Embed Size (px)
Citation preview
WT0965 report
How effective are slurry storage, cover or catch crops, woodland
creation, controlled trafficking or break-up of compacted layers, and
buffer strips as on-farm mitigation measures for delivering an
improved water environment?
Louise M Donnison, Paul J Lewis, Barbara Smith (Game and Wildlife
Conservation Trust) and Nicola P Randall (Lead Reviewer)
Executive summary
Background
Over the last fifty years, European agriculture has become more intensive due to increased
applications of fertilizers and agrochemicals to agricultural land. Currently 50% of the nitrates in
European rivers are estimated to be from agricultural sources. In the UK, agricultural activities are
estimated to contribute 70% of nitrates, 28% of phosphates and 76% of sediments measured in
rivers. River waters of catchments dominated by agricultural land use can have elevated levels of
pesticides and bacterial pathogens.
The aim of this systematic review was to assess the effectiveness of slurry storage, cover/catch
crops, woodland creation, controlled trafficking/break-up of compacted layers and buffer strips, as
on-farm mitigation measures, for delivering an improved water environment.
Methods and outline results
Electronic databases, the internet, and organisational websites were searched to find articles that
investigated the impact of the on-farm mitigation measures on water quality. The searches
identified 146, 941 records (excluding Google Scholar and web searches). The removal of
duplicates and irrelevant articles from the search results left 718 records.
The 718 relevant articles were coded to create a searchable Microsoft Access database (systematic
map) which describes the water quality research to date for the topic specific mitigation measures.
All evidence was coded with country of study, mitigation and water quality measurement, if the
information was missing then not clear was recorded. Additionally full text articles were coded for
study design and those studies without confounding factors were coded for outcome. The database
can be used to sort or filter on category and provide simple numerical counts.
2
The systematic map database was composed of mainly buffer strip (including trees) and cover/catch
crops studies. The map also contained some slurry storage studies which were diverse and often at
least 10 years old. There were only a few woodland creation studies in the map as most studies
composed of trees were categorized under buffer strips, the studies that remained measured water
quality after afforestation on former agricultural soil or planting of tress for biomass. Very little
evidence was found for subsoiling (break up compacted soil) or controlled traffic on grassland.
There were 467 studies coded in the systematic map at full text (including studies with confounding
factors) which were given a value for scientific rigour based on whether they were randomized,
controlled, replicated (spatial or temporal), designed (manipulative, correlative or sampling) and
conducted for longer than a year. These values can be used to provide a rudimentary indication of
the type of research available for each mitigation.
There were 410 studies coded in the systematic map at full text (excluding studies with confounding
factors) which were given a value for effectiveness in reducing N, P, sediment, pesticides or
bacterial pathogens in water. These values were used to provide a rudimentary indication of the
overall effectiveness of each intervention on specified outcomes, based on the available evidence.
A meta-analysis was conducted to assess the effectiveness of cover/catch crops in reducing nitrate
leaching as compared to a fallow no vegetation control. The meta-analysis suggested a consistent
positive effect in of cover/catch crop in reducing leaching Nitrate when compared to a fallow, and
that there was no difference in effectiveness of cereals and brassicas for reducing Nitrate leaching.
Only 10 studies were included in the meta-analysis due to difficulties in extracting data from
primary studies.
3
Buffer strips (Grass and tree buffers)
Buffer strips composed of grass and/or trees are thought to improve water quality by physically
trapping sediments and associated pollutants, and by immobilizing soluble nutrients through plant
uptake or microbial degradation.
Average effectiveness values suggested that buffer strips were most effective for reducing sediment,
followed by pesticides, N, P, and bacterial pathogens (in decreasing order), however these values
should be interpreted within the limitations of the evidence. Pre-existing meta-analyses also found
that buffer strips could be effective in improving water quality.
Buffer strips were the most commonly studied mitigation in the database (225 studies with data that
enabled assessment of effectiveness of the intervention). Over half of the studies were manipulative
(n=147), at least a third were controlled (n=104) and often fully replicated. Nearly half of the
studies were conducted for longer than a year, but not many studies were randomized. Most of the
buffer studies were field or plot based (n=187), often on loam soils (n=121).
