Introduction

The del Plata Basin—the second largest in South America and comprising Argentina, Uruguay, Brazil, Bolivia, and Paraguay—contains as its principal rivers the Paraguay (2459 km long), the Paraná (4352 km long), and the Uruguay (1600 km long), with the last two flowing into the widest estuary of the world, the Río de la Plata (256 km in width). The Paraná is the sixth largest river in the world, with a basin of 1,500,000 km2, a mean annual discharge of 17,000 m3/s, and a suspended load of 118.7 million tons per year (Orfeo and Stevaux 2002). The basin traverses a variety of geological features, including the Andes Mountains, the Chaco-Pampean Plains, the Eastern Plains, the Jurassic-Cretaceous Area, and the Brazilian Shield (Iriondo 1988). The bottom sediments of the rivers are dominated by silt and clay particle sizes (Manassero et al. 2008), with vast amounts of colloids and clay aggregates circulating in the basin (Konta 1985). Most of the Argentine industrial and agricultural activities and population settlements are associated with this basin. Four previous monitoring campaigns (2004, 2005, 2006, and 2009) of the Argentine sector of the Basin have revealed multiple sources of pollution along the Basin (Peluso et al. 2013a; Ronco et al. 2011; SAyDS- OPS- PNA- UNLP 2007). The middle and lower Paraná tributaries traverse areas subjected to extensive agriculture (Marino and Ronco 2005; Ronco et al. 2008, 2011). That Argentina is the tenth largest agricultural nation in the world, with 31 million ha devoted to agriculture that account for 2.2 % of the world’s total area under cultivation is indeed notable (Leguizamón 2014). Pesticide use increased in the last several decades as agriculture became gradually transformed into a system of high technology in order to satisfy growing demands (Heinemann et al. 2013). The adoption of transgenic crops engineered to tolerate the broad-spectrum herbicide glyphosate [N-(phosphonomethyl)glycine] gradually increased over the last two decades, with over 22 million ha of the cultivated land being occupied by glyphosate-tolerant corn and soybeans in the 2012–2013 growing season (CONABIO 2013). Accordingly, the two most common crop rotations of the Argentine pampas (i.e., corn vs. full-season soybeans and corn vs. wheat vs. short-season soybeans) usually require two or three applications of glyphosate (Bindraban et al. 2009). The presence of the herbicide in surface waters near agricultural fields has been documented (Aparicio et al. 2013; Lupi et al. 2015; Peruzzo et al. 2008; Primost 2013; Ronco et al. 2008), but no information on the impact of this compound at the level of the large basins of the region is currently available. Consequently, the present study was conducted first in order to analyze the main water quality parameters and sediment composition of specific sampling locations—with an emphasis on the contents of the herbicide glyphosate and the metabolite aminomethylphosphonic acid (AMPA)—and then to discuss the results obtained. Finally, we analyzed the occurrence of these contaminants in environmental compartments by means of the composition of the sediment-matrix components and regional agricultural activities.

Materials and methods

Study area and sample collection

The sampling campaigns were conducted in the scientific vessel SPA-1 “Dr. Leloir” of the Prefectura Naval Argentina during the months of June through July of 2011 and 2012. Separate water and sediment samples were collected from 23 specific sites in the Paraguay and Paraná rivers and their tributaries (Fig. 1 and Table 1). An Eckman dredge was used to take the sediments—down to an approximate depth of 10 cm, with each composite sample being obtained from at least five different grab samples per site. Twenty of the samples were taken in the confluence of the principal tributaries with the Paraguay or Paraná Rivers, while the rest were from the main Paraná watercourse. Table 1 provides the identification of the sampling sites, the corresponding abbreviations, and a brief description.

