Understanding spatiotemporal variability of in-stream water quality in urban environments – A case study of Melbourne, Australia

Journal of Environmental Management 246 (2019) 203–213

Contents lists available at ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Research article

Understanding spatiotemporal variability of in-stream water quality in urban environments – A case study of Melbourne, Australia

T

Baiqian Shia,∗, Peter M. Bachb,c,d, Anna Linterna,b, Kefeng Zhange, Rhys A. Colemanf, Leon Metzelingg, David T. McCarthya, Ana Deletice a Environmental and Public Health Microbiology Laboratory (EPHM Lab), Department of Civil Engineering, Monash University, Wellington Road, Clayton, VIC, 3800, Australia b Department of Civil Engineering, Monash University, Wellington Road, Clayton, VIC, 3800, Australia c Swiss Federal Institute of Aquatic Science & Technology (Eawag), Überlandstrasse 133, 8600 Dübendorf, Switzerland d Institute of Environmental Engineering, ETH Zürich, 8093 Zürich, Switzerland e UNSW Water Research Centre, School of Civil and Environmental Engineering, The University of New South Wales, NSW, 2052, Australia f Melbourne Water Corporation, La Trobe Street, Docklands, VIC, 3008, Australia g Environment Protection Authority, Victoria, Macleod, 3085, Victoria, Australia

A R T I C LE I N FO

A B S T R A C T

Keywords: Urban water quality Land uses Imperviousness Urban development Environmental management Pollution sources

To support sustainable urban planning and the design of water pollution mitigation strategies, the spatial and temporal trends of water quality in urban streams needs to be further understood. This study analyses over ten years of surface water quality data from 53 upstream catchments (20 of them predominated by a single type of land use) and two lowland sites across Greater Melbourne, Australia. We evaluated the impact of various catchment characteristics, especially urban land uses, on spatial and temporal urban water quality trends. Here, we focused on common urban pollutants: total suspended solids (TSS), total phosphorous (TP), total nitrogen (TN), zinc (Zn), copper (Cu) and nickel (Ni). Site median nutrient and heavy metal concentrations were negatively correlated with the catchment's elevation and its average annual rainfall. Further analysis shows that such trends were driven by the geographical pattern of Melbourne – i.e. low-laying sites tend to have less rainfall and be more urbanised. Only median concentrations of heavy metals (Zn and Cu) were correlated to catchment imperviousness. Further characterising of the urban environment was done into specific land uses (residential, industrial and commercial), yet median concentrations of all pollutants were not significantly correlated with land uses. This is because simple metrics, such as land use proportions, do not adequately reflect the significant variability in pollution sources that can exist even within the same land use type. Indeed, our temporal analysis found that the water quality difference between catchments with similar land uses is likely caused by their sitespecific pollutant sources (construction and illegal discharge) and environmental management actions (wastewater management actions) regardless of similarities in land use. A 3-stage urbanisation cycle (development, operation and renewal) is suggested to further explain the urban water quality variance, but more data from small areas of an urban catchment is required to directly understand the unique impact of each urbanisation stage on water quality.

1. Introduction Water quality is deteriorating rapidly in many urban waterways around the world, which is impairing aquatic ecosystems and threatening public health (Loucks and Beek, 2017; Landsberg, 2002; Gaffield et al., 2003). Anthropogenic activities including continuous urban expansion, agriculture and inadequate wastewater management, have been identified as primary reasons for water quality degradation in

∗

major rivers (Chang, 2008; Jarvie et al., 1998; Riechel et al., 2016). In response, several countries around the world are progressively implementing more sustainable landscape planning and pollution mitigation strategies including both structural and non-structural tools – e.g. stormwater treatment wetlands and government enforcement (Wong, 2006; Dietz, 2007; Liu, 2016). However, to ensure that management strategies are successful at improving water quality, we must first understand the factors affecting spatial and temporal variability of

Corresponding author. E-mail address: [email protected] (B. Shi).

https://doi.org/10.1016/j.jenvman.2019.06.006 Received 9 February 2019; Received in revised form 28 May 2019; Accepted 2 June 2019 Available online 06 June 2019 0301-4797/ © 2019 Elsevier Ltd. All rights reserved.

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

water quality in stormwater and urban waterways (Chang, 2008; Lintern et al., 2018a). Various catchment characteristics including topography (Lintern et al., 2018b), scale (Blöschl and Sivapalan, 1995), rainfall (Liu et al., 2013; Shinya et al., 2003; Borris et al., 2014), construction activities (Sillanpää and Koivusalo, 2015) and stormwater treatment techniques (Bratieres et al., 2008) could potentially have an effect on receiving water quality. Previous statistical studies of in-stream water quality also focused on the relationships between water quality and land uses and identified that urbanisation is responsible for increased pollution concentrations of sediments, nutrients and heavy metals (Chang, 2008; Pratt and Chang, 2012, Sliva and Dudley Williams, 2001; Ahearn et al., 2005; Li et al., 2008; Bu et al., 2014; Ai et al., 2015). However, these studies did not cover detailed urban land use classifications (e.g., residential, industrial, commercial land uses), and instead considered the urban environment as a whole. Some earlier studies on the quality of stormwater discharges (e.g., Duncan, 1999), summarised stormwater pollution concentrations of five major land use groups (e.g. roads, roofs and high-density urban area). Since this statistical analysis, a few studies have explored the impacts of specific urban land use types on spatial variations of water or sediment quality in receiving waters of urban catchments (Zhao et al., 2015; Sharley et al., 2017). However, these studies used data encompassing a few months to a few years, which would make it challenging to capture the dynamic nature of urban environments and the impacts of urban activities on long-term water quality trends. As such, our understanding of how urban water quality responds to specific urban land use types is currently limited (Bach et al., 2015). Therefore, this study aims to understand the key parameters that describe the spatial and temporal trends of pollutant concentrations in urban streams by using a water quality data set for Greater Melbourne from the 1990s to 2016. Since urban stormwater has been recognised as a major threat to urban streams (Duncan, 1999; Francey et al., 2010; Chow et al., 2013; Lee and Bang, 2000; Ren et al., 2008), the understanding of urban water quality based on Melbourne can also inform the urban water management of other cities or watersheds with separate stormwater and sewer systems. The main objectives of this study are:

Fig. 1. The locations and catchment boundaries of 53 selected upstream monitoring sites, and the locations of two selected lowland sites (background map - Esri (2017b), sampling sites - Melbourne Water (2017c) and waterways Melbourne Water (2017a)).