Limitations of the evidence base
Studies were often at a field scale which may not capture the effects of preferential flow paths or
buffer strip placement on buffer strip performance. Studies were often on loam or unknown soil
types, which may not capture the effect of soil particle size on buffer strip performance. Studies
often assessed effectiveness over short periods of time, which may not capture changes in buffer
strip effectiveness over time. Buffer strip effectiveness may depend on experimental factors such as
vegetation types, but this was not investigated. Only a third of the studies had data for all four
seasons, yet season may have an impact on effectiveness due to seasonal differences in plant growth
and nutrient uptake.4
Nitrogen
61% of buffer strip studies investigated the effectiveness of buffers for reducing N of buffer studies,
(n=139).
Authors indicated that buffer strips are generally effective for reducing at least one type of N (72%
of buffer studies measuring N, n=100), but that this varied for different forms. Authors indicated
that buffers strips were more effective at reducing Total-N (74% of buffer studies measuring Total-
N, n= 29) and nitrate-N (67% of buffer studies measuring nitrate, n= 80), than ammonium-N (50%
of studies measuring ammonium, n=23).
Sediment
44% of buffer strip studies investigated the effectiveness of buffers for reducing sediments (n=98).
Authors indicated that buffer strips are generally effective for reducing sediments (87% of buffer
studies measuring sediments, n=85).
Phosphate
42% of buffer strip studies investigated the effectiveness of buffers for reducing P (n=94).
Authors indicated that buffer strips could be effective for reducing at least one type of P (65% of
studies measuring P, n=61) but that this varied for different forms of P. Buffers strips appeared to be
more effective at reducing total-P (73% of buffer studies measuring total- P, n= 46), than
orthophosphate-P (55% of buffer studies measuring orthophosphate, n=23) or soluble-P (26% of
buffer studies measuring soluble P, n=5).
Pesticides
17% of buffer strip studies investigated the effectiveness of buffers for reducing pesticides (n=38),
often using atrazine (68% of buffer studies measuring pesticide, n=26) or metolachlor (32% of 5
buffer studies measuring pesticide, n=12).
Authors indicated that buffer strips are generally effective for reducing at least one of the 38
pesticides measured (71% of studies measuring pesticide, n=27).
Bacterial pathogen counts
Only 8% of buffer strip studies investigated the effectiveness of buffers for reducing bacterial
pathogen counts (n=19). 63% of studies measuring bacterial pathogen counts were effective at
reducing at least one of the bacterial pathogen count measurements (n=12).
Cover/catch crops
Fast-growing cover or catch crops, planted over the winter months are designed to improve water
quality by protecting the soil against erosion thereby minimizing the risk of runoff, and reducing the
risk that nutrients are leached from the root zone.
The Evidence indicated that cover crops are most effective at reducing leaching of N and of sedi-
ments into water courses.
Cover/catch crops were the second most commonly studied mitigation (n=132 studies scored for
effectiveness). Most studies were manipulative (n=125), controlled (n=115), fully replicated and
conducted for longer than a year and sometimes randomized. 84% of cover/catch crop studies were
field or plot based (n=111) often on loam soils (54% of cover/catch crop studies, n=71).
Limitations of the evidence base
Studies were mainly sampled at a field scale. The one study that made measurements within a river
system over 17 years, did not find an agreement between field and river data. Cover catch crop
studies were often conducted on loam or unknown soil types, which may not capture differences 6
between soil types and nutrient leaching (e.g. sandy soils). Only a quarter of the studies assessed
effectiveness across all 4 seasons. Although some studies were of long duration (up to 30 years), the
effect of stopping cover/catch cropping on effectiveness was not studied that often, one study
suggested that nutrients caught by cover catch crops can be leached in subsequent years if no
cover/catch crop is planted. Climatic data was often difficult to extract from studies, however some
studies reported year to year variation in effectiveness depending upon the date when autumn rains
started.
Nitrogen
86% of cover/catch crop studies investigated the effectiveness of cover/catch crops for reducing N
(n=114), mainly measured as nitrate (95% of studies measuring N, n=108).
72% of cover/catch crop studies were reported by authors to be generally effective for reducing at
least one form of N (n=82).
A meta-analysis on a subset of data (n=10), suggested that cover/catch crops are effective at
reducing N compared to a fallow control (Z = 7.869, P = <0.001), but that there was significant
variation between the studies (Q = 131.31, df =10, P = <.001).