Fig. 1
figure 1

Study area and sampling site locations. Table 1 lists the names of the sites and their principal geographic and ecologic features

Full size image
Table 1 Description of the sampling sites, in situ measured water quality parameters and solids (average data of the campaigns 2011–2012). Tributaries of Paraguay and Paraná rivers were sampled nearby each respective confluence
Full size table

Whole water and sediment samples to be used for the main matrix composition parameters were kept in a cooler at 4 °C during transfer to the laboratory. The analysis of glyphosate and AMPA was performed on that water and also on a suspended fraction (in the sampling campaign of 2012). The whole water was placed into 100-mL polypropylene bottles and spiked with 500 ng of 13C,15N-glyphosate (13C, 15N-GLY). Filters containing the suspended fraction (obtained from the filtration of 100 mL whole water through 45-mm, 0.45-μm nitrocellulose filters in situ during the sampling) along with 10 g of sediment samples were also spiked with 30 or 250 ng of the 13C,15N-GLY, respectively. Upon arrival, all samples for chemical analysis of the herbicide were stored at −20 °C until the time of processing.

Physical-chemical analysis

The physical and chemical variables of the water column from each sampling site (i.e., conductivity, transparency, pH, temperature, and dissolved oxygen) were measured in situ at each sampling station (multiparameter water quality monitor HORIBA U-52™). Water analyses were conducted in the laboratory according to the following methodology (APHA 1998): Alkalinity was determined by titration (method 2320); chlorides by potentiometry (method 4500-Cl-D); sulfates by turbidimetric analysis (method 4500-SO42-E); nitrate by cadmium reduction (method 4500-NO3-E); total suspended solids and dissolved solids by methods 2540-D and C, respectively; calcium by EDTA titration (method 3500-Ca-B); and magnesium by calculation (method 3500-Mg-B). In nitric acid-digested samples (method 3030-E) sodium plus potassium along with iron, manganese, and zinc were determined by atomic absorption spectrometry (direct air-acetylene flame; methods 3111 and 3500, respectively).

The physical characteristics of the sediment samples recorded consisted in grain size and organic matter. Sieving and settling velocity with previous cement removal (Day 1965) was employed for grain size analysis after sediments were passed through a set of standard sieves larger than 63 μm to separate the sands. The organic matter content in the sediment was determined by calcination (loss on ignition, LOI) in a muffle furnace at 550 °C (APHA 1998; Heiri et al. 2001). Sulfides were analyzed according to method 9030 (USEPA 1996), while the total phosphorus content was measured by a colorimetric method following ignition of the sample (Andersen 1976). Finally, an analysis of the metal composition was conducted by atomic absorption spectrophotometry (as in the water analyses) following acid digestion of the samples (method 3050, USEPA 1996).

Analysis of glyphosate and AMPA

Glyphosate and AMPA were analyzed by high-performance liquid chromatography and mass spectrometry (HPLC-MS) following derivatization with 9-fluorenylmethoxycarbonyl chloride (FMOC-CL; Sancho et al. 1996). Water samples were pretreated by adjusting 2-mL aliquots to pH  9 with sodium tetraborate (40 mM), followed by the immediate addition of 2 mL of FMOC-Cl solution in acetonitrile (1 mg/mL). The solution was then kept overnight in the dark at room temperature (Ibáñez et al. 2005). The extraction of the nitrocellulose filters containing the suspended fraction was effected by sonication in 3 mL of 100 mM K2HPO4 (Miles and Moye 1988), followed by centrifugation. Next, 2 mL of the extract were used for derivatization with FMOC-Cl as in the water samples. Of the wet sediment, 10 g was extracted with 25 mL of 100 mM K2HPO4 by sonication for 15 min. The derivatization with FMOC-Cl was performed on a 2-mL centrifuged sample adjusted to pH 9. The extractions and derivatizations of the different types of samples were carried out in polypropylene tubes (Miles and Moye 1988). The preparation of standard solutions for constructing calibration curves was done under operational conditions equivalent to those used for the testing of the samples. All derivatized samples were finally extracted with 5 mL dichloromethane and centrifuged and the aqueous phase passed through 0.45-μm filters for HPLC-MS determination. The runs were carried out with pressure generated from a binary Agilent 1100™ pump (Agilent Technologies Inc., Miami, FL, USA) coupled to a mass-quadrupole VL mass spectrometer with an electrospray ionization source (ESI Agilent Technologies Inc., Miami, FL, USA). Chromatographic separation was performed in a C18 X-SELECT™ column (75 mm × 4.6 mm and 3 mm pore size, from Waters Corp., Milford, MA, USA) kept at 25 ± 1 °C. A methanol-water gradient was used (with mobile phases previously conditioned with 5 mM ammonium acetate) at a flow rate of 0.5 mL/min. Nitrogen was applied as an auxiliary gas at 8 L/min and 330 °C. According Meyer et al. (2009), selected ion monitoring in the negative ionization mode was applied for the detection of GLY-FMOC, 13C,15N-GLY-FMOC, and AMPA-FMOC. The ion settings corresponded to the deprotonated compounds and two daughter ions for quantification and identification, respectively. Data acquisition and analysis were conducted by means of Agilent Chemstation LC-MSD Rev 10A.02.

