Main authors: Meindert Commelin, Jantiene Baartman, Piet Groenendijk, Oene Oenema, Susanne Klages, Isobel Wright, Tommy Dalgaard, Morten Graversgaard, Jenny Rowbottom, Irina Calciu, Sonja Schimmelpfennig, Nicola Surdyk, Antonio Ferreira, Violette Geissen
FAIRWAYiS Editor: Jane Brandt
Source document: »Commelin, M. et al. 2018. Review of measures to decrease pesticide pollution of drinking water sources. FAIRWAY Project Deliverable 4.2, 79 pp

 

Here we present a brief overview of the process and procedures related to the execution of the review. A total of 16 institutions across EU-28 (including those running the FAIRWAY case studies) have been involved in the review process.


Contents table
1. Literature review procedures 
2. Quantitative analysis of the effect size of measures 

1. Literature review procedures

Measures to prevent and reduce the risk of surface runoff and leaching can be categorized according to the source-pathway-receptor concept, i.e. there are

  1. source-based measures,
  2. pathway-based measures, and
  3. receptor or effects-based measures.

Examples of source-based measures are appropriate storage of pesticides, and prohibition periods for and restrictions on the application of pesticides. Examples of pathway-based measures are buffer strips, green covers, terracing etc. Examples of receptor or effects-based measures are dredging and creation of riparian zones, etc.

This review focusses on source-based measures and pathway-based measures, receptor-based measure are not anymore related to agricultural management and practises, so outside the scope of this review. At the start of the review a protocol was written and discussed by all partners involved in the review. The purpose of the protocol was ‘to provide guidance for a uniform, effective and efficient literature review and assessment of measures aimed at decreasing pollution of drinking water resources by pesticides’. Two types of reviews were made

  1. a qualitative review of measures, practices and factors that affect pesticide pollution of groundwater and surface waters, and
  2. a quantitative review of the effectiveness and efficiency of measures, based on experimental studies in the field.

The qualitative review focussed on the processes and factors that control the pollution of groundwater and surface waters with pesticides from agricultural sources. This review yielded an overview of controlling factors and a qualitative overview of possible measures to reduce pesticide pollution of groundwater and surface waters. The encountered measures were characterized using a common format (Table 1).

Table 1: Format for the description of measures of the so-called longlist

Name of the measure One sentence
Description Brief characterization of the measure in maximal three sentences; what is (are) the action(s) of the land manager/farmer/citizen
Mode of action

Brief description of the mechanism(s) of the measure in maximum three sentences, addressing the following possible mechanisms:

  • Reduction / substitution of contaminant input
  • Modification of pollution pathway
  • Re-design of the system
Expected effectiveness

Decrease of pollution (concentration or load); select one answer out of five options:

  • High: >25% decrease in concentration/load
  • Moderate: 10-25% decrease in concentration/load
  • Low: 5-10% decrease in concentration/load
  • Insignificant: <5% decrease in concentration/load
  • Unknown
Expected implementation cost

Economic cost, in euro per ha of utilized agricultural land; select one answer out of five options:

  • Low: < 10 euro per ha
  • Moderate: 10-50 euro per ha
  • High: 50-100 euro per ha
  • Very high: >100 euro per ha
  • Unknown
Underpinning of the measure

Is the measure well examined, as shown by various reports; select one answer out of four options:

  • Yes (> 5 reports)
  • Partly (1-5 reports)
  • No (≤ 1 report)
  • Unknown
Applicability of the measure

Is the measure widely applicable; select one answer out of four options:

  • Yes (on more than 75% of the agricultural land)
  • Partly (on 25-75% of the agricultural land)
  • No (on <25% of the agricultural land)
  • Unknown
Adoptability of the measure

Do the land managers/farmers/citizen adopt the measure easily; select one answer out of four options:

  • Yes (more than 75% of the addressees)
  • Partly (on 25-75% of the addressees)
  • No (on <25% of the addressees)
  • Unknown
Other benefits

Does the measure contribute to beneficial side-effects; select one or more answers out of four options:

  • Yes, decreases energy costs
  • Yes, contributes to landscape diversity
  • No
  • Unknown
  • Other: please specify
Disadvantages (other than implementation costs and labour)