Sediment
Only 14% of cover/catch crop studies investigated the effectiveness of cover/catch crops for
reducing sediments (n=19). Authors indicated that cover/catch crops were generally effective at
reducing sediment in 68% of the studies (n=13).
Phosphate
10% of cover/catch crop studies investigated the effectiveness of cover/catch crops for reducing P
(n=14). Of these 14 studies, only 3 were effective at reducing any type of P.
7
Slurry storage
Slurry storage and altering timing of slurry application to crops can impact on water quality by
ensuring that slurry applications are timed to improve uptake of nutrients by crops.
This review did not directly address the question, ‘does alteration of slurry timing impact on water
quality?’, but instead investigated the value of slurry storage for improving water quality. The
evidence was diverse, being mainly composed of studies that measured slurry leakage, or die-off of
pathogens in slurry during storage, but a few studies investigating the timing of slurry applications
to match plant uptake were found. A separate study (a rapid evidence assessment) has been
commissioned to specifically investigate the impact of altering timing of slurry application on water
quality.
With regard to the question asked in this Systematic Review, the value of slurry storage, the
evidence was variable, but indicated that storage can reduce levels of bacterial pathogens in slurry.
A disproportionate amount of studies had confounding factors, particularly at a catchment level and
were excluded from effectiveness assessment. 42 studies were found that could be included in an
assessment of the effectiveness of slurry storage. Under half were manipulative (n=18), with only a
third of the studies controlled (n=13), studies were often not always fully replicated, often of short
duration and not randomized.
Limitations of the evidence base
Many of the studies were more than 15 years old, and some referred to slurry storage using earth
lined stores which may not meet current legislation. Much of the evidence for N and P was based on
detection of slurry leakage rather than water quality which makes it difficult to compare the results
for slurry storage to other mitigation measures. Many studies were not of the highest scientific
rigour, and often did not have pre-slurry storage baseline data. Some authors suggested that results
for leakage may have been due to experimental error e.g. slurry stores being completely emptied, 8
resulting in clay soils cracking. One author had concerns that it was not possible to identify if the
slurry had leaked as part of the initial sealing or much later when to storage was operational. Most
studies were conducted for less than 2 years therefore the effect over time e.g. age of slurry storage
may not have been accurately assessed. Only 10 studies investigated the effect of P.
Nitrogen
71% of slurry storage studies investigated the effectiveness of slurry storage for reducing N (n=30).
Authors indicated that slurry storage was often not effective for reducing or preventing leakage of N
for at least one form of N (17% of slurry storage studies that measured N were effective, n=7).
Bacterial pathogen counts
45% of slurry storage studies investigated the effectiveness of slurry storage for reducing bacterial
pathogen counts (n=19).
68% of studies found that slurry storage was generally effective for reducing bacterial pathogen
counts in stored slurry for at least one form of bacterial pathogen count (n=13).
Phosphate
Only 24% of slurry storage studies investigated the effectiveness of slurry storage for reducing P
(n=10). Only 2 of these studies found that slurry storage was effective for reducing any form of P or
leakage of P.
Woodland creation (excluding tree buffer studies)
Woodland creation has the potential to improve water quality by improving water infiltration
through soil, thereby reducing runoff and the risk of pollutants entering water sources. Woodland
may also uptake nutrients, which would otherwise be lost to water sources.
9
Buffer strip studies with a tree component were not categorized under woodland creation, but were
instead categorized as buffer studies. 48% of buffer studies had a tree component (n=107).
Other woodland creation studies were limited, as most research falls outside the direct scope of the
fairly narrow focus of the question addressed here, and the total number of studies found (n=12)
was lower than originally anticipated. The woodland studies included were quite diverse consisting
of studies of afforestation on former agricultural land, or studies of trees grown for biomass.
Effectiveness of woodland creation was difficult to assess due to variations in the type and design of
studies and a relatively small sample size.
Some afforestation studies did not have a non-woodland control, but instead measured changes in
water quality over different aged woodlands making it difficult to ascertain if woodland had im-
proved water quality compared to agricultural land. Some biomass studies did not have a non-
woodland control, but instead used a non-fertilized treatment as a control. Most of the woodland
creation studies measured N (92% of woodland creation studies, n=11). Only 1 study measured sed-
iments, bacterial pathogen counts or P (n=1).
