Aldrin, Dieldrin, Endrin, Endrin Ketone, Endrin Aldehyde, and DDT isomers (selected OCPs) were widely used as pesticides between the 1950s and 1970s [1] . Concerns about toxicity, persistence, and bioaccumulation of OCPs motivated bans in various locations around the world [2] . OCPs have high carbon/water partition coefficients, high hydrophobicity, strong affinity for organic matter, and rapid sorption kinetics with soil particulates. Residual concentrations of OCPs are commonly found in soils and groundwater at sites where pesticides have been formulated and stored and from agricultural runoff where soil erosion facilitates migration into proximal surface water bodies.

This study examines ambient background of OCPs (specifically Dieldrin, Endrin, and DDT) in water samples collected from the Atibaia River in Sao Paulo, Brazil. Regular monitoring of the Atibaia River water quality demonstrated little to no historical OCP detections for several years [3] . Starting in 2021, the laboratory adopted more modern instrumentation and detectors for water analysis using EPA Method 8081B [4] for quantification of OCPs at lower reporting limits in order to comply with the regulatory surface water quality criteria ( Table 1 ). The conventional quantification limits were reduced from 0.010 µg/L to 0.001 µg/L for Aldrin, Dieldrin and DDT and to 0.004 µg/L for Endrin. The lower detection limit method resulted in more frequent trace detections for OCPs among the surface water samples and motivated this study.

The Atibaia River basin drains an area of 2837.3 km² with the total population of the river basin of approximately 370,000 inhabitants [5] . The 7Q10 minimum flow of this river is 902,016 m³/d [5] with peak flows approaching 23,000,000 m³/d after heavy rainfall [6] . The river basin supports a wide range of land uses ranging from relatively pristine natural environments to extremely urbanized areas, including industrial and agricultural land use [7] . Runoff and the high energy flow readily erodes soil and resuspends sediment, giving the surface water a “muddy” appearance with high suspended solids ( Figure 1(A-B) ). Preliminary inspection of the raw laboratory instrument data files from the 2022 sampling results ( Figure 2 ) indicated a noisy chromatographic baseline that suggested the possible presence of non-pesticide interferences associated with chemicals in the resuspended river particles that recently became more apparent when the detection limits were lowered. In order to be considered a positive match, the peak of interest needs to be 3 - 5 times above the fluctuation of the instrument detector, which is referred to as the baseline variability of the instrument. The peak identified as Endrin on both columns is <3x the baseline noise ( Figure 2 ). In addition, Endrin on Column 1 appears as a shoulder, indicating a possible matrix interference. Although reliable, Method 8081B can be prone to interferences, particularly at low levels, even when the laboratory conducts the analyses using commonly used standard processes [4] .

The purpose of this study was to determine if substances in the river water could positively bias the detection of organochlorine pesticides at low concentrations.

A series of scientific experiments were designed to evaluate whether a matrix interference, present in the water and suspended particles (referred to herein as particulates), potentially creates increased and false positive pesticide detections. Water samples were collected from three locations in Rio Atibaia within the region of Campinas, Brasil.

The river samples were collected on 6^th February 2023 ( Figure 1 ), and the samples were sent via courier to two separate laboratories for confirmatory pesticide analysis using multiple methods. The water samples were analyzed using a certified in-country laboratory (Lab 1), and split samples were collected by an additional external forensic laboratory (Lab 2).

The following procedure was performed by Lab 1 on the river water samples from each location:

All samples arrived at both laboratories intact and in good condition between 2˚ and 6˚C. Water samples were analyzed for organochlorine pesticides ( Table 2 ) by EPA Method 8081B [4] by both laboratories; additionally, after passing the water samples through a 0.7 µm filter, the particulate matter from the filtrate was analyzed by Lab 2. The reporting limits achieved by the laboratories are presented ( Table 3 ).

The capillary column configuration used by Lab 1 included a guard column (Agilent J & W DB-1 100% Dimethylpolysiloxane with length 2 m, diameter 0.53 mm, and film thickness 0.1 µm). Lab 1’s primary column was Agilent J & W DB-35 MS 35%-Phenyl]-methyl polysiloxane with length 30 m × inner diameter 0.53 mm film thickness 1.0 µm. Its secondary column was Agilent J & W DB-5 with length 30 m × inner diameter 0.53 mm × film thickness 1.50 µm. The initial calibration curve acceptance criteria were R² > 0.99 or RSD < 20% depending on the analytes. Target analytes were quantified by external calibration factors. The quantitation limit was the sample equivalent of the lowest initial calibration standard.

