Evaluation of the Concentration-Addition Approach for Pesticide Mixture Risk Assessment in Agricultural Watersheds

Abstract

The environment is a complex system composed of various creatures and pollutants caused by human activity. In this study, a mixture toxicity assessment based on the concentration-addition (CA) model was applied to 51 types of pesticides detected in the Saemangeum Watershed (the Mangyeong and Dongjin Rivers) for one year to evaluate pollution levels. A mixture toxicity assessment was conducted based on the measured environmental concentration (MEC) and the predictive values of basic ecotoxicity data (algae, daphnids, and fish). The risk and the trophic levels that showed the highest sensitivity varied depending on the sampling time and point for each mode of action (MoA) group of pesticides. In particular, the mixture risk for chloroacetamide and thiocarbamate pesticides was high (risk quotient > 1), and the potential risk was highest in autumn. In addition, the driving forces of seasonal changes were examined for samples that exhibited potential risks, and it was found that one to four substances contributed the most to the risk. The results of this study show that an initial risk assessment based on a mixture toxicity assessment can help in managing pesticide pollution in rivers.

1. Introduction

Generally, ecosystems are exposed to a mixture of multiple chemicals rather than to a single chemical [1]. Thus, most streams and rivers are threatened by numerous chemicals from various sources [2,3]. The rapid growth of liquid chromatography–high-resolution mass spectrometry (LC-HRMS) technology has increased awareness of numerous new pollutants [4,5]. Considering the presence of unknown pollutants, many chemicals inevitably exist in the environment as unintended compounds. Conventional risk assessment generally only estimates the risks of single chemicals by monitoring the amounts, frequencies, and impacts of individual substances or by compliance with environmental quality standards (EQSs) [1,6,7,8,9,10]. However, exposure to chemicals with concentrations below the no-observed-effect concentration (NOEC) can have complex effects [1,11,12]. Malaj et al. [2] reported an increase in environmental risk according to the number of chemicals analyzed in a river. Therefore, risk assessments based on a single chemical may underestimate the risks in the actual environment [1,11,13].

The worldwide use of pesticides, which are hazardous contaminants, amounts to 3.7 million tons [14]. Pesticides may also affect nontarget organisms, such as amphibians, arthropods, animals, and green plants, raising concerns regarding their management. In addition, various aquatic species living in river water can be toxic, even at low concentrations [15]. In the case of pesticides, co-application is more common than the use of one pesticide alone, as it improves their efficacy; for example, a pesticide for controlling pests and a fungicide for increasing crop production may be used simultaneously. Therefore, concerns regarding pesticide mixtures are increasing [16,17]. In rural areas, amphibian populations have been decreasing due to pesticide mixtures [11]. Weisner et al. [18] stated that the estimated insecticide mixture risk in streams was 3.2 times higher than that predicted based on single pesticides. In addition, mixtures of carbaryl and carbofuran, which are carbamate insecticides, and diazinon, an organophosphorus insecticide (carbaryl + diazinon and carbofuran + diazinon), showed synergistic toxicity [19]. Therefore, it is necessary to conduct assessments considering the mixture risks to aquatic ecosystems.

The basic concept of a mixture risk assessment is to use evaluation variables from previous studies to avoid unnecessary experiments. Because generating a large amount of experimental data is expensive and requires considerable time, toxicity calculations based on established concepts, such as concentration addition (CA) and independent action (IA), have been verified as efficient risk management techniques [17,20]. Assessing the toxicity of a mixture involves two approaches: whole-mixture and component-based approaches [21]. The component-based approach is more commonly used than the whole-mixture approach because it is difficult to assess the components of a mixture [22,23]. For the component-based method, (1) CA (also known as dose addition) and (2) IA (also known as response addition) have been widely used for mixture assessments [24,25,26]. CA has been established as a reliable and widely applicable primary assessment method for mixture ecotoxicity and risks [27,28]. CA assumes that the concentration of a mixture can be predicted from the sum of its sub-components after adjusting the relative toxicity value [7,23]. Based on this concept, Backhaus and Faust [25] proposed a tiered approach to estimate the risk index for a mixture using a basic toxicity dataset of single substances. However, the CA model has also been applied to compounds with similar modes of action (MoA). Therefore, the MoA of chemicals must be monitored for a more accurate assessment of the risk index of an environment in a mixed state [9,26,29].