Modelling studies were excluded from the review, however they are useful for woodland studies
which experimentally can take years to assess. A recent Forestry Commission review has provided a
comprehensive literature review of the effects of woodland creation on water quality at a broader
level, including modelling studies and considering land use and air pollution, topics which were ex-
cluded for this review.
Subsoiling and controlled traffic on grasslands
The confinement of farm machinery to certain areas of a field (controlled trafficking) or the
breaking up of compacted layers (subsoiling) by a mechanical soil treatment may improve water
10
quality by reducing soil compaction to improve soil infiltration and root penetration, which may
reduce the risk of runoff containing pollutants entering watercourses.
There was little evidence found for the direct impact on water quality of subsoiling or controlled
traffic on grasslands (n=5). However, studies that included related evidence, e.g. studies that
measured improvements in soil water infiltration, were not included in this review. Therefore the
lack of evidence may be artificial and that a question phrased as “What effect does subsoiling have
on soil infiltration” may have been more appropriate for this mitigation.
Conclusion
Buffer strips (including woodland buffers) were the most commonly studied intervention. N was the
most commonly measured indicator, and most evidence came from loam or unknown soil types.
Approximately a quarter of the studies made measurements in all four seasons.
Overall, study authors suggested that cover/catch crops and buffers strips can be effective for
improving water quality. However, the evidence is generally based on short-term studies conducted
at field scale, and there was not enough evidence recorded in the systematic map to assess
mitigation effectiveness at a catchment scale. Most evidence was from loam or unknown soil types.
On average cover/catch crops studies were slightly more rigorously executed than those of buffer
strips.
Implications for policy and management
Most evidence was drawn from journal articles, despite the search strategy being designed to
capture unpublished evidence. Although several projects were found on websites, little information
could be used in the systematic map. The allocation of resources to improving project
documentation and archiving can be invaluable for improving the evidence base for a given topic.
11
The systematic map provides a large database of research on the primary topic that can be used to
filter information by mitigation or water quality measurement, which should help enable decision
makers and delivery agencies to better facilitate catchment planning.
Generally, the evidence supports existing guidance for the use of buffer strips alongside water
courses to improve water quality, although the research illustrates a wide variation in buffer strip
implementation design and management. The evidence also generally supports the implementation
of cover crops for reducing pollutants into water bodies. Further evidence is needed to support the
other interventions investigated, and this may take the form of refocused evidence syntheses that
more effectively address the questions posed, or further primary research.
Implications for water quality research
Studies designed with controls, and pre and post water quality measurements would improve the
quality of the evidence base.
Multiple sampling points from both within field and rivers would provide greater insight into the
impact of preferential flow paths, upswellings of groundwater and critical points in river systems.
Long term studies with seasonal data would allow the effects of full crop rotations and degradation
of mitigation effectiveness over time to be assessed. The effects of vegetation type may only
become apparent over time, tree buffers would potentially have a longer actively growing life span
than grass buffers.
Standard reporting of statistics with fields for summary data that include an intuitive metric,
associated sample size and a measure of dispersion such as confidence intervals or standard
12
deviation would enhance the evidence base. Submission of data with journal papers would ensure
that water quality data is not lost to science
It would be useful to use further, more focused, evidence collation and syntheses to investigate
under which conditions mitigations perform best. An iinvestigation into the the impact of altering
timing of slurry applications for reducing water pollution was thought to be of particular potential
value, and since the completion of this work has been commissioned as a rapid eveidence
assesssment.
Future evidence syntheses into the water quality benefits of woodland creation and of soil
management methods such as controlled trafficking and subsoiling, are likely to find more
evidence, if the questions are refocused.
13
Background
Intensification of European agriculture over the last 50 years, has resulted in increased usage of
fertilizers and agrochemicals [1]. Soil compaction and reductions in organic matter content,
resulting from the intensification of agricultural practices, have increased the risk of soil erosion
and water run-off. Nutrient applications in excess of plant needs, coupled with increased run-off
from agricultural land, has contributed to a decline in water quality [2].
Nitrate levels across Europe exceeded European water quality standards (50mg/litre) in 15% of
groundwater monitoring stations and 3% of surface stations in the period between 2004 and 2007.