The objective for Lab 2 was to analyze the particulate and dissolved water fractions as separate fractions to determine the contributions of each fraction to the total target analyte concentrations and NTA interferences, if detected. The water and particulates were separated by passing the water through a 0.7 µm filter and extracting each separately. Laboratory chemists fortified the water samples with surrogates to monitor extraction efficiency and were solvent-extracted at a neutral pH by EPA Method 3510 [9] . The filtered particulates (and filter paper) were fortified with surrogates and extracted by EPA Method 3570 [10] . The extracts were concentrated, cleaned using activated copper (EPA method 3660 [11] ) and silica gel (EPA Method 3630 [12] ), and spiked with internal standards. The water and particulate extracts were analyzed EPA Method 8081 [4] and reported separately.

Lab 2 utilized a deactivated fuse silica phenyl-methyl, 5 m × 0.32 mm ID Restek Guard Column followed by a RTX-5 Stationary Phase, 60 m length × 0.25 mm ID, 0.25 mm film thickness column. Its confirmatory column was RTX-CL Pesticide II Stationary Phase, 60 m length × 0.25 mm ID, 0.2 mm film thickness confirmation column. The calibration standard concentrations ranged from 0.5 µg/L to 200 µg/L and produced response factors with less than 20% relative standard deviation (RSD) acceptance criteria. Samples were quantified using average response factors, and analyte concentrations were determined versus internal standards, which were added prior to analysis.

The water samples analyzed by Lab 1 contained no detectable pesticides above the reporting limit for whole water and filtered water samples from all locations ( Table 4 ). The raw instrument data exhibited a variable baseline with many peaks eluting near the target analytes ( Figure 3 ). Lab 1 GC/ECD pesticide chromatogram for AT07 whole water includes particulate and dissolved fractions ( Figure 3(A) ), and the filtered water passing through a 0.22 µm filter represents the dissolved fraction only ( Figure 3(B) ). The unfiltered water ( Figure 3(A) ) exhibited significantly more peaks than the filtered water ( Figure 3(B) ). This observation demonstrated that filtration helped remove significant interferences associated with suspended particulates. However, filtering did not eliminate all of the interferences. The remaining peaks ( Figure 3(B) ) are associated with interferences in the dissolved fraction. Little difference was observed between the level of interferences observed in the water filtered through the 0.22 and 0.45 µm filters. This observation demonstrated that either the 0.22 or 0.45 µm filters effectively reduced the particulate interferences, while allowing the dissolved phase interferences to remain in the filtered water sample. The variable baseline in the unfiltered and filtered samples indicated the presence of complex particulate-phase and dissolved-phase organics commonly observed in urban water samples [13] [14] . This same trend showing variable baseline in the unfiltered and filtered reduction also appears in AT01 and AT09 ( Figure 4 ).

Lab 1 instrument chromatograms (column 1 and column 2) ( Figure 4 ) for the retention time in the whole water (unfiltered) and filtered samples from all three sample locations for Dieldrin. The double vertical lines and shaded bars indicate the expected target analyte retention times based on the calibration standard. The symbol (*) identifies NTAs that co-elute with target analytes based on the retention time shifts, and the absence of these target analytes at the Method Detection Limit (MDL) in confirmation runs on a GC equipped with a mass spectrometer (GC/MS). A similar behavior is observed for Endrin ( Figure 5 ). These data demonstrate that baseline variability is caused by non-target analytes (i.e., interferences) in the particulate-phase and dissolved-phase organics in the river samples, regardless of the location and filtering.

Lab 2 confirmed that no river samples contained pesticides above the reporting limit in both the particulate and dissolved fractions ( Table 5 ). In this case, Lab 2 analyzed the water and particulate portion separately, which resulted in different reporting units for the particulates (0.22 µg/kg) and dissolved water (0.00047 µg/L) samples due to differing samples sizes (~1 g particulate and ~1000 mL water). These data complemented the Lab 1 results by directly comparing the particulate and dissolved fractions in separate extracts.

The Lab 2 raw instrument data file for dual column GC/ECD were analyzed with the calibration spike to estimate the magnitude of the false positive bias due to the baseline variability ( Figure 6 , Figure 7 , Figure 8 ). Figures 4-5 are an example of a matrix interference at low level detection for the AT07 water ( Figure 6 ) and associated particulates ( Figure 7 ). Although Endrin met retention criteria, the peak is shifted to the left on column #1 and to the right on column #2 of the calibration standard, as shown in Figures 4-5 . This indicates the peak chosen as Endrin was an interferent compound.

