A multi-pronged approach to source attribution and apportionment of heavy metals in urban rivers

Abstract

Heavy metal (HM) contamination of water bodies is caused by both first generation (industries) and second generation (distributed sources, domestic sewage, sediments) sources. We applied a multi-pronged approach to quantify the contribution of first and second generation sources to the HM load in a stream located in an industrialised catchment. We found that, despite strict regulation, first generation sources contributed significantly to the HM load (60%–80%), showing the ineffectiveness of current regulation. Domestic sewage contributed significantly to Cu, Ni, and Mn load (15%–20%). The contribution of distributed sources and sediments to HM load is insignificant. In a 24-hour cycle, HM concentrations frequently exceeded FAO’s irrigation water quality standards, with the highest concentrations observed at night. Empirically, the study highlights the continued plight of urban streams in rapidly industrialising centers and the failure to regulate first-generation sources. Methodologically, it demonstrates the importance of temporally intensive measurement of contaminant concentration and load. Policy implications include the need for ambient water quality standards, inclusion of HMs in such standards, load-based regulation, and a problem-oriented monitoring and enforcement approach.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s13280-022-01734-y.

Introduction

The wave of industrialisation in the nineteenth and twentieth centuries led to pollution of rivers across the world, with contamination by heavy metals (HMs) being one of the most insidious and toxic outcomes (Black et al., 1980; Jacob et al., 2018). In India, the pollution of the Palar river and the Ganga by tanneries in Vellore and Kanpur, respectively, are notorious cases (Paul 2017). The risk to public health is aggravated when freshwater used for irrigation is replaced by or augmented by wastewater due to the rapid urbanisation of river catchments. Reuse of industrially contaminated wastewater results in HMs entering the agricultural food produces, thereby threatening the health of farmers and consumers (Feenstra et al. 2000; Khan et al. 2013; Gaurav et al. 2018).

Governments have responded to the cases of untreated effluent discharges (what we call ‘first generation’ cases) by mandating the installation of effluent treatment plants (ETPs) and demanding changes in technologies at the point sources. This effort has partially managed to mitigate the problem (Menon 1988). Nevertheless, HMs presence in water and aquatic life continues to be reported, especially in countries like China and India (Suthar et al. 2009; Li and Zhang 2010; Sen et al. 2011). Multiple ‘second generation’ sources are hypothesised to explain this: HMs may come from domestic sewage, or non-point sources such as car washes (Yadav et al. 2002; Rule et al. 2006; McKenzie et al. 2009), or from the resuspension of contaminants that had settled in sediments (Ahipathy and Puttaiah 2007; Chen et al. 2016). Nevertheless, it is also possible that the enforcement of pollution regulations against the ‘first generation’ polluters is ineffective. Scientists and regulators are grappling with apportioning the contribution of first and second generation sources, especially in rapidly urbanising river basins in developing countries, where multiple modes of contamination are present, and the monitoring capacities are weak.

Methodologically speaking, studies seeking to pinpoint the source of HMs in rivers and streams have used different approaches, depending upon whether they focus on a single source or multiple sources. Khwaja et al. (2001) used a spatial monitoring approach to pinpoint the sources of HM pollution in the Ganga river. The river water was sampled upstream and downstream of Kanpur city. The ten-fold increase observed in chromium (Cr) concentrations was attributed to untreated industrial effluent discharges from the city. Audry et al. (2004) applied a mass balance approach to identify and quantify the HM load from mining tailing to Lot-Garonne fluvial system (France), particularly focusing on the contribution of molecular diffusion from sediments to the dissolved HM flux in the river. When multiple sources of contamination are likely, studies have employed statistical approaches such as factor analysis and principal component analysis to identify sources of pollution in the catchment (Rawat et al. 2003; Huang et al. 2010; Li and Zhang 2010; Iqbal et al. 2013; Chen et al. 2016). However, these approaches only broadly identify the potential sources of pollution and their rank ordering without specifying the quantum of HM load contributed by various sources. Quantifying pollution load rather than only concentrations is useful in identifying major polluters and focusing regulatory attention on them (Wainger 2012).

From the above, it appears that when multiple sources and forms of pollution are present in complex geography, a multi-pronged approach is required and may yield important policy pointers. Our goal was to demonstrate the usefulness of such an approach using the case of the Vrishabhavathy River in Bengaluru city, India. A review of the literature on the Vrishabhavathy River and news stories suggested that this river is highly contaminated with HMs (Ahipathy and Puttaiah 2006, 2007; Lélé et al. 2013; Madhukar and Srikantaswamy 2013; Sequeira 2016) and that the likely source was industries. However, in an expert consultation held in November 2012 to discuss our review, representatives of the state-level pollution regulator, Karnataka State Pollution Control Board (KSPCB), attended the meeting. They insisted that first generation problems of untreated effluents from industrial point sources had been addressed. The observed pollution in the river was contributed by dispersed sources that are hard to regulate. Other experts suggested the possibility of resuspension of past pollutants from the river sediments and the possibility of domestic sewage being another source of HMs.

We, therefore, launched a systematic multi-pronged effort to examine this problem. We hypothesised that both first and second generation sources contribute to the HM load, and the existing monitoring strategies were inadequate to capture contamination levels and locate the sources accurately. Our objective, therefore, was (a) identify and quantify the HM loads from first and second generation sources in different segments of the heavily industrialised sub-catchment of the river to (b) identify the geographic locations of first generation sources contributing to the HM load after controlling for the contribution from second generation pollution sources, and (c) understand what gaps in the official monitoring and legal standards result in continued contamination. We did this using a combination of spatially distributed and temporally intensive in-stream monitoring, estimation of sediment contributions, and statistical analysis. Our results highlight the level and nature of HM contamination and have important implications for research methods that should be deployed in such situations and for the monitoring and enforcement policy to be adopted in such contexts.

Materials and methods

Study site

Bengaluru city is a 13 million-strong urban hub for various industrial and commercial activities in southern India (see Fig. 1a, b). The Vrishabhavathy River, which originates in a residential part of Bengaluru, has been long known to carry Bengaluru’s sewage into irrigation tanks downstream (see Fig. 1c). However, the industrialisation of its catchment in the 1980s has added industrial pollutants to the problem of pollution. The western arm of the Vrishabhavathy—hereinafter called the Peenya stream–has a length of ~ 13 km and a catchment of ~ 32 km² and is the most industrialised catchment of the Vrishabhavathy. It can be divided into three broad segments. The upper sub-catchment is heavily industrialised, overlapping with the Peenya Industrial Area (PIA) and two neighbouring industrial areas (see Fig. 1d). PIA is one of the largest industrial areas in Asia, established in the late 1970s, explicitly for small-scale industries (www.peenyaindustries.org). The area houses more than 3500 large, medium and small-scale industries, ranging from machine tools and hydraulics to textile dyeing, electroplating and rubber moulding. The other industrial areas also contain a similar density of industrial units.