Particularly high levels of nitrate were found in the surface waters of England, Belgium (Flanders),
Netherlands, France (Brittany), Estonia, Northern Italy, North East Spain and Slovakia [3]. The
levels of 500 different chemicals in 4 European river basins (Elbe, Danube, Schelde and Lobregat)
were measured in a recent study, which found that 40 chemicals, were at levels harmful to
organisms 75% of which were pesticides, [4]. It is estimated that each year 200 million cubic metres
of sediment are dredged from European rivers [5]. Agricultural activities are estimated to be the
source of 28% of phosphates, 70% of nitrates and 76% of sediments in UK rivers [6, 7]. UK
Catchments dominated by agricultural land use have elevated levels of bacterial pathogen counts
[8].
A decline in water quality (including sediment) has increased water cleaning costs, reduced
reservoir capacities and can have negative impacts on wildlife and flood defences [9]. Climate
change scenarios suggest that the UK will experience wetter winters, and warmer, drier summers,
which could impact on water quality. Increased extreme weather events may increase the likelihood
of heavy rains washing soil and pollutants into river systems, and drier summers will concentrate
levels of pollutants in rivers [10].
14
European member states have a policy commitment to tackle water pollution through a number of
directives namely the Water Framework Directive (WFD), the Nitrates Directives, the Ground
Water Directive and the Bathing Water Directive. In the UK, Nitrate Vulnerable Zones (NVZs) are
used to implement some of this policy nationally [10]. During the last 10 years the UK Department
for Environment Food and Rural Affairs (Defra) and the Environment Agency (England and Wales)
have funded, at a cost of around 70 million pounds, 200 catchment projects of which Defra funded
178 [11]. Much of this was spent on studies that assessed the efficacy of mitigation measures in
delivering an improved water environment [12].
Objective of the Review
In order to inform future decision-making, a need was identified by funders, to evaluate the
evidence for the effectiveness of five on-farm mitigation measures that may improve or affect
environmental water quality: slurry storage; cover/catch crops; woodland creation; break-up of
compacted layers/controlled trafficking; and buffer strips [13].
Slurry storage may reduce pollution incidents caused by spills and leaks, and timing of
slurry applications to improve uptake of nutrients by crops can also reduce water pollution
[14].
Fast-growing cover or catch crops, planted over the winter months, can protect the soil
against erosion, minimize the risk of runoff, and ensure that nutrients stay in the root zone
[15-17].
Woodland creation can improve soil structure which aids soil water infiltration thereby
reducing water runoff and the risk of pollutants entering water sources [18, 19].
The confinement of farm machinery to certain areas of a field (controlled trafficking) or the
breaking up of compacted soil layers (subsoiling) could reduce soil erosion, soil compaction
and water runoff [20].
Buffer strips composed of grass and/or trees can physically trap sediments and associated
15
pollutants and immobilize soluble nutrients through plant uptake or microbial degradation
which can result in an improved water quality [21, 22].
Primary Objective
The study design was discussed at a series of meetings held with a stakeholder group comprising;
Defra, the UK Natural Environment Research Council (NERC), the Environment Agency (UK), and
the Forestry Commission (UK). The review aimed to describe and evaluate the evidence for the
effectiveness of slurry storage, cover/catch crops, woodland creation, controlled trafficking on
grasslands/break-up of compacted layers (subsoiling) and buffer strips as on-farm mitigation
measures for delivering an improved water environment. Improvements in water quality were
defined as reductions in levels of N (all forms of N), P (all forms of P), sediments, bacterial
pathogen counts and pesticides.
The aim was to produce three outputs:
A searchable systematic map database of published and unpublished studies in the subject
area
Provide values to indicate the type of available evidence, and the overall level of
effectiveness for each intervention.
A meta-analysis of the effectiveness of cover/catch crops in reducing nitrate leaching when
compared to a fallow control
Secondary Objectives
The secondary objectives were to:
Provide an overview of published research and grey literature in the subject area for use by
practitioners, policy makers, researchers and the public.
