Journal list menu

Volume 50, Issue 5 p. 1110-1122
Open Access

Application of the Danish pesticide load indicator to arable agriculture in the United Kingdom

Kathleen Lewis

Corresponding Author

Kathleen Lewis

Agriculture and Environment Research Unit, School of Life and Medical Sciences, Univ. of Hertfordshire, Herts, Hatfield, AL10 9AB United Kingdom


Kathleen Lewis, Agriculture and Environment Research Unit, School of Life and Medical Sciences, Univ. of Hertfordshire, Hatfield, Herts, AL10 9AB United Kingdom

Email: [email protected]

Contribution: Conceptualization, ​Investigation, Methodology, Project administration, Software, Validation, Writing - original draft

Search for more papers by this author
James Rainford

James Rainford

FERA Science Ltd, York Biotech Campus, Sand Hutton, York, YO42, 1LZ United Kingdom

Contribution: Conceptualization, Formal analysis, Methodology, Project administration, Software, Validation, Visualization, Writing - original draft

Search for more papers by this author
John Tzilivakis

John Tzilivakis

Agriculture and Environment Research Unit, School of Life and Medical Sciences, Univ. of Hertfordshire, Herts, Hatfield, AL10 9AB United Kingdom

Contribution: Conceptualization, Formal analysis, ​Investigation, Methodology, Software, Writing - original draft

Search for more papers by this author
David Garthwaite

David Garthwaite

FERA Science Ltd, York Biotech Campus, Sand Hutton, York, YO42, 1LZ United Kingdom

Contribution: Conceptualization, Formal analysis, Funding acquisition, Methodology, Resources, Supervision, Writing - original draft

Search for more papers by this author
First published: 04 July 2021
Citations: 3

Assigned to Associate Editor Ke Sun.


Pesticides are an important component of worldwide agriculture systems and have contributed to significant increases in crop quality and yields and therefore to food security. However, despite their societal benefits, pesticides can be hazardous to humans and the environment. Therefore, effective pesticide polices are needed that balance the societal and economic benefits with the unintentional and undesirable environmental and health impacts. As a result, there has been consistent policy interest in pragmatic and practical techniques that are suitable for assessing the environmental and human health implications of agricultural pesticide use from a national perspective for assisting in the development of policy initiatives and for communicating policy outcomes to the public. The work described herein explored the appropriateness of the Danish Pesticide Load Indictor for assessing agricultural pesticides applied in the United Kingdom from 2016 and 2018. The findings for the two datasets appear broadly comparable, suggesting that the overall environmental load from pesticides on the U.K. environment remained relatively constant during this period. Regional differences in environmental load and the major contributing substances were identified. Where large differences between the two years were seen, regulatory interventions appear to have been the cause. Overall, the indicator behaves as expected and appears to be sufficiently responsive to changes in pesticide use. However, various concerns were identified that may lead to modifications in how the indicator is calculated and what parameters are included to make it better able to deliver U.K. policy objectives.


  • DT50
  • soil half-life
  • EIQ
  • Environmental Impact Quotient
  • NAP
  • U.K. National Action Plan for the Sustainable Use of Pesticides
  • PL
  • pesticide load
  • PLI
  • pesticide load indicator
  • Pesticide Occupational and Environmental Risk
  • PPDB
  • Pesticide Properties Database
  • PUS
  • Pesticide Usage Survey.

    Pesticides are an important component of worldwide agriculture systems and have contributed to significant increases in crop quality and yields. According to the Food and Agricultural Organization (FAO, 2020), the production of major crops has more than tripled since 1960, largely due to pesticides; therefore, these chemicals have improved food security and the viability of farming businesses (Cooper & Dobson, 2007; Verger & Boobis, 2013). However, despite their societal benefits, pesticides can be hazardous to humans and the environment. When used for agriculture, pesticides tend to be applied over large areas of land and thus have the potential to disperse in the environment and contaminate nontarget areas and poison nontarget species. The consequences are well documented (Bourguet & Guillemaud, 2016; Dereumeaux et al., 2020; Kim et al., 2017; Lee et al., 2019; Sánchez-Bayo & Wyckhuys, 2019; Sud, 2020), and therefore effective pesticide regulatory processes and sound policies are needed that balance the societal and economic benefits with the unintentional and undesirable environmental and health impacts.