This study aimed to evaluate the environmental risks of the pesticides detected in the Saemangeum Watershed, as in a previous study [30], using a mixture toxicity assessment. An ecotoxicity dataset of single compounds for algae, daphnids, and fish was used as a basic set, and the potential mixture risk index (RQ_mix) was calculated using (1) the RQ-based and (2) toxic unit (TU)-based assessment methods proposed by Backhaus and Faust [25]. Selection of the main pollutants in the Saemangeum area was also attempted using a mixture toxicity assessment. A mixture risk assessment can be helpful in comprehensively understanding the ecotoxicity risks of pesticide runoff in the Saemangeum Watershed.

2. Materials and Methods

2.1. Sampling Sites and Sample Preparation

Surface water samples were collected from six sampling points in the Mangyeong River (M1, M2, and M3) and Dongjin River (D1, D2, and D3) in the Saemangeum Watershed, as reported in a previous study [30]. The detailed sampling points (Figure 1 and Table S1) were divided into upstream (M1 (Gosan) and D1 (Dongjin 1)), midstream (M2 (Samrye) and D2 (Dongjin 2)), and downstream (M3 and D3) points. Sampling was performed monthly from March 2021 to February 2022. The instrumental analysis conditions were described in a previous study [30]. The details are listed in Table S2, and they can be summarized as follows: The collected samples were pretreated using an Equan MAX online-SPE system (Thermo Fisher Scientific, Waltham, MA, USA). In this case, 1 mL of each sample was injected into the system. Liquid Chromatography–Tandem Mass Spectrometric Analysis (LC-MS/MS) was conducted using a high-resolution mass analyzer (Orbitrap Q Exactive Plus, Waltham, MA, USA), and the electrospray ionization (ESI) method was used for ionization. The mass resolution was 70,000, the mass accuracy was ≤5 ppm, and measurements were performed in full scan/data-dependent MS2 (ddMS2) analysis mode in the mass range of m/z 100–1500.

2.2. Mode of Action Classification for Pesticide Selection

In a previous study, 308 pesticides managed by the maximum residue limits were analyzed. Of these, 171 were identified within the study area. When the detected pesticides were classified according to their MoA, 40 were identified. There were 24 types of substances whose MoA could not be accurately identified or were unclassified. The four most frequently detected MoA groups were selected for further investigation (

Table 1. List of target compounds for mixture toxicity assessment.

Group	Compound
1	Alachlor, butachlor, cafenstrole, dimethenamid, esprocarb, fentrazamide, mefenacet, metolachlor, napropamide, pretilachlor
2	Acephate, cadusafos, β-chlorfenvinphos, diazinon, dichlorvos, dimethoate, dimethomorph, dimethylvinphos, ethoprophos, imicyafos, malathion, phenthoate, profenofos, tebupirimfos
3	Cyproconazole, difenoconazole, diniconazole, epoxiconazole, fenarimol, fenbuconazole, hexaconazole, ipconazole, metconazole, myclobutanil, penconazole, prochloraz, propiconazole, tebuconazole, tetraconazole, triadimefon
4	Azimsulfuron, bensulfuron-methyl, cyclosulfamuron, flucetosulfuron, halosulfuron-methyl, imazosulfuron, metazosulfuron, penoxsulam, pyriftalid, pyriminobac-methyl, pyrimisulfan

). Group 1 (HRAC code: K3) inhibits the synthesis of very-long-chain fatty acids (VLCFAs) and is involved in cellular metabolism. Group 2 (IRAC code: 1B) is an organic phosphate that acts as an acetylcholinesterase (AChE) inhibitor. Group 3 (FRAC code: G1) inhibits sterol biosynthesis in the cell membrane by inhibiting C14-demethylase in sterol biosynthesis. Group 4 (HRAC code: B) inhibits acetolactate synthase. Finally, 51 pesticides were selected, the details of which are listed in Table S2.