The solvents used in the chromatography were of HPLC grade, while the salts were analytical grade (JT Baker-Mallinckrodt Baker Inc., USA). Nanopure water was obtained in the laboratory by means of a Sartorius Arium water purification system (Sartorius AG, Göttingen, the Netherlands). Standards of glyphosate (99 %) and AMPA (98.5 %) were acquired from Sigma Aldrich (St. Louis, MO, USA).

Quality controls and quality assurance

Quality controls during the analysis of the major components involved the use of reagent blanks, duplicate samples, and certified reference material (Pond Sediment 2, National Institute for Environmental Studies, Yatabe, Tsukuba, Ibaraki, Japan). The analysis of reference materials provided results at an accuracy between 80 and 95 %. The chemicals for sample treatments or the analysis of major matrix components were of analytical grade. The certified standards of metals (1.00 g/L standard stock solutions) were obtained from Accu Standard, Inc. (New Haven, CT, USA).

The quality control and assurance in the analysis of glyphosate and AMPA was done according to the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use guidelines (ICH 2005). The linearity, reproducibility, detection and quantification limits, matrix effect, and recovery were accordingly tested. Isotopically labeled glyphosate (13C,15N-GLY) was used to evaluate the holding time and recovery for the complete procedure.

Analysis of data

The principal water components were plotted in a Piper diagram by means of the program Diagrammes: Logiciel d’hydrochimie (http://www.lha.univ-avignon.fr). The type of distribution of sediment data was verified using a normality test followed by principal components analysis (PCA) of the chemical contents of the matrix components Fe, Mn, Cu, total phosphorous, sulfides, fines, LOI, and the concentrations of glyphosate and AMPA. Additionally, Varimax rotation to increase the participation of the variables with higher contribution and by simultaneously reducing that of the variables with lesser contribution was applied (Helena et al. 2000). The PCAs were done by loading the main variables and using the biplot of factor scores for the sampling sites in order to correlate both types of data. Significant factors were selected based on the Kaiser principle of accepting Eigen values >1 (Quinn and Keough 2002), and the component loading of the matrix correlation was considered to be significant for values that exceeded a critical level for n = 23. Concentrations below the detection limit were replaced with a value of one-half of the corresponding limit. Statistical analysis was performed by means of the XL-STAT software (Addinsoft 2005, version 7.5.3).