    To facilitate risk mitigation, the nature of risks themselves in terms of their occurrence, severity, and location need to be understood. However, the factors that affect these issues are multiple and highly varied, including the physicochemical properties of the active substance (such as its rate of degradation and its mobility, solubility, and volatility) and its toxicity to nontarget biodiversity. In addition, the active substance formulation, application method, timing, and dosage as well as the environmental conditions (including soil characteristics, weather, and topography) are important. Therefore, risks arising for a specific active substance are location specific as well as species specific, making it extremely difficult to accurately quantify these risks from a regulatory perspective. Within the European Union (EU), several standard scenarios (e.g., Centofanti et al., 2008; Pereira et al., 2017), known as the FOCUS (the FOrum for the Co-ordination of pesticide fate models and their Use) scenarios, are used to evaluate risks to specific environmental compartments (e.g., surface water, groundwater) using regulatory approved and validated mathematical models. There are many such models. For example, PEARL, a model for Pesticide Emission Assessment at Regional and Local scales, describes the fate of a pesticide in the soil-plant system (Tiktac et al., 2012); PELMO, the Pesticide Leaching Model, simulates the vertical movement of pesticides in soil and can be used to predict pesticide leaching (Klein et al., 1997); PRZM, the Pesticide Root Zone Model, is often used to predict runoff from agricultural fields (Marin-Benito et al., 2020); and MACRO, which is a preferential flow model used to assess pesticide leaching and drainage (Marin-Benito et al., 2020; Nolan et al., 2008). There are also other models used to consider wider concerns. For example, PRIMo is the European Food Safety Authority pesticide residue intake model used to perform dietary risk assessment for pesticide residues (EFSA et al., 2018), and the GLEAMS model simulates different management systems, including crop rotations, tillage practices, conservation practices, irrigation, drainage, fertilizer practices, and pesticide treatments, among others (Leonard et al., 1995; Rekolainen et al., 2000).

    Although these models are undoubtedly invaluable for pesticide regulation, they are less useful for assessing the environmental and human health implications of pesticide use from a national perspective, for devising policy initiatives, or for communicating policy outcomes to the public. These types of models tend to consider the risks arising from of a single pesticide applied to a specific crop rather than the multitude applied across a nation and do not consider the scale of use; these issues are highly pertinent from a policy perspective. Consequently, there has been consistent policy interest in the development of more appropriate techniques for evaluating and monitoring the potential nontarget effects to enable temporal trends to be identified. Establishing policies that promote a reduction in the impacts of pesticides is far from straightforward, and there are many challenges for the development of these types of tools. Not only do they need to be reflective of the differences between the multitude of pesticide active substances available in terms of their environmental and human health impacts and be sensitive to changes in the quantities applied, they need to account for the diverse and ever-changing composition of pesticides being applied due to the development of new active substances and regulatory changes, such as the removal from the market of older active substances with known harmful impacts (Milner & Boyd, 2017; Schäfer et al., 2019).

    Core Ideas

    • The Danish Pesticide Load Indicator is sensitive to U.K. regulatory interventions.
    • Minimal change in the PLI was seen between 2016 and 2018, but regional differences were noted.
    • Results show a PLI decline associated with the use of insecticides between 2016 and 2018.
    • Issues relating to data confidence and linearity of the indicator are highlighted.

    To assess the multi-dimensional risks around pesticide usage in a consistent, albeit simplified, manner, a variety of indicator frameworks have been developed, some of which are hazard based, whereas others attempt to consider risk. One of the earliest and simplest but perhaps the most used to date is the pesticide Environmental Impact Quotient (EIQ) (Kovach et al., 1992). This hazard-based indicator uses a scoring approach to determine the effect of pesticides on humans, groundwater, and biodiversity. It has been applied to a range of different applications, including cropping (Cross et al., 2006; Gallivan et al., 2001; Kleter et al., 2007; Kromann et al., 2011), comparing the impact of different herbicides (Kniss & Coburn, 2015), and tomato (Solanum lycopersicum L.) processing (Bues et al., 2004). Another well-cited approach is the Environmental Yardstick for Pesticides, which quantifies the risks of pesticide use at field, regional, and national levels. For each active substance the yardstick allocates environmental impact points for the risk to groundwater, aquatic species, and soil organisms (Bressers, 2014; Reus et al., 2000, 2002). In support of the pesticide reduction program of Flanders, Belgium, the Pesticide Occupational and Environmental Risk indicator (POCER) was developed (Claeys et al., 2005; Vercruysse & Steurbaut, 2002). This is comprised of several modules reflecting the risk to humans from occupational, nondietary exposure, and the risk to the environment. In Germany the SYNOPS indicator (Strassemeyer & Gutsche, 2010; Strassemeyer et al., 2017) supports the National Action Plan on the Sustainable Use of Plant Protection Products (a requirement under EU Directive 2009/128/EC, which aims to achieve more sustainable pesticide use across the EU) and assesses the acute and chronic pesticide risks to soil, surface water, groundwater, and pollinators. Many other indicators have been developed to support national pesticide risk reduction programs (e.g., Dosemeci et al., 2002; Kookana et al., 2005; Trevisan et al., 2009; Tsaboula et al., 2016).