Fig. 1.

Location of the study site: a Karnataka state within India, b Bengaluru within Karnataka, c the Bengaluru city, the Vrishabhavathy river catchment and the Peenya catchment within it, and d the distribution of industrial areas and sampling locations on the Peenya stream

Downstream of PIA, for 3 km, the Peenya stream passes through mixed-use areas that include residential, commercial and scattered small-scale industrial units. The final 5 km stretch of the stream is through the Bangalore University campus, which is a green zone with almost no contribution of industrial effluents to the stream before it meets the eastern branch of the Vrishabhavathy. We used this geographical variation to set up a multi-pronged monitoring protocol to quantify the contribution of various sources to stream pollution loads.

Nature of potential polluters and pollutants

In India, for pollution regulation, industries are classified into ‘red’, ‘orange’ and ‘green’ categories, wherein ‘red’ category industries are those producing the most significant quantity and type of toxic effluents and therefore subject to the most stringent regulations, followed by orange and green (EMPRI 2008). Large polluting industries such as petrochemicals, paper & pulp, sugar, or distilleries are not present in metropolitan areas. However, small-scale industrial units carrying out electroplating, dyeing, alloy smelters are present in large numbers in industrial estates such as PIA and its surroundings. An official survey in 2008 reported around 3700 industries, out of which 372 red, 112 orange, 715 green category industries are located in the Peenya industrial area as a whole (EMPRI 2008). More recent estimates (2021) from Peenya Industries Association (PIA) indicate 5040 industries in the area (PIA 2021). There are additional red category industries in neighbouring industrial areas.

Large and medium scale industries are required to treat toxic effluents individually in the ETPs located within their premises before releasing them into the stormwater drain or sewage lines (both of which end up in the river, in the absence of any sewage treatment plants in the vicinity). Small-scale industries are permitted to collect and transport their effluents to common effluent treatment plants (CETPs) for treatment and disposal. However, they may be tempted to release their effluents clandestinely into stormwater drains. Therefore, all these industrial units are the ‘point sources’ of pollution. In theory, they are easy to identify and regulate, especially in designated ‘industrial areas’ such as PIA and its neighbouring Rajajinagar and Yeshwanthpur industrial areas.

In addition to point sources, several unregulated manufacturing and service industries within the catchment contribute to the effluents in the stream (Chukwu et al. 2008; Tajuddin et al. 2020). These units (incense-manufacturing units, car-washing units, scrap metal processing and solid waste recycling units) are dispersed across the Peenya catchment, making identification and regulation of such small size units difficult. For ease of understanding, we labelled them as ‘distributed’ sources of pollution. Wastewater from these units is directly discharged into the streams that lead to the Vrishabhavathy River.

Finally, due to the lack of sewage treatment plants and last-mile sewer-line connectivity in the catchment, large quantities of untreated sewage is discharged into the stream at many locations (Shankar et al. 2008). It may also be noted that, since the stream originates within Bengaluru city that experiences at least 4–5 dry or very low rainfall months (January–May), the summer flows in the stream consist of wastewater diluted marginally by baseflows.

While industrial pollution takes many forms, we chose to focus on HMs, specifically Chromium (Total Cr), Copper (Cu), Lead (Pb), Nickel (Ni), and Manganese (Mn). While other non-conservative contaminants degrade over a few kilometres before the river reaches the farming area, HMs are persistent and subject to bio-magnification from water to fodder to cattle to milk or water to fruit/vegetable products (Lokeshwari and Chandrappa 2006; Lokhande et al. 2011). The specific choice of HMs to be monitored was guided by the nature of the industrial activities in the catchment and prior literature citing HM presence (Lélé et al. 2013).

Mapping pollution sources and selection of monitoring locations

The PIA and neighbouring industrial areas constitute a clearly demarcated industrial zone that houses most of the red-category industries in the catchment. Given difficulties securing permissions to monitor effluents from industries, we treated the entire zone as a single large point source. However, compared to the industries located in the industrial zone, the middle segment of the catchment houses several small-scale industries (point sources) and service units (such as car washing-distributed sources) that are neither identified nor monitored by KSPCB. We, therefore, produced a revised map of potentially polluting (point and distributed) sources in the catchment. First, we used the list of industries in the Peenya catchment from the KSPCB database to geotag industries within the catchment boundary. Second, we added missing industries to the database (ones on Google Maps but missing from the KSPCB list). Third, we did ground-truthing across the Peenya catchment to confirm the presence of these added industries. As per our revised database, the total number of red, orange and green category industries in the Peenya catchment was 104, 24 and 75, respectively. For distributed sources, we identified, mapped and surveyed all car-washing units, scrap metal processing units, petrol pump stations (without carwash units) and incense making units in the catchment.

Spatially, based on the distribution of potential industrial polluters and sites accessible for stream sampling, we selected three sampling locations, including one at Chowdeshwari Nagar (CHO) (immediately downstream of PIA), one at Sumanahalli (SUM) and one at the downstream end of the green zone in Bengaluru University (VRH), with nested catchment sizes of 18, 22 and 32 km² (see Fig. 1d). However, we faced several logistic problems while carrying out sampling at SUM, so full-year sampling was possible only at CHO and VRH. Nevertheless, the 3-month data from SUM (including one dry season reading) provided important additional data useful for source apportionment.

Temporally, it was important to capture both seasonal variations and diurnal variations. Monthly variations are important to capture the variability in flows introduced during the monsoon season because of high surface flows. Diurnal variations were considered important because anecdotal information suggested that some industries may resort to the clandestine discharge of untreated effluents at night. We, therefore, adopted a temporally intensive monitoring strategy: (a) monthly sampling and (b) in each month one 24-h cycle of hourly sampling and flow measurement. We sampled over 15 months (between May 2014 and August 2015), but sampling efforts failed or were incomplete during a few months because of rainstorms and treacherous conditions at certain sites. We thus ended up with 12 months × 24-h datasets at CHO and VRH and a 3-month × 24-h dataset at SUM.