Results

Physicochemical parameters

The water composition of the tributaries (Table 1 and Fig. 2) exhibited a wide variability in the total dissolved solids at from <100 mg/L (the Paraguay and Pilcomayo rivers) up to 2000 mg/L (the Salado and Carcarañá rivers). The piper diagram in Fig. 2 shows the percentages of the main ions and their classification according to distribution, with those plots mostly indicating a central tendency for bicarbonated, sulfated, and chlorinated water and for calcium- and sodium-rich water. The concentration ranges in milligram per liter were as follows: bicarbonate, 24–326; sulfates, 11–1800; chlorides, 2–1250; nitrates, 0.05–2.8; calcium, 7.7–170; magnesium, 0.9–15.4; sodium, 3.3–1100; and potassium, 2–25. The extreme low and high values correspond to the Paraguay-Paraná main course and the Salado-Carcarañá, respectively. The suspended solids were ubiquitous in the basin, with most tributaries and the main Paraná watercourse containing over 100 mg/L and with the Arrecifes or Bermejo rivers reaching almost 1000 mg/L. The Bermejo River was the tributary that contributed the most solids to the basin owing to the river’s high flow rate and type of provenance materials within its upper stretch (Manassero et al. 2008). That the Carcarañá River exhibited the lowest content of total suspended solids (at 25 mg/L), with most being present in the dissolved form, is noteworthy. Bicarbonate dominated the tributaries of the lower stretch, although that anion was variable and lower in other sectors relative to the chlorides and sulfates (e.g., in the Salado and Carcarañá rivers). The content of iron and manganese (in whole water) was between 0.6 to 8.9 and 0.005 to 0.3 mg/L, respectively, with the Bermejo River having the highest contents of both metals. All values detected were within those observed in the earlier monitoring campaign that we carried out (Ronco et al. 2011) and the data reported from recent studies on the middle stretch of the Paraná (Campodonico et al. 2015).

Fig. 2
figure 2

Piper diagram of major water components

Full size image

Table 2 summarizes the data on sediment characterization. Most samples were composed of fine particles (over 50 % content and up to 99 %), with a dominance of silts that could be classified as silty loam. The exceptions were samples from the main Paraná watercourse: those exhibited lower contents of fine particles, organic matter, and sulfides. Samples from the sandy banks of the left margin tributaries also exhibited low levels of those parameters. The average content of organic matter varied from 0.2 to 9 % according to the LOI values. The sampling sites corresponding to the upper stretches of the basin tributaries tended to exhibit LOI values below 3 %, with the sole exception of S8. The tributaries of the lower stretch of the Paraná contained higher levels of LOI that became as high as 9 %. A similar pattern was observed in the sulfide contents of the sediments, where lower levels were present in the tributaries of the upper stretch of the Paraná (with almost all having <50 mg/kg) and higher levels in the lower stretch. A very high content of sulfides was detected in the mouth of the Río Saladillo, where the values reached 1907 mg/kg. The total phosphorus content did not follow any clear pattern, with concentrations ranging between 10 and 1162 mg/kg. The iron composition of the samples ranged from 1 to 15 g/kg, followed by Mn and Cu, at concentrations between 24 and 480, and <1 and 22 mg/kg, respectively. Lower levels of metals were observed in sediments from the main course of the Paraná and from the sandy banks of the left-margin tributaries (i.e., the Corrientes and Guayquiraró rivers).

Table 2 Physical-chemical measurements of grain size (%), organic matter (OM, %), and sulfide, total phosphorus (TP) and metals (mg/kg dw) in bottom sediments (average values from the two campaigns)
Full size table

Glyphosate and AMPA

The method of HPLC-MS for the analysis of glyphosate and AMPA in the environmental matrices studied produced satisfactory results (giving linear values between 1 and 2000 μg/L, with a correlation coefficient > 0.99, a relative standard deviation between 2 and 10 % for both analytes, and levels of detection (Table 3) comparable to those reported for similar types of samples (Aparicio et al. 2013; Primost 2013). The recoveries of 13C,15N-GLY were over 95 % for whole/surface water, between 90 and 95 % for water-suspended particles (over the entire procedure), and between 75 and 85 % for the sediments. The results provided in the tables were corrected by each corresponding recovery factor upon consideration of the expected and measured concentrations of 13C,15N-GLY. The sediment samples containing higher levels of organic matter affected the detection of glyphosate and AMPA during the ionization step of the analytical procedure causing a suppression of the signal down to 80 % (Taylor 2005); but the addition of the 13C,15N-GLY internal standard to each sample enabled an effective correction to be introduced for that damping of the detector response (Ibáñez et al. 2005).

