Critical assessment of pendimethalin in terms of persistence, bioaccumulation, toxicity, and potential for long-range transport

ABSTRACT

Pendimethalin (PND, CAS registry number 40487-42-1) is a dinitroaniline herbicide that selectively controls broad-leaf and grassy weeds in a variety of crops and in noncrop areas. It has been on the market for about 30 yr and is currently under review for properties related to persistence (P), bioaccumulation (B), and toxicity (T) in the European Union (EU). A critical review of these properties as well as potential for long-range transport (LRT) was conducted. Pendimethalin has a geometric mean (GM) half-life of 76–98 d in agriculturally relevant soils under aerobic conditions in the lab. The anaerobic half-life was 12 d. The GM for field half-lives was 72 d. The GM half-life for sediment-water tests in the lab was 20 d and that in field aquatic cosms ranged from 45 to 90 d. From these data PND is not persistent as defined in the Annex II of EC regulation 1107/2009. The GM bioconcentration factor for PND was 1878, less than the criterion value. This was consistent with lack of biomagnification or accumulation in aquatic and terrestrial food chains. The GM no-observed-effect concentration (NOEC) value for fish was 43 µg/L, and 11 µg/L for algae. These do not trigger the criterion value for toxicity. In air, the DT50 of PND was estimated to be 0.35 d, which is well below the criterion of 2 d for LRT under the United Nations Economic Commission for Europe (UNECE) Aarhus protocol. Modeling confirmed lack of LRT. Because of its volatility, PND may be transported over short distances in air and was found in samples in local and semiremote regions; however, these concentrations are not of toxicological concern. Unlike other current-use pesticides, PND has not been found in samples from remote regions since 2000 and there is no apparent evidence that this herbicide accumulates in food chains in the Arctic.

Pendimethalin (N-(1-ethylpropyl)-3,4-dimethyl-2,6-dinitrobenzenamine, PND; CAS Registry Number 40487-42-1, Figure 1) is a dinitroaniline herbicide that selectively controls certain broad-leaf and grassy weeds in a variety of crops and in noncrop areas. Pendimethalin is predominantly applied to soil as a preplant, preemergence, but sometimes postemergence herbicide and has been on the market for approximately 30 yr. Similar to all dinitroaniline herbicides, PND is an inhibitor of cell division and targets the microtubules involved in division of plant cells. In the European Union (EU), PND is used on fruit, grapes, vegetables, oil seed, cereals, tobacco, and ornamentals (European Community 2003). The maximal application rate in the EU was 2 kg/ha but as of 2013–2014 is 1.6 kg/ha. In the United States, the current maximal agricultural label rate is 6.7 kg/ha (6 lb/acre) and that for nonagricultural uses is 4.6 kg/ha (4.1 lb/acre) (U.S. Environmental Protection Agency [EPA] 2012). PND is currently being reassessed according to regulation, and European Food Safety Authority [EFSA] (2016) has published a peer review of the pesticide risk assessment of the active substance pendimethalin.

Figure 1. Structure of pendimethalin.

The focus of this document is on persistence (P), potential for bioaccumulation (B), toxicity (T), and short-range transport as well as long-range transport (LRT) of PND in the environment. The document presents a refined characterization of these properties in the context of EC 1107/2009 (European Community 2009) and current uses in the EU, and on the results of recent studies reported in the literature and conducted by the registrant (see Supplemental Information [SI] for a discussion of the assessment endpoints). In general, guidance for evaluation of pesticides from DG SANCO (2012) was followed.

Key Properties and Uses of Pendimethalin

For the purposes of this characterization, this review focuses on new information that became available since the review of PND undertaken by the EC in 2003 (European Community 2003). However, the key properties related to P, B, and T and LRT of PND are presented in Table 1 as a point of reference for further analysis of data in this report. The criteria for classification as PBT are listed in Table SI-1.

Table 1. Key properties of pendimethalin used in this assessment.

Property	Value
Molecular weight	281.3
Vapor pressure	1.94 × 10^–3 Pa at 25°C
Henry’s law constant	2.728 × 10^–3 (kPa × m^3 /mol) 25°C
Solubility in water	pH 4: 0.54 mg L^−1 at 20°C
	pH 7: 0.33 mg L^−1 at 20°C
	pH 10: 0.44 mg L^−1 at 20°C
Partition coefficient (log KOW)	pH 7: log KOW = 5.2
Log KOC (geometric mean; see SI)	Log KOC = 4.11
Acid–base dissociation constant (pKa)	2.8
Hydrolytic stability (DT50)	Stable from pH 4 to 9 (>5 d, 50°C)
Photostability in water (DT50)	DT50 = 21 d (under 24-h daily exposure) pH not stated
Data from (European Community 2003)

Pendimethalin is registered for use in a large number of countries in every continent, each of which has a range of growing conditions for crops on which PND is utilized. These regions vary in terms of temperature, rainfall, and soil types. Although the intended focus of EC 1107/2009 is on Europe, this did not restrict our collection of data to Europe only. Where data on PND were available for regions outside of the EU where equivalent crops were grown, some of these data were considered in the assessment. This is based on the premise that, if the same crops are grown, then the behavior of PND in the environment is representative of what might be found in the EU, even if these specific scenarios have not been monitored or investigated.

In assessing P and B for chemicals, the concern is for the general environment, not a particular local scenario. Logistically, all localized scenarios cannot be assessed and protection of the environment is better directed to the average condition. Extreme values that might be observed in specific situations are not representative of all locations, and the best values to use in assessing P, B and T, are mean values and their ranges. Because many of the processes leading to P or B are driven by first-order kinetics, the geometric mean (GM) value is the most appropriate for comparison to the criteria for classification; however, upper bounds for P and B and lower bounds for T may be employed to characterize uncertainty. Since more data on environmental fate and effects are available for pesticides such as PND, it was possible to use these additional data as lines of evidence to further refine the assessment and characterization of PND. Thus, one did not have to rely solely on the simple criteria for classification that are outlined in EC 1107/2009 (Table SI-1). In addition to evaluating PBT, the potential for LRT of PND was also assessed using criteria of the United Nations Economic Commission for Europe (UNECE) (United Nations Economic Commission for Europe 1998). In addition to data from published scientific literature, findings from reports submitted by the manufacturer to various regulatory agencies were utilized.

Persistence (P)

Solomon, Matthies, and Vighi (2013) critically assessed the criteria and process used in the categorization of plant protection products (PPPs) as persistent under EC 1107/2009. DG SANCO (2012) developed recommendations by publishing the working document “Evidence Needed to Identify POP, PBT and vPvB Properties for Pesticides,” where they summarized conclusions on various issue identified in P assessment. Both documents are the primary basis for the evaluation of PND as a persistent compound. Since the review of PND in 2003 (European Community 2003), many new studies have been conducted testing PND on its abiotic and biotic degradation in air, water, soil, and sediment. Data already used in the review (“old studies”) are employed for comparison with observations of “new studies.” Classification endpoints are time for 50% and 90% degradation (DT50 and DT90) derived from kinetic modeling of time series from lab and field studies according to the FOCUS kinetic guidelines (FOCUS 2006). DT50 values are identical to half-lives if a single-first-order (SFO) kinetic is the best kinetic model.