basis. A total of 23 water quality variables were measured, which can be grouped into seven major categories including water temperature, dissolved oxygen, electrical conductivity, pH, nutrients, faecal contamination and metals. In this study, we focused on Total Suspended Solids (TSS), Total Phosphorus (TP), Total Nitrogen (TN) and three heavy metals (Zn, Cu, Ni) due to their recognised impacts on aquatic ecosystems (e.g. algal blooms and toxicity) and typical association with urban areas across the region (Victorian Stormwater Committee, 1999; SEPP, 1999). Analyses were undertaken by accredited laboratories by using standard methods and procedures: TSS and TN - APWA-AMWA (1989), TP and heavy metals – method 365.4 and 200.8 of US EPA (1983). 2.1.2. Data on site characteristics From the in-stream water quality dataset of Melbourne, we selected 53 upstream sites (labelled as red dots in Fig. 1), which are located in the upland area of each waterway for two reasons: 1) understanding of upstream sites are more relevant to urban landscape planning (Ding et al., 2016) and 2) ensuring that land uses across the catchment is uniform. As an additional analysis, we picked two lowland catchments to demonstrate the impact of environmental management actions (labelled as blue dots in Fig. 1). In addition to the water quality data set, we collected 15 other types of data (summarised in Table 1) on planning, infrastructure, land uses, topography and climate indicators at each catchment to serve as our independent variables for the analysis. Digital spatial maps (in Geographic Information System – GIS format) of catchment boundaries of the 53 selected upstream sites were created based on Melbourne Water's derived drainage catchment boundary and stormwater infrastructure network. For each site, the catchment characteristics (see Table S1) data were then clipped to the derived catchment boundaries using ESRI ArcMap (Esri, 2017a). For the spatial variation of water quality, catchment characteristics are considered as static parameters by averaging data from multiple years. The catchment characteristics were highly variable across the selected sites (see Table S2), which formed a diverse data set for the analysis.

1) Understand what catchment characteristics (topography, climate, urban development and specific land uses) can potentially explain the spatial pollutant variation in urban streams; 2) Determine the significance of site-unique pollution sources (wastewater management, construction and illegal discharges) and environmental management actions on the temporal water quality trend in urban catchments. 2. Methods and materials 2.1. Data collection and preparation Greater Melbourne, the second largest city in Australia and the capital of the state of Victoria, with a population of just over 5 million, was selected as our case study. It incorporates both densely populated regions and also less dense regions that are dominated by agriculture or parks. Over the past decade, rapid urban growth has resulted in the urbanisation of large areas of rural land along the urban fringe as well as major residential, commercial and industrial redevelopment in existing urban areas (DELWP, 2017; Committee of Melbourne, 2010). With broad spatial coverage and long timespan, this data set is ideal for spatial and temporal trend analysis of urban water quality. 2.1.1. Water quality data collection Water quality sampling was mostly conducted at a monthly frequency from 1998 onwards with the collection of grab samples at a depth of 0.1 m from the surface during both dry and wet weather. Before 1998, water quality was monitored on a weekly or bi-weekly

2.1.2.1. Topography and climate. Catchment size, length of waterway reaches, and elevation data (see Fig. 2(a)) were collected from online sources (NEDF, 2010; Melbourne Water, 2017a) and used as static parameters in the spatial trend analysis. For the climatic data, gridded 204

Journal of Environmental Management 246 (2019) 203–213

annual rainfall depth (5 × 5 km) based on data between 2007 and 2016 (see Fig. 2(b)) was utilised for each site and considered in the spatial analysis. For the temporal trend analyses, we obtained daily rainfall data over the water quality monitoring period from the nearest rain gauges and antecedent dry weather period (ADWP).

Land use percentage for each available year

2.1.2.2. Development and infrastructure. Imperviousness has been considered a key indicator to assess the impacts of urbanisation on waterways (Brabec et al., 2002). For the spatial variation analysis, the imperviousness map generated by Grace Detailed-Gis Services (2012) was used to represent the intensity of the urban environment of all selected sampling sites. For the temporal analysis of selected catchments, satellite images ranging from 2009 to 2017 were captured and compared to understand land activities during that time. The population density (see Fig. 2(c)) was also used as an indirect indicator of urbanisation, as was the proportion of road to verify the influence of traffic-related activities on water quality. To test the potential of other indicators on effectively describing the impacts of urbanisation on water quality, drainage and sewerage infrastructure maps were obtained (Melbourne Water, 2017a). The sewerage network and the number of septics of exurban catchments were also considered to understand the impact of cross-connections and onsite wastewater management. Finally, age of development (Kuller et al., 2018) was incorporated into our study, given the likelihood that catchments developed in different periods would exhibit unique characteristics, such as roof material, construction activities and crossconnection rates. 2.1.2.3. Land uses. To refine the urban environment in this analysis, we adopted the land use classification of Victorian Land Use Information System - Vluis (2017) which has nine primary categories including residential, commercial, industrial, extractive industries, agricultural, infrastructure and utilities, community services, sports heritage, and forest reserves. This dataset (see Fig. 2(d)) collected information at the land parcel level, which is the finest resolution of land use currently accessible across the study area. The dataset was cleaned by removing duplicate polygons and calculated the percentage of each land use for the 53 upstream sites. As the dataset was revised every two years, the same data processing method was repeated for the datasets published in 2006, 2008, 2010, 2012 and 2014. For spatial water quality trend analyses, average land use proportions were calculated based on all years of data to represent the average condition during the monitoring period. For temporal trend analyses, the land use changes between 2006 and 2014 were used to interpret the observed trends in water quality. 2.2. Data analysis

Land use [%]

For the spatial water quality variability analyses, ten years of data between 2007 and 2016 were used to achieve an equal sampling period across monitoring sites. For the temporal variability analysis, the full length of the dataset for each of the selected monitoring sites was also used to capture the longer-term water quality trend. 2.2.1. Correlation analysis to detect spatial trends For a general understanding of how surface water quality differs spatially, we used the median pollutant concentration to represent the overall water quality status over the monitoring period. The Spearman Rank Correlation coefficient (Well and Myers, 2003) between median pollutant concentrations and temporally averaged catchment characteristics (See Table S2) was calculated to identify which independent variables could explain the observed spatial variance.

Land use

WSUD Technologies [No.]

Septic tanks [No.]

Sewerage system [km]

2

Population density [pp/km ] Age of development [years]

Number of permanent residents from Census 2016 Time since the first development in a selected area

Averaged land use percentage based on data from 2006 to 2014; predominant land use catchments are selected based on the result

ABS (2017) Unpublished data from Melbourne Museum Vluis (2017)

Kuller et al. (2018)

The time that sewer backlog program connects properties to the trunk network Year of Water Sensitive Urban Design techniques was constructed by governments Number of failing septic and offsite discharges

Unpublished data from Melbourne Water Manningham City Council (2011)

Vluis (2017) Melbourne Water (2017a) Road [%] Drainage Infrastructure [km]

Imperviousness [%]

Climate

Development & infrastructure

Percentage of the catchment that is covered by an impervious surface based on 2004 aerial imagery Major and minor road layout of Melbourne region Total drainage distance, including natural waterway, channels and underground pipes Total sewerage network distance

Daily rainfall from closest rain gauges Number of days without rain events before the sample collection day Satellite imagery of selected catchments for the sampling period (NEARMAP, 2017)

Melbourne Water (2017b) NEDF (2010) Melbourne Water (2017a) NEDF (2010) Bom (2017) Calculated based on daily rainfall data (Grace Detailed-Gis Services, 2012) Calculated size based on defined catchment boundary DEM-9s data, approx. 250m grid The length of the main drainage network through the catchment The average slope of the catchment surface Gridded annual rainfall between 2007 and 2016 Catchment size [km2] Elevation [m] Catchment length [m] Catchment slope [%] Rainfall depth [mm] ADWP (days) Topography

Spatial trend analysis Independent Variables Categorises

Table 1 Independent variables selected for the statistical analysis of urban water quality.

