci. 64:771–785. http://dx.doi.org/10.1007/s12665-010-0898-3.