Soil

Soil is the primary compartment for PND due to its application as a herbicide to cereals, grape, fruit, vegetables, oil seed, tobacco, and ornamentals. DT50 and DT90 values are the endpoints of aerobic and anaerobic soil degradation studies in the lab and field. Photolysis of PND applied on soil surface might also be a relevant degradation process. All other dissipation processes (volatilization, leaching, runoff, or uptake into plants) are not subjects of soil degradation studies and are excluded from P assessment. Guidelines are available from various institutions for standardization of testing in lab as well as field studies (Solomon, Matthies, and Vighi 2013).

Older aerobic soil degradation lab studies before 2000 (n = 7) provided DT50 values from 72 to 172 d with an arithmetic mean of 123 d, and GM of 118 d (European Community 2003). Since 2000, three soils from the United States and five soils from the EU have been investigated in lab simulation investigations. Table SI-2 provides an overview on the soil characteristics for lab experiments carried out since 2000. These cover a broad range of European soils and three acidic U.S. soils. Temperature was 20°C in all studies, but moisture differed from 40 to 50% of maximal water-holding capacity.

Aerobic Degradation

American Cyanamid (2000b) selected three acidic soils having a pH (CaCl₂) value from 4.5 to 4.9. Two of them (Mississippi and Louisiana) exhibited rather large DT50 values of more than 300 d, which are outside the range of all other soils investigated thus far. The soil from North Carolina had a DT50 of 167 d (Table SI-2). BASF (2013c) compared the soils to those from EU with the same soil characteristic and found that the pH values of the two soils from Mississippi and Louisiana are outside the pH range of EU soils (Figure SI-1). BASF (2013c) concluded that these two soils are not representative for the EU and thus should not to be included in the PBT assessment. The soil from North Carolina had a pH within the range of EU soils of the same characteristics. The DT50 values from the nonrepresentative soils in Mississippi and Louisiana were included in this review for comparison but excluded for P assessment for the EU.

Standard EU soils were selected (LUFA 2.2 and 2.3, LUFA 5M), as well as soils often used for degradation studies (Speyerer Wald, Bruch West). The biological activity of the treated soil from Speyerer Wald fell from 28.1 to 4.1 mg C/100 g dry soil at the end of the incubation study (122 d) (BASF 2011a), which might be the reason for decline in degradation and/or formation of bound residue. Since the reduced biological activity precluded biodegradation testing, this DT50 was excluded from soil P assessment. Thus, DT50 values were available from five lab soil degradation studies. These represent a wide variety of soils used for agricultural purposes including one acidic soil from the United States.

The standard soil LUFA 2.2 was employed in another study by BASF (2001a). The dissipation of PND might better be described by a kinetic of double first order in parallel (DFOP) rather than SFO kinetics if the whole time scale of 211 d is considered. An initial fast decline is followed by a slow disappearance over the long term. A DT50 value of 41 d was calculated. However, according to FOCUS kinetics, the rate constant of the slow process needs to be used for calculation of DT50, which provides a value of more than 200 d. Microbial biomass declined from approximately 40 at the start of the incubation to 17 after 120 d and 14 mg C/100 g dry weight after 211 d at the end of the study (BASF 2001a). As regards the initial establishment of a list of candidates for substitution, DG SANCO (2012) recommends that “the cut-off value should be compared to the DT90 value in the list of endpoints divided by 3.32, provided the DT90 value is calculated and not estimated.” A DT50 value of 122 d would then be derived by this approach, which is more than the DT50 value of 41 d for SFO kinetics but less than that of 201 d of the FOCUS approach. This value for P assessment was not utilized because it provided no new information.

Normalization to Temperature and Moisture

Normalization to reference conditions (i.e., temperature of 20°C and soil moisture at pF 2) as proposed by FOCUS (2006) was recommended by Solomon, Matthies, and Vighi (2013). Since all lab studies were carried out at 20°C, no normalization to temperature was required. DG SANCO (2012) provides no recommendation on normalization for moisture. Only normalized DT50 values were used for P assessment of PND.

DG SANCO stated that “Data on metabolites should not be included in the assessment against the criteria laid down in point 3.7 of Annex II to Regulation (EC) No 1107/2009” (DG SANCO 2012). None of the lab or degradation investigations in field soil showed formation of metabolite(s) of PND at more than 10% of total radioactive residue (TAR). The major metabolite identified was 4-[(1-ethylpropyl)amino]-2-methyl-3,5-dinitrobenzoic acid. One of the methyl-groups on the benzene ring is oxidized to a carboxylic acid group, which enhances solubility and availability for further degradation.

The DT50 for tested soils (Table SI-3) ranged from 41 to 201 d or even more than 300 d if acidic U.S. soils were included. This variability is not surprising because soils cover a wide range of soil properties, for example, content of organic carbon (OC), texture, cation exchange capacity (CEC), pH, and biological activity. Some of the DT50 values are above and others below the cutoff value of 120 d as specified in Annex II to Reg. 1107/2009, point 3.7.1.1 (see Table SI-1). Solomon, Matthies, and Vighi (2013) proposed the GM because it provides the best estimate of the midpoint of a range of values derived from first-order kinetics (log-normal). The same approach was recommended by DG SANCO (2012) for aggregation of DT50 values from different studies as regards the initial establishment of a list of candidates for substitution.

Geometric means ranged from 76 to 98 d for 5 relevant soils depending upon various calculations and normalizations (Table SI-3). Including the 2 acidic U.S. soils provided ranges from 109 to 130 d. Normalization to standard moisture content resulted in smaller DT50 values, except for soil 7, and thus also a smaller GM (Table 2). The most appropriate GM with normalization was 76 d, where DT50 of study 7 was calculated assuming SFO kinetics within the first 120 d of incubation. The GM would be 98 d if the DT50 value of 210 d according to the FOCUS approach (2006) is used. Thus, for both cases, the GM DT50 for degradation in soils representative of the EU is less than the cutoff value for classification as P.

Table 2. Summary statistics for laboratory soil aerobic degradation studies.

Statistics	DT50 (d), normalizeda
	All soils	Relevant soils
Count	7	5
Geometric mean	129.9	97.5
Median	146	97
Std. dev.	91.2	62.6
CV	59%	56%

^a See Table SI-3 for additional details.

Anaerobic Degradation

PND degrades rapidly under anaerobic conditions (DT50 = 11.7 d) (BASF 2011b). One of the nitro groups is substituted by an amino group, which results in the major anaerobic metabolite 4,5-dimethyl-3-nitro-N2-(pentan-3-yl)benzene-1,2-diamine. This metabolite is further degraded under aerobic as well as anaerobic conditions. Anaerobic data need to be used for PBT/vPvB assessment (where v refers to “very”), but only as additional information (SANCO 2012)). Given the small value, this additional information also does not support classification as P.

Soil Photolysis

PND is susceptible to photolysis by solar irradiation. Two new studies confirmed older observations that PND is degraded on the surface of the soil. An experiment in which PND was applied to LUFA 2.2 soil and irradiated continuously with solar light resulted in a soil photolysis half-life of 46 d (BASF 2001d). The study was repeated with loamy sand soil from Bruch west and the half-life of 46 d was confirmed (BASF 2010c). In both investigations, the dark control displayed less than 10% degradation after 15 d. DG SANCO (2012) concluded that soil surface photolysis is relevant for P/vP assessment. Thus, in addition to the results of field experiments with sand coverage (measuring only biodegradation), soil photolysis needs to be taken into account for P/vP assessment.