2.3. Environmental Toxicity Information

The L(E)C₅₀ values for aquatic organisms, including the EC₅₀ of green algae (96 h), the LC₅₀ of daphnids (48 h), and the LC₅₀ of fish (96 h), were estimated using Ecological Structure-Activity Relationships (ECOSAR) software (version 2.3). This program estimates data related to acute toxicity based on molecular structures [12]. The ECOSAR software (version 2.3) system is one of the most reliable predictive tools in this field, and it is maintained and developed by the US Environmental Protection Agency (EPA) [31]. Molecular descriptor information in the Simplified Molecular Input Line Entry System (SMILES) strings and the CAS number were used to predict the expected aquatic toxicity. In the event of multiple predictions, the lowest value was selected. The selected values are listed in Table S3.

2.4. Environmentally Realistic Mixture Risk Assessment

In accordance with the tiered approach proposed by Backhaus and Faust [25], the environmental pesticide mixture risk was evaluated using two CA-based approaches and calculated as follows:

The first approach for assessing an environmental mixture was based on the risk quotient (RQ). The RQ was calculated by dividing the predicted environmental concentration (PEC) or the measured environmental concentration (MEC) by the predicted no-effect concentration (PNEC). The MEC of each sample was used to conduct an ecotoxicological assessment of the study area. The PNEC value was calculated by selecting the lowest value among the acute toxicity data for algae, daphnids, and fish and dividing it by the assessment factor (AF) (100) [32]. Subsequently, the RQs of the individual substances in each sample were added to obtain the sum of the risk quotients (SRQ) for each individual substance in each sample.

$$\mathrm{R}\mathrm{Q}={\displaystyle \frac{\mathrm{M}\mathrm{E}\mathrm{C}}{\begin{array}{c}\mathrm{PNEC}\end{array}}}$$

(1)

$$\begin{gathered} \mathrm{S}\mathrm{R}\mathrm{Q}=\sum _{i=1}^n{\displaystyle \frac{{\mathrm{M}\mathrm{E}\mathrm{C}}_i}{{\mathrm{P}\mathrm{N}\mathrm{E}\mathrm{C}}_i}} \\ ={\sum }_{i=1}^n{\displaystyle \frac{{\mathrm{M}\mathrm{E}\mathrm{C}}_i}{\min ({\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i,\mathrm{a}\mathrm{l}\mathrm{g}\mathrm{a}\mathrm{e}},{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i,\mathrm{d}\mathrm{a}\mathrm{p}\mathrm{h}\mathrm{n}\mathrm{i}\mathrm{d}},{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i,\mathrm{f}\mathrm{i}\mathrm{s}\mathrm{h}})/\mathrm{A}\mathrm{F}}} \end{gathered}$$

(2)

The second approach was based on the toxicity unit (TU) approach. TUs were calculated by dividing the MEC by the acute toxicity data for each trophic level (algae, daphnids, and fish). The resulting TUs were then added to calculate the sum of the toxicity unit (STU). The RQ_STU was calculated by multiplying the maximum of the calculated STUs for each trophic level by the AF (100).

$$\mathrm{T}\mathrm{U}={\displaystyle \frac{\mathrm{M}\mathrm{E}\mathrm{C}}{{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50}}}$$

(3)

$$\mathrm{S}\mathrm{T}\mathrm{U}={\sum }_{i=1}^n{\displaystyle \frac{{\mathrm{M}\mathrm{E}\mathrm{C}}_i}{{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i}}}$$

(4)

$$\begin{gathered} {\mathrm{R}\mathrm{Q}}_{\mathrm{S}\mathrm{T}\mathrm{U}}={\mathrm{S}\mathrm{T}\mathrm{U}}_{\mathrm{M}\mathrm{a}\mathrm{x}}\times \mathrm{A}\mathrm{F} \\ =\mathrm{M}\mathrm{a}\mathrm{x}\left({\mathrm{S}\mathrm{T}\mathrm{U}}_{\mathrm{a}\mathrm{l}\mathrm{g}\mathrm{a}\mathrm{e}},{\mathrm{S}\mathrm{T}\mathrm{U}}_{\mathrm{d}\mathrm{a}\mathrm{p}\mathrm{h}\mathrm{n}\mathrm{i}\mathrm{d}},{\mathrm{S}\mathrm{T}\mathrm{U}}_{\mathrm{f}\mathrm{i}\mathrm{s}\mathrm{h}}\right)\times \mathrm{A}\mathrm{F} \\ =\mathrm{Max}\left({\sum }_{i=1}^n{\displaystyle \frac{{\mathrm{M}\mathrm{E}\mathrm{C}}_i}{{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i,\mathrm{a}\mathrm{l}\mathrm{g}\mathrm{a}\mathrm{e}}}},{\sum }_{i=1}^n{\displaystyle \frac{{\mathrm{M}\mathrm{E}\mathrm{C}}_i}{{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i,\mathrm{d}\mathrm{a}\mathrm{p}\mathrm{h}\mathrm{n}\mathrm{i}\mathrm{d}}}},{\sum }_{i=1}^n{\displaystyle \frac{{\mathrm{M}\mathrm{E}\mathrm{C}}_i}{{\mathrm{E}\left(\mathrm{L}\right)\mathrm{C}}_{50,i,\mathrm{f}\mathrm{i}\mathrm{s}\mathrm{h}}}}\right)\times \mathrm{A}\mathrm{F} \end{gathered}$$