    Perhaps two of the most notable indicators from a policy perspective are those of Norway and Denmark, with both forming the basis of a national pesticide tax system (Böcker & Finger, 2016; Finger et al., 2017; Sud, 2020). The Norwegian system (Spikkerud, 2005) is based on a “cumulative environmental index” across all pesticide active substances used in the country annually to demonstrate the change in risk over time. It considers various physicochemical and fate properties of the pesticides applied and the ecotoxicity of these pesticides to a variety of terrestrial and aquatic taxa together with a surrogate measure of exposure, which is assessed according to the formulation type, the application method, the total application area, and a Standard Area Dose; the latter is set by the national pesticide authority and is based on the specific pesticide's maximum application rate for each crop it is applied to. The determined risk is then used to place the pesticide product into one of seven bands, which each have a set tax rate per treated area. The Danish system is described as an “environmental load” or pressure indicator (Kudsk et al., 2018; Miljøstyrelsen, 2012). The term “load” can be ambiguous in the context of assessing the environmental impact of pesticides. The Danish indicator does not try to account for actual harm or damage but aims to reflect the relative environmental pressure that occurs due to the differing hazardous nature of the pesticides used and the variability in quantities applied. The concept is that active substances that are used in large amounts and that are persistent, are mobile, bioaccumulate, and are ecotoxic to many species should have a higher load than those that are not. The indicator is comprised of three subindicators that aim to measure the potential pressure on human health, environmental fate, and ecotoxicity. These subindicators are then (a) used to determine the level of taxation and (b) combined with national usage data to enable the monitoring of usage trends and environmental load over time.

    In contrast to some countries, the U.K. National Action Plan for the Sustainable Use of Pesticides (NAP; Defra, 2013) has, to date, not made use of an indicator system such as those described above. Instead, a suite of simpler measures is used to monitor how pesticides are used and their impact. These measures include surveys of pesticide use; results of farm inspections; cropping statistics and rates of adoption and impact of industry initiatives including product stewardship, training, knowledge transfer; and monitoring the impacts of pesticide use on human health and the environment (Defra, 2013). The 2013 NAP is now being revised, with one of several aims being the reduction of risks associated with pesticides by improving metrics and indicators. Hence, a more detailed look at existing pesticide risk indicators and their appropriateness to the United Kingdom has been undertaken.

    The aim of the study described here was to explore, using a case study, the appropriateness of the Danish approach to the United Kingdom based on pesticide use on arable crops during 2016 and 2018. Selection of the Danish approach for this study was made at the request of the U.K. policy team, based on the Danish indicators proven track record in pesticide policy (Pendersen et al., 2015; Sud, 2020) as well as recent evidence of robustness when predicting extreme risks relative to other quantity-based indicators (Möhring et al., 2019). More complex indicators, such as the SYNOPS model or the POCER indicator, mentioned previously, have been discussed in a U.K. context but are reliant on localized data (e.g., relating to weather and soil conditions at the time of application) that cannot be meaningfully aggregated at the national scale for historic U.K. periods based on existing data collections (de Baan, 2020). By contrast, the structure of the pesticide load indicator (PLI) and its close links to existing resources, such as the Pesticide Properties Database (PPDB) (Lewis et al., 2016), mean that it can be straightforwardly and transparently adapted to a novel national context with minimal changes or requirements for further large-scale data collection. Other indicators, such as the EIQ (Kovach et al., 1992), were not seen as offering sufficient breadth in terms of coverage of environmental issues or ability to discriminate between active substances (Kniss & Coburn, 2015).


    2.1 The Danish pesticide load indicator

    The calculation of the Danish PLI has been described in detail previously (Kudsk et al., 2018; Miljøstyrelsen, 2012). In summary, a pesticide load (PL) is calculated for three subindicators (human health, environmental fate, and ecotoxicity) and expressed as the PL per unit of commercial product (e.g., liter, kilogram, standard dose, capsule, or tablet). This is shown schematically in Figure 1. The human health indicator focuses largely on operator exposure and is determined by assigning a score between 10 and 100 for the risk phrases for each pesticide active substance (defined according to the Classification, Labelling and Packaging Regulation No. 1907/2006 [EC, 2008]), with the highest score allocated to those substances most damaging to health. For example, a score of 10 is given if the active substance is a skin irritant or harmful if swallowed but a score of 100 if the active substance is fatal if swallowed or may cause genetic defects. Human health scores are summed and converted to “pesticide loading points” by dividing the total points by 300. The environmental fate indicator considers soil degradation expressed as the soil half-life (DT50), the SCI-GROW index (which is an indicator of mobility and leaching risk; USEPA [2007]), and bioaccumulation using the bioconcentration factor. The ecotoxicity subindicator is determined using acute and chronic toxicity threshold data (e.g., the dosage of a chemical that kills 50% of a sample population [LD50], the concentration of a chemical that kills 50% of a sample population [LC50], no observed effect concentration [NOEC], and lowest observed effect concentration [LOEC]) for a range of species (Figure 1). Data for the fate and ecotoxicity subindicators are taken from the PPDB (Lewis et al., 2016). The PPDB is a comprehensive relational database of pesticide chemical identity, physicochemical, human health, and ecotoxicological data collated from regulatory studies and peer-reviewed literature that is maintained by the University of Hertfordshire, U.K.

