The study area is located in West Kalimantan, Indonesia, near the city of Pontianak (0.09∘ N, 109.24∘ E) on the island of Borneo (Fig. 1). The climate is humid equatorial, with 2953±564 mm of rainfall and a mean annual temperature of 27 ∘C (1985–2017 data). The average monthly rainfall ranges from 170±126 mm (August) to 349±98 mm (November). The highest rainfall is measured from October to January. The mean rainfall is 274±123 mm for January and 199±106 mm for June (Fig. S1 in the Supplement). The study focused on the Ambawang River, which flows into the Landak River, which in turn flows into Kapuas Kecil River. It is a blackwater river draining a watershed (approximatively 706 km2) entirely covered with peatlands. This river was selected to represent water of exclusively peatland origin. All peatlands in the sampling area have been drained and converted to agriculture. Current land use consists of small-scale rubber plantations, secondary forest, oil palm plantations and human settlements.

Two sampling campaigns were conducted in June 2013 (drier period) and January 2014 (wetter period). Using a boat, samples were collected in the center of the river, from the origin of the Ambawang River (BR – blackwater river site) to its downstream confluence with the Landak and Kapuas Kecil (WR – whitewater river site). Whitewater river samples were collected upstream of the confluence with the whitewater river (WRu – whitewater river upstream). Drainage canals (DC) flowing into the blackwater river were also sampled during the second sampling campaign (Fig. 1). In January 2014, a rain collector was installed on the roof of the Pontianak's meteorological station to collect rain samples for lead isotopic analysis. In situ parameters (pH, conductivity and dissolved oxygen) were measured using a multiparameter probe (WTW, Germany). Depth profiles of dissolved oxygen in the blackwater river were also measured with an oxygen microelectrode (MI-730 dip-type micro-oxygen electrode and O2-ADPT adapter; Microelectrodes, Inc., Bedford, NH, USA). Frequent calibration was performed with a zero-oxygen solution and distilled water equilibrated to ambient oxygen concentrations, where temperature was carefully monitored. To create the zero-oxygen solution, 1 g of sodium sulfite (Na2SO3) and a few crystals (∼1 mg) of cobalt chloride (CoC12) were dissolved in 1 L of distilled water. For measurements of additional parameters, a larger volume of water was collected for further analysis. Samples were filtered immediately following collection on the boat using a portable peristaltic pump (Geotech, USA) and prebaked (5 h, 450 ∘C) and pre-weighed GF/F Whatman filters (0.7 µm); stored in glass bottles for DOC, δ13C-DOC and optical properties of DOM analysis; and acidified with HCl for DOC and δ13C-DOC. Samples were filtered with cellulose acetate filters (0.22 µm), acidified with HNO3, and stored in polypropylene vials for analysis of major nutrients and trace elements. DOC analysis was repeated on the cellulose acetate samples, and DOC concentrations did not differ significantly based on filtration at 0.2 or 0.7 µm. In January 2014, at selected sites (eight), samples were first filtered with GF/D filters (2.7 µm) to assess to the coarse colloidal fraction of trace metals and DOC.

Non-purgeable organic carbon (NPOC, referred to hereafter as DOC) was analyzed on filtered samples after acidification to pH 2 (HCl) with a TOC-V CSH analyzer (Shimadzu, Japan), with a quantification limit of 1 mg L-1. Major cations and anions were analyzed using high-performance liquid chromatography (HPLC; Dionex, USA). The quantification limit was 0.5 mg L-1 for chloride, nitrate and sulfates and 0.025 mg L-1 for ammonium, potassium, magnesium and calcium. Certified material (ion 915 and ion 96.4, Environment and Climate Change Canada, Canada) was included in the analytical loop, and recovery was >95 % of the certified value. For trace element analysis, samples were acidified with ultrapure HNO3 prior to inductively coupled plasma mass spectrometry (ICP-MS; 7500ce, Agilent Technologies) analysis. 115In was used as an internal standard. SLRS-4 (river water certified for trace elements) was used as a reference material on every run and accuracy (recovery > 95 %) was checked. Determination limits were <0.5 µg L-1 for Fe and Al; <0.05 µg L-1 for Ni, Cu and Zn; and <0.005 µg L-1 for Cd and Pb. Pb isotope ratios (206Pb/207Pb; 208Pb/206Pb) in water samples were analyzed using high-resolution ICP-MS (Thermo Element II XR; OMP service ICP-MS, Toulouse, France). Measurements were corrected for mass bias using individual sample bracketing with certified and adequately diluted NIST NBS-981 (100 to 500 ng L-1) according to Krachler et al. (2004).