(5)

The SRQ generally had a higher value than the RQ_STU. If the SRQ value exceeded a threshold of 1, which was indicative of potential risk, the RQ_STU value for the respective sample was used in the final calculation of RQ_mix. Consequently, in cases where the SRQ exceeded 1, the calculation was conducted using the TU-based method. The level of risk was evaluated as follows: high risk (RQ ≥ 1), moderate risk (0.1 ≤ RQ < 1), low risk (0.01 ≤ RQ < 0.1), and rare risk (RQ < 0.01).

$${\mathrm{R}\mathrm{Q}}_{\mathrm{m}\mathrm{i}\mathrm{x}}=\mathrm{S}\mathrm{R}\mathrm{Q} \mathrm{o}\mathrm{r} {\mathrm{R}\mathrm{Q}}_{\mathrm{S}\mathrm{T}\mathrm{U}}$$

(6)

3. Results and Discussion

In this study, a mixture toxicity assessment was conducted for 51 pesticide types detected in the Mangyeong and Dongjin Rivers based on a previous study [30]. The MoA of the detected pesticides were identified based on previously reported information to apply the CA-based model, which provided reliable estimates for a mixture toxicity assessment [33,34,35]. A mixture toxicity assessment was conducted for the four groups that exhibited the highest frequencies and concentrations (Table S4).

3.1. Mixture Risk Assessment Based on RQ

To conduct a mixture toxicity assessment in accordance with the methodology described by Backhaus and Faust [25], the RQ was first calculated using the MEC values based on the acute toxicity data for single substances (algae, invertebrates, and fish). Mixture toxicity assessment using the SRQ is common because it can present the most conservative results in terms of regulations [5]. The results of calculating the SRQ for each group using 72 samples are shown in Figure 2, Figures S1 and S2. Figure 2 shows the SRQ values at points M3 and D3, which are the representative points of the Mangyeong and Dongjin Rivers, respectively. As shown in Figure S2, the percentages of the samples that exhibited SRQ values ≥ 1 were 84.21% for Group 1 (K3; VLCFA inhibitors), 52% for Group 2 (1B; acetylcholinesterase inhibitors (ACIs)), 3.2% for Group 3 (G1; inhibitors of C14-demethylase in sterol biosynthesis (SBIs)), and 6.45% for Group 4 (B; acetolactate synthase inhibitors). This indicates the presence of potential risk. The risk distribution pattern for each group was similar based on the sampling point. Group 1 had a high risk during most periods except for August. Group 2 showed a high risk in April, October, and January; Group 3 had a high risk in October and January; and Group 4 had a high risk from June to July. These differences in risk by group, depending on the sampling time, appear to be affected by the application periods of pesticides or the target crops [2,36]. For pesticides, the target crops may vary depending on the MoA, and the application period may differ. Pesticides in Groups 1 and 3 are mainly used as herbicides, those in Group 2 as insecticides, and those in Group 4 as fungicides. Li et al. [36] reported that herbicides, insecticides, and fungicides showed various detection characteristics depending on the season. Moreover, Herrero-Hernández et al. [37] observed that the herbicide atrazine exhibited the highest concentrations during the spring, yet its degradation products were detected at their peak in June, when insecticides were predominantly identified. Considering that the residual concentration and risk of each chemical are significantly affected by the target crop and the application period, an assessment based on the annual worst-case scenario may be inappropriate.