Field Studies

Solomon, Matthies, and Vighi (2013) recommended that “half-lives from field studies should be included, if available, conducted under standardized conditions, and evaluated as proposed by EFSA (2010).” DG SANCO (2012) concluded that “field dissipation studies should be included in the assessment if it is possible to derive degradation half-lives from them.” However, in contrast to Solomon, Matthies, and Vighi (2013), normalization to standard temperature and moisture is not recommended for field studies. In total, 19 field investigations have been conducted since 1999 in various European countries (Table SI-4). These cover several soil types, climate regimes, application times, and formulations of PND. Most of them (13) were field dissipation experiments and some are also accumulation studies (6). In the most recent investigations (numbers 16–19 in Table SI-4), soil was covered by a layer of sand to avoid volatilization from and photolysis on surface soil according to the scientific opinion of EFSA (2010). Two studies with U.S. soils were also available, but were not further used since crops were grown in fields, which may have contributed to dissipation (BASF 2001c).

In the field experiments, PND was applied in spring or autumn to bare soil (no plants) and the length of the study was usually 1 yr (365 d). Several formulations were used, including CS, which possesses reduced volatilization due to encapsulation. Table SI-5 depicts DT50 for all field studies. The DT50 values in the field investigations ranged from 39 to 187 d, and except for 2 studies, a single first order (SFO) was the best kinetic model.

Statistical analyses revealed little variance in values (Table 3). Arithmetic mean, GM, and median of nonnormalized DT50 were similar; the coefficient of variation (CV) was 0.51 and standard deviation (SD) 47 d. The GM was 83 d for all studies with SFO kinetics, which is considerably less than the cutoff value of 120 d. DT50 values for the sand-cover experiments (numbers 16–19 in Table SI-5) with the CS formulation ranged from 40 to 121 d with a GM of 68 d. Whether this was due to the sand cover, formulation, or greater environmental temperatures in the areas where the studies were conducted is uncertain.

Table 3. Summary of DT50 values for pendimethalin from field studies in soil.

Statistics	DT50 (d), FOCUS, nonnormalized^a	DT50 (d), normalized
Arithmetic mean	93.9	109.7
Geometric mean	83.2	72.0
Median	87.1	66.0
SD	47.6	114.1
CV	0.51	1.04

^aSee Table SI-5 for additional details and citations; n = 19 soil types.

DT50s were normalized according to the approach described in EFSA (2010). The GM for all 19 tested field soils was 72 d (Table 3), which is considerably lower than the GM for lab soil studies. The variance was higher, that is, 1.04, than for the nonnormalized DT50, indicating natural variability of degradation in a wide variety of European soils under field conditions.

Some of the field experiments were run longer than 1 yr to investigate potential accumulation of PND in soil due to annually repeated application in the spring. No accumulation was observed over 4 yr. In the Dutch investigations (numbers 5 and 6 in Table SI-5), DT50 derived from the first year might also be used to describe the degradation behavior of the following years. In the Italian experiments (numbers 7–10 in Table SI-5), DT50s for the second year were more representative for all years than those from the first year. These studies did not show year-to-year accumulation in soil and, although accumulation in soil is not used for assessment of P/vP, these findings provide further evidence that PND is not P in soil. No major metabolites (more than 10% active residue [AR]) were found in all field studies, which confirm observations from lab studies.

The GM DT50 for aerobic degradation of PND in soil in lab simulation tests (Table 2) was less than the criterion value of 120 d for P. The most appropriate GM DT50 is 98 d, taking into account all valid studies with relevant soils. Moreover, there is sufficient evidence from a large number of field investigations (Table 3) that shows that DT50 of PND is less than the criterion value (GM normalized DT50 of 72 d) and it is therefore not flagged as a being P or vP. It can thus be concluded that PND is not P in soil as specified in Annex II to Reg. 1107/2009, point 3.7.1.1 (European Community 2009).

Sediment

Degradation in freshwater sediment was investigated with standardized lab water/sediment incubation experiments. No marine sediment studies were available. Guidelines recommend two different water/sediment systems (Organization for Economic Cooperation and Development [OECD] test 308, OECD 2002) to determine the DT50 of the total system as a conservative value for the half-life in sediment. Sediment is usually part of aquatic cosms, which are designed for determination of B and T (see later discussion) and used to ground-truth the outcome of lab studies under realistic field conditions.

Degradation in sediment was examined in different systems by various authors (Table SI-6). In all studies, samples were taken from two different freshwater systems and run in parallel. Table SI-6 shows that the variability in texture of sediment was rather large, but also other properties (pH, content of OC, and redox potential) differed considerably. The first study (BASF 1992) was conducted according to an earlier guideline from the BBA Guideline, whereas the three latter studies followed OECD Test Guideline 308 (OECD 2002). The most recent study (BASF 2012) used the same system to repeat the earlier study (BASF 2004b).

Table SI-7 reports observations of all lab water/sediment studies. The kinetics of the total system degradation follows a SFO process in all experiments. Mineralization was low (<10%) in all studies. The major fate process is the formation of NER (up to 79% TAR). Large amounts of volatile PND were found in systems 1 and 2 (BASF 1992). The same metabolites detected in anaerobic soil were also noted in the sediment of systems 3 and 4 (BASF 2002b) indicating partly anaerobic conditions. One study (number 5 in Table SI-7) was not fully valid but could be modeled by SFO kinetics.

DT50 values for total system degradation ranged from 4 to 103 d with a GM of 19.7 d. Even the maximal DT50 value was below the criterion value of 120 d for freshwater sediment. Not all investigations were fully valid but were used for statistical analysis to demonstrate the experimental and natural variability of degradation in sediment (CV = 1.14; Table 4). The arithmetic mean reflects experimental variability and is 36 d. The GM reflects natural variability and is 20 d, which is far below the cutoff value of 120 d for sediment.

Table 4. Summary statistics for the results of water/sediment studies with pendimethalin.

Statistics	DT50 total system (d)^a	DisT50 water (d)
Arithmetic mean	36.4	0.86
Geometric mean	19.7	0.75
Median	20.3	0.80
SD	41.5	0.47
CV	1.14	0.55

^aSee Table SI-7 for additional details and citations; n = 8 tests.

Higher tier experiments (aquatic ecosystems studies) are usually carried out for B and T assessment under realistic conditions (see later discussion). These also provide a line of evidence to evaluate relevance of lab studies on degradation. Three aquatic cosm investigations were conducted with PND, one by Isensee and Dubey (1983), one by BASF (2001b), and a more recent study by Bromilow et al. (2006). Isensee and Dubey (1983) applied PND adsorbed to soil particles to a cosm containing fish (Gambusia), algae, daphnids, and snails and investigated the distribution in 30 d. The sediment-to-water ratio was 400 g soil in a 20-L vessel, that is, 1:50, which is smaller than that in the lab water/sediment study (approximately 1:4). Little of the adsorbed PND desorbed and less than 1% accumulated in biota. The amount of total extractable radioactivity declined from about 92% at d 2 to 60% at the end of the experiment (d 30). Assuming a SFO kinetic a DT50 of approximately 45 d can be calculated, which confirms the DT50 values found in the lab water/sediment experiments.