HM load estimation and source apportionment

Our objective was to estimate the HM concentrations and their respective loads and then estimate the contribution of industrial sources (supposedly regulated) and distributed sources (unregulated) highlighted through expert consultations to the HM load in the Peenya stream. We present below the methods used in these two stages.

Measuring HM concentration and load

Estimation of the HM load at a location in the stream involved two steps: (a) measurement of the HM concentration and (b) measurement of the flow rate at the time of sample collection. The daily HM load at that location was then the product of the concentration and the flow summed over 24 hours, i.e.,

$${\text{HM}}_{i}load=\sum _{i=1}^{24}{Q}_t\ast C_{\mathit{it}}$$ 1

where Q_t is the stream flow at time t (m³/hour) and C_it is the concentration of ith HM at time t (kg/m³).

We collected a grab sample every hour to estimate C_it using an Automatic Water Sampler (Hach Sigma 900MAX Portable Sampler). Given only one sampler and staffing constraints, the dates on which samples for a particular month were collected differed slightly for each location. Hence such measured loads cannot be directly correlated with loads across multiple locations during the same month.

Standard protocols were followed for digesting water samples (APHA 2005), and concentrations of each HM in both digested and filtered samples were determined using a Perkin-Elmer Atomic Absorption Spectrometer (see Appendix S1). In addition to HMs, water samples were also analysed for conventional water quality parameters, including pH, conductivity, total suspended solids (TSS), biochemical oxygen demand (BOD₅), and chemical oxygen demand (COD) using standard protocols (APHA 2005).

Measuring hourly flow in the river was essential to estimate the total load of HMs in a stream that was highly seasonal, and experienced variations in diurnal flows. The standard procedure is to measure the height of the flow in the channel (‘stage’) and then use a ‘stage-discharge rating curve’ for a given location to estimate the flow rate (Seibert and Beven 2009). Given that no stage-discharge curves were available for the stream at any location, we had to generate them ourselves. The stage was measured using an Odyssey Capacitance Water Level Logger at 10-min intervals over 24 h. We measured the average velocity of the stream flow along the width of the channel at a particular location using a velocity meter (HACH velocity flow meter—Model No. FH950). We repeated these velocity measurements at significantly different stage values and thereby estimated a stage-discharge curve for that location. We estimated hourly flow based on this curve (see Appendix S1). Due to the lack of a fixed stage recorder at the sampling locations, we repeated the procedure of estimating stage-discharge curves every month for all three locations.

Distributed sources

The distributed sources mapped in the catchment included car washing units, petrol pumps (without carwash units), scrap metal dealers, plastic recycling and incense making units. We also collected data on water consumption, presence or absence of the wastewater treatment facility and discharge point for effluents, and volume of activity for each unit. Only the car washing units reported significant effluent discharge, other units used ‘dry’ production practices. Hence, we estimated their HM contribution using literature on the HM pollution concentrations in effluent released from car washing facilities (Oknich and District 2002; Sorme and Lagerkvist 2002) and the reported water use.

Sewage

Untreated sewage enters the Peenya stream at multiple points, and neither sample collection at these points nor flow measurement was physically possible. Given that the composition of sewage is unlikely to vary much from location to location or over seasons, we collected samples for one 24-h period from the inlet of a sewage treatment plant (STP) located in the Indian Institute of Science (IISc) campus in Bengaluru (outside the Peenya catchment). The STP receives 500 Kilo litres of raw sewage daily and is operated for 19 h daily with an average hourly flow of 20,834 L. Samples were kept cool using pre-cooled silica gel ice packs. After completing a 24-h sampling round, the samples were transported to the water and soil Lab at ATREE and analysed to determine the flow weighted average HM concentration in sewage. This flow weighted average concentration was then multiplied by the sewage flow likely to be generated by the human population in each sub-catchment. This flow consists of two components a) generated from residential units and b) commercial, institutional, and educational establishments. For domestic sewage, we estimated the human population using fine-scale (census enumeration block-level) population data from Census 2011 and rastered the same to 30 × 30 m pixels. Assuming an average domestic water consumption of 100 lpcd and therefore a domestic sewage outflow of 80 lpcd, we estimated domestic sewage contribution of each sub-catchment: 24 MLD at CHO, an additional 20 MLD at SUM, and an additional 15 MLD at VRH. Further, based on a survey conducted by ATREE (quantification of the water supply across various sectors in Bangalore), we estimated an additional contribution of 15 MLD from the commercial and other establishments for the whole catchment, i.e., VRH (Apoorva et al. 2021).

Sediment contribution through diffusion and resuspension

The magnitude of diffusive flux from sediments was estimated using Fick’s Law of diffusion modified for appropriate use in sediments, whereby

$$J_x=-\varnothing D_{\mathrm{s}}{(\delta c|\delta z)}_{z=0}$$ 2

where J_x is the diffusive flux (mg cm⁻² s⁻¹) with negative values indicating an outward flux from the sediment. ∅ is the porosity at the sediment–water interface (measured to be 0.37 and 0.48 at CHO and VRH, respectively. D_s is the molecular diffusion coefficient (cm² s⁻¹), and ${(\delta c|\delta z)}_{z=0}$ is the concentration gradient (mg cm⁻⁴) at the sediment–water interface. As the molecular diffusion coefficient depends on temperature, the values of D_s used were corrected for the average temperature observed during the monitoring period (Audry et al. 2004). The corrected sediment diffusion coefficients are shown in Table 1.

Table 1.

Molecular diffusion coefficients corrected for the temperature at VRH

Molecular diffusion co-efficient	Cr	Cu	Pb	Ni	Mn
(Ds) (10−6 cm2/s)	5.94	8.30	9.45	6.79	6.88

Sediment porosity (∅) and HM concentration gradient $\delta c$ was estimated by collecting sediment samples from the streambed at the three locations monthly twice for two months using a Nornico Van Veen Box Corer. The samples were placed in the zip lock bags, stored below 4⁰C and were transported to the ATREE water and soil Lab within 24 h from the time of collection. Sediment porewater was extracted by centrifuging 100 g of sediments for 30 min at 6000 RPM. The sediments (digestion with nitric acid) and the porewater samples were analysed for HMs. One hundred grams of sediment samples from four sites were dried at 110⁰C for 24 h, and porosity was estimated (Batley and Simpson 2016).