    Details are in the caption following the image
    Structure of the Danish Pesticide Load Indicator (PLI). Values given are numerical constants (weighting factors) used to aggregate standardized values. Where multiple values are listed the second applies to seed treatment, and the first to all other treatments. GW, groundwater

    For the environmental fate indicators, the “raw” data, as extracted from the PPDB, are converted to “pesticide loading points” by first identifying a reference pesticide, defined as the most harmful active substance for each parameter (e.g., the longest soil half-life), and all other substances are expressed relative to this reference active substance (i.e., the value for the pesticide divided by the value of the reference substance to give a score between 0 and 1). For the ecotoxicity indicators, the process is similar, except the reference pesticide is the one with the lowest value (i.e., greatest toxicity) and an inverse relationship is used to derive the loading points (i.e., the load value of the other substances is derived using the equation: 1/[active substance value/reference active substance]). This defines a standardization “curve” for each indicator that converts the raw data to an index value. The “pesticide loading points” for each measure are then determined by multiplying this value by a weighting factor, which is essentially the maximum number of PL loading points per unit of commercial product for that particular parameter. Because the weighting value varies from parameter to parameter (Figure 1), these weighting values allow issues or policy concerns, such as loss of pollinators or groundwater contamination, to be given greater relative significance within the subindicators. The weighting factors are also different depending on whether the product is applied as a field application or as treated seed (Figure 1). Each of the three subindicators is expressed as the sum of the weighted values of the underlying measures and given equal significance in the overall PLI.

    The final step is to take the pattern of usage into account. In Denmark farmers are obliged to report their pesticide usage for each growing season to authorities. These data are used to estimate a treatment frequency index by dividing the total amounts of active substances used in each crop by the standard doses assigned to each use of the active substance (Kudsk, 1989) for the whole country. This data are then combined with the total “pesticide loading points” to provide an estimate of the pressures placed on the nation from pesticide use.

    2.2 Application of the Danish PLI to the United Kingdom

    For this initial evaluation of the PLI in the U.K. context, the Danish approach was applied exactly in terms of input parameters and weightings to arable production in the United Kingdom. Consequently, the first stage of this work was to identify the pesticides used in this context. The parameters used and the identified reference active substances are shown in Table 1.

    TABLE 1. Fate and environmental parameters used in the U.K. Pesticide Load Indicator
    Subindicator Parameter U.K. Reference substance (value)
    Environmental fate
    Soil degradation soil half-life (DT50 as days) diquat (5,500 d)
    Soil mobility SCI-GROW: calculated from DT50 and the organic carbon sorption constant (ml g−1) flutriafol (5.13)
    Bioaccumulation bioconcentration factor (L kg⁻¹) pendimethalin (5,100 l kg⁻¹)
    Birds acute LD50 oxamyl (3.16 mg kg⁻¹)
    Mammals acute oral LD50 oxamyl (2.5 mg kg⁻¹)
    Fish acute 96 h LC50 tefluthrin (0.00006 mg l⁻¹)
    Fish chronic 21 d NOEC tefluthrin (0.000004 mg l⁻¹)
    Daphnia acute 48 h EC50 tefluthrin (0.00008 mg l⁻¹)
    Daphnia chronic 21 d NOEC lambda-cyhalothrin (0.000002 mg l⁻¹)
    Algae acute 72 h EC50 picolinafen/bifenox (0.00018 mg l⁻¹)
    Aquatic plants acute 7 d EC50 clodinafop-propargyl (0.00019 mg l⁻¹)
    Earthworms acute 14 d LC50 beta-cyfluthrin (0.565 mg kg⁻¹ soil)
    Earthworms chronic 14 d NOEC imazaquin (0.028 mg kg⁻¹ soil)
    Honeybees acute 48 h LD50 deltamethrin (0.0015 μg bee⁻¹)
    • Note. LC50, concentration of a chemical that kills 50% of a sample population; LD50, dosage of a chemical that kills 50% of a sample population ; NOEC, no observed effect concentration.

    In contrast to Denmark, U.K. farmers are not required to report their pesticide usage to authorities. Instead, biennial agricultural; horticultural; and four yearly grassland, fodder, and amenity pesticide usage surveys (PUSs) are conducted for a representative subset of holdings to estimate the mass of individual actives applied within regions of the United Kingdom (Thomas, 2001). These data (in their aggregated form) are accredited National Statistics and can be accessed from the U.K. public repository (Fera, 2021). A list of pesticides used, both as sprays and seed treatments, was extracted from the two most recently published arable surveys from 2016 and 2018 (Garthwaite et al., 2018, 2019). For each of the pesticides identified from the surveys, data were extracted from the PPDB (Lewis et al., 2016), and the reference active substance was identified (Table 1). For each parameter, missing data were given default values based on a “reasonable worse case” taken as the 95th percentile (or 5th percentile for ecotoxicity where the lower the value the more toxic the substance) of all data for that parameter. Active substances with a soil degradation time of <1 d were represented as the weighted sum of the properties of their major metabolites, based on the formation fraction listed in the PPDB (Miljøstyrelsen, 2012).