As shown in Figure 2 and Table S5, the maximum observed concentration in Group 4 (B) was higher than that in Group 1 (K3; the Mangyeong River); however, its SRQ was approximately 300 times lower. Group 2 (1B) showed concentrations and detection distributions similar to the maximum concentration observed in Group 3 (G1); however, its SRQ was approximately 70 times higher. The RQ distribution results by group were obtained from the differences in the toxicity data for the substances constituting each group. Group 1 herbicides (chloroacetamides and thiocarbamates) are widely used as pre-emergence pesticides [38], and because they are highly hydrophilic, they have a relatively high risk of accumulation and distribution in water environments and have been identified as substances with genetic toxicity, endocrine-disrupting factors, and carcinogenicity in some animal studies [39]. Organic phosphates (Group 2) are pesticides that inhibit acetylcholinesterase (AChE). They act on the nervous system and may cause neurotoxicity in nontarget organisms [40].

Furthermore, as summarized in Table S4, the PNEC and RQ had wide distribution ranges that were different from the results of the concentration and detection frequency, which are the factors used in conventional environmental monitoring. These results demonstrate that individual RQs for ecotoxicological hazard assessment are constrained by their utilization as representative values for risk assessments of actual environments in the presence of mixtures of various substances. Therefore, a comprehensive aquatic environment monitoring step that considers mixture risks is required. Because the RQ is generally used to predict toxicity assessments for single compounds, its data can be easily secured. In addition, synergism and antagonism are less likely to occur at the concentrations detected in the environment [28]. Assuming that there is no interaction between mixtures, the SRQ, which determines the mixture risk using the exposure assessment factor (the PEC or MEC) and effect assessment factor (the PNEC), can be a useful method for initial risk screening.

3.2. Mixture Toxicity Assessment Based on TUs

The SRQ may overestimate mixture toxicity owing to the sum of the toxicity values for different species because it uses PNEC values based on the toxicity data for the most sensitive species. Therefore, if the SRQ was > 1 (the threshold), RQ_mix was calculated considering the AF after selecting the highest sum (STU_max) based on the TU for each trophic level (algae, daphnids, and fish). Algae, daphnids, and fish are nontarget species that represent different trophic levels of aquatic ecosystems [41], and the TU-based approach is trophic level-specific. Therefore, STU values for each trophic level provide a mixture risk assessment for various organisms [5]. Figure 3 shows the results of RQ_STU (algae, daphnids, and fish) by group calculated using an AF value of 100. In Group 1 (K3) (Figure 3a), RQ_STU (algae) showed a high risk in 82.5% of the detected samples. For RQ_STU (daphnids), 24.6% of the samples were classified as high-risk.

For RQ_STU (fish), 38.6% of the samples were evaluated as moderate- or high-risk, which was similar to the result (36.8%) for daphnids. In Group 3 (1B) (Figure 3b), 52% of the samples showed a high risk for RQ_STU (daphnids), which was the most sensitive trophic level (daphnids > fish > algae). In Group 3 (G1) (Figure 3c), RQ_STU (daphnids) and RQ_STU (fish) exhibited similar risk tendencies, and only one fish sample showed a high risk. For RQ_STU (algae), the proportion with rare risk decreased and that with low or moderate risk increased compared with daphnids and fish. Only 3.23% (n = 2) of the samples exhibited high risk, but algae were the most sensitive to Group 3 (G1). In Group 4 (B) (Figure 3d), RQ_STU (daphnids) and RQ_STU (fish) exhibited almost no risk (< 0.01) for all detected samples, whereas low (45.2%), moderate (29%), and high (6.5%) risks were observed for RQ_STU (algae). In Group 4 (B), algae were also the most sensitive.

These results were consistent with the general characteristics of each group. The pesticides in Groups 1 (K3) and 4 (B), which are herbicides, are the most harmful to algae, and those in Group 3 (G1), which are fungicides, are the most harmful to fish [2,42]. Group 2 (1B), which represents organic phosphates, is more harmful to invertebrates and fish [43], and it was also more toxic to daphnids in the present study. Some substances and certain MoA may depend on certain species [25,44]. For example, Casillas et al. [45] calculated RQ_STU results for Chironomus spp. and Americamys bahia, which are substitutes for Daphnia, in accordance with the Plant Product Regulation, and reported differences in risk levels. Because the tiered methodology presented by Backhaus and Faust [25] assumes that only minimal basic ecotoxicity data are available for mixture risk assessment, caution must be exercised when data on various trophic levels (primary producers, primary consumers, and secondary consumers) are analyzed in cases dependent on specific species.