The multi-location sampling was useful for estimating the possible contribution from the resuspension of sediments. The upstream location (at CHO) was steep enough to prevent sedimentation, whereas the lowest stretch (between VRH and SUM) was flatter and contained more sediments that could get re-suspended. We hypothesised that resuspension contributes significantly to the HM flux. Therefore daily peak flows in the stream will cause an unexplainable increase in HM concentrations at VRH compared to CHO. Statistical analysis, including descriptive statistics, paired sample t tests and graphs, were analysed using Microsoft Excel 2013 (Microsoft Corporation, Redmond, WA; USA) and IBM SPSS Statistics 23.0 (Armonk, NY: IBM Corp.).

Results

We begin by examining unacceptably high HM contamination in the stream. We examine the seasonal and hourly HM concentrations at CHO, the site immediately downstream of the major industrial areas, and compare these values with ambient water quality and irrigation water quality standards. We then present estimates of the HM load at all three sampling locations. Finally, we develop estimates of HM load contribution from distributed sources, sewage and sediments (diffusion and resuspension). This will enable us to estimate industrial point source contributions of HM load in a nested manner and then compare the relative contributions of all these sources at VRH.

Mean and seasonal variations at CHO

The 24-h average of HM concentrations across the 15 months at CHO are shown in Fig. 2. During the monitoring period, the 24-h average concentrations in mg/l ranged from 0.03–0.46 for Cr, 0.04–0.27 for Cu, 0.01–0.12 for Pb, 0.02–0.06 for Ni and 0.37–0.73 for Mn. To assess the toxicity levels, we compared the HM concentrations with the standards set by various regulatory agencies.

Fig. 2.

(a–e) Hourly HM concentrations at CHO and comparison with the WHO irrigation water quality standards (y-axis for Cr has been truncated). f  shows the color change and sharp increase in HM concentrations in the late-night samples

The Environment (Protection) Rules 1986, notified by the Government of India, neither specify ambient water quality standards for HMs (Jamwal et al. 2016) nor specify quality standards for different uses of water. However, Central Pollution Control Board (CPCB) use ‘water quality criteria’ to determine whether a water body is suitable for a particular use. The lowest possible use is irrigation (CPCB 2011). Conventional water quality parameters such as BOD₅, COD, Dissolved Oxygen and fecal coliform (FC) are used to develop the water quality criteria. Additional parameters such as conductivity, salinity, boron content, and pH are included for irrigation water quality. Heavy metals are not mentioned in the list of parameters (Jamwal et al. 2016).

For comparison, we used the ambient water quality standards notified by the US EPA as the ‘fishing and swimming’ standard for fresh and saltwater sources (US EPA 2015). This standard’s acceptable concentrations of HMs range from 2.5 µg/l for Pb to 74 µg/l for Cr, which are significantly lower than those observed at CHO. Clearly, in addition to the sewage contamination and high concentrations of HM, the Vrishabhavathy River does not meet swimmable or fishable standards.

It may, however, be argued that applying a ‘swimmable/fishable’ standard in a seasonal urban river that, during the summer months, is essentially a drain carrying only sewage and treated industrial effluent would be unrealistic. Therefore, given that the Vrishabhavathy waters are used for irrigation at many points along its course downstream of VRH, we compared the HM concentrations at CHO to the irrigation water quality guideline set by the World Health Organisation (WHO) (Feenstra et al. 2000; Jeong et al. 2016; Mahfooz et al. 2020). These standards are far less stringent than the US EPA ambient water quality standards. Figure 2 shows the comparison between HM concentrations observed and the irrigation water quality guidelines set by WHO. Except for Pb (WHO standard (irrigation) for Pb is exceptionally high i.e. 5 mg/l) the average HM concentrations at CHO exceeded WHO’s irrigation water quality standards : either always (in the case of Mn) or frequently (in the case of Cr and Cu) or occasionally (in the case of Ni) (see Appendix S2). This is even visible from the colour of the sample itself (Fig. 2f). The data strongly suggests that discharges of untreated and partially treated industrial effluents from the industrial cluster located u/s of CHO sampling location enter the stream, and the discharge standards are inappropriate for a stream diluted by sewage during the summer season.

The seasonal pattern is as expected: the concentrations drop during the monsoon months of July-Dec (although irregularly, because rains are intermittent) and peak during late summer (April–May). The extreme peaks in the case of Cr suggest episodic discharge of untreated effluents from (say) electroplating units.

We then examined the variations in concentrations and loads within the 24-h cycle. Figure 3 shows the daily variations in HM loads and flows during the four low-flow dates at CHO: May 2014, December 2014, January 2015 and July 2015. In each case, we observed peaks in HMs load during the night (~ 10 pm to 2 am) and early morning hours (6 am to 11 am). It is expected that sewage flows would peak during the morning hours, and the graph of Q (streamflow) over the 24-h period in Fig. 3f confirms that. However, neither the steady declines in total flows at night nor the increased flow in the early morning hours explain the sudden steep rises in HM load during that period. The pattern suggests discharge of industrial effluents either during the late night because it is invisible or during the early morning hours hoping that it will get diluted.

Fig. 3.

Diurnal variations in HM load (a–e) and streamflow (f) for four sampling months at CHO

The average load of HMs observed during the morning (8:00–10:00) and night (23:00–2:00) hours was greater than the load observed during the afternoon (15:00–18:00) hours. However, the difference was statistically significant for loads observed for all HMs between morning and afternoon hours (paired t test, p < 0.05). Except for Pb and Ni significant difference was observed between the HM loads during the afternoon and night hours (paired t test, p < 0.05). The hourly load observed in the morning was twice as observed in the afternoon hours (Appendix S2). Given these patterns and the quantum of contamination, the only possible explanation is the discharge of untreated effluents by industrial sources.

HM load and its spatial distribution

Using hourly HM concentrations and streamflow measurements with Eq. 1, we estimated the daily HM load at the three sites for each sampling date. Table 2 presents the averages and standard deviations of the flows (dry and wet weather) and the HM loads. For ease of comparison, the average HM loads are plotted in Fig. 4.

Table 2.