In the study by BASF (2001b), PND was applied in dissolved form to a cosm containing fish (golden orfe), daphnids, mayflies, phytoplankton, and periphyton. The water/sediment ratio was 125 cm water column to 15 cm sediment layer height, that is, 1:8.3. DT50 for total system degradation was estimated to be approximately 80–90 d, which confirms recent water/sediment studies. Bromilow et al. (2006) examined the fate of eight pesticides including PND between water and sediment in outdoor cosms with and without the aquatic plant Lagarosiphon major. PND displayed a rapid degradation with a half-life of 5–7 d, and was not detectable after 56 d. In sediments, the majority of the compound was found in the top layer with a maximum concentration of 15% of applied.

Sediment is the relevant compartment for PND in water/sediment studies because PND partitions rapidly from the water phase into the sediment. DT50 of total system degradation (GM of 20 d) is less than the criterion value of 120 d for all experiments. Aquatic cosm studies confirm rates of degradation measured in water–sediment lab investigations.

Water

Pendimethalin is not readily biodegradable in OECD 301B test protocol (BASF 2013e). PND is not hydrolyzed at environmentally relevant pH values (BASF 1992). However, several new studies conducted since 2002 demonstrated that PND is susceptible to photolytic degradation in water. The main metabolite is 2,6-dinitro-3,4-dimethylaniline—that is, the ethylpropyl group at the 1-N atom is abstracted. DT50 values range from 1.5 to 5 d with no pH dependence. DG SANCO (2012) concluded that “data on photolysis should also be considered, when relevant.” However, it is not clear what is meant by relevant aqueous photolysis. Solomon, Matthies, and Vighi (2013) recommended that lab tests on direct and/or indirect aquatic photolysis be used to determine abiotic half-life in the water phase.

Mineralization of PND was tested using the OECD 309 guideline (OECD 2004). Preliminary results showed no significant mineralization and/or degradation under the standard conditions of OECD 309 (no or little sediment content, no light). The experiment was carried out at two concentrations. Mean recovery rates for the two concentrations were 85%. More than 30% of the applied radioactivity adsorbed to the glass wall of the flask in the high-concentration treatment, and 9.7% in the low. The initial mean concentration of PND decreased from 96.5% to 74.1% after 63 d. The resulting half-life at 20°C was 193 d (EFSA 2016).

Water/sediment investigations for PND were previously described that were performed in the dark to avoid aqueous photolysis. Solomon, Matthies, and Vighi (2013) recommended that “if a PPP is photolytically and hydrolytically stable, i.e., failed the P/vP criteria in water, partitioning into sediment may be rapid enough to allow degradation in sediment.” Equilibration into sediment is rapid with 50% transfer to sediment (DisT50) in 0.4–1.6 d with a CV of 0.55 (Table 4). The GM of all studies is less than 1 d.

The percent AR of PND in pond A4 of the BASF cosm study (BASF 2001b) fell rather rapidly due to partitioning into the sediment. DisT50 was approximately 3 d, which is longer than in the lab water/sediment studies because of the greater water-to-sediment ratio. Other ponds exhibited DisT50 values <2 d. These observations clearly confirm data from lab findings. It is not known whether PND is also photolytically degraded and this could not be assessed, as metabolites found in the cosm study could not be identified.

Aqueous photolysis is the only significant degradation process of PND in freshwater. DT50 values in artificial as well as natural waters are less than the criterion value of 40 d. Further, PND rapidly partitions to sediment, where it is degraded and/or bound to sediment. Both processes reduce bioavailability for aquatic organisms, which was noted in cosm experiments. Based on the recommendations of Solomon, Matthies, and Vighi (2013) and suggestions in the DG SANCO Working Document “Evidence Needed to Identify POP, PBT and vPvB Properties for Pesticides” (SANCO 2012), PND is not P or vP as specified in Annex II to Reg. 1107/2009, point 3.7.1.1 (European Community 2009).

Bioaccumulation (B)

When characterizing bioaccumulation (B), a number of lines of evidence were used and refined tests included. The first of these was the bioconcentration factor (BCF) criterion as required in EC 1107/2009; however, biomagnification factors (BMF), biota–sediment bioaccumulation factor (BSAF), trophic magnification factor (TMF), and measures of concentrations in organisms in the natural environment were utilized as additional lines of evidence. Available data are discussed in the SI and summarized in Table 5.

Table 5. Bioaccumulation/biomagnification endpoints for pendimethalin in aquatic organisms.

Species	BCF^a	Kinetic corrected	Growth corrected	Lipid normalized	Reference
Oncorhynchus mykiss	931	Y	Y	Y	(BASF 2013f)
Danio rerio	1138	Y	Y	Y	(BASF 2003)
Lepomis macrochirus	3300	Y	N	N	(American Cyanamid 1986)
Ictalurus punctatus	1446	N	N	N	(American Cyanamid 1974)
Geometric mean of the BCFs in fish	1878
Procambarus simulans	9.1	N	N	N	(American Cyanamid 1980)
Procambarus clarkii	1.02	N	N	N	(BASF 2005)
Daphnia	35	N	N	N	(Isensee and Dubey 1983)
Helisoma sp.	145	N	N	N	(Isensee and Dubey 1983)
Gambusia affinis	20	N	N	N	(Isensee and Dubey 1983)
Eisenia fetida (soil)	<1	N	N	N	(BASF 2000)
Bufo americanus (soil)	<1	N	N	N	(Van Meter et al. 2016)
	BMF
Oncorhynchus mykiss	0.040	Y	Y	Y	(BASF 2013f)
Lepomis macrochirus	0.105	Y	Y	Y	(BASF 2013h)
	BSAF	OC/lipid corrected
Tubifex tubifex	0.5	N			(BASF 2016)
Tubifex tubifex (PND metabolites)	2.6	Y			(BASF 2016)
Tubifex tubifex	0.6	N			(BASF 2016)
Tubifex tubifex (PND metabolites)	3.3	Y			(BASF 2016)
	TMF
Algae/Daphnia	0.16				(Isensee and Dubey 1983)
Daphnia/Gambusia	0.7				(Isensee and Dubey 1983)
Vegetation-caribou-wolf food chain in the Canadian Arctic	Not observed	N	N	Y	(Morris et al. 2014; 2016)

^aSee SI for more detailed discussion of the data.

Although PND has been used worldwide for about 35 yr, there is no apparent evidence of its presence in living organisms through environmental monitoring. A report from the U.S. EPA on residues of chemicals in lake fish (U.S. EPA 2009) included PND among the “rarely detected chemicals,” with a frequency of detection <1% of predatory fish and <5% frequency in bottom dwelling fish. The limit of detection (LOD) was 6.2 µg/kg, and 486 samples from predatory fish and 395 bottom-dwelling fish were analyzed. For predatory fish, the maximal concentration was 13 µg PND/kg and the 95th centile was <6.2 µg PND/kg. For bottom dwellers, the maximal concentration was 33 µg PND/kg and the 95th centile was <6.2 µg PND/kg.