3.3. Environmental Risk Assessment with Mixture Toxicity Assessment

Table 2. A summary of the RQ_mix results based on the SRQ and toxicity unit (TU) approaches.

Group	River a	RQmix			Ratio of Exceedance of SRQ b (%)	Ratio of Exceedance of RQSTU c (%)	Trophic Level d			Maximum Ratio of SRQ to RQSTU
							Algae	Daphnids	Fish
		Min	Max	Median			(%)	(%)	(%)
Group 1	M	3.62 × 10−1	4.06 × 102	1.84 × 101	72.2%	100%	100	0	0	1.03
	D	3.74 × 10−1	1.65 × 103	1.45 × 101	61.1%	95.5%	100	0	0	1.39
Group 2	M	6.71 × 10−6	1.31 × 102	1.38	41.7%	100%	0	100	0	1.01
	D	7.63 × 10−3	1.80 × 102	1.12	30.5%	100%	0	100	0	1.01
Group 3	M	9.12 × 10−4	1.83	7.70 × 10−3	2.8%	100%	100	0	0	1.05
	D	1.61 × 10−3	2.56	1.56 × 10−2	2.8%	100%	100	0	0	1.10
Group 4	M	1.88 × 10−3	2.62 × 10−1	2.10 × 10−2	0%	-	-	-	-	1.00
	D	1.09 × 10−3	1.36	3.83 × 10−2	5.6%	100%	100	0	0	1.00

^a M, the Mangyeong River; D, the Dongjin River. ^b The ratio of SRQ > 1 detected in the surface water of the Mangyeong and Dongjin Rivers. ^c The ratio of exceedance of RQ_STU = n/N, where n is the number of samples with RQ_STU > 1 and N is the total number of samples with SRQ > 1. ^d The ratio of RQ_STU = n/N, where n is the number of samples with RQ_STU for algae, daphnids, and fish and N is the total number of samples with RQ_STU > 1.

shows the RQ_mix results for the study area calculated based on the approaches in Section 3.1 and Section 3.2. When the SRQ was > 1 and RQ_STU was > 1, the RQ_STU value was used as the RQ_mix value. RQ_STU was > 1 in most cases when SRQ was > 1 (except for D2 in Group 1 in November), and one species was 100% dominant by group for RQ_STU. In this study, the maximum ratio of SRQ to RQ_STU for the samples with SRQ values ≥ 1 was 1.39 (in Group 1). According to Backhaus and Faust [25], the ratio of SRQ to RQ_STU varies according to the number of trophic levels used, and the difference increases as the proportion of substances with strong species specificity in the mixture increases. In addition, the SRQ and RQ_STU exhibited similar predictive values when the substances that constituted the mixture were similar or had similar toxicities for the species. Zheng et al. [1] reported that the maximum ratio of SRQ to RQ_STU was 1.91 when 15 organic chlorinated pesticides were targeted. Backhaus and Karlsson [46] reported that the maximum ratio was 1.3 in pharmaceutical mixtures. When herbicides, insecticides, and metabolites were targeted in major Portuguese rivers, the maximum ratio between the two mixture toxicity indicators (the SRQ and RQ_STU) was 2.055 because the target substances showed different sensitivities at the three trophic levels [47]. The substances targeted in this study were grouped according to the MoA, and trophic level-specific sensitivity was observed according to the groups. Therefore, the differences between the SRQ and RQ_STU values for each group were not significant. SRQ mixture toxicity assessment is used within a legal framework, such as during the registration and authorization of chemicals [5]. Because it uses PNEC values, which are the data submitted for the registration of chemicals, the time and effort required to secure additional data are low [48]. The STU method is more reasonable for following the basic concept of CA; however, the SRQ method, with a relatively simple calculation, is favorable for rapid risk assessment. In addition, the SRQ and RQ_STU showed similar results in driver substance assessments [8]. These results show that individual RQ data used for single-compound toxicity assessment can also be used in the initial stage of mixture toxicity assessment for groups with similar MoA through the SRQ method.