Average stream flows and HM load at three sampling sites across 12 monthly sampling dates

Monitoring site	Average flow (MLD) ± SD	Average dry weather flow (MLD) ± SD	Average wet weather flow (MLD) ± SD	Cr (kg/day) ± SD	Cu (kg/day) ± SD	Pb (kg/day) ± SD	Ni (kg/day) ± SD	Mn (kg/day) ± SD
CHO (n = 12)	76 ± 41	49 ± 13	95 ± 16	3.62 ± 2.03	11.65 ± 4.76	2.68 ± 1.82	2.52 ± 1.31	36.34 ± 17.23
SUM (n = 3)	87 ± 30	53	98 ± 24	7.33 ± 0.76	19.00 ± 3.19	2.85 ± 1.06	3.50 ± 0.08	51.54 ± 17.34
VRH (n = 9)	126 ± 23	109 ± 18	147 ± 5	7.63 ± 1.66	20.60 ± 8.09	4.54 ± 3.03	3.88 ± 1.06	61.51 ± 12.19

Fig. 4.

Average HMs load estimated at three monitoring sites on Peenya stream

Several points are noteworthy. First, the uppermost (and most industrialised) sub-catchment (CHO) contributes half or more of the load of all HMs. Second, the sub-catchment between CHO and SUM contributes a significant additional HM load, which contains several small-scale polluting industries (as shown in the map in Fig. 5). On the other hand, the lowest HM load is contributed by the sub-catchment (either green space or has residential or educational uses) between SUM and VRH, except in the case of Pb. This pattern of spatial contributions further supports the earlier inference that first generation sources (industries in designated zones and therefore supposedly well-regulated) are the major contributors to the HM pollution (see Fig. 5).

Fig. 5.

Location of various industries in Peenya catchment

Having established a prima facie attribution to industrial point sources, in the following section, we use a multi-pronged approach to quantify the contribution of each source to HM load at VRH.

Source apportionment of load observed at VRH

We first estimate the daily HM load contribution at VRH from distributed sources, untreated domestic sewage and sediments, which enables us to estimate the contribution of an industrial point source as a residual from the total load estimates in Table 2. We then compare and discuss the estimated shares of all these sources.

Distributed sources

Figure 6 shows the location of the possible distributed sources of pollution (metal/incense making/plastic recycling units) in the Peenya catchment. A summary of their typical operations is presented in Appendix S2.

Fig. 6.

Distributed sources across Peenya catchment

For the carwash units, the variations in average numbers of cars washed per day were 3 to 15 per unit indicated by primary survey data. The amount of water used for washing varies from 150 to 250 L per car. Thus, the estimated minimum and maximum wastewater generation is 14,850 L/day and 123,750 L/day. Using the data on water consumption for car washing per day, we estimated the actual wastewater discharges from all units during the study period (see Appendix S2). The wastewater (~ 32 000 L/day) from these units is directly released into the stormwater drains, which finally joins the Peenya stream. Using HM concentrations reported by (Sorme and Lagerkvist 2002), we estimated the contribution by car wash units total HM load as given in Table 3.

Table 3.

HM load estimation from car-washing units

Emissions	Cr	Cu	Pb	Ni	Zn
HM emission/vehicle/litre of water used (µg/l) (Sorme and Lagerkvist 2002)	29	203	162	26	1574
Net load at VRH (kg/day)	0.001	0.006	0.005	0.001	0.050

Sewage contribution

The contribution of domestic and commercial establishment to the HM load was quantified by multiplying the estimated sewage generated in the catchment with the concentrations observed for raw sewage at the IISc STP inlet (see Appendix S3).

Using the population estimates from each sub-catchments corresponding domestic sewage flows were estimated (see Appendix S3) (Apoorva et al. 2021). The total sewage generation from Peenya catchment, including CII sources’ contributions, is estimated as 74 MLD. We observed relative concentrations of HMs in following order Mn > Cu > Ni > Cr ≥ Pb (Table 4).

Table 4.

Estimated HM load contributions by sewage at VRH

Metals	Flow weighted average HM concentration in raw sewage (µg/l) (n = 19)	Estimated contribution of sewage contribution to total HM load (kg/day) at VRH*
Cr	8	0.56
Cu	70	5.16
Pb	5	0.38
Ni	8	0.59
Mn	173	12.79

*Based on average estimated sewage flow of 74 MLD at VRH

Contribution of sediments through diffusion and resuspension

Our estimates of the contribution of sediments via diffusion is given in Table 5, which includes estimates of HM concentrations in the porewater and sediments at various locations on Peenya stream. We estimated the diffusive flux for the stretch between CHO and VRH by taking the average of δc/δz observed at CHO and VRH (Eq. 2).

Table 5.

Diffusive flux contribution to HM load in stream

HMs	Average dissolved HMs concentrations in water column (ppb) (Cd)		HM concentrationsporewater (ppb) (Cp)		Δc = (Cp − Cd) (ppb)		δc/δz (δz = 0.5 cm) (mg/l/cm)		Molecular diffusion co-efficient (10–6 cm2/s)	Mean diffusive HM flux (between CHO and VRH) (kg/day)
	VRH	CHO	VRH	CHO	VRH	CHO	VRH	CHO
Cr	11.2	5.7	106.0	155.0	94.8	149.3	189.5	298.7	5.94	− 0.06
Cu	10.0	8.8	30.0	45.0	20.0	36.2	40.1	72.4	8.30	− 0.02
Pb	3.2	4.3	125.0	237.0	121.8	232.7	243.7	465.5	9.45	− 0.15
Ni	11.2	14.0	46.0	94.0	34.8	80.0	69.7	160.0	6.79	− 0.04
Mn	373.1	359.8	479.0	771.0	105.9	411.3	211.8	822.5	6.88	− 0.18

*Area of the river bed between CHO and VRH is approximately 0.124 km² based on an average width of 8.1 m

**Negative value indicates diffusion from sediment to the water column

The highest diffusive fluxes were observed for Mn and Pb followed by Cr, Ni and Cu. Higher fluxes could be attributed to the greater differences observed between the HM concentrations in porewater and in the overlying water (dissolved HM concentrations). Our results on the relative distribution of HMs in porewater and sediments are similar to the data reported from studies conducted in Netherlands and China (Van Den Berg et al. 1999; Ji et al. 2018).