Does the measure contribute to negative side-effects: select one or more answers out of four options:

  • Yes, decreases crop yield
  • Yes, decreases crop quality
  • Yes, decreases soil quality and biodiversity
  • Yes, contributes to (more) pest and diseases
  • No
  • Unknown
References Provide up to three key literature references

Based on the qualitative review, a tentative list of key measures was established, with the objective to collect quantitative data and information about these measures for a quantitative assessment. The subsequent quantitative review provided the basis for a meta-analysis of the effectiveness and efficiency of measures, and for the identification of most promising measures. A systematic search was performed through online databases, and a local/expert based search was done throughout Europe. The aim of the local search was to find high quality studies which are not easily accessible through online databases, but which contain valuable data. The criteria used for this search were;

  1. well documented (peer reviewed or reports),
  2. the study should be about an measure to decrease pesticide transport/pollution,
  3. the study must be an experiment, with quantitative data presented in the source, so a meta-analysis is possible.

For the online systematic search three online databases were used; Scopus, Ovid and Web of Science. The following search formula was used in these databases:

  • IN TITLE: (pesticid* OR herbicid*) AND (leaching OR runof* OR overland flow OR drift OR spray drift OR infiltration) AND (effect* OR impact OR influence OR reduc* OR decreas*)) NOT (model* OR industr*))
  • AND IN ABSTRACT: (agricult* OR farm* OR field* OR crop*)

This resulted in 180 unique records. In Web of Science the formula was slightly different, ‘IN ABSTRACT’ was changed for ‘TOPIC’ which also includes title and keywords, this is done because ‘IN ABSTRACT’ is not available.

In addition, University and Institute libraries were examined in Member States of the European Union, because a significant fraction of the research on measures to reduce pesticide leaching and surface runoff has been conducted before the 1990s and 2000s when it was still common to publish the results in reports and documents. These reports and documents quite often have not been digitalized and made available to the international scientific audience and as such are not traced by the search machines of Google Scholar and Scopus.

Data and results of reviewed reports and articles were collected in Excel spreadsheets in a uniform manner. The Excell spreadsheets were subsequently transferred to a flat csv database for statistical analyses.

D42 fig01
Figure 1

The flowchart in Figure 1 shows the general lay-out of the protocol of the review. Each block represents a set of questions, as described here further below:

  1. Contributor: information on person(s) who did the data collection
  2. Reference: Two option available, 1) peer reviewed articles, and 2) book or report. This last category includes so-called ‘grey literature’.
  3. Number of measures: the number of measures described in the literature source.
  4. Pollution type: Nitrate or pesticides or both.
  5. General information: Data about the location, land use, soil type etc. This information is used to categorize and specify the results (and effectiveness of the measure).
  6. Control treatment: Describe the characteristics of the reference or control situation. This information is essential for estimating the effectiveness and efficiency of the measure(s).
  7. Measure: Describe briefly the characteristics of the tested measure.
  8. Effectiveness: Describe the test results, in terms of reduced leaching and/or loading of the pollutant.
  9. Economic cost: Describe the operational (running) economic cost of the tested measure, in euro per ha per year, compared to the control (reference) treatment.

In the review, common definitions were used, as follows:

Measure: an agro-management technique, or a change in an agro-management technique, applied at field, farm, landscape and/or water basin levels. A measure often involves a plan or action to achieve a particular purpose. Measures may relate to (changes in) crop types, rotations, cover crops, soil tillage and cultivation, fertilization, irrigation, drainage, pest and disease management, weed management, harvesting, machines and trafficking, landscape management, etc.

Effectiveness: The extent to which the objectives have been achieved, i.e., the extent to which the pollution of drinking water resources by pesticides has decreased. The effectiveness can be expressed in different units; here we used the decrease in pollutant concentration (mg/l, or µg/l) or pollutant load (kg/ha/year or g/ha/yr), depending on the results available in the literature source.

Efficiency: The extent to which the desired effects are achieved per unit of cost. The term refers also to “cost effectiveness”, which is expressed as ratio of the effect achieved and the costs required (e.g. µg pesticides per litre per euro).

