Main authors: Meindert Commelin, Shaun Coutts, Jantiene Baartman, Isobel Wright, Antonio Ferreira, Gerard Velthof, Oene Oenema and Violette Geissen
FAIRWAYiS Editor: Jane Brandt
Source document: »Commelin, M. et al. 2020. Identification of most promising measures and practices: 1. Reduction of diffuse pesticide transport from agricultural land to groundwater and surface waters bymanagement practices. FAIRWAY Project Deliverable 4.3, 72 pp


Contents table
1. Data sources 
2. Data description 
3. Meta-analysis 
4. Case studies

1. Data sources

As a starting point for the literature synthesis and meta-analysis, initial data sources and literature were collected in two ways.

  1. A systematic search was performed through online databases, and
  2. a local/expert based search was done throughout Europe.

The purpose of the local search was to find studies containing valuable data which were not easily accessible through online databases. The selection criteria for this search were;

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

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

  • 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*) In Web of Science the formula was slightly different, ‘IN ABSTRACT’ was not available and ‘TOPIC’ was used, which includes abstract, title and keywords.

The local and online search, combined with an added ‘snowball’ search in relevant reviews resulted in 270 unique titles (Figure 1.1). These papers included four meta-analysis papers and seventeen reviews. The results of these papers were summarized and combined with the results of the meta-analysis. The reviews were analyzed by abstracting the answers to several questions, for both transport to ground and surface water:

  • How effective is the described measure to reduce transport of pesticides to ground and surface water?
  • What is the influence of explanatory variables, like climate, type of agriculture and soil type?
  • Are there known side effects of the measure which might influence the effectiveness on the long term or cause different problems?

D4.3 figb01
Figure 1.1

Within the collected sources for the meta-analysis there was a lot of heterogeneity between studies in terms of experimental design and measured data. For the meta-analysis, only sources were included that presented sufficient statistical data (sample size, standard deviation and clear treatment and control). Based on this criteria 31 papers were included in the database for further analysis (Figure 1.1).

2. Data description

The collected data for meta-analysis mainly covered four measures; soil tillage practices, vegetative filter strips (VFS), application management and drift reduction. The most extensive data set was available for tillage practices. Within most studies multiple comparisons or treatments were done, leading to a total of 141 experiments in the database (Table 1.1).

For tillage practices 41 experimental comparisons were used for the analysis. Because there is a high diversity in tillage practices, the dataset was homogenized by selecting all data comparing no-till (NT) with conventional (plow) tillage (CT). Other measures in the dataset included disc tillage and mulch tillage, however the number of studies was too low to perform a proper statistical analysis. For the general analysis the data is separated into two groups; pesticide transport to groundwater and to surface waters. In the dataset 38 comparisons were found. Recent papers used more data (Reichenberger et al., 2019), however the statistical constraints of the meta-analysis strongly reduced the available data.

Of the included studies eighteen were conducted in the United States, twelve in Europe and one in China. The studies covered a variety in climates including humid continental, Mediterranean and oceanic climates. The main soil types were clay and loam (including silt clay loam, silty clay, clay loam and silt loam) two studies were conducted on sandy.

Table 1.1: Measures included in the dataset

Measure Number of experiments Number of papers
Tillage 47 8
VFS 40 10
Application rate 19 5
Drift 18 3
Other 17 5
Total 141 31

Eighteen different pesticides were included in the studies, fifteen of these were herbicides and three insecticides. Of the 141 experiments 127 used herbicides and three studied insecticides (eleven spray drift studies did not specify pesticide type). Most studies covered pesticides that are no longer approved in Europe (83), while 47 studies did use currently approved pesticides. The main reason for this is that many older (pre-2000) studies are included. However, these studies are still valuable for understanding how different measures affect the of transport pesticides with different characteristics.

The risk of pesticide transport is mainly influenced by three characteristics of each pesticide; water solubility (Sw), adsorption coefficient (Koc) and half-life time (DT50). Table 1.2 shows the value ranges related to each parameter (Lefrancq, Jadas-Hécart, La Jeunesse, Landry, & Payraudeau, 2017; Tang et al., 2012; Young & Fry, 2019).

Table 1.2: Characteristics of pesticides in the dataset

Characteristic Value Classification Occurrence in dataset (total n = 130)
Water solubility (Sw) – mg L-1 <50 Low 53
50 - 500 Moderate 30
> 500 High 47
Adsorption coefficient (Koc) < 75 Low 4
75 – 500 Moderate 107
> 500 High 9
Unknown   10
Half-life time (DT50) – days <30 Non persistent 98
30 - 100 Moderately persistent 31
100 – 365 Persistent 1
> 365 Very persistent 0

Half of the studies had a block design with a comparison over the same time period (spatial replication). The other studies were time-split where the effect of a measure was studied over time (temporal replication). The duration of most studies was between 1 and 5 years, but one study run for eleven years. A few studies (drift or VFS related) covered several days. Pesticide transport was described by the load, or mass per area per year in runoff. However if total loads were not available, concentrations of pesticides were used. Locations of measurements varied from soils, runoff water and lysimeter leachates to concentrations in larger water bodies like rivers or groundwater inlet points.

3. Meta-analysis

To perform the meta-analysis the R-package ‘metafor’ was used (Viechtbauer, 2010). The goal of a meta-analysis is to combine all quantitative data from the collected studies and draw an overall conclusion on the effectiveness of a specific measure. In the reviewed studies the effect of a treatment was shown with different values and units. For a meta-analysis these different designs, units and approaches have to be normalized so they can be compared. To be able to compare effect sizes between studies all data was recalculated to the response ratio (R):

D4.3 eq01 [1]

Where (XT ) ̅ represents the means of the treatment group and (XC ) ̅ the means of the control (Borenstein, Hedges, Higgins, & Rothstein, 2009). For each study the mean, standard deviation and sample size was recorded. On several occasions statistical recalculations were done to obtain comparable statistics from each study (Lajeunesse, 2011). The distribution of R cannot be assumed to be normal, so to do statistical analyses it is preferable to use the natural logarithm of R (Borenstein et al., 2009; Hedges, Gurevitch, & Curtis, 1999). The variance of ln(R), needed to derive the uncertainty of a study, is calculated with:

D4.3 eq02 [2]

with SDT and SDC the standard deviations for the treatment and control groups respectively, and nT and nC the sample sizes of the groups.

A random effects model was used to combine the estimated effect sizes for all studies within one group (e.g. tillage measures). The model accounted for within study effects when multiple treatments from one study were used. The resulting weighted means and summary effect sizes were transformed back to percentage response effects. A 95% confidence interval (CI) was calculated, and the effectiveness of a measure is considered significant when there is no overlap with a response effect of 0%, indicating ‘no effect’. To understand mechanisms of effectiveness better sub group analysis is done based on pesticide characteristics as presented in Table 1.2.

4. Case studies

We also utilize insights from nine FAIRWAY case studies located across Europe that are investigating measures to minimize pollution of ground and surface drinking water resources by pesticides. The case studies reflect different pedo-climatic zones and assess the effectiveness of different measures, their cost, adoptability and applicability for farmers. We collected data from all case study leaders through a questionnaire (»Annex 1). The respondents were experts who are in close contact with land managers who apply the measures. The collected data include

  1. the measures in the region, and
  2. the evaluation of the measure in terms of effectiveness, cost and applicability by farmers.



For full references to papers quoted in this article see »References

Download the full report for Annex 1


Go To Top