Temporal trend analysis

Source

B. Shi, et al.

2.2.2. Water quality variation between different predominant land-use catchments The water quality monitoring sites were narrowed down to 53 205

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

Fig. 2. Catchment characteristics data: (a) elevation, (b) annual rainfall, (c) population density and (d) land uses.

Yue et al., 2002; Hirsch et al., 1982; Chang, 2008). Results were compared between sampling stations to understand why water quality may have improved or deteriorated at the l upstream sites. To further understand the long-term water quality variations, detailed time series analyses were carried out for three upstream catchments: Cherry Creek (55% of industrial and 40% imperviousness), Anderson Creek (69% low density residential and 17% imperviousness) and Brushy Creek (53% low density residential and 25% imperviousness). These three case study catchments were selected because they showed significant historical water quality variations, and could shed light on the possible factors governing temporal trends in other catchments across the region. For these three representative catchments, both temporal changes in land use and the implementation of environmental management actions (e.g. land use, construction of Water Sensitive Urban Design facilities, wastewater management, termination of cross-connections) were analysed to clarify whether these factors could describe the long-term water quality tendency of each catchment. A Spearman Rank correlation analysis was also conducted based on the entire sampling period to understand how pollutants correlate with each other, thereby indicating the potential causes of the temporal trends. We also analysed the historical water quality data of two lowland sampling sites to understand the effects of decommissioning aging wastewater treatment plants within the monitoring period. Box plots were used to compare the water quality before and after the treatment plant's decommissioning, and subsequent connection to a centralised wastewater treatment facility.

upstream catchments of varying land uses to a sub-sample of 20 catchments, where a single land use was predominant (i.e. > 50% for industrial and > 65% for other land uses) to investigate 1) how urban water quality relates to different land use types, and 2) the variance of water quality within the same land use type. Instead of only using the median pollution concentration, the distribution of pollutant concentrations for each catchment was compared by using box plots to cover the data range over the monitoring period. Within these, two catchments were classified as industrial catchments, which have 55% and 26% of industrial land use cover respectively. The site with the second highest industrial land use was also included, to avoid only having one single industrial catchment. From the 12 residential sites, six catchments were defined as low-density residential with imperviousness lower than 30%, and another six catchments with higher impervious coverage are grouped as the median density residential. There are also two agricultural catchments (> 70%) and four forested sites (> 90%). We were unable to identify any commercially predominant catchments to sub-sample as these activities are typically surrounded by large residential regions or located in highdensity mixed-use developments. 2.2.3. Temporal trends of surface water quality The non-parametric method, Mann-Kendall's test (Mann, 1945; Kendall, 1975) was used to assess if there is a monotonic trend (i.e. increase or decrease through time but the trend may not be linear) for each pollutant. The normalised test statistics were calculated, a negative value indicating a decreasing trend and a positive value representing an upward trend (α = 0.05). This method has been widely used in previous studies to analyse water quality trends (Yu et al., 1993; 206

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

Fig. 3. Spearman rank correlation analysis results between catchments' median pollution concentration and catchment characteristics including development & infrastructure, land uses, topography and climate (detailed results are shown in Table S3).

3. Results and discussion

3.1.2. Influence of urban development Catchment imperviousness, as a representation of urban development, showed a weak correlation with TSS concentrations. Although Duncan (1999) reported that urbanised basins contribute a higher level of sediments than rural land uses, urban catchments with higher imperviousness in this study tend to have similar or even lower TSS levels than most rural catchments (Fig. S1). In addition, TP and TN, demonstrated a weak positive correlation with imperviousness (ρ = 0.31 and 0.27, p < 0.05, respectively). Previous studies such as Hatt et al. (2004) and Chang (2008) reported the significance of urbanisation rates on determining the level of TN and TP, because of the high amount of nutrients from fertilisers and animal faeces in surface runoff. However, they also indicated that inadequate sewage management, such as septic tanks, cause even higher concentrations of nutrients in exurban catchments with low imperviousness. In this sense, the weak correlation between nutrients and imperviousness might be significantly affected by wastewater-related pollution sources that could not be directly represented by catchment’ imperviousness. A detailed comparison between urban catchments is provided in Section 3.1.3. Our study also attempted to include the coverage of sewerage and stormwater system in the correlation analysis to test the impact of water infrastructure on spatial variability, but these characteristics all showed similar results to imperviousness (see Fig. 3) due to their strong cross-correlation, which demonstrates that they cannot explain the spatial water quality trend better than imperviousness. Unlike sediments and nutrients, Zn and Cu strongly correlated with catchment imperviousness (ρ = 0.81, p < 0.001 and 0.80, p < 0.001 in Fig. 3). Such a relationship was expected as major urban heavy metal sources, such as brake pad wear, tire wear (Van Bohemen and Van De Laak, 2003), roof materials (Robert-Sainte et al., 2009) and motor oils (Ball et al., 1998; Davis et al., 2001), are widely spread over urban impervious surfaces (Petrucci et al., 2014). The results suggest that imperviousness is a strong indicator for the spatial variation of Zn and Cu. All other urban developments and infrastructure characteristics showed a similar correlation with Zn and Cu as imperviousness. The correlation between Ni and urban development is not as significant as Zn and Cu, because geology characteristics (i.e. basalt or sedimentary) are more predominant for describing the spatial trend of Ni in Melbourne (Sharley et al., 2017).

3.1. Spatial variance of water quality Pollution concentrations in urban streams varied greatly across the study region (see Fig. S1). Highly urbanised areas (i.e. inner suburbs) tended to have higher levels of Zn and Cu than forested and agricultural sites. The concentrations of Ni from some agricultural and industrial sites were two to three times higher than that of residential catchments of inner Melbourne. TSS levels had an irregular spatial pattern across the entire study region due to substantial data variability between catchments with similar land uses. While TP and TN in the inner urbanised areas were high, some exurban catchments with both residential and agricultural land uses discharged even higher levels of nutrients to receiving waters. 3.1.1. Impact of topography and climate conditions In our study, median TSS concentrations were positively correlated with both catchment size and length (ρ = 0.38, p < 0.01 and 0.33, p < 0.05 in Table S3 and Fig. 3). The correlation analysis indicated that larger catchments tend to have less impervious coverage (ρ = −0.56, p < 0.01) but higher agricultural and forest proportions, which suggest that the difference of median TSS between catchments in this case study could be driven by the urbanisation rather than catchment size and length. Other urban stormwater quality characterisation studies, such as Francey et al. (2010) and Lee and Bang (2000), also indicated that the catchment size itself could not be considered as a deterministic factor to explain the spatial variance of sediments and nutrients in urban water. Results from our study showed that all pollutants tend to negatively correlate with topographic characteristics including mean catchment slope and elevation (Fig. 3). However, previous studies found catchment slope positively correlated with the level of sediments and heavy metals. This is likely due to the higher rate of erosion from steeper land (Pratt and Chang, 2012; Sliva and Dudley Williams, 2001). These divergent findings may be the result of unique geographical patterns. Our study and other similar studies that reported a negative correlation, such as Chang (2008), Brett et al. (2005) and Lintern et al. (2018b), showed a cross-correlation between land use and catchment elevation/ slope – i.e. mountainous areas are mostly forested with low pollutant concentrations and low-lying areas tend to be agricultural and developed with poorer water quality. The negative correlation between averaged rainfall and all median pollutant concentrations could also be caused by the fact that lower annual rainfalls occur in the urbanised/ highly-impervious areas of Melbourne (ρ = −0.66, p < 0.001). In this sense, catchment averaged annual rainfall and elevation may not be considered to describe the spatial variation of pollution concentrations in urban streams.