Finally, in the study area, the MoA groups were assessed using the CA-based mixture toxicity method (CA-based assessment), but there were several samples that were evaluated to have potential risks. Therefore, it was necessary to examine the safety of the aquatic ecosystems in the watersheds through further risk assessments [25].

3.4. Substances Driving Seasonal Ecotoxicological Risk

Other studies that conducted research on environmental samples reported that only some of the components were chemicals driving mixture toxicity. Tian et al. [49] reported that 8 of 38 pesticides contributed to 95.5–99.8% of the mixture toxicity. Markert et al. [8] confirmed that only 1 to 3 substances among the 153 pollutant substances in the Erft River accounted for 90% of the mixture toxicity. Among the 15 organic chlorinated pesticides detected in surface water in China, only 1 or 2 substances contributed to more than 50% of the total toxicity [1]. As pesticides exhibit different seasonal patterns depending on the crop and application situation [50], their contributions to the mixture toxicity were examined for each season (Figure 4 and Figure S3). As shown in Figure 4, seasonal risk factors were identified for samples that had SRQ values > 1 using the highest MEC detected in each season at points M3 and D3 for seasonal distribution assessment under the worst-case scenario.

In Group 1, pretilachlor, metolachlor, alachlor, and butachlor showed similar contribution rates in spring at M3, and the risk of pretilachlor was dominant at D3 (Figure 4). In summer, the risks associated with pretilachlor use accounted for 82.75% and 86.49% of the risk at M3 and D3, respectively. In autumn and winter, the contribution rate of butachlor was approximately 70% at M3 and D3, and that of pretilachlor was approximately 25%. Pretilachlor was one of the most frequently detected substances in the study area (with a detection rate of 38.89%), and it is mainly used as a rice paddy herbicide. Butachlor is an herbicide used in rice paddies and fields, and its detection rate in the study area was 34.72%. The characteristics of butachlor, such as low soil adsorption and application to control annual weeds, appear to have continuously contributed to its risk in autumn and winter [15]. Metolachlor exhibited the highest detection frequency in Group 1 (K3), with a detection rate of 69.44%; however, its contribution to the total risk was low except in spring at M3. Pretilachlor, metolachlor, alachlor, and butachlor are α-chloroacetamide herbicides. Metolachlor, alachlor, and butachlor are priority substances that are managed in Korea. Although few studies have been conducted on pretilachlor compared with chloroacetamide herbicides, it has been reported that pretilachlor exhibits toxicity in fish by inducing behavioral changes [51]. Junghans et al. [52] reported that a mixture of eight chloroacetamide herbicides exhibited much higher toxicity than individual substances in Scenedesmus vacuolatus (algae) and that CA can be properly used for herbicide groups with similar MoA. Because mixtures are likely to have higher risks despite differences in contribution depending on the season and location, continuous management is required.

In Group 2 (1B), risks were observed in all cases except in summer at M3 (Figure 3 and Figure S3b). Group 2 (1B) showed differences in driver substances depending on the season and location. Higher levels of diazinon compared to dimethoate were observed in spring at M3, and higher diazinon risk was observed at D1 and D2 in the Dongjin River. In summer (Figure S3b), diazinon risk was observed at M1 and D2. In the Mangyeong River, the contribution rate of dimethoate increased downstream. In the autumn season, when the RQ was the highest, tebupirimfos exhibited the highest risk, followed by diazinon and phenthoate at both M3 and D3, and the contribution of tebupirimfos to the risk was more than 50%. Tebupirimfos and phenthoate were only detected in autumn and winter. Both substances are used as horticultural pesticides. For the risks in summer at M3, winter at M3, and winter at D3, when diazinon was not detected, the RQ decreased by a factor of approximately 14 compared with the previous season, when there was a risk of diazinon. Continuous management of diazinon, as a priority substance in the watershed, is required.