The UV absorption spectra of pore water were measured with a spectrophotometer (Secoman UVi-light XT5) from 190 to 700 nm in a 1 cm quartz cell. The specific UV absorbance (SUVA) at 254 nm (L mg-1 m-1) was calculated as follows: SUVA = A254/(b⋅DOC) (Weishaar et al., 2003), where A254 is the sample absorbance at 254 nm (non-dimensional), b is the optical path length (m) and DOC (mg L-1). The baseline was determined with ultra-pure water. Potential additional absorbance related to Fe content was examined following the procedure described by Poulin et al. (2014). The additional absorbance was small and represented only 3.6±1.4 % of the total absorbance across all samples and was therefore neglected.

Emission excitation matrices (EEMs) were acquired using a Hitachi F4500 fluorescence spectrometer, and instrument-specific correction was applied. Emission spectra were acquired from 250 to 550 nm for excitation ranging from 250 to 550 nm. The slits were set to 5 nm for both the excitation and emission monochromators. The scan speed was 2400 nm min-1 and the integration response was 0.1 s. Fluorescence intensity was corrected from the excitation beam to ensure stability. The inner-filter effect water was taken into account using a dilution approach as developed by Luciani et al. (2009). The fluorescence index was calculated as defined by McKnight et al. (2001) and adapted by Jaffé et al. (2008): by the ratio of the fluorescence intensity at 470 nm to the fluorescence intensity at 520 nm for a 370 nm excitation. The shape of the excitation spectra was checked following the recommendation of Cory et al. (2010). The PARAFAC analysis (parallel factor analysis; Bro, 1997) was performed on all samples using the PROGMEEF program in MATLAB (Luciani et al., 2008).

The isotopic composition (δ13C) of DOC was determined at the UC Davis Stable Isotope Facility, following the described procedure (http://stableisotopefacility.ucdavis.edu/doc.html, last access: 30 June 2019). Briefly, a TOC analyzer (OI Analytical, College Station, TX) is interfaced to a PDZ Europa 20–20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK) utilizing a GD100 Gas Trap Interface (Graden Instruments).

Statistical analysis was performed using R (R Core Team, 2019) and the R studio software (Version 1.2.1335), using ggplot (Wickham, 2016), dplyr (Wickham et al., 2019) and dunn.test (Dinno, 2017) packages. Significant differences (p<0.05) between groups were evaluated using Kruskal–Wallis and Dunn's post hoc multiple test.

The observed water chemistry of the Ambawang River and drainage canals is typical of blackwater rivers draining peatlands (Table 1; Fig. 2) and does not show significant differences between the two sampling seasons. It is acidic, with a pH of 3.2±0.6 and 3.5±0.3 in the drainage canals and the blackwater river, respectively, has a low conductivity (DC: 89.8±21.4 µS cm-1; BR: 85.2±21.6 µS cm-1), is hypoxic (DC: 2.3±0.3; BR: 1.9±0.7 mg L-1), and has low nutrient concentrations (dissolved inorganic nitrogen < 0.3 mg L-1 and P-PO4 < 0.015 mg L-1) but high DOC concentrations (DC: 35.2±5.9; BR: 35.8±3.5 mg L-1). The Cl- concentrations are low and homogeneous (DC: 2.6±0.7; BR: 2.4±0.6 mg L-1). After the confluence with the whitewater river, the chemistry of the river radically changes. An abrupt increase in pH is observed (WR: 5.3±0.7). The dissolved oxygen concentration increases to 3.7±1.0 mg L-1, while DOC concentrations drop sharply to 9.2±3.2 mg L-1. We also observe a slight increase in NO3- and decrease in PO42-. Across all samples, the DOC concentrations show a significant negative correlation with DO concentrations (r2=0.63, n=40, p<4×109). In contrast, no increase in Cl- concentration is observed until close to the ocean (three samples corresponding to ocean water intrusion were excluded from Fig. 2).

No systematic differences are observed for the DOM characteristics between the two sampling campaigns. The δ13C signature of DOC (Fig. 3a) is very negative, reaching -30.3±0.4 ‰ in the drainage canals. It gradually and continually increases along the continuum from upstream in the blackwater river to the ocean (Fig. 3a). As a result, the δ13C of DOC in the drainage canals and the blackwater river is significantly more depleted than the whitewater river. The δ13C signature of DOC is significantly negatively correlated with DOC concentration (r2=0.68, p<10-4, n=40), with the highest DOC values being associated with the lowest δ13C-DOC values.