3.1.3. Impact of land use types Fig. 3 shows that Zn and Cu are positively correlated with the urban environment including residential, commercial and industrial land uses but the correlations were weaker than imperviousness (ρ = 0.48–0.66 and ρ = 0.56–0.64, p < 0.001). The results suggest that detailed urban land use classifications do not better explain the spatial variation of these two metals. Compared with imperviousness, Ni had a better correlation with industrial coverage (ρ = 0.50, p < 0.001), which 207

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

Fig. 4. Comparison of 20 predominant land use catchments' pollution levels – (a) TSS, (b) TP, (c) TN, (d) Zn, (e) Cu and (f) Ni; residential catchments are arranged in order of imperviousness from left to right in all plots.

high nutrient concentrations of the three low-density residential catchments. Previous studies found that failing septic systems can become chronic sources of nutrients and interact with nearby waterways (Hoxley and Dudding, 1994; Withers et al., 2014). According to Manningham City Council (2011), where the two streams are located, a large proportion of split greywater septic systems also exist, and untreated greywater is permitted to be directly discharged into the creek. This may be having a large impact on nutrient level, due to higher connectivity of highly polluted water. Elevated TN in Brushy Creek was also likely affected by the local wastewater treatment plant, which releases treated sewage to the creek. When wastewater-related point sources appear in a certain catchment, nutrient levels could be even greater than catchments with higher impervious coverage and land use proportions, thereby affecting the significance of these characteristics in describing the spatial trend of water quality (Zhou et al., 2016). Surprisingly, Cherry Creek, an industrial catchment, had the highest TP level of all the predominant land use sites studied. Common TP sources (e.g. fertiliser, animal faeces and organic matter) in urban catchments cannot explain the levels observed at Cherry Creek, as TP concentrations were three times higher than other agricultural and residential catchments. The specific sources of TP are unknown but possibly relate to cross-connections with sewage and industrial discharges. Unlike construction activities which are periodic pollution issues during land development, illegal discharges and cross-connections may randomly occur in an urban environment and these pollutant sources are more difficult to control. Both catchment imperviousness

suggests that Ni is likely to be related to specific industrial activities more than urban impervious surfaces. Sharley et al. (2017) also reported the significant influence of industrial land use on the level of Ni in Melbourne. The spatial variation of TP and TN was not better explained by residential and commercial land uses (less significant correlation compared with imperviousness in Fig. 3). The poor correlation between urban land uses and site median pollutant concentrations indicate that specific time-averaged land use classifications are not as significant as imperviousness for characterising water quality in urban streams. The significant water quality variation within different predominant land use groups could partially explain the weak correlation observed in the Spearman Rank Correlation analysis. Fig. 4(a) showed that industrial, residential and agricultural catchments tend to have similar TSS median concentrations. There was some variance between catchments within the same land use type (e.g. two low-density residential catchments demonstrated higher TSS concentrations than their counterparts). Such variability is likely caused by differences in land management methods and the occurrence of certain activities within each catchment (e.g., type and amount of construction/renewal activities). Goonetilleke et al. (2005), in their study on urban stormwater quality, also showed two similar residential catchments with the same impervious proportion had highly different TSS mean concentrations. TP and TN levels within low-density residential catchments also varied (Fig. 4(b) and (c)). The existence of septic tank systems in catchments such as Anderson Creek and Brushy Creek could explain the 208

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

and land uses are insufficient to describe the appearance of these point sources in our catchments. Unlike sediments and nutrients, (d) and 4(e) show that Zn and Cu tend to rise with increasing impervious cover, and this may also explain why Cherry Creek had higher Zn and Cu than the other industrial site. This further emphasises the significance of imperviousness in describing the spatial variation of Zn and Cu in urban streams. As addressed in the Spearman Correlation analysis, the levels of Ni (see Fig. 4(f)) within predominantly industrial catchments are generally higher than other land use types. One medium-density residential site also showed wide Ni data ranges, but the reason for this is unknown based on the catchment characteristics considered in this study.

Table 2 Spearman rank correlation coefficients testing the temporal relationship between pollution parameters of Cherry Creek (p < 0.001 for all results). ρ

Turb

TSS

Cu

Ni

Zn

Turb TSS Cu Ni Zn

1.00 0.83 0.75 0.54 0.68

1.00 0.62 0.53 0.49

1.00 0.49 0.82

1.00 0.41

1.00

and because Cherry Creek (Fig. 5) was the only one that exhibited an increasing trend in both Zn and Cu (z = 4.06, p < 0.001 and 4.58, p < 0.001 in Table S4), it was selected as a case study to understand the causes of increasing heavy metals at this particular site. TSS at Cherry Creek has also slightly increased since 2010 (z = 1.95 p = 0.05). Heavy metals were significantly correlated with sediments in Cherry Creek (see Table 2). The strong correlation between sediments and heavy metals suggests that they may be derived from the same sources or transport pathways. By comparing the historical land use proportions (see Fig. S3), industrial land use increased from 51% in 2008 to 58% in 2010. This significant land use change would have occurred alongside intensive construction activities (the box shaded area and impervious maps in Fig. 5), which could potentially explain the increased TSS and heavy metals. In residential catchments, TSS in Brushy Creek significantly increased (z = 8.34, p < 0.001) since 2009 and remained at a high level at the end of the monitoring period (see Fig. S4). Residential land use increased from 48% to 62% from 2010 to 2012. Satellite images also suggest that intensive residential constructions occurred within just 300 m from the creek (Fig. S4). The water quality trends and the development trends at Cherry Creek and Brushy Creek, suggest that construction and land use change are factors that may lead to increasing TSS concentrations in urban streams. Atasoy et al. (2006) also

3.2. Explaining the temporal variability of water quality The study catchments exhibited differing temporal trends in water quality over time (see Fig. S2 and Table S4). TSS concentrations decreased in 28% of sites which are mainly located in inner urban areas. TP and TN levels decreased in 36% and 32% of all catchments respectively and increased in only 6 sites. Anderson Creek had the strongest downward trend in TP and TN (z = −13.42, p < 0.001 and −7.55, p < 0.001). Also, Cu and Zn show downward trends in 72% and 57% of stations respectively including all types of land uses and land covers, which suggests that discharge of Cu and Zn has been decreasing during the sampling period. Only a few sampling sites demonstrated increasing trends in Cu and Zn. Cherry Creek was selected as a case study because both Cu and Zn increased. For the temporal trend of Ni, 58% of stations showed downward trends, while 42% of sites have no significant temporal trend. Both daily rainfall and ADWPs had no strong correlation with pollution concentrations, which cannot describe the historical trend of water quality parameters. 3.2.1. Impact of construction activities Only a limited number of sites show increasing heavy metals levels,