    To estimate usage in the United Kingdom, 2016 and 2018 PUS data for arable cropping (Garthwaite et al., 2018, 2019) were used to estimate the amount of each active substance applied in each region and standardized holding size group (“strata” as defined in Thomas [1999]). The PUS data are collected as pesticide application records at field level derived from the recorded rate of application of each product (including seed treatments) and formulation used by a farmer. These data were then used to provide regional estimates of pesticide usage via a conservative bootstrap approximation, wherein the population of unsampled holdings for each active substance (taken from the Defra “June survey” [Defra, 2021] and assumed to be known without error) were modeled using random draws of the rates from those holdings sampled during the PUS. The bootstrap is explicitly conservative (in terms of proclivity to underestimate the use of a specific active) in that, where there were no sampled holdings within a strata where a particular active was applied, this was considered insufficient to provide an (implicitly errorless) estimate of the potential usage of that active for that population. In those cases, bootstrap sampling was instead based on a “conservative reference population,” comprising either on all sampled holdings within the region (regardless of size group) or sampled across different regions within each bootstrap replicate (such that every active had a change of a non-zero record of use in the population used to represent the unsampled holdings). Presented estimates of uncertainty correspond to the cumulative uncertainty on the aggregated value of the indicator given the 95% confidence interval for each active substance on each holding size-group within a region.


    As might have been expected due to the relatively short time between the two surveys, when the overall regional U.K. PLI values are calculated, there is limited evidence for significant differences between estimates for 2016 and 2018. The estimates of application mass across all pesticides combined are quite similar for the two years and are within the estimated margin of uncertainty for all regions (Figure 2).

    Details are in the caption following the image
    Estimated raw mass of pesticide application for the different U.K. regions for the two case study years. Confidence limits (95%) are based on the cumulative estimate of the conservative bootstrap of mass of application based on data from the U.K. Pesticide Usage Survey

    Estimates of regional variation in the value of U.K. PLI are largely in line with the mass of pesticide application, with the highest overall loadings associated with the more intensively farmed East Midlands and Eastern regions (Figure 3).

    Details are in the caption following the image
    Total calculated U.K. Pesticide Load Indicator per U.K. region for arable cropping in 2016 and 2018. Confidence limits (95%) are based on the cumulative estimate of the conservative bootstrap of mass of application based on data from the U.K. Pesticide Usage Survey multiplied by total “pesticide loading points” for each substance across all measures

    Figure 4 provides a breakdown of the contribution of different pesticide groups to the components of the indicator when aggregated nationally. Herbicides are the most important group overall (comprising 78% of the overall value in 2016) in determining the value of the U.K. PLI, followed by fungicides and insecticides. This pattern was also observed within the majority of years in Denmark (Miljøstyrelsen, 2012).

    Details are in the caption following the image
    Contribution of pesticide groups to the value of the U.K. Pesticide Load Indicator (PLI) across years. Values are given as the mean estimate of the mass applied

    Insecticides are the group that makes the largest contribution to the indicator calculation relative to their mass of application and are the group undergoing the largest change between the two years sampled. In both datasets, insecticides make up approximately 1% of the mass of pesticide application, but their contribution to the U.K. PLI value was approximately 10 and 7% for 2016 and 2018, respectively. Given their reason for use, it is unsurprising that the component most affected by insecticides is environmental toxicity (Figure 4, Ecotoxicology), which shows a notable decline in both the total value (4,260,723 PLI units in 2018 compared with 4,883,026 PLI units in 2016) and in the contribution of insecticides (28.4% in 2016 compared with 19.1% in 2018). This suggests a decline in the hazard associated with the use of insecticides between the two surveys.

    Fungicides are most significant for their contribution to the human health component of the index, wherein they comprise over half of the load associated with operator exposure. Beyond reinforcing the significance of fungicides when considering mitigation of operator exposure (particularly given the increasing importance of tank mixing as a technique for combating development of resistance), there is limited evidence for major changes in fungicide application or loading observed between the two surveys.

    Most of the minor components contributing to the U.K. PLI show patterns of change largely in proportion to their mass of application, suggesting minimal shifts in the relative loading associated with the composition of active substances in use. Interesting patterns can be observed with respect to molluscicides in that an increase in overall application mass in 2018 is not associated with a notable increase in contribution to the U.K. PLI. In-depth investigations indicate this may be associated with a shift in compound use away from the high-impact actives, such as metaldehyde, and toward lower-impact alternatives, notably ferric phosphate.

    Identifying the specific active substances responsible for driving the changes in the U.K. PLI showed that, among the major contributors (here defined as any active substance that individually contributes ≥20% to the U.K. PLI or any of its subindicators), the largest absolute changes between 2016 and 2018 are associated with declines in the application of chlorpyrifos and cypermethrin, additional to a smaller-scale increase in the application of clothianidin seed treatments (Figure 5). Chlorpyrifos is an organophosphate compound with long-standing concerns over toxicity to humans, bees, and aquatic invertebrates (e.g., Rauh et al., 2012). The majority of products containing this compound were officially withdrawn for use within the United Kingdom in 2016, which was therefore the last year where wide-scale use of chlorpyrifos would be expected to be observed within the PUS, providing evidence that the PLI as defined is sensitive to the withdrawal of a high-impact active substance.