The SUVA index (Fig. 3b) has high values in the blackwater river (5.3±1.2), with the highest values measured upstream. A wide range of values is measured in the drainage canals (4.3±1.4). The SUVA values of the blackwater river are significantly higher than those measured in the Kapuas Kecil (4.5±0.3) and its tributaries (4.2±0.2). The fluorescence index has relatively high values for tropical peatlands, where most of the DOM is of terrestrial origin (Gandois et al., 2014; Zhou et al., 2019). The FI values vary widely in the drainage canals (1.60±0.17) and blackwater river (1.55±0.06) but are more uniform in the whitewater river, both upstream (1.55±0.03) and downstream (1.55±0.05) of the confluence with the blackwater river. Despite the high FI values, these two optical indices show coherent spatial patterns within the blackwater river and drainage canals (Fig. 3b and d). For example, lower SUVA values are associated with higher FI values in three drainage canals and in the blackwater river close to their connection, sampled during the second sampling campaign. Across all samples, a significant correlation is observed between FI and SUVA values (r2=0.37, p<10-4, n=41).

The EEMs of all water samples have two main peaks (Fig. S2). The primary peak (λex=250 nm, λem=460 nm) is coupled with a less intense peak (λex=350 nm, λem=460 nm). The peaks are typical of high molecular weight and aromatic molecules, which have been observed in wetlands (Fellman et al., 2009). The PARAFAC analysis reveals two fluorophores: C1 (λex=255 nm, λem=450 nm) and C2 (λex=285 nm, λem=485 nm; Fig. S3) The first component constitutes 60 %–73 % of the total fluorescence of samples. The relative contribution of these two fluorophores evolves along the sampled continuum, with the lowest values measured upstream in the blackwater river (Fig. 3c). The spatial evolutions of the C1 / C2 ratios and the δ13C-DOC values show consistent trends. A significant (r2=0.43, p<0.001, n=41) relationship is observed across all the samples between these two indicators. A stronger relationship (r2=0.85, p<0.001, n=5) is observed when the drainage canal samples alone are considered.

Blackwater rivers originating from drained peatlands have a unique composition of inorganic elements. The concentrations of trace metals (Pb, Ni, Zn, Cd) as well as Al and Fe are significantly higher in the blackwater river and drainage canals than in the whitewater river (Table 2; Fig. 4). For Al, Fe and As, high concentrations are measured in the blackwater river during the first sampling campaign (drier conditions). In contrast to other TMs, higher Cu concentrations are measured in the whitewater river. A principal component analysis (PCA; Fig. 5) of TM concentration and DOM properties reveals specific associations between DOC, Fe and As and to a lesser extent Zn and Cd, while another group is formed by Al, Pb and Ni. Copper shows no association with DOM but does show increased concentrations with higher FI. The first axis of PCA (load of DOC, Fe, As) strongly discriminates the blackwater river and drainage canal samples from the whitewater river.

The distributions of DOC and TMs are presented in Table 3. Dissolved organic carbon is mostly (>98 %) dissolved or in the form of fine colloids (<0.22 µm) along the entirety of the studied continuum. Iron and As are mostly present in dissolved form or as fine colloids in the blackwater river and drainage canals (>96 %). However, after transfer to the whitewater river, half of Fe and a third of As are present in the coarse colloidal form. Zinc and Cd do not show similar patterns. Aluminum is mostly present in the coarse colloidal phase (>60 %) in the blackwater river and drainage canals, and this proportion further increases in the whitewater river (>80 %). Lead is mostly present in the dissolved and fine colloid phase (>75 %) in the drainage canals and blackwater river and shifts to coarse colloidal (>60 %) forms after the confluence with the whitewater river. Nickel and Cu are mostly present in the dissolved and fine colloidal phase in the DC and BR but almost entirely in the coarse colloidal fraction in the whitewater river.

We observe distinct differences between the lead isotope ratios in the whitewater river and those in the blackwater river and drainage canals. A decrease in the 206Pb/207Pb isotopic ratio is observed with increasing Pb concentrations in the blackwater river but not the whitewater river (Fig. 6a). Furthermore, the biplot of the 206Pb/207Pb and the 208Pb/206Pb signatures illustrates significant differences between the white water and blackwater river and drainage canal groups (Fig. 6b).