To assess the contribution from resuspension of sediments, we hypothesise that the lower reaches (CHO to VRH) of the catchment are either in steady-state (sediment inflow equals sediment outflow) or (being flatter) are net recipients of sediment deposits (implying sediment inflow would be greater than sediment outflow). In either case, there would be no positive contribution from sediment resuspension to the HM flux at VRH. The exception to this could be storm events or peak flow intervals, during which sharp increase in TSS and HM concentrations are observed simultaneously. Considering the higher concentrations of HMs in sediments we hypothesise that, during peak flows the resuspension of bottom sediments will contribute to a peak HM concentration and TSS levels at VRH (see Appendix S3). In Fig. 7a–e we present the data on diurnal variations in average HM concentrations at CHO and VRH. The maximum flow velocity observed during peak flows at CHO and VRH is 0.91 ± 0.14 m/s and 0.70 ± 0.09 m/s respectively. The minimum velocity observed at CHO and VRH is 0.17 ± 0.03 m/s and 0.14 ± 0.06 m/s respectively. During peak flows the average TSS levels observed at VRH are significantly lower than CHO (see Fig. 7). In addition, the average TSS concentrations at VRH during peak flows (see Fig. 7f) corresponds to the TSS concentrations reported in raw sewage, indicating contribution from raw sewage influx rather than resuspension of bed sediments (Muserere et al. 2014).

Fig. 7.

Average HM concentration at CHO and VRH and corresponding TSS concentrations at CHO and VRH

In Table 6, we present the average daily TSS and HM concentrations observed at CHO and VRH. The data shows that the average daily HM concentrations observed at VRH were either less (Cr or Cu) or equal to concentrations observed at CHO. Also, no significant difference was observed in the average daily TSS concentrations at CHO and VRH (p > 0.05). This further supports our argument that the resuspension of sediments to HM load in the Peenya stream is negligible.

Table 6.

Average daily TSS and HM concentrations in water samples at CHO and VRH

Heavy metal	CHO (Mean ± SD)	VRH (Mean ± SD)
TSS (mg/l)	458 ± 110	484 ± 131
Cr (mg/l)	0.100 ± 0.12	0.068 ± 0.015
Cu (mg/l)	0.198 ± 0.06	0.160 ± 0.036
Pb (mg/l)	0.039 ± 0.03	0.033 ± 0.008
Ni (mg/l)	0.033 ± 0.01	0.030 ± 0.006
Mn (mg/l)	0.507 ± 0.12	0.490 ± 0.196

Pollution sources: apportionment

Based on the above, the industrial point source contribution in various sub-catchments was estimated as follows:

1. At CHO, assuming no contribution from sediments (diffusion or resuspension), the average estimated loads from distributed sources and sewage were subtracted from the average total load to estimate the contribution of industrial point sources.

2. At SUM, the contribution of HM from industrial sources at CHO, additional sewage flows in the SUM sub-catchment, distributed sources and diffusive flux (for the 5 km length) was subtracted from the average total load to estimate the contribution of industries located between CHO and SUM to HM load.

3. Finally, at VRH, the estimated total contribution of industrial point sources at SUM, diffusive flux (for the 13 km length between CHO and VRH), dispersed sources, and additional sewage in the VRH sub-catchment was subtracted to estimate the contribution of industrial point sources in the VRH sub-catchment.

The resulting source-wise estimated contributions at VRH are plotted in Fig. 8 as pie charts. The findings are quite startling. First, industrial point sources located in designated industrial areas in the CHO sub-catchment contributed 45% to 67% (depending on the HM) of the total HM load at VRH. Second, industrial point sources in the sub-catchment between CHO and SUM also contributed significantly (between 20% to 40%) to the HM load at VRH, except for Pb and Ni (< 10%). Third, the industrial points sources in the relatively unindustrialised sub-catchment between SUM and VRH contributed marginally to HMs except for Pb flux (~ 31%). Fourth, sewage contributed a non-negligible fraction (between 14% to 19%) to Ni, Mn and Cu load at VRH. Fifth, sediment diffusion contributed a negligible fraction (between 1% to 3%) with no significant contribution from sediment resuspension. Finally, the contribution of the distributed (minor) sources to HM load is completely negligible.

Fig. 8.

Relative contribution of various sources to HM load at VRH

Out of the first-generation and second-generation sources, industrial activities contributed significantly to HM load at VRH. Depending upon the HM, the share of industrial point sources (u/s SUM) remained between 60 and 80%, with a greater proportion of the industrial clusters located u/s of CHO. The findings are also supported by the HM concentrations and loads observed during the wee hours at the CHO sampling location (see Figs. 2 and 3).

Discussion

Industrial pollutants in the urban river: what standards?

Our results show that the Peenya stream carries a large year-round load of HMs discharged to Vrishabhavathy River, which finally flows into the Byranmangala irrigation reservoir downstream (see Fig. 1c). The concentrations observed at CHO are orders of magnitude greater than USEPA’s ambient water quality standards for a river designated for ‘fishing and bathing’ use. Even the irrigation standards set by WHO are exceeded frequently at CHO. However, the concentrations may dilute as the stream merges with the Vrishabhavathy River and flows downstream.

Given the persistent nature, HMs either tend to stay in the sediments or enter groundwater, thereby contaminating the aquifers (Shankar 2019). Research by our team and several scientists reported accumulation of HMs in the food chain, including milk, vegetables and fruit that are produced in the command area of Byranmangala reservoir (Sekhar et al. 2005; Jamwal and Lele 2017), posing a significant threat to the health of consumers of those products and the farmers themselves.

These findings point to a glaring policy gap in the Indian context, viz., the absence of ambient water quality standards in general and the absence of specific standards for persistent industrial contaminants such as HMs. Although the concept of ‘designated best use’ has been proposed, it has not been implemented in Indian water pollution law (Jamwal et al. 2016; CPCB 2018) (see Lele et al. 2021 for a more detailed discussion).

Furthermore, our findings point to the incompatibility of current discharge standards with ambient quality or use standards (such as the WHO use standard) for urban seasonal rivers such as the Vrishabhavathy. A comparison of the HM discharge standards set for electroplating and other such industries in India with the WHO irrigation quality standard is given in Table 7. We note that for regulators to adopt the WHO irrigation quality standard (which is the lowest acceptable standard for water use in CPCB’s water quality criteria) as ambient water quality critera for irrigation, the treated effluents from point sources require a dilution by a factor of 10 to 30 (except for Pb) to be compatible with this ambient/use critera.

Table 7.