    Details are in the caption following the image
    Contribution of individual insecticides to the U.K. Pesticide Load Indicator (PLI) within this chemical group. Values are given as the mean estimate of the mass applied

    Cypermethrin is a nonsystemic pyrethroid that remains authorized for use in the United Kingdom on a wide range of agricultural crops as well as in veterinary medicine. Declines in cypermethrin usage are possibly related to substitution with the synthetic pyrethroid lambda-cyhalothrin, as well as shifts in the isomer used to favor the more rapidly acting zeta-cypermethrin.

    The only insecticide to show significant increases in overall usage between the two years is the neonicotinoid clothianidin when used as a seed treatment (Figure 5). This was mainly for wheat and to a lesser degree was also used with oilseed rape (Brassica napus L.; 2016 only), sugar beet (Beta vulgaris L.), and other winter cereals. There has been considerable controversy in recent years around the use of neonicotinoids, particularly with respect to impacts on pollinators (Lu et al., 2020; Osterman et al., 2019), and U.K. guidance now prohibits the outdoor use of clothianidin. Longer-term trends in neonicotinoid usage in the United Kingdom suggest a gradual increase in clothianidin usage since 2000 (Budge et al., 2015), but it is unclear from the 2016 and 2018 data what precise impact the change in regulations has had.

    The carbamate insecticide oxamyl is used for the control of various soil-based invertebrates and nematodes, particularly on potatoes (Solanum tuberosum L.) and sugar beet. An increase in contribution to the U.K. PLI can be seen in Figure 5 when comparing the data for the two surveys. However, data for the West Midlands region (Figure 6), which is subject to an overall significant change in U.K. PLI value (Figure 2), show a significant decrease in the contribution of oxamyl to the indicator value, suggesting that the overall patterns of change in the PLI are best understood at a regional level. Oxamyl was withdrawn from use on most arable crops in the United Kingdom in 2020. Lambda-cyhalothrin was also identified as undergoing inconsistent patterns of change across regions. This complex pattern of change highlights one of the key challenges facing the indicator approach as implemented as a policy tool in that, by nature, a single indicator aggregates over a wide range of different and potentially conflicting changes within the composition of pesticide usage. Understanding this underlying complexity is particularly important where the goal is to support intervention.

    Details are in the caption following the image
    Contribution of individual insecticides to the U.K. Pesticide Load Indicator (PLI) subindictors and the total. Values are given as the mean estimate of the mass applied

    Ethoprophos is a soil-incorporated insecticide used to control wireworm and potato cyst nematode infestations. It is considered to be an acetylcholinesterase inhibitor and may also be genotoxic (EFSA, 2018). Hence, as may be expected, it has a relatively high human health contribution to the U.K PLI. A notable decline in usage occurred between 2016 and 2018.  Although it was important for the human health component in both years, the decline in 2018 changed the contribution to the overall U.K. PLI only slightly. Ethoprophos no longer has approval for use in the United Kingdom, having been withdrawn in 2019.


    Broadly speaking, the calculated U.K. PLI behaves as expected and provides a good visual representation of the impacts of arable agricultural use. When comparing the two datasets, reasonable explanations were evident when significant changes in the indicator value were seen. However, several observations were made during the study highlighting issues of concern that may mean modifications to the Danish methodology are required if it is to be used in the U.K. policy context.

    A key challenge is ensuring the U.K. PLI is suited to the U.K. geological and hydrological landscape. The Danish PLI is heavily focused on the protection of groundwater, which reflects the fact that more than 99% of water use in Denmark has groundwater as its source (Jørgensen & Stockmarr, 2009); hence the use of the SCI-GROW parameter. However, in the United Kingdom, surface water is the main (∼68%) source of tap water, with only about a third being extracted from groundwater; this value is much less in Scotland (3%) and in Northern Ireland (6%) (Water UK, 2021). Consequently, some rebalancing in favor of protecting surface water quality may be required within the U.K. PLI. There is also an issue with the SCI-GROW calculation in that data are sometimes not available to facilitate the calculation, meaning that a default is used and thus uncertainty is created. The Groundwater Ubiquity Score index (Gustafson, 1989) may be an alternative to SCI-GROW because there are fewer substances for which the Groundwater Ubiquity Score cannot be calculated.

    Resolving the implications of uncertainty within the index is a further challenge for the U.K. PLI. Care needs to be taken to ensure that uncertainty is appropriately represented, particularly given their comparatively large size relative to the observed trends between years. Although the principal source of the observed uncertainties is related to the need to estimate usage overall based on the finite sample of holdings recorded as part of the U.K. PUS, there are also concerns around parameter uncertainty. Some of the parameters, such as DT50 and soil adsorption coefficient, are known to vary substantially. The values held within the PPDB and used here (Lewis et al., 2016) are the geometric mean estimated across all available experimental data. For many active substances there are data for a comprehensive range of soils, but for others the parameter value may be based on just one or two soil types and so may not be truly representative. Improving the data and how it is handled and/or determining a measure of uncertainty may be needed.