Applicability: Applicability is the extent to which a measure can be implemented in practice (without the special provisions that can be made during a research or experiment). Applicability is expressed in the percentage of the area where the measure can be implemented in practice without much difficulty.

Willingness: the extent to which stakeholders implement the measures without additional incentives and, if necessary, maintain the extra facilities that have to be taken. Willingness is expressed in the percentage of stakeholders who implemented the measure(s) without external incentives.

The literature review was divided among the FAIRWAY partners involved, according to regions. Five regions have been distinguished, as follows:

  • Central EU: Czech, Slovakia, Hungary, Romania, Bulgaria, Slovenia, Croatia, Bosnia, Serbia
  • Central – northern EU: Poland, Germany, Austria, Schwitzerland, Baltic States
  • Mediterranean: Andorra, Portugal, Spain, Italy, Greece,
  • Scandinavia: Denmark, Norway, Sweden, Finland, Iceland
  • Western Europe: Ireland, United Kingdom, Netherlands, Belgium, France
  • The world outside EU: America, Australia

2. Quantitative analysis of the effect size of measures

The results discussed in this report are based on literature study and statistical analyses. There are three approaches to express the effects of measures.

The first approach applied in this report through simple response ratios, which is the pesticide pollution from a treatment measure divided by the pesticide pollution of the reference treatment (control treatment), according to

RR = YT/YR

where RR is the response ratio (dimensionless; or percentage), YT is the measured result of the treatment measure, and YR is the measured result of the reference treatment. The latter is usually current practice or conventional practice. The ratio may vary from 0 to more than 1; a value smaller than 1 indicates that the treatment measure decreases the pesticide transport relative to the reference treatment. A ratio of 1 means no effect. Instead of a relative comparison of pesticide losses, the response ratio was sometimes derived from a comparison of pesticide concentration in waterbodies or from the content of pesticides in the soil between treatments, depending on the availability of the data in the reviewed publications.

The second approach is to express the effectiveness in terms of relative effects, i.e., the ratio of the treatment measures, corrected for the reference treatment, and the reference treatment according to

D42 eqn1

where ES is the effect size (dimensionless; or percentage). In case a treatment measure does not result in a (significant) different outcome than the reference treatment, then ES = 0. For YT > YC this results in ES > 0, and vice-versa.

The third approach is the one used in most meta-analyses studies; the means and standard deviations of the effects are determined based on ln-transformed ratio’s (following the protocol of Hedges et al (1999) as given by

D42 eqn2

Once the ln-transformed average ratio (and standard deviation) are known, it can be back-transformed to obtain the average effect size according to

D42 eqn3

Similarly the confidence interval for ES can be determined by back-transforming the confidence interval limits for L. The reported average ES is significant when the available confidence interval (based on standard deviation) does not include the value zero. Meta-analysis studies often are based on the ln-transformed approach, whereas single studies and some reviews mostly consider the effect size or the response ratio RR=YT/YC. In this report, we estimated and used RR (see »Quantitative analysis of measures and practices).

The data as collected through the structured data review from the excel sheets was processed in R and also manually to obtain a good quality uniform database. The main focus during the processing was on homogenizing units of measurement and setting the right reference treatment. This was done to optimize the calculation of the response ratio for each treatment in each study.

The collected data was divided in categories based on the already identified measures in the shortlist. For each category of measures the reference was defined and this was applied to all individual treatments, in this way the uniformity between studies was optimized. As reference the conventional of standard management practise was used.

As a general analysis, the response ratios for each study within a category were combined and a summary effect ratio was calculated for each measure. In the case of input control there was a clear relation between effectiveness and amount of reduction, so a linear regression was applied to study the relation. Further analysis of co variables and the fitting of a random effect model will be carried out as next step in this research to identify the most promising measure included in the database.

It was intended to analyse the efficiency of the measures as a combination of effectiveness and costs. However the amount of data included in the database about costs of application and maintenance were too scarce to make any general calculations. This will be taken into account during further research analysis.

 


Note: For full references to papers quoted in this article see

»References

 

Go To Top