Fig. 5. Time-series plots of the water quality of Cherry Creek from 2007 to 2016 including TP, TSS and Zn (left); satellite imagery of the catchment at 2016 – red and yellow shaded areas represent industrial and residential land uses respectively (right); box shaded area had continuous construction activities since 2010 - Nearmap satellite imagery (bottom). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

209

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

through laundry, kitchen and bathroom, managing the greywater discharge by continuous septic maintenance and replacing split greywater systems with new septic systems could also contribute to long-term TP reduction. Recently, countries around the world including the European Union, the United States and Australia have banned the use of phosphates in household cleaning products which could explain the TP reduction in recent years if a catchment has industrial discharges to the drainage network. For example, Cherry Creek's TP concentration slightly decreased since 2013, with almost no concentrations above 1 mg/L (see Fig. 5). This timing aligns with the 2011 ban on phosphate in detergents and an all-out ban of phosphates in 2014 in Australia. The effluents from wastewater treatment facilities can degrade water quality in urban streams as shown in the results of previous studies (Gibson and Meyer, 2007; Andersen et al., 2004), but their impact is being increasingly managed by introducing advanced treatment techniques or terminating discharges altogether (Carey and Migliaccio, 2009). The two lowland sampling sites on Dandenong Creek and Maribyrnong River, for instance, had wastewater treatment plants that discharged treated wastewater from upstream until the plants were decommissioned during 1994 and 1998 respectively, and their wastewater was redirected to other treatment facilities. After decommissioning these plants, the concentrations of TP and TN substantially declined to a level equivalent to other urban/suburban catchments, i.e. TP < 0.25 mg/L and TN < 2 mg/L (see Fig. 7 and Fig. S6). Therefore, the closing down of these treatment plants, as an environmental management action was the main driver of observed decreases in TP and TN. Similar to construction activities, environmental management actions such as decommissioning wastewater treatment plant would also affect the spatial analysis of water quality trends.

suggested that new construction can affect TSS levels by tracing residential development and change in land use to different water quality measures in North Carolina, USA. 3.2.2. Impact of wastewater-related pollutant sources As discussed in the land use analysis (see Section 3.1.3), Cherry Creek had the highest level of TP in all predominant land use catchments. By splitting the dataset into dry and wet weather samples, we found that dry weather TP concentrations were more variable and had higher median concentrations than wet weather concentrations (see Fig. S5). This contrasts with many urban stormwater quality studies (Lee and Bang, 2000) and even other catchments in this study. Rather than surface runoff, the high levels of TP in Cherry Creek before 2012 is likely to be caused by industrial discharges, since wastewater could contribute up to more than 1000 mg/L of phosphorous from certain types of industries (Noukeu et al., 2016; Britz et al., 2004). Similarly, Anderson Creek was affected by sewage sources such as failing septics and direct greywater discharges (see Section 3.1.3). Previous studies reported that blackwater and greywater contained around 105 mg/L and 10 mg/L of phosphorous, respectively (Eriksson et al., 2002; Muserere et al., 2014). A similar issue has been reported by Bach et al. (2010) for another Melbourne catchment with on-site septic systems. 3.2.3. Influence of environmental management actions TP levels in Anderson Creek have decreased significantly since 1995 (z = −13.42, p < 0.001in Table S4 and see time series plot in Fig. 6). Over this period, only 12 small rain gardens were constructed which only treated 0.2% of the catchment area, and the whole catchment remained low-density residential without significant change in land use (see Fig. S3). This suggests that surface runoff management was not the main contributor to water quality improvements. According to Manningham City Council (2011), the sewer backlog program, as an infrastructural renewal, led to the removal of many direct greywater discharges and fixing of leakages from failing septic tanks by bringing the number of septics from 1486 in 2007 down to 791 in 2016, which potentially explained the decreasing TP. Although over the long term (1994–2016), TN had a decreasing trend, TN levels remained largely stable between 2000 and 2014. This suggests that the causes of TP reduction were not very effective on TN sources. Tjandraatmadja et al. (2010) analysed the nutrient level of household products in wastewater and concluded that their phosphorous contribution could be up to 180% of the anthropogenic load, while their TN levels were estimated to be 7% of the anthropogenic load. Since household products such as toilet cleaner, liquid detergent and shampoo are mainly discharged

3.3. Urbanisation cycle – driving the spatiotemporal variability of urban water quality The statistical analysis of this study suggests that time-averaged land use itself is not sufficient to explain the spatial trend of urban stream water quality. One possible reason is that site-specific pollution sources within the same land use type (e.g. constructions and industrial discharges) and environmental management actions (e.g. change of wastewater management) can affect the pollutant levels, thereby impacting the relationship between median pollutant concentrations and time-averaged catchment characteristics. Case study catchments including Cherry Creek, Anderson Creek and Brushy Creek are some examples that demonstrate the influence of site-specific pollutant sources Fig. 6. Time-series plots of the surface water quality of Anderson Creek from 1994 to 2016 for TP and TN (left); satellite imagery of the catchment at 2016 (right) – yellow shaded areas represent the area covered by residential land use, and red dots represent the on-site septic tanks and direct greywater discharge. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

210

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

Fig. 7. TP and TN levels of Dandenong Creek Site and Maribyrnong River Site before and after the wastewater treatment plant was decommissioned.

stormwater network when the sewer has exceeded a flow threshold) and groundwater leaking into drainage pipes. Previous studies have reported that dry weather flows from drainage pipes could contribute up to 80% of the total annual nutrient load in some separate-sewer catchments (Deffontis et al., 2013; Meng et al., 2009). Ellis and Butler (2015) suggest that urban catchments could have a cross-connection rate ranging from 1% to 30%. These point wastewater-related sources stochastically occurred during construction and on-going infill development stages and caused water quality degradation during the long operation phase. During the urban renewal phase, some countries are trying to introduce an array of policies and guidelines to promote sustainable development and pollution mitigation techniques (e.g. Epa Victoria, 1996; US EPA, 2005). For example, low impact development and naturebased solutions for stormwater treatment such as wetlands, biofilters, swales and pervious pavements have been adopted for more than a decade and are also increasingly being considered in the initial development stages (Martin et al., 2000; Dietz, 2007; Urbonas and Stahre, 1993). Other environmental management actions could also be conducted to target subsurface pollutant sources that occur during site operation. In this study, sewering of residential areas that previously relied on septic systems and decommissioning or upgrading existing waste management facilities such as wastewater treatment plants, proved to be effective actions. While research has been done to understand the performance of structural and non-structural management measures (Davis, 2007; Carleton et al., 2000), few of them have studied the influence of management activities on urban streams at the catchment scale (Walsh et al., 2015).