    Other issues that were noted included the lack of chronic toxicity data for mammals and birds within the Danish PLI. At the time the Danish PLI was established, these data were not included within the PPDB, and, because no alternative suitable data source was found, these parameters were excluded from the Danish PLI. However, these parameters have now been added to the PPDB and so could quite easily be added to the indicator. Although it would be desirable to include chronic toxicity data for earthworms and wild bees (e.g., bumble bees and solitary bees), insufficient data are available in regulatory documents or peer-reviewed literature to make this exercise worthwhile, although it is expected that this issue will resolve in time as more data are generated for regulatory purposes (Lewis & Tzilivakis, 2019).

    One area where a potential solution is yet to be identified is the lack of coverage of the risks to consumers. Although the inherent toxicological properties of pesticides are the same regardless of the receptor, using the active substance risk phrases focuses largely on the operator. Consumer risks are not directly covered and are complex in that dietary exposure is dependent on another set of variables, such as dietary intake and consumer profile (e.g., age, cultural factors, underlying health issues, preparation/cooking/processing, etc.). Dietary risks also need to be combined with the need to consider surface water quality as the major drinking water source in the United Kingdom. There is, of course, the argument that as an “environmental load” consumer impacts are outside of the scope of such an indicator, given that this could implicitly lead to a tradeoff of human health and environmental concerns (Maud et al., 2001), but there is also a counter argument that consumers are the end receptors in the same way as wildlife are exposed when they feed on pesticide-treated crops. Resolving this issue, particularly in response to the different systems of regulation around these types of risk in the United Kingdom, is an area to be explored in future work.

    Finally, a major issue around the calculation of the PLI, albeit one shared with many other pesticide indicator methodologies (Reus et al., 2002), is the implied linear relationship between the weighted measures of impact and the mass of pesticide application. Under the current model the overall U.K. PLI for a given active substance in a region is simply a multiplication of the calculated indicator value and the mass of application; thus, a doubling of the application mass has the effect of doubling the indicator value. This procedure has the advantage of transparency and thus simplicity for policy but does mean that, to varying degrees, certain measures have the potential to be misrepresented in terms of their assumed impact. The extent to which this is a concern for the value of the indicator is dependent on whether in-field application rates lie outside the effectively linear part of the dose response curve, which may vary between actives.

    This linearity also has consequences for the role played by the reference active substances in determining the overall distribution of scores across the set of included pesticides. Because reference active substances are those that have the highest potential for impact per kilogram “load” with respect to a specific measure, it follows that the value of this active substance when compared to the remaining distribution can have a significant effect on the relative scoring of active substances, and thus on the relative contribution of measures, on the overall indicator value. There is also the potential for regulatory change to affect what is assigned as the reference substance. For example, if diquat were withdrawn in the United Kingdom, its role as the reference active substance for soil degradation would be called into question. If a new worse-case active substance is used, comparison with previous years would no longer be valid. However, if the original reference active substance is kept and other active substances with relatively high loadings are removed over time, then the reference active substance becomes more and more extreme, thus introducing the potential for distortions in the standardization “curve.” This can result in many typical active substances having extremely low score values, resulting in an indicator that disproportionally prioritizes a small number of high-impact pesticides. For example, diquat has a DT50 of 5,500 d, whereas the median across all active substances is 20 d; therefore, diquat would be associated with an unweighted score of 1, and the median substance would have an unweighted score of 0.064. As another example, an active substance that has a soil DT₅₀ of 366 d has a resulting load index of 0.07, but the regulatory interpretation of this value would be that it is very persistent; hence, there is a disparity between the load indicator value and the regulatory interpretation of the soil DT₅₀. The extent of deviation between the value of the reference active substance and those of more typical active substances is known to vary across different measures, which has the effect that their relative contribution for typical pesticides may be strongly distorted by the choice of reference substance.

    Finding a solution to the issues caused by linearity is not easy, but one potential way forward is to introduce “reference points” into the definition of the standardization “curve” for each load metric. The reference active substance would still be used, but regulatory threshold values could be introduced that set reference points between 1 and 0 to ensure the load metric corresponds with a suitable load value. For example, general interpretation of soil degradation values would consider a soil DT50 of <30 d as nonpersistent, 30–100 d as moderately persistent, 100–365 d as persistent, and >365 d as very persistent. These thresholds could be set to correspond to load values of 0–0.25, 0.25–0.5, 0.5–0.75, and 0.75–1.0. Thus, rather than a straight line between 0 and 5,500 d corresponding to a load index of 0–1, a series of straight lines define the standardization “curve,” resulting in load index values that are more in line with the regulatory interpretation of the metric. Using these thresholds, a soil DT₅₀ of 366 d would be given a load index of 0.75, compared with 0.07.