Table 3 Glyphosate and AMPA concentrations in the sediments, whole water, and suspended matter of the sampling sites
Full size table

Only 15 % of the samples contained detectable concentrations of glyphosate in the water at an average value of 0.60 μg/L (Table 3). AMPA was not detected in these samples. The detection sensitivity in water-suspended matter, being about one order of magnitude more sensitive than for the whole water, enabled the quantification of glyphosate and AMPA in 39 % and 9 % of the samples, respectively, in the 2012 sampling campaign. In this instance, the average detectable concentrations were 0.1 and 0.04 μg/L for the two compounds, respectively.

Analysis of the sediment samples (Table 3) indicated detectable levels of both compounds mostly in the middle and lower stretch tributaries of the basin sectors under investigation, with 39 and 17 % of the samples being positive in the 2011 campaign and 35 and 17 % in the 2012 for glyphosate and AMPA, respectively. The mean concentrations detected were 742 and 521 μg/kg, respectively; while the maximum, minimum, and mean glyphosate-to-AMPA ratios were 7.80, 0.06, and 2.76, respectively. The highest concentration was detected in the Luján River, containing over 3000 μg/kg of glyphosate and 5000 μg/kg of AMPA.

Analysis of data

Since the sediment data set obtained did not adjust to a normal distribution (α = 0.05), nonparametric statistics were used and the PCA was solved through the use of a Spearman correlation matrix at a threshold value of 0.415. The variables contributing most to factor 1 were glyphosate, AMPA, sulfides, organic matter, Cu, and Mn (Table 4), with those parameters explaining 36.8 % of the variance. Figure 3 shows that the majority of the sediment samples grouped together and were distributed along factor 2, except for S16 and S23, which sites were sequestered along factor 1. These latter two samples exhibited the highest contents of glyphosate, AMPA, sulfide, and Cu, thus differentiating them from the other much greater group of sediment samples—which, for their part, were mostly correlated with the contents of fine particles, total phosphorus, and Fe. In addition, Table 5 indicates a satisfactory correlation between the content of glyphosate plus AMPA and the high concentrations of sulfides, Mn, and Cu in the sediments (cf. rows 1 and 2 of the table vs. columns 5, 8, and 9). The pairing of Cu and sulfide, moreover, exhibited a substantial correlation at a value of 0.815.

Table 4 Coordinates of the variables analyzed for the first two factors after Varimax rotation, the Eigen values of each factor, and the percentages of the explanation and the accumulated variance
Full size table
Fig. 3
figure 3

Principal components analysis (with Varimax rotation) biplot of variables and the study sites for the first two meaningful principal components

Full size image
Table 5 Correlation parameters of concentrations data. Numbers in italics indicate significant differences for an α = 0.050 (bilateral test)
Full size table

Discussion

The Paraná Basin is primarily affected by pollution in the middle and lower stretches of the Argentine portion, during which sectors the most industrial activities and urbanization predominate. Within this part of the basin tributaries run through areas of extensive agriculture with genetically modified Roundup-resistant crops—with the production in tonnage being 10,000-fold higher than those of the upper stretch. Previous studies on a local scale in surface water streams of this core soybean production area revealed glyphosate levels ranging from undetectable to up to 700 μg/L in water and from undetectable to as much as 1850 μg/kg in sediments (Peruzzo et al. 2008). Recently, Primost (2013) detected glyphosate and AMPA in a study on a local scale in small surface water streams traversing soybean-cultivated areas flowing into the de Galeguay River of the Entre Ríos Province within this same basin, with the research being performed during the spraying season. For the two compounds, the authors observed the following concentration ranges in water and sediments, respectively: 0.2–1.8 for glyphosate and 0.1–1.9 μg/L for AMPA and 52–3294 for glyphosate and 180–7220 μg/kg for AMPA. That these latter data from streams influenced by direct farming activities exhibited similar concentration levels of the herbicide in the sediments to those detected in the distal-tributary sediments in which our study demonstrates the impact of agriculture on the main Paraná watercourse. In addition, the monitoring of several rivers, canals, and streams from the southeastern sector of the Buenos Aires Province, an agricultural area within the region, revealed detectable levels of glyphosate and AMPA in 35 and 33 % of the water samples, respectively, and in 66 % of the sediments, at respective concentration ranges of 0.5–4 and 0.5–2.3 μg/L for glyphosate and 5.7–221 and 5.1–235 μg/kg for AMPA (Aparicio et al. 2013). Furthermore, Lupi et al. (2015) found similar frequencies of detection in water and sediment samples in the same region, although with concentration levels of both glyphosate and AMPA in water ranging from nondetectable up to 0.5 μg/L. Sediments from the preapplication period indicated relatively lower pesticide levels, with concentrations of glyphosate ranging from 7 to 14 μg/kg and AMPA from 2 to 13 μg/kg, consistent with the less intensive agriculture at that time.