and management actions that influence pollutant concentrations in urban waterways thereby disrupting the simple link between water quality and land-use. To generalise, at an urban parcel scale, the urbanisation processes can be represented as a loop of three main phases: development, operation and renewal (Whitehand, 1994; Barras, 1987). Because of the continuous and dynamic nature of urban environments, a variety of specific pollutant sources and management activities are uniquely associated with each urbanisation phase, which forms the framework illustrated in Fig. 8. For example, at the development stage, construction activities produce higher concentrations of sediments, heavy metals and litter to the drainage system, similar to that found in Cherry Creek and Brushy Creek in this study and previous studies such as Sillanpää and Koivusalo (2015), but the impact is transient and usually reduced once the site is in the post-development phase. During the operation stage, surface runoff from established surfaces including roads, roofs and parking lots has been considered as the major source of pollutants in urban streams (Bannerman et al., 1993; Vaze and Chiew, 2002; Egodawatta et al., 2012). For separate-sewer catchments, Zhang et al. (2019) outlined various non-conventional pollutant sources could occur on urban surface or underground as well. In our study, illegal discharges (e.g. car washing, surface dumping and cooling water) and on-site wastewater discharge are found to be critical urban water quality threats. Other pollution sources have also been noted but less studied, for example, chemical spills from industrial land, discharge from emergency accidents such as fire-fighting water (Victorian Stormwater Committee, 1999)cross-connections, emergency sewer overflows (i.e. engineered discharges from the sewer network into the

Fig. 8. Urbanisation cycle: development, operation and renewal – with associated water pollution sources and management strategies during each stage. 211

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

Appendix A. Supplementary data

Due to the limitations of the Melbourne urban stream water quality data set used in this paper, it was difficult to trace the change of pollutant concentrations to each development stage. A catchment could be under multiple development stages within different sub-catchment areas. Furthermore, the impact of different development stages of small land parcels may not be significant enough to be detected by analysing the water quality at the catchment outlet. The negative impact of crossconnections is an example of wastewater-related pollutant sources during operation stage that cannot be directly detected and understood at the catchment scale. As most of these non-conventional pollution sources are not driven by rainfall but highly relate to human behaviours, detecting urban dry weather flows and linking their flow rate and pollution concentrations to different land uses and development stages could be helpful to explain the spatiotemporal water quality trend at catchment scales. However, currently, the high cost of setting up urban water monitoring sites including the use of flow meters and autosamplers hinders researchers gathering highly spatially-distributed data within a catchment. Therefore, the development of a cheap sensor network will be crucial to the future research of water quality prediction and trend analysis.

Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jenvman.2019.06.006. References ABS, 2017. In: STATISTICS, A.B.O. (Ed.), Data Packs of 2016 Census. Ahearn, D.S., Sheibley, R.W., Dahlgren, R.A., Anderson, M., Johnson, J., Tate, K.W., 2005. Land use and land cover influence on water quality in the last free-flowing river draining the western Sierra Nevada, California. J. Hydrol. 313, 234–247. Ai, L., Shi, Z., Yin, W., Huang, X., 2015. Spatial and seasonal patterns in stream water contamination across mountainous watersheds: linkage with landscape characteristics. J. Hydrol. 523, 398–408. Andersen, C.B., Lewis, G.P., Sargent, K.A., 2004. Influence of wastewater-treatment effluent on concentrations and fluxes of solutes in the Bush River, South Carolina, during extreme drought conditions. Environ. Geosci. 11, 28–41. APWA-AMWA, 1989. Standard Methods for the Examination of Water and Wastewater. American public health association. Atasoy, M., Palmquist, R.B., Phaneuf, D.J., 2006. Estimating the effects of urban residential development on water quality using microdata. J. Environ. Manag. 79, 399–408. Bach, P.M., Mccarthy, D.T., Deletic, A., 2010. Redefining the stormwater first flush phenomenon. Water Res. 44, 2487–2498. Bach, P.M., Staalesen, S., Mccarthy, D.T., Deletic, A., 2015. Revisiting land use classification and spatial aggregation for modelling integrated urban water systems. Landsc. Urban Plan. 143, 43–55. Ball, J., Jenks, R., Aubourg, D., 1998. An assessment of the availability of pollutant constituents on road surfaces. Sci. Total Environ. 209, 243–254. Bannerman, R.T., Owens, D.W., Dodds, R.B., Homewer, N.J., 1993. Sources of pollutants in Wisconsin stormwater. Water Sci. Technol. 28, 241–259. Barras, R., 1987. Technical change and the urban development cycle. Urban Stud. 24, 5–30. Blöschl, G., Sivapalan, M., 1995. Scale issues in hydrological modelling: a review. Hydrol. Process. 9, 251–290. Bom, 2017. Climate Data Online. [Online]. Available: http://www.bom.gov.au/climate/ data/, Accessed date: 22 November 2017. Borris, M., Viklander, M., Gustafsson, A.-M., Marsalek, J., 2014. Modelling the effects of changes in rainfall event characteristics on TSS loads in urban runoff. Hydrol. Process. 28, 1787–1796. Brabec, E., Schulte, S., Richards, P.L., 2002. Impervious surfaces and water quality: a review of current literature and its implications for watershed planning. J. Plan. Lit. 16, 499–514. Bratieres, K., Fletcher, T.D., Deletic, A., Zinger, Y., 2008. Nutrient and sediment removal by stormwater biofilters: a large-scale design optimisation study. Water Res. 42, 3930–3940. Brett, M.T., Arhonditsis, G.B., Mueller, S.E., Hartley, D.M., Frodge, J.D., Funke, D.E., 2005. Non-point-source impacts on stream nutrient concentrations along a forest to urban gradient. Environ. Manag. 35, 330–342. Britz, T.J., Schalkwyk, C.V., Hung, Y.-T., 2004. Treatment of Dairy Processing Wastewaters. Handbook of Industrial and Hazardous Wastes Treatment. Marcel Dekker. Bu, H., Meng, W., Zhang, Y., Wan, J., 2014. Relationships between land use patterns and water quality in the Taizi River basin, China. Ecol. Indicat. 41, 187–197. Carey, R.O., Migliaccio, K.W., 2009. Contribution of wastewater treatment plant effluents to nutrient dynamics in aquatic systems: a review. Environ. Manag. 44, 205–217. Carleton, J.N., Grizzard, T.J., Godrej, A.N., Post, H.E., Lampe, L., Kenel, P.P., 2000. Performance of a constructed wetlands in treating urban stormwater runoff. Water Environ. Res. 72, 295–304. Chang, H., 2008. Spatial analysis of water quality trends in the Han River basin, South Korea. Water Res. 42, 3285–3304. Chow, M.F., Yusop, Z., Shirazi, S.M., 2013. Storm runoff quality and pollutant loading from commercial, residential, and industrial catchments in the tropic. Environ. Monit. Assess. 185, 8321–8331. Committee of Melbourne, 2010. Melbourne beyond 5 Million. Getting Better As We Get Bigger, Melbourne. Davis, A.P., 2007. Field performance of bioretention: water quality. Environ. Eng. Sci. 24, 1048–1064. Davis, A.P., Shokouhian, M., Ni, S., 2001. Loading estimates of lead, copper, cadmium, and zinc in urban runoff from specific sources. Chemosphere 44, 997–1009. Deffontis, S., Breton, A., Vialle, C., Montrejaud-Vignoles, M., Vignoles, C., Sablayrolles, C., 2013. Impact of dry weather discharges on annual pollution from a separate storm sewer in Toulouse, France. Sci. Total Environ. 452–453, 394–403. DELWP, 2017. Metropolitan Melbourne Redevelopment 2017. Urban Development Program. Land Use and Population Research, DELWP, Melbourne. Dietz, M.E., 2007. Low impact development practices: a review of current research and recommendations for future directions. Water Air Soil Pollut. 186, 351–363. Ding, J., Jiang, Y., Liu, Q., Hou, Z., Liao, J., Fu, L., Peng, Q., 2016. Influences of the land use pattern on water quality in low-order streams of the Dongjiang River basin, China: a multi-scale analysis. Sci. Total Environ. 551–552, 205–216. Duncan, H.P., 1999. Urban Stormwater Quality: A Statistical Overview. Cooperative Research Centre for Catchment Hydrology, Melbourne. Egodawatta, P., Miguntanna, N.S., Goonetilleke, A., 2012. Impact of roof surface runoff on urban water quality. Water Sci. Technol. 66, 1527–1533.