    The sensitivity of the Danish PLI methodology when adapted to a novel context can be viewed as the combined impact of missing/estimated data, linearity in response to change, and the manner in which weighting values are defined. A challenge for conventional sensitivity analysis is that, unlike other indicators previously explored in this way, such as PURE (Zhan & Zhung, 2013) and SYNOPS (de Baan, 2020), not all the parameters in the PLI have a meaningful and objective parameter space over which the performance of the indicator can be assessed (de Baan, 2020). For example, the weighting values defined in Denmark are used here without modification, in part due to a lack of consensus among consulted U.K. policy makers, as to how to define an appropriate parameter space for those that are inferred to reflect socio-political concern for the various elements of pesticide load (Miljøstyrelsen, 2012; a similar issue arises when considering how to vary the scores assigned to different risk phrases in the human health component). In the future it may be possible to define a measurable basis for weighting via calculating measures of revealed preference and willingness to pay (e.g., Travisi & Nijkamp, 2008; Travisi et al., 2006), although, due to the range of potential stakeholders involved, this may raise further political complications relating to the adoption and interpretation of the indicator. In contrast with PURE (Zhan & Zhung, 2013) and SYNOPS, the Danish PLI (excluding human health) is a fully linear aggregation and thus has no expectation of interactions or nonlinear responses among input factors beyond that defined by the standardization curve. Intrinsic correlations between underlying indicators may exist and be of interest in certain policy contexts but are of unclear practical relevance in the scope of a national monitoring tool.

    In terms of how the U.K. PLI could be used in practice, the analysis undertaken focused primarily on its utility as a comparative tool; that is, in reflecting differences in the inferred load on ecological and human systems across time and space. This is also reflected in the original development context for the Danish PLI, where the primary aim was differentiating between active substances so that they could be taxed at varying levels. However, the limitations of these types of indicators do need to be understood. Ultimately the validity of any indicator use for comparison or benchmarking is strongly dependent on maintaining a consistent baseline against which to compare change. In the context of the PLI, this baseline depends on a single reference active substance and a consistent set of weighting values, and these cannot change without recalculating all previous years; this itself would affect the indictor transparency. Likewise, because pesticide usage in terms of active substances applied is different in the United Kingdom compared with that of Denmark, the reference active substances are different for the two countries, and therefore the pesticide loadings cannot be compared with other countries or other indicators.


    The work described herein explored the appropriateness of the approach used for the Danish pesticide load indicator to agricultural pesticides applied in the United Kingdom, largely for policy purposes. In summary, the indicator values for the 2016 and 2018 datasets appear broadly comparable, suggesting that the overall environmental load from pesticides on the U.K. environment remained relatively constant during this period. For the differences that were identified, logical reasons were seen that were mainly due to regulatory interventions, and the sensitivity of the indicator to such changes was clearly demonstrated. The U.K. PLI also appears to facilitate transparent communication of the impact of agricultural pesticide use. As noted previously, the United Kingdom does not currently have a widely recognized policy tool equivalent to the PLI; that role is played by a combination of the PUS and more targeted indicators defined under the U.K. NAP (Defra, 2013). The role the PLI may play in U.K. policy development remains to be clearly defined, although it is likely that, due to differences in the underlying data structures, this will be different from that used in Denmark.

    Some shortcomings were identified in the PLI methodology in terms of its practical application and the suitability of the parameters included (or excluded) such that if the decision is taken to use the indicator in the United Kingdom, adjustments might be needed to better reflect U.K. conditions and policy objectives. There may also be the need to resolve, or at least identify a means of communicating, the uncertainties associated with the usage data that in the United Kingdom are based on a voluntary PUS from a subset of holdings, unlike the statutory obligation on the part of Danish farmers to report their pesticide use to government. There is no doubt that solutions still need to be found and research undertaken to show that the changes appropriately reflect the environmental and human health implications of agriculture pesticide use.

    This work was undertaken to consider the suitability of the Danish PLI to the United Kingdom specifically, but there are several observations that might be useful to highlight if a similar approach were to be taken in other countries. The availability of suitable quality data relating to pesticide usage is a critical factor, particularly if regional interpretations are to be made. The greater the uncertainty relating to usage, the less reliable the indicator will be at describing trends. The inclusion of data relating to human health also needs careful consideration because such data have significant potential to be misinterpreted and/or under-represented due to the absence of a consumer element. In addition, issues relating to linearity and the lack of correlation with regulatory data interpretation are important for transparency. Nevertheless, the Danish Pesticide Load indicator may provide a sound and adaptable starting point for a U.K. national pesticide policy indicator.


    Funding for this study was provided by the U.K. Department for Food, Environment and Rural Affairs (DEFRA). The authors thank colleagues at DEFRA and members of the project steering group for their support, advice, and invaluable input.


      This manuscript was written by K. Lewis and J. Rainford with support and input from other authors. The indicator calculation script and statistical analysis was conducted by J. Rainford. Data extraction from the PPDB and its validation was done by K. Lewis and J. Tzilivakis. Data from the PUS was formatted and provided by D. Garthwaite. Technical developments, interpretation and analysis were a shared team responsibility. Kathleen Lewis: Conceptualization; Formal analysis; Methodology; Project administration; Software; Validation; Visualization; Writing-original draft. James Rainford: Conceptualization; Formal analysis; Methodology; Project administration; Software; Validation; Visualization; Writing-original draft. John Tzilivakis: Conceptualization; Formal analysis; Investigation; Methodology; Software; Writing-original draft. David Garthwaite: Conceptualization; Formal analysis; Funding acquisition; Methodology; Resources; Supervision; Writing-original draft.


      The authors declare that there are no conflicts of interest.