It is widely acknowledged that BCF is not the most suitable parameter for quantifying biomagnification in the aquatic food web, and BMF or the TMF is preferred as the best measure (Borgå et al. 2012; Conder et al. 2012; Gobas et al. 2009; Solomon, Matthies, and Vighi 2013; Weisbrod et al. 2009). Bioaccumulation through food and biomagnification via the food chain are the most relevant routes of exposures in fish (Nichols et al. 2009) and become more important as log K_OW increases above 5, as in the case of PND. In EC 1107/2009 there are no criteria for BMF or TMF or BSAF. However, in a lines-of-evidence approach, considering this information is useful for interpretation of results from lab BCF tests that may sometimes display environmentally unrealistic exposures. From data in Table 5, bioconcentration does occur in fish and some of the BCF values exceed the criterion of 2000; however, the GM was 1878, less than the criterion. In addition, elimination and metabolism was rapid (see SI). Biomagnification through diet in fish did not occur (BMF <1). No bioconcentration was observed in sediment-dwelling organisms or from soil into organisms, and accumulation in species at higher levels was not observed in terrestrial or aquatic food chains in studies conducted in the Canadian Arctic (Morris et al. 2014, 2016). A recent assessment of secondary poisoning reported that “a low risk to fish-eating birds and mammals could be concluded” (EFSA 2016). All these lines of evidence indicate that the potential for bioconcentration of PND is moderate and that the compound has no apparent potential for biomagnification. Therefore, it should not be classified as bioaccumulative under EC 1107/2009.

Toxicity (T)

Acute and chronic tests were considered in the characterization of toxicity of PND. These data are summarized in the following subsections. In all cases, toxicity values are expressed as PND active ingredient, even if formulations were utilized for testing.

Acute Toxicity

The classification of T under EC 1107/2009 is based on chronic toxicity to aquatic organisms; however, acute toxicity data are useful for illuminating chronic data, which are usually less available. Acute toxicity data were obtained from reports from the registrant and published findings. Although tests for unicellular algae are relatively short, there were viewed as chronic because they span several life cycles and are included in the section on chronic toxicity. Acute toxicity data for invertebrates and fish from tests conducted by the BASF (Table 6) were used to characterize acute toxicity. Although there is no criterion for acute toxicity to nontarget organisms in EC 1107/2009, these data may be useful for a preliminary assessment of sensitivity and for informing the assessment of chronic toxicity.

Table 6. Acute toxicity for pendimethalin in invertebrates and fish.

Scientific name of species	LC50 (μg/L)	Duration (h)	Design	Reference
Invertebrates
Daphnia magna	>1000	48	Static	(American Cyanamid 1981)
Daphnia magna	400	48	Static	(BASF 2001f)
Penaeus duorarum	1600	96	Static	(Mai et al. 2013)
Crassostrea virginica	210	48	Static	(American Cyanamid 2000a)
Geometric mean	580
Fish
Pimephales promelas	>240	96	Flow-through	(BASF 2010a)
Pimephales promelas	332	144	Flow-through (non GLP)	(American Cyanamid 1993)
Pimephales promelas	137	14 d	Flow-through (non GLP)	(American Cyanamid 1993)
Ictalurus punctatus	418	96	Static	(American Cyanamid 1972)
Cyprinodon variegatus	707	96	Static	(American Cyanamid 1983)
Oncorhynchus mykiss	890	96	Static	(American Cyanamid 2000c)
Oncorhynchus mykiss	138	96	Static	(American Cyanamid 1972)
Geometric mean	344

Few additional data on acute toxicity are available in the open literature. Bringolf et al. (2007) examined T in early life stages of several species of freshwater mussels (unionids) and found no marked effects at the greatest concentration tested, corresponding to the water solubility (3000 μg/L). Kyriakopoulou, Anastasiadou, and Machera (2009) reported EC₅₀ values in various nontarget aquatic invertebrates that were orders of magnitude above the maximum water solubility of PND, which were excluded from this paper.

Chronic Toxicity

Short-term toxicity tests with algae were included in the characterization of chronic data. Algae reproduce rapidly and the so-called acute tests are actually multigenerational studies and best characterized as chronic responses. Some data are summarized in Table 7 where EC₅₀ and EC₁₀ values were reported. For algae, EC₁₀ values are currently assumed as equivalent to NOEC (Crane and Giddings 2004).

Table 7. Toxicity for pendimethalin in algae and aquatic plants from the open literature.

Scientific name of species	EC50 (μg/L)	EC10 (μg/L)	Duration (h)	Reference
Chlorella vulgaris	281	85	96	(Ma et al. 2002)
Chlorella pyrenoidosa	394	14	96	(Ma 2002)
Scenedesmus obliquus	490	137	96	(Ma 2002)
Raphidocelis subcapitata	179	34	96	(Ma et al. 2006)
Lemna minor^a	634	90	168	(Cedergreen et al. 2005)
Scenedesmus vacuolatus	240	69	24	(Junghans et al. 2006)
Geometric mean	342	49

^aAll EC50s based on yield (number of cells) except for L. minor, which was based on growth rate. The data are homogeneous so the use of the geometric mean was judged to be appropriate.

Higher toxicity for algae and plants is to be expected for a herbicide. Nevertheless, all published NOEC data available are above the threshold of 10 μg/L, the T criterion used in EC 1107/2009. The GM value for the NOEC was 49 μg PND/L. Chronic toxicity values for algae and plants from studies conducted by BASF are summarized in Table 8.

Table 8. Chronic toxicity values for pendimethalin in algae and aquatic plants from BASF studies.

Scientific name of species	EC50 (μg/L)	EC10 (μg/L)	Duration (h)	Calculation of EC50	Reference
Skeletonema costatum	13	4	72		(American Cyanamid 1992a) with (BASF 2013a)
Raphidocelis subcapitata [formerly Selenastrum capricornutum]	32 (GM of 24, 34, 41)	9.4 (13, 4, 16)	72 and 120	Geometric mean of three separate studies	(American Cyanamid 1992e) with (BASF 2013b), (American Cyanamid 2000d), (American Cyanamid 1992d) with (BASF 2013b)
Navicula pelliculosa	35	12.2	120		(American Cyanamid 1992b) with (BASF 2013d)
Scenedesmus subspicatus	55	Not given	120	Raw data not available, 72-h EC50 not calculated	(Cyanamid International Corp 1985)
Anabaena flos-aquae	> 174	174	120	No effect at the highest concentration tested	(American Cyanamid 1992c)
Lemna minor	102	3.2 (LOEC)	14 d		(BASF 2002a)
Lemna gibba	43	7.5	7 d		(BASF 2013g)
Geometric mean	49	11	The data are homogeneous so the use of the geometric mean was judged to be appropriate

In addition to observations for algae, two chronic (7 and 14 d) EC₅₀ values were noted for Lemna gibba and Lemna minor of 43 and 102 μg/L, respectively (BASF 2002a; 2013g). The lowest chronic NOEC/lowest-observed-effect concentration (LOEC) was observed for Lemna minor. However, in the study by BASF (2002a), recovery of the plant was also observed. After a 7-d recovery period, frond growth was similar to control and in all treatments, including the greatest tested concentration of 200 μg PND/L.

No reliable published chronic toxicity data for fish and invertebrates were found; however, findings were available from tests conducted by BASF (Table 9). The three life-cycle experiments in fish were in reliable agreement with NOEC reported from 20 to 50 μg PND/L and GM of 43 μg PND/L. The experiment on Pimephales promelas was not performed under Good Laboratory Practices (GLP). However, the experimental design was interesting, allowing the observation of several endpoints over two generations of fish. The most sensitive endpoint observed was survival of fry during the first 30 d of exposure. It should be noted that a continuous exposure for 180–288 d is highly unrealistic for a pesticide in the aquatic environment, even for relatively persistent chemicals. It is also worth noting that, in all bioconcentration experiments discussed in the preceding, no adverse effects were noted in chronic exposure concentrations up to 20 μg/L, confirming results of life-cycle experiments.