Comparison of average HM concentrations at CHO with Indian industrial discharge standards and WHO irrigation standards

Heavy metal	Mean ± SD (mg/l) (n = 15)	Range (ATREE)	Range (KSPCB)	Indian discharge standard (mg/l)	WHO Irrigation standard (mg/l)	Ratio of irrigation standard to discharge standard
Cr	0.100 ± 0.12	0.015–0.764	Not detected	2.00	0.10	20
Cu	0.198 ± 0.06	0.007–0.784	0.006–0.580	3.00	0.20	15
Pb	0.039 ± 0.03	0.003–0.232	BDL- 0.004	0.10	5.00	0.02
Ni	0.033 ± 0.01	0.005–0.126	0.010–0.037	3.00	0.10	30
Mn	0.507 ± 0.12	0.213–4.238	Not tested	2.00	0.20	10

In the case of an urban and seasonal river such as the Vrishabhavathy, the only dilution in the dry season comes from sewage, which may be insufficient and carry an additional HM load. Our observations above indicate that the sewage contribution to total flow at CHO is only about 50% to 80% (a dilution of 1 to 4 times), against the required dilution of 10–30 times. This highlights the problem of inadequate dilution and, therefore, the inadequate attention given to setting context-sensitive discharge standards, or preferably load-based standards.

In a context where possibly hundreds of industrial units are discharging their effluents into a river with no summer freshwater flows, even individual compliance with emission standards will lead to a large pollutant flux in the ‘river’. The absence of an ambient standard and the consequent lack of load-based regulations means reluctance on the side of regulators to estimate stream flows and industrial load contributions before setting emission standards. One-size-fits-all emission standards are insufficient, especially when water is being (re-)used downstream for irrigating food crops.

First-generation (point) industrial sources still pollute

Notwithstanding the ambiguity about what constitutes acceptable ambient water quality, and notwithstanding the claims by the environmental regulator about pollution from clandestine or unregulated sources, our results show that effluents from designated industrial areas are the major contributors to HM pollution in the Peenya stream. The concentrations observed are significantly greater than emission standards. The significant contribution of HMs from the industrialised CHO sub-catchment, unexplainable peaks in the HMs load, and marginal contribution from distributed sources indicate that the so-called ‘regulated’ industrial areas are the primary sources of pollution. It also indicates a failure of the environmental regulator to effectively monitor and enforce pollution regulations on these (known and registered) polluters.

Past studies reported high concentrations of HMs in the Vrishabhavathy River (Jan et al. 2008; Jayadev and Puttaiah 2013; Madhukar and Srikantaswamy 2013). However, the agency could evade the implications of poor regulation by suggesting scattered units located outside designated industrial areas are the primary contributors or that one-time monitoring data do not indicate persistent violations. Through its spatially and temporally intensive sampling and systematic exclusion of other sources, our study demonstrates that these episodes are persistent and significant and are directly and mainly attributable to effluents from industrial point sources. This points to a regulatory failure that needs to be addressed before one tackles second generation sources (Jones et al. 2017).

Domestic sewage: an important second generation source?

Although poorly regulated industrial units remain the major sources of HM pollution in the Peenya stream, our results also show that in the case of Mn, Cu and to an extent Ni, domestic sewage contributes an estimated 14–16% of the load in a highly industrialised catchment. Several studies have reported the presence of HMs in the sludge generated from the STPs (Madoni et al. 1996). More recently, high concentrations of Mn and Zn (Lu et al. 2016) and Cu and Zn have also been reported from studies conducted in Australia, France and the United Kingdom (Hargreaves et al. 2017; Drozdova et al. 2019). These increasing concentrations are broadly attributed to the rising use of chemicals and industrial products in the household.

However, in the case of Bengaluru, Mn is the most prominent HM in domestic sewage. Groundwater is a significant source of domestic water use (~ 40%) (Malghan et al. 2016) and industrial water use in Bengaluru (Apoorva et al. 2021). This groundwater contains high concentrations of Mn (Malini et al. 2003) therefore contributing significantly to high Mn load in domestic sewage. Small fractions of Cu and Ni may result from the increasing use of industrial products in households.

It must also be noted that the relatively high contribution of domestic sewage to the HM load could be attributed to the discharge of large quantum (~ 100 MLD at VRH) of untreated domestic sewage into the stream. Though conventional STPs are not designed to treat HMs explicitly, studies suggest that a large fraction of the HMs get concentrated in the sludge, preventing HMs from entering the surface water bodies downstream. Alternatively, the presence of non-negligible quantities of Cu, Mn, and Ni in domestic sewage alerts us to the need for better fecal sludge management in the STPs deployed for domestic sewage treatment in the future.

Monitoring protocols

While monitoring individual industrial units through regular or even surprise inspections is challenging, the problem of illegal effluent releases can be addressed (by KSPCB) by deploying the monitoring approach presented in this study.

We compared the ambient water quality monitoring data collected by KSPCB’s at sampling points closest to CHO (2011–12) with our results for two metals: Cr and Cu (see Appendix S3). The average and the maximum concentrations observed by KSPCB are significantly lower than the concentrations we observed during the study period.

Further, we found that the monitoring protocol adopted by KSPCB suggests collecting one grab sample during working hours every four months. We observed the lowest concentrations of HM in the samples collected during the daytime (11 am-6 pm) of the 24-h sample collection cycle. While additional errors in sampling and calibration cannot be ruled out, the data suggests that a biased monitoring protocol further undermines the regulator’s chances of assessing the impact of unit-level monitoring and enforcement strategies on the water quality of receiving waterbody. These insights are relevant in other parts of India, where the CPCB’s monitoring protocol to identify ‘polluted river stretches’ involves once-in-3-months grab sampling. The water quality criteria is based on the conventional water quality parameters with no coverage for various HMs. In addition, the protocol does not indicate the importance of data collection on stream flows and pollution source identification. Therefore we suggest the up-gradation of the existing protocol to capture the variations in HM concentrations from an urbanised catchment (Allan et al. 2008; Shaw and Mueller 2009).

Conclusions

Following the passing of legislation in the 1970s to regulate first generation (industrial point) sources of HMs and other industrial contaminants, the attention of scientists in the last few decades has focused on assessing the impact of second generation sources (surface runoff, domestic sewage, distributed sources, and sediments) on HM concentrations in water sources. As a result, despite the presence of point sources, regulators tend to attribute HM pollution in surface water bodies to second generation sources.

We propose a spatially distributed approach that accounts for both types of pollution sources and addresses the possibility of clandestine discharges through temporally intensive concentrations and load measurements. This protocol has enabled us to estimate the relative contribution of multiple sources of HMs in an urban stream for possibly the first time in India.