Provide a map that is searchable by topic
16
Inform future research syntheses, reviews and meta-analyses
Identify knowledge gaps in order to inform future research
Compare the effectiveness in reducing nitrate leaching of different catch/cover crops grown
on different soil types
Methods
The methods used in the development of the systematic map and subsequent systematic review
analyses were adapted from the Collaboration for Environmental Evidence Systematic Review
Guidelines [23] and from an existing systematic map report [24]. A scoping search was performed
to validate the methodology, and is detailed in a review protocol [25], which was used to inform the
final methodology outlined here.
Searches
A comprehensive search of multiple information sources attempted to capture an un-biased sample
of literature to encompass both published and grey literature.
The following online databases were searched to identify relevant literature: ISI Web of Knowledge
involving the following products: ISI Web of Science; ISI Proceedings , Science Direct, Wiley
Online Library, Ingenta Connect, Index to Theses Online, CAB Abstracts, Agricola, Copac and
Directory of Open Access Journals.
An internet search was conducted using the following organisational websites: Defra online
databases, Environment Agency, NERC Open Research Archive, Forestry Commission/Forestry
Research, Centre for Ecology and Hydrology, Natural England , Countryside Council for Wales,
Scottish Natural heritage, Scottish Environment Agency, Northern Ireland Environment Agency,
European Environment Agency, European Commission Joint Research Centre, Finnish Environment
17
Agency, Ministry of Agriculture and Forestry (Finland), Swedish Environment Agency, Danish
Environment Agency, Ministry of Food, Agriculture and Fisheries (Denmark), Government Norway
Portal, Flemish Environment Agency, Agriculture and Agri-Food Canada, Environment Canada, US
Department of Agriculture, US Environment Protection Agency, Agency of the Environment and
Energy (France), Federal Environment Agency (Germany), Federal Ministry of Food, Agriculture
and Consumer Protection (Germany), Netherlands Environmental Assessment Agency, Department
for the Environment, Transport, Energy and Communication (Switzerland), Environmental
Protection Authority (New Zealand), Ministry of Agriculture and Fisheries (New Zealand), Food
and Agriculture Organization of the United Nations, Ecologic Institute and EU Cost (European
Cooperation in Science and Technology). The EU Water Framework Directive and Controlled traffic
farming sites (European site) were not searched as a search box could not be found.
Further internet searches were performed using the search engines: http://www.Scirus and
http://scholar.google.com. The first 50 hits from organization web sites and search engine searches
(.doc .txt.xls and .pdf documents where this could be separated) were examined for appropriate
data.
Database and repository searches were conducted in the English language. Therefore any European
Environment Agency or Agricultural Department website which was not searchable in English was
excluded. The potential language bias associated with this strategy was discussed with funders and
stakeholders at an initial inception meeting, and was considered acceptable for this review.
The search terms used for the database and web searches are listed in Table 1. A wildcard (*) was
used where accepted by a database/search engine to pick up multiple word endings. For example,
'pollut* matches pollutant or pollution. A keyword was made more restrictive by the addition of a
qualifier e.g. (slurr* stor* AND water qualit*), (slurr* stor* AND water pollut*). The combination
of qualifiers and keywords varied for each intervention. Where not already used as a qualifier, the
18
search string was appended with ‘AND water’ if more than 900 search results were retrieved.
A record of each search was made so that when necessary a search could be re-run. The following
data were recorded: date when search conducted, database name, search term, number of hits and
notes. The exact keyword and qualifier combinations used for each database or website were
recorded in a spread sheet [Additional file 1].
Topic specific bibliographies of meta-analyses and reviews were searched for relevant articles
missed by the previous searches [18, 21, 26-28], as well as reference lists e.g. the list of buffer strip
studies maintained by Corell [29] (http://www.unl.edu/nac/riparianbibliography.htm ). Recognised
experts, practitioners and authors were contacted for further recommendations and the provision of
relevant unpublished material or missing data.
The results of each search were imported into a separate EndNote X2TM library file and a record
made of the number of references captured. At the end of the search process, endnote files were
collated into a single database library and duplicates removed using the automatic function in the
EndNote X2TM software. Google Scholar and organizational web search results were imported into
spread sheets.
Study inclusion and exclusion criteria
All retrieved articles were assessed for relevance using the following inclusion criteria, which were
developed in collaboration with funders, stakeholders and subject experts.