Table 9. Chronic toxicity values for pendimethalin in invertebrates and fish.

Scientific name of species	NOEC (μg/L)	Duration (d)	Design	Reference
Daphnia magna	15	21	Flow-through	(American Cyanamid 1981)
Chironomus riparius	138	30	Static	(American Cyanamid 1994b)
Oncorhynchus mykiss	50	21	Static renewal	(American Cyanamid 1994a)
Pimephales promelas	22–36	288	Full life cycle. Flow-through (non GLP)	(American Cyanamid 1993)
Danio rerio	50	184	Full life cycle, static	(BASF 2001e)
Danio rerio	20	179	Full life cycle, static sediment present	(AGAN 2012)
Geometric mean	43	The data are homogeneous and the responses in the full life cycle studies are based on the early life-stage so the use of the geometric mean was judged to be appropriate

It is not surprising that a herbicide is toxic for primary producers (algae and plants). Nevertheless, the GM of available data (11 μg/L, Table 8) showed a chronic NOEC value greater than the threshold of 10 μg/L indicated as the T criterion by EC 1107/2009. Only some tests developed by BASF indicated lower NOEC values for Lemna. However, it is important to note that, even at the highest concentrations tested (up to 200 μg PND/L), recovery of the plants was complete and rapid, reaching the same rate of growth as the control after 1 wk. This is particularly important considering the exposure patterns of pesticides in surface water, characterized by event-driven peaks. Toxicity in animals was lower, always above the criterion, even in tests performed using unrealistically long exposure. These lines of evidence indicate that PND would not be considered as fulfilling the criterion of T for the aquatic ecosystem as a whole.

Studies in Cosms

Several investigations on the influence of PND have been conducted in aquatic cosms (Table 10). Despite the complexity of the cosm studies, the results were all within a similar range and consistent with findings of T tests conducted in the lab. This provides strong support for reliability of the results obtained in these studies and clearly reduces the uncertainty with respect to risk of PND to aquatic ecosystems. Using a conservative approach, a GM value of 13.8 µg PND/L is suggested on the basis of the no-observed-adverse-effect concentration (NOAEC) of cosm studies. Considering the rapid recovery observed in the cosms, any potential short-term effects on a few aquatic organisms—should they occur—will not have significant ecological consequences.

Table 10. Summary of the results obtained in the cosm studies (without fish).

Test system	Duration (d)	Endpoints (μg/L)a	Reference
Aquatic community in outdoor cosm	128	NOEC = 0.23	(BASF 2001b)
		LOEC = 1.1
		NOEAEC = 150
		NOEAEC (conserv.) = 4.9
Aquatic community in outdoor cosm	70	NOAEC = 10	(BASF 2004a)
Aquatic community in outdoor cosm	84	NOEC = 1	(BASF 2008)
		NOEAEC = 16
Aquatic community in outdoor cosm	140	NOEC = 3.8	(BASF 2010b)
		NOEAEC = 46

Note. See SI for detailed discussion of the data from the mesocosms.

^aNOEC = no observed effect concentration; LOEC = lowest observed effect concentration; NOEAEC = no observed ecologically adverse effect concentration.

Long- and Medium-Range Transport

Regulation EC 1107/2009 explicitly addresses long-range environmental transport (LRT) in Article 3.7.1.3. The criteria are the same as those defined in the UNECE Aarhus protocol (United Nations Economic Commission for Europe 1998) and the United Nations Environmental Programme (UNEP) Stockholm Convention (UNEP 2001) on Persistent Organic Pollutants (POPs): “DT50 > 2 days in air or monitoring data or environmental fate properties and/or model results that shows long-range transport.”

Fate and Transport in Air

PND is moderately volatile and has the potential for migrating significantly through the air. For such substances, the LRT criterion of DT50 in air is >2 d. The major degradation pathway of PND in air is the reaction with •OH-radicals in the gas phase. A reaction rate constant of 3.03 × 10^–11 was calculated by AOPWIN (U.S. EPA 2016). Assuming a global •OH half-day concentration of 1.5 × 10⁶, the DT50 of PND is estimated to be 4.24 h or 0.35 d, which is well below the criterion of 2 d for LRT. Direct photolysis, degradation by •OH and ozone were experimentally determined in an inert solvent perfluorohexane, which was employed to mimic reactions that might occur in the atmosphere (BASF 2001f). DT50 values of 12 h due to direct photolysis and 8.5 h due to reaction with •OH were found; the latter is two-fold longer than the value calculated with AOPWIN.

PND was observed to be partially adsorbed to atmospheric particles depending on ambient conditions, for example, air temperature and humidity, particle concentration, and composition. Mai et al. (2013) quantified a particle-associated PND fraction of 13% in the atmospheric boundary layer of the North Sea at 11–16°C. A comparison with observations of land-based sampling campaigns in winter (−3 to 6°C) demonstrated that low ambient temperatures favored adsorption of PND to atmospheric particles with a maximum of 38% at −3°C. Partitioning of hydrophobic substances to aerosols can be estimated by correlations to the vapor pressure (Pankow 1987) or the octanol–air partition coefficient K_oa (Harner and Bidleman 1998) or more mechanistically by polyparameter linear free energy relationships (pp-LFER) (Endo and Goss 2014). Goetz et al. (2007) suggested for polar chemicals such as PND that the pp-LFER approach be adopted whenever solvation parameters are available (Stenzel, Goss, and Endo 2013 for PND). Partition coefficients log K_aerosol of PND calculated with the pp-LFER approach at 15°C (http://www.ufz.de/index.php?en=31698) lie between 2.51 and 3.49 m³/g aerosol for various European locations and seasons with an arithmetic mean of 2.97 ± 0.27 m³ air/g aerosol. Log K_aerosol values for spring and fall are near the mean, and lowest for summer and highest for winter.

The fraction of adsorbed substance mass Θ can be calculated by

with TSP the total solid particulate matter (µg/m³).

Assuming typical concentrations of aerosols in rural and global environments of 30 μg/m³ and a temperature of 15°C for the PND application time in spring and autumn, a particle-associated fraction of 2.7% can be calculated using log K_aerosol estimated with the pp-LFER approach.

Particle-adsorbed substances are sheltered against photochemical reactions due to competing reactions with organic material and water film of aerosols (Goetz et al. 2008). Socorro et al. (2015) confirmed these observations for PND and found lifetimes of 53.8 d for reactions of PND coated on silica particles with •OH and ozone, which is lower than in the gaseous phase. However, data indicating that PND is persistent in real environmental atmosphere and thus subject to LRT are not valid since only a fraction of PND is attached to aerosols and aerosols are effectively deposited to water and land surfaces. Mai et al. (2013) reported deposition of PND to seawater in the German Bight. The fugacity ratio (f_water/f_air) is a measure for the net gaseous flux between water and air (Bidleman and McConnell 1995). Bidleman and McConnell (1995) found values in the range of 0.2–0.9, indicating at least partially net gaseous deposition.

Modeling of Medium- and Long-Range Transport

Various models for LRT have been developed since the mid 1990s, and differ in their spatial and temporal resolution (Fenner et al. 2005). ELPOS as a regional multimedia model (Beyer and Matthies 2002; Matthies et al. 2009) was selected. In addition, the OECD P_ov and LRTP Screening Tool (Wegmann et al. 2009) was utilized as a standard generic model for global transport. Both models are freely available: ELPOS 2.1 (ELPOS 2016) and OECD Tool (OECD 2016).