Our results indicate that (a) concentrations of HMs in the stream are consistently high, and concentrations exceed both the USEPA ambient quality standards and WHO’s irrigation water quality standard, (b) industrial point sources continue to be the major source of HMs, (c) surprisingly, untreated sewage, because of its high volume, is a significant (even if minor) contributor to the HM load, except for (Mn) that could be attributed to geogenic pollutants in groundwater, (d) contribution of distributed sources such as carwashes and sediments to HM load were found to be insignificant, (e) absence of HMs in defining ambient water quality standards and criteria for irrigation use of surface water, and (f) the protocols adopted for monitoring of HMs in streams by regulating agencies fail to capture peak in HMs concentrations at odd hours. In addition to having implications in the particular context of the Vrishabhavathy River in Bengaluru, this study significantly contributes to the research and environmental policy narratives in the rapidly urbanising and industrialising parts of the global South.

Empirically, this work highlights the plight of urban rivers in rapidly industrialising and urbanising parts of the global South. Our results for the Vrishabhavathy River may be somewhat extreme because the river originates within the city itself. In addition to Vrishabhavathy River, similar studies reporting the presence of HMs and other industrial contaminants have been reported in urban stretches of the Ganga (Aktar et al. 2010), Yamuna (Jamwal et al. 2011; Sen et al. 2011), Hindon (Suthar et al. 2009) and Musi (Sekhar et al. 2005) rivers. The contaminated water from these rivers is reused for irrigation downstream (Feenstra et al. 2000; Khan et al. 2008; Iqbal et al. 2013). Such results also continue to be reported elsewhere, including Sri Lanka (Herath and Amaresekera 2007).

From a policy perspective, our study highlights gaps at three levels: standard setting, monitoring, and enforcement. First, the results call for defining ambient water quality standards and fixing the “HM gap” in water use standards in India and elsewhere (Zhen-guang et al. 2013; for a detailed review, see Vareda et al. 2019). The results also demonstrate the need to rework discharge standards and set load-based standards for urban streams with zero natural flows in the dry season (Gunawardena et al. 2018). Second, the results indicate that beyond the lack of effective monitoring of individual industry discharges, there is a major loophole in ‘outcome monitoring’, i.e., measuring ambient levels of HM and other industrial contaminants in the river to be protected. Interestingly, while many studies report high levels of HMs or other industrial contaminants in rivers (Jamwal et al. 2016), few studies seek to explain why this continues despite monitoring by state agencies. If the purpose of regulation is to ensure healthy rivers, then (even if ‘healthy’ remains ill-defined in the law) the Vrishabhavathy River case epitomises the situation where poor (predictable, single-sample based, daytime) monitoring of outcomes hides the fact that ‘the patient is dead’. Monitoring protocols must provide for the possibility to capture the impact of the irregular and clandestine release of effluents, and rigorous ‘outcome’ (i.e., ambient water quality) monitoring can help identify locations that require greater enforcement and attention.

Third, environmental regulators in the global South should continue focusing their attention on the first generation sources, i.e., industries. The World Water Assessment Program in 2009 reported that around 70% of industrial wastewater in developing countries remains untreated (WWAP 2009, p. 141); our study suggests that since then situations may not have changed significantly. The second generation sources will become relevant only when the pollution from first generation sources is sufficiently addressed; highlighting/flagging them in the current scenario diverts the attention of practitioners and policymakers.

In rapidly and haphazardly industrialising towns/cities, regulating pollution from the thousands of small-scale industries in designated areas as well as outside presents a challenge for regulators. However, regulators tend to focus on ex-ante consent processing rather than ex-post monitoring of violations and prosecution (Lele et al. 2016). Identification of unregistered industries and greater inspection and prosecution, along with citizen participation (Rajaram and Das 2008), must be combined with reducing the compliance burden on small firms. This can be achieved by starting (or, in the Vrishabhavathy case, restarting) publicly managed CETPs and promoting small-sized heavy metal recovery technologies that simultaneously provide cost-saving and pollution reductions.

Methodologically, the study shows the usefulness of multi-pronged approach that (a) covered multiple seasons, day-night hours, multiple locations (covering both concentration and load estimation), (b) disaggregated multiple sources using a combination of mapping, source characterisation, and (c) analysed sediment samples to identify and quantify the contribution of sources (point and distributed) to HM load in the stream in the context of a developing country. This approach is especially important where the officially available data on ambient quality and pollution sources are unreliable, and sewage and industrial effluents are often untreated or partially treated. Implementing 24-h quality and flow monitoring is challenging and points out the need to develop innovative methods for real-time monitoring of HMs and other industrial contaminants. This could include using electrical conductivity as a proxy for HMs in real-time monitoring (Loock et al. 2015) and passive sampling devices such as Chemcatchers (Shaw and Mueller 2009) for measuring concentrations in streams. Multi-location monitoring, coupled with round-the-clock sampling, can be a parsimonious way of narrowing down the locations of non-compliant sources.

Supplementary Information

Below is the link to the electronic supplementary material.

Biographies

Priyanka Jamwal

is a fellow at ATREE. She broadly works in the area of water resource management with a focus on water quality. Her work focuses on identification of contaminant sources, modeling the fate and transport of contaminants in urban hydrological systems and assessing the risk to human health due to exposure to contaminants. Her empirical work has focused on quantification of microbial load from point and non-point sources in urbanising watersheds.

Divya Nayak

was a Research assistant at ATREE. Her research interests include Integrated Water Management and has supported projects on Contaminated Site Management, and Transaction Services.

Praveen Raje Urs

was a Lab analyst at ATREE. His research interests include monitoring fate and transport of contaminants in hydrological systems.

Mohamed Zuhail Thatey

was a Research Assistant at ATREE. His research interests include water quality, ecosystem services, forest and biodiversity management.

Malavika Gopinath

was a Research Assistant at ATREE. Her research interests include modelling the fate and transport of contaminants in surface and groundwater and sustainability science.

Mohammad Idris

was a Research Assistant at ATREE. His research interests include modelling the fate and transport of contaminants in surface and groundwater, Health risk assessment.

Sharachchandra Lele

is a Distinguish fellow at ATREE. His research interests include conceptual issues in sustainable development and sustainability and analyses of institutional, economic, ecological, and technological issues in forest, energy, and water resource management. He attempts to incorporate strong interdisciplinarity in his own research and teaching, which straddles ecology, economics, and political science.

Author contributions

PJ: Conceptualisation, data curation and analysis, writing—review and editing. DN: sample analysis, data curation and analysis: PRU: fieldwork and sample collection and analysis. MZ: fieldwork and sample collection and analysis. MG: fieldwork and sample collection and analysis, MI: fieldwork and sample collection and analysis, SL: conceptualisation, writing—review and editing, funding acquisition.

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.