Relevant subject(s) and Geographic area:
Studies that investigated some aspect of water quality improvement by one of the on-farm
mitigation measures, irrespective of scale.
19
Stakeholders agreed that the review should focus on temperate countries with similar farming
systems to the UK. Those countries were: UK, Ireland, France, Belgium, Switzerland, Germany,
Holland, Luxembourg, Liechtenstein, Denmark, Sweden, Norway, Finland, Austria, Slovakia,
Poland, Hungary, Czech Republic, Romania, Lithuania, Latvia, Estonia, Belarus, Ukraine, Canada
and New Zealand and northern states of the USA (all states that were entirely above the bottom of
Oklahoma), which excluded states such as Georgia, Mississippi, Texas and California.
Types of intervention (mitigation measure):
The following types of studies measuring the effectiveness of on-farm interventions in improving
water quality were included:
Buffer strips:
Studies measuring the impact on water quality of buffer strips composed of
trees/grass/shrubs. This also included shelterbelts and hedges. However, studies of wetlands
(unless wetland adjacent to buffer strip) or floodplains were excluded.
Slurry storage:
Studies measuring seepage of slurry from slurry storage. However, studies of solid manure
storage were excluded.
Studies measuring changes in bacterial pathogen counts over time with slurry storage
(excludes changes in N or P or air pollution studies).
Studies measuring the impact on water quality of the timing and amount of slurry
applications.
Cover/catch crops:
Studies of cover/catch crops or crops grown for winter cover and effects on water quality.
Winter wheat or volunteer weeds were categorized as cover/catch crops if they provided
20
ground cover in the same manner as a traditional cover/catch crop.
Woodland creation:
Studies measuring changes in water quality after afforestation of former agricultural land
were included. Studies were excluded that compared water quality between different land
uses (forest, urban, arable, grassland) or measured changes in soil nutrient cycling after
afforestation.
Studies growing trees for biomass and testing their potential in cleaning waste water.
Studies measuring the impact of crops intercropped with trees on water quality.
Woodland buffer strip studies were excluded from this intervention as they were considered
instead under the intervention ‘buffer strips’
Subsoiling/controlled trafficking on grasslands.
Subsoiling studies that measured water quality.
Studies that measured water quality after the break up/loosening of compacted soil layers.
Studies that measured the effect on water quality of controlled traffic on grasslands.
Types of outcome:
Water quality was measured by changes in the levels of any form of:
nitrogen
phosphorous
bacterial pathogen counts
pesticides
sediments
Studies were included that estimated water quality from soil samples taken at different depths or 21
that measured slurry leakage or changes in bacterial pathogen counts over time in slurry. Studies
that measured soil infiltration rates, crop yields, plant biomass, denitrification rates, mineralization
of soil N and pesticide drift surface deposits were excluded as the effect on water quality could only
be inferred. Some examples of studies excluded at full text are listed in Additional file 2.
Types of study:
Only studies that reported primary research investigating the effect of an intervention on water
quality was considered for inclusion in the review, which therefore excluded review articles and
modelling studies. A list of systematic reviews and meta-analysis found as part of the search process
are attached as Additional file 3.
Types of comparator:
No restriction was made on the type of comparator. However, studies with a no mitigation treatment
(e.g. cropped or bare ground plots) were categorized as controlled studies, whereas studies using
measurements over time and space were categorized as not controlled, but with comparator.
Language:
Studies published in English.
Date:
No date restrictions were applied.
Study exclusion:
The initial Endnote file contained a large number of irrelevant articles, therefore a list of keywords
was drawn up to use as exclusion terms based on discussions between reviewers. Targeted keyword
searches were used to filter out articles relating to non-relevant subjects such as mining, transport,
22
medicine, cell biology, oceans, palaeontology and energy. Studies from irrelevant geographical
zones were removed using general (e.g. tropical, Africa) and specific keywords (Australia, India).
Articles from journals specializing in non-topic subjects were checked for irrelevance and then
removed (e.g. Lancet). A second stage of keyword searching was used to filter out studies that were
more closely related to the target topic areas but still irrelevant to the question, e.g. biodiversity,
zoology, soil biology, plant pathology, plant physiology, sewage, air pollution and southern Europe.
All articles that were excluded from the second stage of keyword searching were manually
examined by at least by title before finally being excluded.