ELPOS is based on the regional multimedia model, which is used for the PEC_regional as part of the exposure assessment under REACH. Additional features are described in Matthies et al. (2009). ELPOS calculates the characteristic travel distance (CTD, Bennett et al. 1998) as a metric of long-range transport potential (LRTP) of a chemical, either emitted into air or water. The OECD Tool is similar to ELPOS but assumes a global geometry without marine sediment. Two indicators, CTD and surface transfer efficiency, are calculated as metrics for LRTP.

Table 11 presents results of model calculations for PND with the OECD Tool and ELPOS at two temperatures. Input parameters are partition coefficients from Table 1, and the GM of degradation half-lives in field soil (72 d, Table 3) and in total water–sediment systems (19.7 d, Table 4). Half-life in air was assumed to be 4.2 h as determined by AOPWIN (see earlier discussion), and half-life in fresh water (193 d) was taken as a conservative value from the OECD 309 testing in the dark (EFSA 2016). All calculated values are far less than bounds derived for POPs and demonstrate that PND is only transported in air and water over small- and medium-range distances. Both models deliver similar results for the CTD in air and water at 20°C. It is interesting that CTD in water is greater than in air at 20°C for both models, which is due to the large half-life in water compared to the short half-life in air. Aqueous photolysis was not considered and would result in a reduction of estimated transport distances, particularly in water. Even if, in ELPOS, the particle-associated fraction is enhanced to 22%, the CTD in air at 12°C increases up to 150 km, which is still medium-range transport. In addition, doubling of the photodegradation half-life would result in a CTD in air of 286 km at 12°C (BASF 2001f). It should be noted that, in both models, degradation in air is applied to the gas phase only; that is, particle-bound substance is assumed to be totally shielded against photodegradation.

Table 11. Results from models and model scenarios for atmospheric and aquatic transport of PND.

Emission to	Model	POV (d)	CTD in air (km)	CTD in water (km)	Surface transfer efficiency (%)
Air	OECD Tool 20°C	0	87	–	1.28 × 10^–5
Water	OECD Tool 20°C	93	–	160	5.58 × 10^–6
Soil	OECD Tool 20°C	60	–	–	2.33 × 10^–6
Air	ELPOS 20°C	1	89	–	–
Water	ELPOS 20°C	32	–	96	–
Soil	ELPOS 20°C	104	–	–	–
Air	ELPOS 12°C	1	126	–	–
Water	ELPOS 12°C	51		139	–
Soil	ELPOS 12°C	147	–	–	–

In conclusion, there is no apparent evidence that PND is a potential persistent organic pollutant (POP) under the criteria for classification in the Stockholm Convention (United Nations Environmental Programme 2001); however, transport over moderate and long distances (United Nations Economic Commission for Europe 1998) is considered as a criterion under EC 1107/2009. Short- and medium-range range transport is largely driven by volatility of PND but is mitigated by the rapid rate of degradation in the atmosphere. The model results are supported by findings of Mai et al. (2013), who attributed airborne PND concentrations in the German Bight to emissions in England during spring and fall by calculating backward trajectories. However, as mentioned in Sections 6 and 7 of the Review Report for the Active Substance Pendimethalin (European Community 2003) and in the recently published EFSA Conclusion on PND (EFSA 2016), a separate analysis of concentrations in semiremote locations and close to area of use was provided (see next section).

Concentrations Measured in Local, Semiremote, and Remote Locations

One of the criteria for LRT is the presence of a chemical in remote locations (United Nations Economic Commission for Europe 1998). The following subsections discuss the results of studies on concentrations of PND in local, semiremote, and remote locations.

Concentrations Close to Areas of Use: Air and Rainwater

A number of investigations measured levels of PND in areas close to the main regions of use (Table 12). Mai et al. (2013) noted peak concentrations in air declined rapidly, likely through deposition from the gas phase and/or degradation (not investigated in the study). Overall, the quantities measured in air were small in comparison to those at sites of application (discussed earlier) and do not present a risk to humans or environment.

Table 12. Summary of concentrations of pendimethalin measured in air and rainwater in areas close to locations of use.

Date of sampling	Location	Concentration measureda	Comment	Reference
January 2000 and July 2001,	Roskilde and Oure in Denmark	0.019 to 0.386 μg/L	Rainwater samples	(Asman et al. 2005)
2004, and 2007	Area of high use of PND in France	Maximum concentration for all years and sites was 117 ng/m^3	Air samples (n = 1157) taken in 31 locations	(BASF 2009)
October 2005 to October 2006	Cloud forest in Costa Rica	<LOD (0.00025)–0.030 ng/m^3	Air samples from passive samplers	(Shunthirasingham et al. 2011)
October 17, 2005, and September 11, 2006	Belen, Costa Rica	0.001 to 0.03 ng/m^3	Air samples (24 h) from a high-volume pumped sampler	(Gouin et al. 2008)
2006 to 2008	Centre Region of France, three rural and two urban sites	0.0007 to 0.117 ng/m^3	Air samples from a low-volume sampler	(Coscollà et al. 2010)
2009 and 2010	German Bight	0.0005–0.014 ng/m^3 in 2009 and 0.003–0.026 ng/m^3 in 2010	Land-based passive samplers	(Mai et al. 2013)
2009 and 2010	German Bight and North Sea	0.0005–0.011 ng/m^3 in 2009	Ship-based high-volume samplers	(Mai et al. 2013)
2009 April	Baltic Sea	0.0001 to 0.01 ng/m^3	Ship-based high-volume samplers	(Mai 2012)

^aSee SI for more details on the sampling and analysis.

Concentrations in Surface Waters and Groundwaters

Several studies determined the concentrations of PND in surface waters and groundwaters and are summarized in Table 13 and 14. Based on data for levels of PND in surface waters and groundwaters in all U.S. states from the National Water Quality Monitoring Council (NWQMC 2016), concentrations of PND in surface waters and groundwaters (Table 14) were generally small and not of biological relevance. The few large values as indicated by the maxima are likely due to spills or misuse and do not represent general use or good agricultural practice.

Table 13. Summary of concentrations of pendimethalin measured in surface waters in areas close to locations of use.

Date of sampling	Location	Concentration measureda	Comment	Reference
October 2005 to October 2006	Cloud forest in Costa Rica	<LOD (unstated)	Four water samples from mountain lakes and one from epiphytes in trees.	(Shunthirasingham et al. 2011)
2009	Jiulong River estuary in China	<LOD (0.06) to 0.98 μg/L	Water samples at 35 sites	(Zheng et al. 2016)
2008	Baltic Sea	0.014–0.133 μg/L	Twenty samples at 6 m depth	(Mai 2012)
2009 and 2010	German Bight	<LOQ–0.00021 μg/L in 2009 and 0.0001–0.0004 μg/L in 2010	Grab samples at 5 m depth	(Mai et al. 2013)
2009	North Sea	<LOQ–0.00005 μg/L	Grab samples at 5 m depth	(Mai et al. 2013)
June–July and September–October 2014	Glacial-fed streams of the Italian Alps	<LOD (0.01 μg/L)	Surface water grab samples	C. Ferrario, 2016, personal communication

^aSee SI for more details on the sampling and analysis.