Figure 8 showed interference for 4,4’-DDT in the water associated with sample AT01. The peak chosen for 4,4’-DDT matches on column #1, however the peak is shifted to left on the confirmation column #2, indicating a possible interference is present at low levels and not verifying the column #1 match. 4,4’-DDT was not detected in the associated AT01 sample’s particulates ( Table 5 ).

Further interferences examples can be found ( Figure 9 ), which depicts the Lab 2 raw instrument chromatograms (column 1 and column 2) for the retention time of Endrin in the filtered water and filtered particulate samples from all three river sampling locations. The double vertical lines on Figure 9 chromatograms indicate Endrin's expected retention time based on the calibration standard. The red lines are the integration boundaries for the non-target interference peak used to estimate the magnitude of the false positive bias. The symbol (*) identifies potential interferences with a target analyte, Endrin. Regardless of the location and whether the samples represent dissolved phase or particulates, the baseline variability indicates the presence of unidentified organic substances at low levels.

Lab 2 data were reintegrated, combining water and particulates in the associated water, to recalculate pesticide concentrations on a µg/L basis by using the associated water sample size (1.05 L), which is reported for each sample location ( Table 6 ). On Table 6 , analytes shaded in blue exhibited noisy chromatographic baselines signifying matrix interference as shown in Figures 4-9 . To be clear, these data ( Table 6 ) are not detected pesticides; but rather, an estimate of the magnitude of the combined interfering peak in the particulate and dissolved samples. This procedure differs from Lab 2’s standard operating procedure (SOP) because it integrates peaks that are below the detection limit to make this evaluation of potential interferences at low levels. Based on the combined particulate and water results, sample AT07 would have reported 4,4’-DDT (0.00058 µg/L) and sample AT09 would have reported Endrin (0.00055 µg/L), all above the reporting limit ( Table 2 ); however, as discussed, these concentrations are positively biased by the presence of NTA interferences. This observation strongly suggests the need for careful inspection of the peak retention time and baseline variability if low level detections occur.

Note: U = Non-detected at 3 - 5 times signal to noise.

Bolded analyte concentration is above the water reporting limit for the reintegrated dissolved water and particulate sample. Note that the dissolved water results were below the reporting limit for all samples.

This study reviewed the electronic instrumental data files for the whole and filtered surface water samples for spilt samples sent to two laboratories for evidence of matrix interferences that could result in false positive and/or biased high results and to evaluate the effects of filtering. This review confirmed the potential for matrix interferences associated with low levels for organochlorine pesticides (i.e., Endrin, Dieldrin, and isomers of DDT) at all three river sampling locations evaluated (AT-01, AT-07, and AT-09). In addition, the interference can be seen in the whole water as well as both filtered samples (0.45 and 0.22 µm), which indicates these interferences are not specifically isolated to the particulates but are also present in the dissolved water fraction. EPA Method 8081 [4] , as normally applied, generates accurate and precise data with sufficient sensitivity to confidently measure pesticide analytes with detection limits below the applicable regulatory compliance limits, and it is well suited for future analyses of river water. However, as demonstrated in the previous sections, it may exhibit a false positive bias at low levels due to the presence of natural and anthropogenic substances in urban waterways. This bias can be reduced by the application of the following procedures:

Mass Spectrometry Confirmation Analysis. Sample extracts can be re-analyzed on a GC/MS instrument if future analyses using EPA Method 8081 [4] detect pesticides with a variable baseline. High-resolution GC/MS instruments work best when NTAs coelute with low level OCPs. However, a conventional quadrupole GC/MS operated in scanning and selected ion monitoring (SIM) modes may also confirm pesticide detections at moderate to high concentrations. Most environmental laboratories possess quadrupole GC/MS instruments for the analysis of semivolatile hydrocarbons using EPA Method 8270 [15] . It is recommended that the instrument run time be no shorter than 30 minutes with a constant temperature ramp from approximately 65˚C to 300˚C to maximize analyte separation. These operating conditions will promote peak separation, minimize coelutions with NTA interferences, and enable retention time indexing. The scanning mode can help identify the identity of interfering compound or class of NTA interferences, while the SIM run can confirm the identity of the target pesticide analyte. This analysis should include the analysis of the low calibration standard from the initial calibration curve to accurately identify the target analyte retention time and approximate response factor.