4. Conclusions This study analysed over ten years of water quality data from 55 stations of Melbourne, Australia to enhance our understanding of the spatiotemporal variability of urban water quality from separate-sewer catchments. Various catchment characteristics including imperviousness, specific land uses, rainfall, elevation and catchment size were considered in the spatial analysis. Dynamic characteristics such as a change in land use, point pollutant sources and environmental management actions were addressed in the temporal trend analysis. The key findings include:

• The spatial variation of median Zn and Cu concentrations can be described by catchment imperviousness. • Median concentrations of all pollutants did not strongly correlate • •

with land use types (e.g. residential and industrial), since catchments under same land use type even have highly variable pollutant sources (e.g. constructions, wastewater discharges and industrial spills); The site-specific pollutant sources are continuously changing due to environmental management actions (e.g. sewer upgrade and banning phosphates in detergents), which significantly improved urban water quality at some urban catchments. A 3-stage urbanisation cycle (i.e. development, operation and renewal) is suggested by linking specific pollutant sources and management actions to each development phase to study spatial and temporal water quality variation together. High spatial resolution water quality data that isolates the relative impacts of land use changes and environmental management actions is critical to understand water quality threats and management opportunities in urban streams.

Acknowledgement This project is funded by the Australian Research Council (ARC), Linkage Project LP160100241, titled “Advancing water pollution emissions modelling in cities of the future”, EPA Victoria, Australia, Melbourne Water, Australia and Know City Council, Australia. We also appreciate the help and support from Caroline Carvalho, Daniella Gerente and Johnathan Wright of Knox City Council, Trish Grant of Melbourne Water, Alice Phung of EPA Victoria and Andrew Allan of Manningham City Council.

212

Journal of Environmental Management 246 (2019) 203–213

B. Shi, et al.

Noukeu, N.A., Gouado, I., Priso, R.J., Ndongo, D., Taffouo, V.D., Dibong, S.D., Ekodeck, G.E., 2016. Characterization of effluent from food processing industries and stillage treatment trial with Eichhornia crassipes (Mart.) and Panicum maximum (Jacq.). Water Resour. Ind. 16, 1–18. Petrucci, G., Gromaire, M.C., Shorshani, M.F., Chebbo, G., 2014. Nonpoint source pollution of urban stormwater runoff: a methodology for source analysis. Environ. Sci. Pollut. Control Ser. 21, 10225–10242. Pratt, B., Chang, H., 2012. Effects of land cover, topography, and built structure on seasonal water quality at multiple spatial scales. J. Hazard Mater. 209, 48–58. Ren, Y., Wang, X., Ouyang, Z., Zheng, H., Duan, X., Miao, H., 2008. Stormwater runoff quality from different surfaces in an urban catchment in Beijing, China. Water Environ. Res. 80, 719–724. Riechel, M., Matzinger, A., Pawlowsky-Reusing, E., Sonnenberg, H., Uldack, M., Heinzmann, B., Caradot, N., Von Seggern, D., Rouault, P., 2016. Impacts of combined sewer overflows on a large urban river – understanding the effect of different management strategies. Water Res. 105, 264–273. Robert-Sainte, P., Gromaire, M.-C., De Gouvello, B., SAAD, M., CHEBBO, G., 2009. Annual metallic flows in roof runoff from different materials: test-bed scale in Paris conurbation. Environ. Sci. Technol. 43, 5612–5618. SEPP, 1999. Variation of the State Environment Protection Policy (Waters of Victoria) Insertion of Schedule F7. Waters of the Yarra Catchmen. Sharley, D.J., Sharp, S.M., Marshall, S., Jeppe, K., Pettigrove, V.J., 2017. Linking urban land use to pollutants in constructed wetlands: implications for stormwater and urban planning. Landsc. Urban Plan. 162, 80–91. Shinya, M., Tsuruho, K., Konishi, T., Ishikawa, M., 2003. Evaluation of factors influencing diffusion of pollutant loads in urban highway runoff. Water Sci. Technol. 47, 227–232. Sillanpää, N., Koivusalo, H., 2015. Stormwater quality during residential construction activities: influential variables. Hydrol. Process. 29, 4238–4251. Sliva, L., Dudley Williams, D., 2001. Buffer zone versus whole catchment approaches to studying land use impact on river water quality. Water Res. 35, 3462–3472. Tjandraatmadja, G., Pollard, C., Sheedy, C., Gozukara, Y., 2010. Sources of Contaminants in Domestic Wastewater: Nutrients and Additional Elements from Household Products. Water for a Healthy Country Flagship Report. CSIRO, Canberra Google Scholar. Urbonas, B., Stahre, P., 1993. Stormwater: Best Management Practices and Detention for Water Quality, Drainage, and CSO Management. US EPA, 1983. Methods for Chemical Analysis of Water and Wastes. Environmental Monitoring and Support Laboratory, Office of Research and Development, Cincinnati. US EPA, 2005. National Management Measures to Control Nonpoint Source Pollution from. Urban Areas, Washington. Van Bohemen, H., Van De Laak, W.J., 2003. The influence of road infrastructure and traffic on soil, water, and air quality. Environ. Manag. 31, 0050–0068. Vaze, J., Chiew, F.H., 2002. Experimental study of pollutant accumulation on an urban road surface. Urban Water 4, 379–389. Victorian Stormwater Committee, 1999. Urban Stormwater: Best-Practice Environmental Management Guidelines. CSIRO publishing. Vluis, 2017. Victorian land use information (2006 - 2014). In: Department of Economic Development, J., Transport and Resources, Melbourne). Walsh, C.J., Fletcher, T.D., Bos, D.G., Imberger, S.J., 2015. Restoring a stream through retention of urban stormwater runoff: a catchment-scale experiment in a social–ecological system. Freshw. Sci. 34, 1161–1168. Well, A.D., Myers, J.L., 2003. Research Design & Statistical Analysis. Psychology Press. Whitehand, J.W., 1994. Development cycles and urban landscapes. Geography 3–17. Withers, P.J., Jordan, P., May, L., Jarvie, H.P., Deal, N.E., 2014. Do septic tank systems pose a hidden threat to water quality? Front. Ecol. Environ. 12, 123–130. Wong, T.H., 2006. Water sensitive urban design-the journey thus far. Aust. J. Water Resour. 10, 213–222. Yu, Y.-S., Zou, S., Whittemore, D., 1993. Non-parametric trend analysis of water quality data of rivers in Kansas. J. Hydrol. 150, 61–80. Yue, S., Pilon, P., Cavadias, G., 2002. Power of the Mann–Kendall and Spearman's rho tests for detecting monotonic trends in hydrological series. J. Hydrol. 259, 254–271. Zhang, K., Deletic, A., Bach, P.M., Shi, B., Hathaway, J.M., Mccarthy, D.T., 2019. Testing of new stormwater pollution build-up algorithms informed by a genetic programming approach. J. Environ. Manag. 241, 12–21. Zhao, J., Lin, L., Yang, K., Liu, Q., Qian, G., 2015. Influences of land use on water quality in a reticular river network area: a case study in Shanghai, China. Landsc. Urban Plan. 137, 20–29. Zhou, P., Huang, J., Pontius JR., R.G., Hong, H., 2016. New insight into the correlations between land use and water quality in a coastal watershed of China: does point source pollution weaken it? Sci. Total Environ. 543, 591–600.