We observe in-stream processing of DOM, but the total DOM exported from tropical peatlands exceeds the processing capacity of the rivers which drain them and a large proportion of DOM is transported to the ocean. We find persistently high DOC concentrations in both drainage canals and blackwater rivers draining degraded peatlands, consistent with the range of previously reported values in Borneo (Moore et al., 2011; Cook et al., 2018) and in the upper range of blackwater rivers in Sumatra (Baum et al., 2007; Rixen et al., 2008). We also find indicators of in-stream processing of DOM. The transformation of DOM we observe along the continuum is likely primarily due to photo-oxidation, with a smaller contribution from microbial processing. We observe an increase in the δ13C-DOC values along the studied continuum (Fig. 3a). This shift toward higher δ13C-DOC is correlated with an increase in the C1/C2 ratio of PARAFAC fluorophores (Fig. 3c). The two fluorophores are typical of terrestrial input of DOM (Yamashita et al., 2008) and similar to observed fluorophores in other blackwater rivers in Borneo (Harun et al., 2015; Zhou et al., 2019). An increase in this C1/C2 ratio reflects a shift toward lower wavelengths and therefore toward lower aromaticity and lower molecular weight (Austnes et al., 2010; Zhou et al., 2019). Moreover, a decreasing trend in SUVA values is observed along the continuum (Fig. 3b). These observations indicate that at our site, aromatic features are preferentially processed in-stream, consistent with a dominant effect of photo-oxidation (Amon and Benner, 1996; Spencer et al., 2009; Sharpless et al., 2014). This has also been observed in the Congo River, where photo-oxidation led to an increase in δ13C-DOC and a decrease in aromatic features (Spencer et al., 2009).

However, photo-oxidation is not the only process responsible for the processing of DOM. The low oxygen levels in the blackwater river and drainage canals and the significant relationship between DOC and DO concentrations suggest that nearly all oxygen entering the well-mixed water is quickly consumed by DOM oxidation (Fig. 2a and b). Furthermore, the sharply decreasing oxygen profiles measured in the blackwater river suggest that the transformation of DOM is restricted to the shallow surface layers of these waters (Fig. S3). Additionally, localized increases in fluorescence index, coupled with decreases in SUVA (reflecting a higher proportion of microbial derived DOM; Fig. 3d), suggest that microbial processing occurs in some locations in drainage canals. Both photo-oxidation and microbial processing have been quantified in laboratory experiments for DOM originating from tropical peatlands. Martin et al. (2018) found that up to 25 % of riverine DOC from a blackwater river in Sarawak, Malaysia, was lost within 5 d of exposure to natural sunlight. Microbial long-term incubation studies by Rixen et al. (2008) showed that 27 % of DOC was degraded after 2 weeks. In blackwater rivers, it is likely that in-stream microbial processing of DOM is limited by the low oxygen concentrations, low pH and low nutrient levels (especially inorganic nitrogen; Wickland et al., 2012) rather than intrinsic refractory characteristics. Although the precise extent of in-stream processing cannot be quantified here, our results are consistent with in-stream transformation of DOM by photo-oxidation as well as some contribution of microbial degradation in the shallow surface layers. In the future, quantitative assessment of outgassing in tropical peatland drainage canals would improve the evaluation of carbon release following peatland drainage. Overall, more work is needed to understand the extent of upstream processing of peatland DOM.

This study provides the first record of trace metals in blackwater rivers originating from degraded tropical peatlands. We observe strong enrichment of Al and Fe as well as Pb, As, Ni and Cd in peat-draining waters. The measured concentrations are comparable to those measured by Kurasaki et al. (2000) in Borneo rivers for Pb, Zn, Cu and Cd but significantly higher (5 to 10 times) for Fe. The concentration levels, however, remain low compared to highly impacted regions of Indonesia (Arifin et al., 2012). The elevated concentrations of Al and Fe in water draining tropical peatlands is consistent with existing observations of elevated Fe concentrations from blackwater rivers in the tropics (Zhang et al., 2019) and northern peatlands. This enrichment is likely due to the weathering of mineral material under the peat during peat accumulation processes (Tipping et al., 2002; Pokrovsky et al., 2005). As a consequence, in water draining peatlands, strong organo-mineral associations between DOM and Fe (Krachler et al., 2010, 2012; Broder and Biester, 2015) and DOM and Al (Helmer et al., 1990) have been observed. These colloidal associations between DOM and Al and Fe in the form of hydroxides strongly control TM transfer and speciation in peat-draining waters (Tipping et al., 2002). In the present study, specific associations of trace metals with Al and Fe are observed, including strong links between Al and Pb and Ni. However, the lack of a direct relationship between Pb and DOM contrasts with reported observations in the literature (Graham et al., 2006; Pokrovsky et al., 2016; Jeremiason et al., 2018). Despite this, we do observe strong links between Fe, As, Zn, Cd and DOM, which have been previously reported in water draining peatlands (Neubauer et al., 2013; Broder and Biester, 2015; Pokrovsky et al., 2016). The coupled dynamics of Fe and As might be related to similar mobilization processes within the peat column, with the sorption of As to Fe(III)-(oxyhydr)oxides (ThomasArrigo et al., 2014) in anoxic peat water. Widespread drainage of tropical peatlands and the corresponding release of anoxic water to surface water networks could induce a coupled increase in DOM and Fe concentrations, similar to that which has occurred in Sweden (Kritzberg and Ekström, 2012).