Group 3 (G1), which was mainly composed of fungicides, showed potential risks in autumn at both M3 and D3 (Figure 4 and Figure S3c), and the risk ratio according to the substance was also similar. Diniconazole exhibited the highest risk ratio, followed by prochloraz and difenoconazole. However, the contribution rate of diniconazole was 41.5%. Prochloraz also showed a high contribution (52.69% at D3) to the risk in spring. Group 4 (B) exhibited potential risks in summer at D3 (Figure 4b), with imazosulfuron and metazosulfuron being the major pesticides driving mixture toxicity. The detection frequencies of imazosulfuron and metazosulfuron, which are rice paddy herbicides, were 29.17% and 40.28%, respectively. These herbicides were only detected during certain seasons, which may have been because they are mainly used as rice paddy herbicides and their usage and residues are high during busy farming seasons. As the risk of penoxsulam, which had a relatively high detection frequency and maximum residual concentration, was barely observed, it seems necessary to select management priorities that consider the environmental risks of the target substances when monitoring the aquatic environment. The results of this study were similar to those of other studies that reported that only a few substances contributed to mixture risks. This shows that the risks to the entire aquatic environment can be managed by managing a few substances. It was also confirmed that the major pesticides that contribute to the risk differ depending on the location and season. These results indicate that priorities should be distinguished temporally and monitored spatially when evaluating the pollution levels in water environments around farmland. Therefore, continuous management of the major driver pesticides detected in mixture toxicity assessments is expected to be helpful in securing aquatic ecosystem stability in the management area.

Unlike the actual environment, experimental results for the mixture toxicity of pesticides have been evaluated for a small number of mixtures. Despite the increased awareness of numerous trace pollutants in the environment owing to improvements in various analytical technologies and nontarget analytical techniques, toxicity assessment for individual substances remains difficult. In addition, the toxicities of some metabolites may be higher than those of their parent compounds; however, studies on metabolites remain insufficient. Although the metabolite of metolachlor, metolachlor morpholinone, was analyzed in a previous study using a nontarget technique, there is a lack of sufficient toxicity data available for this compound, which may be a factor contributing to the underestimation of the mixed toxicity assessment results. Furthermore, distinctions based on MoA or mechanism, as in this study, are necessary to enhance the reliability of toxicity predictions. However, the process of gathering information for this purpose is challenging, and the classification of similar or dissimilar modes of action can unnecessarily increase the complexity of a mixture toxicity assessment. For example, the same species may exhibit different toxic modes of action depending on the developmental stage or exposure concentration [53]. Conversely, Wood et al. [54] observed that benthic diatom taxa exhibit similar sensitivities to herbicides with different MoA, indicating that distinguishing between MoA groups for mixed toxicity assessments in the real world may be challenging.

This study confirmed that a more rapid initial risk assessment for target waters is possible if a mixture of toxicity assessments that can supplement existing data (e.g., acute toxicity data and the PNEC) and existing monitoring strategies (e.g., the detection frequency and concentration) is utilized. Discovery of the substances driving mixture toxicity in the target watershed and management based on the setting of mixture thresholds for the region will enable the management of the risks of various substances that cannot be detected or evaluated. In addition, using data on actual biological species is expected to be helpful in managing risks in the environment. Therefore, it will assist in preparing standards for assessing aquatic ecosystem risks to water resources and measures for regulatory management.

4. Conclusions

In this study, the pesticide pollution levels of the Mangyeong and Dongjin Rivers in the Saemangeum Watershed were evaluated using a concentration addition (CA)-based two-stage mixture toxicity assessment. The detected pesticides were classified into four groups based on their MoA to improve the accuracy of the mixture toxicity assessment. Pesticides from various points in the watershed were evaluated over one year. The pollution level according to the MoA varied depending on the sampling time and location. Overall, the pollution levels in the watershed were high. In particular, for the samples that exhibited potential risks, pretilachlor, butachlor, and diazinon showed the highest risks. The results of the SRQ and STU methods, which were the first and second stages, respectively, differed by a factor of up to 1.38, and the trophic level that showed the highest sensitivity varied depending on the MoA. Therefore, it was confirmed that the SRQ method can also be used to assess the initial aquatic ecosystem risk in a watershed. Our findings suggest identifying the compounds of concern in a watershed by season and show that monitoring without considering time and space may cause errors in pesticide analysis. However, the data gap between in vivo data and predicted toxicity endpoints may increase the uncertainty of mixture toxicity assessment. Therefore, conducting an aquatic ecosystem risk assessment using alternative in vivo results under the worst-case scenario for each season and identifying the driving substances based on a mixture toxicity assessment will assist in preparing standards for the safe management of water resources.