Ellis, J.B., Butler, D., 2015. Surface water sewer misconnections in England and Wales: pollution sources and impacts. Sci. Total Environ. 526, 98–109. Epa Victoria, 1996. Environmental Guidelines for Major Construction Sites. Melbourne. Eriksson, E., Auffarth, K., Henze, M., Ledin, A., 2002. Characteristics of grey wastewater. Urban Water 4, 85–104. Esri, 2017a. ArcGIS Desktop: Version 10.5. Environmental System Research Institue, CA. Esri, 2017b. Topographic [basemap], Scale Not Given. Francey, M., Fletcher, T.D., Deletic, A., Duncan, H., 2010. New insights into the quality of urban storm water in South Eastern Australia. J. Environ. Eng. 136, 381–390. Gaffield, S.J., Goo, R.L., Richards, L.A., Jackson, R.J., 2003. Public health effects of inadequately managed stormwater runoff. Am. J. Public Health 93, 1527–1533. Gibson, C.A., Meyer, J.L., 2007. Nutrient uptake in a large urban river 1. JAWRA J. Am. Water Resour. Assoc. 43, 576–587. Goonetilleke, A., Thomas, E., Ginn, S., Gilbert, D., 2005. Understanding the role of land use in urban stormwater quality management. J. Environ. Manag. 74, 31–42. Grace Detailed-Gis Services, 2012. Directly Connected Imperviousness Compilation for Melbourne Water Selected Catchments. Report prepared for Melbourne Water, Melbourne. Hatt, B.E., Fletcher, T.D., Walsh, C.J., Taylor, S.L., 2004. The influence of urban density and drainage infrastructure on the concentrations and loads of pollutants in small streams. Environ. Manag. 34, 112–124. Hirsch, R.M., Slack, J.R., Smith, R.A., 1982. Techniques of trend analysis for monthly water quality data. Water Resour. Res. 18, 107–121. Hoxley, G., Dudding, M., 1994. Groundwater Contamination by Septic Tank Effluent: Two Case Studies in Victoria. pp. 145 Australia. Water Down Under 94: Groundwater/ Surface Hydrology Common Interest Papers; Preprints of Papers. Jarvie, H., Whitton, B., Neal, C., 1998. Nitrogen and phosphorus in east coast British rivers: speciation, sources and biological significance. Sci. Total Environ. 210, 79–109. Kendall, M.G., 1975. Rank Correlation Methods, fourth ed. Charles Griffin, London. Kuller, M., Bach, P.M., Ramirez-Lovering, D., Deletic, A., 2018. What drives the location choice for water sensitive infrastructure in Melbourne, Australia? Landsc. Urban Plan. 175, 92–101. Landsberg, J.H., 2002. The effects of harmful algal blooms on aquatic organisms. Rev. Fish. Sci. 10, 113–390. Lee, J.H., Bang, K.W., 2000. Characterization of urban stormwater runoff. Water Res. 34, 1773–1780. Li, S., Gu, S., Liu, W., Han, H., Zhang, Q., 2008. Water quality in relation to land use and land cover in the upper Han River Basin, China. Catena 75, 216–222. Lintern, A., Webb, J., Ryu, D., Liu, S., Waters, D., Leahy, P., Bende‐Michl, U., Western, A., 2018a. What are the key catchment characteristics affecting spatial differences in riverine water quality? Water Resour. Res. 54 (10), 7252–7272. Lintern, A., Webb, J.A., Ryu, D., Liu, S., Bende-Michl, U., Waters, D., Leahy, P., Wilson, P., Western, A.W., 2018b. Key Factors Influencing Differences in Stream Water Quality across Space, vol. 5 Wiley Interdisciplinary Reviews, Water. Liu, A., Egodawatta, P., Guan, Y., Goonetilleke, A., 2013. Influence of rainfall and catchment characteristics on urban stormwater quality. Sci. Total Environ. 444, 255–262. Liu, D., 2016. Water supply: China's sponge cities to soak up rainwater. Nature 537, 307. Loucks, D.P., Beek, E.V., 2017. Water Resource Systems Planning and Management: an Introduction to Methods, Models, and Applications. Deltares & UNESCO-IHE, Paris. Mann, H.B., 1945. Nonparametric tests against trend. Econometrica 13, 245–259. Manningham City Council, 2011. Domestic Wastewater Management Plan. Martin, P., Turner, B., Waddington, K., Pratt, C., Campbell, N., Payne, J., Reed, B., 2000. Sustainable Urban Drainage Systems: Design Manual for Scotland and Northern Ireland. C521. CIRIA, London, UK. Melbourne Water, 2017a. Centerlines of Waterways, Channels and Underground Drains. Melbourne Water, Melbourne. Melbourne Water, 2017b. MWC Catchment [Online]. AURIN Portal Available: https:// portal.aurin.org.au/, Accessed date: 30 March 2017. Melbourne Water, 2017c. Waterway Water Quality Data [Online]. Available: https:// www.melbournewater.com.au/water/health-and-monitoring/river-health-andmonitoring/waterway-water-quality-data, Accessed date: 15 February 2018. Meng, Y., Feng, C., Li, T., Wang, L., 2009. Identifying dry-weather flow and pollution load sources of separate storm sewer systems with different degrees of illicit discharge. Environ. Sci. 30, 3527–3533. Muserere, S.T., Hoko, Z., Nhapi, I., 2014. Characterisation of raw sewage and performance assessment of primary settling tanks at Firle Sewage Treatment Works, Harare, Zimbabwe. Phys. Chem. Earth, Parts A/B/C 67–69, 226–235. NEARMAP, 2017. Nearmap. MapBrowser. NEDF, 2010. In: AUSTRALIA, G. (Ed.), GEODATA 9 Second DEM.

213