The isotopic composition of Pb in peat-draining water strongly suggests that it is of anthropogenic origin. The isotopic signatures measured in river samples are a combination of the signature of undisturbed soils of Borneo (Valentine et al., 2008) and a mix of both present and past anthropogenic inputs. Older anthropogenic inputs are reflected by the signature of atmospheric deposition from Java aerosols (Bollhöfer and Rosman, 2000), while the signature of recent regional anthropogenic inputs was characterized by rain samples collected in Pontianak as part of this study (Fig. 6b). In the blackwater river and drainage canals, the isotopic ratio is close to that of aerosols and recently sampled rainwater and is dominated by anthropogenic inputs, whereas the isotopic ratio in the whitewater river is closer to the natural signal (Fig. 6). This isotopic difference is consistent with the difference between the watersheds drained by these two rivers: tropical peatlands are ombrotrophic systems, and the trace metal content in peat soil is derived from the atmosphere (Weiss et al., 2002), whereas the Kapuas Kecil is recharged from a larger watershed and reflects contribution of mineral soils. Tropical peatlands can serve as secondary sources of atmospheric pollutants to the environment. With peatland drainage, blackwater rivers release the accumulated atmospheric deposition over hundreds of years on much shorter timescales. For example, the isotopic signature observed in the blackwater river reflects anthropogenic sources deposited at different times, including older deposition such as the lead measured in the Java aerosols (Bollhöfer and Rosman, 2000), and more recent deposition following the widespread introduction of unleaded fuel (characterized by samples collected from rainwater during the January 2014 sampling period in this study). This release of lead by degraded tropical peatlands has the potential to impact records from environmental archives, for example the corals of the Singapore Strait (Chen et al., 2015). Although this is the first measurement of the aquatic release of trace metals from tropical peatlands, the role of tropical peatlands as a secondary source of contaminants has also been highlighted by the trace metal content analysis of dust emitted to the atmosphere by peat fires (Betha et al., 2013).

Sharp changes in physico-chemical conditions are observed after the mixing of the black and the whitewater river, including sharp increases in DO concentrations and pH values. This strongly controls the transport of DOM and TMs drained from degraded tropical peatlands. After the confluence with the whitewater river, DOC concentrations decrease abruptly. This decrease primarily results from the dilution of the blackwater river signal. However, the sudden elevation of pH and DO after the confluence might create favorable conditions for microbial processing of DOC, making the mixing zone a likely hotspot of GHG emissions (Palmer et al., 2016). This would also be consistent with the decrease in the SUVA index observed after the confluence. Despite processing of DOM along the continuum, a significant proportion of DOM originating from degraded peatlands actually reaches the ocean. We observe high DOC concentrations at all sampling locations, with concentrations remaining high even close to the ocean (Fig. 2a). Additionally, the results of our physical fractionation show that even close to the estuary, DOC remains in the dissolved and fine colloid form (<0.22 µm) and that flocculation processes might be limited. The large areas of coastal peatlands in the region might explain the relatively high fluvial organic carbon export to the South China Sea (Huang et al., 2017). The decrease in trace metal concentrations after the confluence might be influenced by shifts in physical fractionation and an increased proportion of colloidal form. This is especially true for Al and Pb. Some flocculation at the estuary might limit their transfer to the ocean. For Fe and As, a higher proportion remains in the form of fine colloids after mixing with the whiter river and is still associated with DOC. Similar conservative behavior of low-molecular-weight organic molecules associated with Fe was observed at the outlet of northern peatlands (Krachler et al., 2012) and in Arctic rivers (Pokrovsky et al., 2014). This highlights the finding that dissolved organic molecules derived from tropical peatlands can also act as carriers of trace metals to the ocean.