Degenhardt et al. (2012) have observed in wetland environments that degradation and sorption were the more main pathways for the dissipation of glyphosate from the water column. Our findings also demonstrate a similar pattern in the partitioning of glyphosate and AMPA into the sediments. For example, in the present study, the Luján River, where the respective concentrations of the two compounds were determined for all compartments, the glyphosate levels were three to four orders of magnitude higher in the sediments than in the whole water. Furthermore, statistical analyses demonstrated that the concentrations of glyphosate and AMPA were well correlated with the sulfide levels in the sediment samples (Table 4; Fig. 3; also cf. Table 5, rows 1 and 2 vs. column 4). The inputs of sulfur-containing additives (e.g., ammonium sulfate, gypsum) used with glyphosate during the weed control practices could explain these results (Nalewaja and Matysiak 1993; Faircloth et al. 2004). After entering the environment, sulfates are reduced to sulfides bacteriologically in the anaerobic compartments of the aquatic systems—particularly in those rivers that are recipients of high loads of organic matter such as the Río Saladillo and the Luján tributaries (Peluso et al. 2013a, b; Ronco et al. 2011), where the concentrations of sulfides, glyphosate, and AMPA are accordingly high. Moreover, according to the United States Environmental Protection Agency (USEPA 2008), the half-life of glyphosate in anaerobic sediments is two orders of magnitude longer than in the aqueous compartment, where the dissipation is extremely rapid. This differential degree of degradation is in agreement with our results. Sediments with higher loads of sulfides are necessarily located in the anaerobic compartments that as such provide conditions of oxidative stability and therefore contain higher levels of the herbicide.

An association of the herbicide and its metabolite was also observed in relation to Cu (Table 5, cf. rows 1 and 2 vs. column 9). Morillo et al. (2000) had reported that the presence of Cu enhanced herbicide adsorption to soil particles via Cu-glyphosate complexes. Sediment materials originating from soils containing such complexes could explain the association of the herbicide observed with the bottom sediments (e.g., in the Saladillo and Luján rivers). Furthermore, the organic matter, total phosphorus, Cu, and sulfides in the sediments were well correlated. Peluso et al. (2013a), upon taking into account the combined biologic effects (as assessed by toxicity tests and benthic community indices) along with the physical chemistry of the sediments from the same tributaries of the Paraná Basin, demonstrated the role of sulfides in contributing to toxicity in the samples they tested. The sediment from the Saladillo River, in particular, exhibited a high and acute lethal toxicity against the test organisms used.

Conclusions

This study provides the first survey of glyphosate and its metabolite AMPA in water and sediments from the Paraná River on a basin-wide scale. The results revealed higher levels of these compounds in the middle- and lower-course tributaries in accordance with the intensive agriculture in those regions. Bottom sediments were found to be the main sinks of both compounds in the sampling sites investigated. The high flow rates and dilution capacity of the main watercourse of the Paraná attenuates the tributary inputs, although AMPA could be detected in sediments and suspended particulate matter throughout the river’s lower stretch. These results provide information that is highly relevant to the management programs and environmental control of watercourses in the Paraná Basin, indicating the need for particular attention to agricultural practices associated with the chemical control of pests in the agronomically productive areas of the region.