Table 14. Concentrations of pendimethalin in measured in surface-waters of the United States.

Source	Number of samples	Concentration (μg/L)a			Frequency of detects (%)
		90th centile	99th centile	Maximum
Surface water	62,616	0.0	0.094	255	5.65
Groundwater	23,558	0.0	0.0	150	0.24

^aSee SI for more details on the analysis of data.

Similar to many other pesticides, PND undergoes short-range transport. The concentrations in air, even close to area of application, are small (<120 ng/m³) and not of biological relevance to humans or environment. Levels in air decline rapidly with distance from application, and some deposition via adsorption to particulates might be occurring in regions near and downwind of use areas. Overall, these data are consistent with the physical properties of PND (minimal solubility in water and large K_OC) and do not suggest that PND is entering surface waters or coastal marine systems in amounts that are ecotoxicologically significant or relevant to the environment or humans.

Concentrations in Remote Regions

Pendimethalin was detected in small concentrations in some remote locations but not in others (Table 15). Most of the detections in remote and semiremote regions were in experiments conducted before 2000, even though some of the newer studies that collected samples for the 1990s in ice cores did not report finding PND. Why the frequency of detections decreased in recent years is not clear, as analytical methods have improved; however, amounts used in agriculture and application methods might have changed over that period. Based on results of recent studies, LRT and deposition of PND to remote locations are not occurring. Overall, residues of PND in remote and semiremote areas are either not detectable or minimal. Concentrations of traditional POPs such as the “dirty dozen” are almost always detectable and often present in greater amounts. This line of evidence suggests that PND is not transported to remote regions in biologically significant amounts and that it is not a substance subject to LRT in the sense that several other current-use pesticides (CUPs) are (Zhang et al. 2013). The concentrations in ice cores and in air are almost three orders of magnitude less than those detected in areas close to where PND is used and are not of toxicological concern.

Table 15. Summary of concentrations of pendimethalin measured in remote regions.

Date of deposition	Location	Concentration measured^a	Comment	Reference
1953–2005	Holtedahlfonna (Svalbard)	<LOD (0.000001 μg/L)	Samples of ice cores	(Ruggirello et al. 2010).
1992–1998	Austfonna	0.019 μg/L	Samples of ice cores	(Hermanson et al. 2005)
1993–1994	Russian Arctic	0.31 to 0.92 μg/L	Samples of ice; contamination of samples might have occurred	(Boyd-Boland, Magdic, and Pawliszyn 1996)
1993–2008	Devon Island Ice Cap	<IDL (0.00014 ng/L)	Samples of ice cores	(Zhang et al. 2013)
2008 and 2009	Canadian Arctic	<IDL (0.6 µg/kg wet weight)	Terrestrial vegetation in Nunavut	(Morris et al. 2014)
2007 and 2010	Canadian Arctic	<MDL (117 pg/sample from 183 to >400 L	Marine water in three locations	(Morris et al. 2016)
2005–2008	Global program in all continents	The 90th centile concentration was 0.1 µg/sampler; no detections in remote, high-latitude locations	Air samples collected with passive samplers	(Shunthirasingham et al. 2010)
April–June 2009	Global program in all continents	Of four samples in remote locations, only one from Iceland was >LOD (0.27 ng/m^3)	Air samples collected with passive samplers	(Koblizkova, Lee, and Harner 2012)

^aSee SI for more details on the sampling and analysis.

Conclusions

The results of several studies on P for PND in soil and sediment/water provide evidence that PND should not be classified as P. The GM DT50 for aerobic degradation of PND in soil in lab simulation tests was less than the criterion value of 120 d for P. The most appropriate GM DT50 is 98 d, taking into account all valid investigations with relevant soils. Moreover, there is sufficient evidence from a large number of field experiments that demonstrates that the DT50 of PND is less the criterion value (GM normalized DT50 of 72 d) and is therefore not flagged as being P or vP. For water–sediment systems, DT50 values for total system degradation ranged from 4 to 103 d with GM of 19.7 d. Even the maximal DT50 value was below the criterion value of 120 d for freshwater sediment. This conclusion was corroborated in aquatic cosm studies. For aquatic systems, DT50 values in artificial as well as natural waters were less than the criterion value of 40 d. Further, PND rapidly partitions to sediment, where it is degraded and/or bound to sediment. Both processes reduce the bioavailability for aquatic organisms, which was observed in cosm experiments. Based on these data, PND should not be classified as P or vP in soil, sediment, and aquatic systems.

Bioconcentration of PND does occur in fish and some of the BCF values exceed the criterion of 2000; however, the GM was 1878, less than the criterion for classification as B or vB. In addition, elimination and metabolism was rapid. Biomagnification through diet in fish did not occur (BMF <1). No bioconcentration was observed in sediment-dwelling organisms or from soil into organisms, and no trophic transfer was noted in aquatic or terrestrial food chains. All these lines of evidence indicate that the potential for bioconcentration of PND is moderate with no potential for biomagnification.

As an herbicide, PND is not expected to be T to aquatic animals. Even in aquatic plants, it was not highly T; the GM of available chronic data (NOEC of 11 μg/L) did not trigger the T criterion of 10 μg/L of EC 1107/2009. Toxicity in animals was lower, always above the criterion, even in tests performed using unrealistically long exposure. Studies in aquatic microcosm corroborate lab data; a GM value of 13.8 µg PND/L was calculated for several cosm experiments. Taken together, these lines of evidence indicate that PND do not trigger the criterion of T for the aquatic ecosystem as a whole.

There is no apparent evidence that PND is a potential POP under the criteria for classification in the Stockholm Convention; however, transport over moderate and long distances is considered as a criterion under EC 1107/2009. Based on half-life in air of 0.35 d, which is well below the criterion of 2 d, PND is not classifiable as a substance undergoing LRT under UNECE. Results of studies in remote regions suggest that PND is not transported to these locations in biologically significant amounts and that it is not a substance subject to LRT in the sense that several other legacy and CUPs are. The few measured concentrations in older ice cores and in air are almost three orders of magnitude less than those detected at sites close to where PND is used and are not of biological concern.

Short- and medium-range range transport is largely driven by volatility of PND but is mitigated by the rapid rate of degradation in the atmosphere. Transport over moderate and long distances is considered as a criterion under EC 1107/2009. Similar to many other PPP, PND undergoes short- and medium-range transport. This is evidenced in models and values measured in the environment. Concentrations in air decline rapidly with distance from application, and some deposition via adsorption to particulates might be occurring in regions near and downwind of use-areas. Overall, these data are consistent with the physical properties of PND (minimal solubility in water and large K_OC) and do not suggest that PND is entering surface waters or coastal marine systems in amounts that are ecotoxicologically significant or relevant to the environment or humans.

Availability of Unpublished Reports

The data in unpublished reports have been summarized in the peer review of EFSA (EFSA 2016). Additional information on the unpublished reports is available in the dossiers and their summaries as submitted to EFSA. See the website http://registerofquestions.efsa.europa.eu and search for “Pesticide Dossier Pendimethalin.”

Acknowledgments

The authors thank BASF for providing unpublished reports and acknowledge funding for the study from BASF. The approaches used and the opinions in this article are those of the authors only and do not represent those of BASF or the author’s institutions.

SUPPLEMENTAL DATA

Supplemental data for this article can be accessed on the publisher’s website.