The remaining articles were examined at title, and then title and abstract level for relevance. Full
text screening was used to produce the final reference list for the systematic map database. An
article was passed to the next inclusion stage if there was doubt about its relevance. At least 2
reviewers checked an articles relevance when there was doubt over the application of inclusion
criteria. The number of references retained and excluded at each stage of the screening process was
recorded. A Kappa analysis was performed following the keyword exclusion stage on 50 randomly
selected articles, read at title or abstract level to assess the agreement between 2 reviewers in the
application of inclusion criteria to the next stage. The kappa statistic was calculated using the online
calculator at: http://www.graphpad.com/quickcalcs/kappa1.cfm.
Duplicates and irrelevant articles were removed from Google search results using the procedure
outlined for the main search results. Search results from organizational web sites were checked by
title for relevance. Those that passed the inclusion criteria were then examined at abstract/full text
by following the web links. The remaining Google scholar and web site search results were
combined with the main search results before the final stage of screening at full text, and any
duplicates removed.
23
Coding for the systematic map
Key wording was used to describe, categorise and code articles in the systematic map database.
Keywords were generated from the primary question, topics reported in articles, existing systematic
maps and expert knowledge. Articles were either coded on full text, abstract or title depending upon
the availability of text (recorded in map under text read). Although literature searches were
performed in English, some translated foreign language texts were included in the database. The
definitions of the categories and codes used in the systematic map are detailed in Additional file 4.
For some categories more than one code was applied to an article, for example articles that reported
results from more than one country or had multiple water quality measurements.
Coding was moderated between reviewers to ensure consensus. Reviewers met at least weekly to
review progress and to clarify any ambiguities. Any uncertainty in the application of a code to a
specific article was flagged and discussed.
In summary the following information was recorded (full details are in Additional file 4):
Bibliographic information: first author, title, year of publication, full reference and article
type.
Linked study: articles reporting on the same study were cross referenced by a number
(linked study) e.g. early study finding or journal articles linked to reports, thesis or
conference papers. Where possible, both were coded, but only one was used for further
analysis. The article for exclusion from further analysis was marked as [Dup] based on the
following criteria: data were not extractable for meta-analysis; the study length was shorter;
less water quality measurements recorded.
General study information: intervention; country of study; length of study; scale of study.
Study design: replicates (temporal or spatial); randomized; control and/or comparator; study
type. The study type was categorized as either manipulative (the intervention was applied by
the investigator e.g. different rates of fertilizer applied, buffer strip vegetation planted),
24
correlative (the intervention may have been existing, but a comparator/control was always
employed), monitoring (intervention effectiveness was validated against a standard or value
e.g. drinking water standards) or sampling (samples taken from study area, but no
control/comparator employed).
Sampling information: time of year measurements taken; sampling location (e.g. lab,
mesocosm, plot, river/stream); sampling method (soil cores different depths, lysimeters,
ceramic cups, stream samples).
Confounding factors: A study where an outcome could not be definitively apportioned to
one intervention e.g. outcomes from studies of best management plans composed of multiple
mitigation measures including one or more of the review-specific interventions combined
with others (e.g. fencing streams to deny access to cattle, implementation of farm nutrient
plans).
Topic specific: fertilizer (organic or inorganic); flow path (surface, subsurface,
groundwater); soil texture/geology. Subsurface was the default coding when measurements
were taken below ground (e.g. ceramic cup), but the flow path was not stated by the authors.
Intervention-specific information: Buffer type (vegetation composition); tree type
(deciduous or conifer); cover/catch crop grouped under cereal (e.g. barley), grass (e.g.
annual rye grass), legume (e.g. vetch), brassica (e.g. mustard), volunteer weeds and winter
wheat (classed separately due to ambiguity in definition); slurry storage location (above or
below ground) and construction material (concrete, steel, earth lined).
Water quality measurement: The water quality measurement used in the study (N, P,
sediment, pesticide and bacterial pathogen counts).
Study outcome water quality: An overall outcome for the effectiveness of a study in
improving water quality based on reviewers interpretation of authors conclusions (no
statistical checks). There were 3 possible outcomes: yes pollutant reduced (clear statement
by author that pollutant was reduced); no pollutant reduced (clear statement by author
25
pollutant not reduced); not clear pollutant reduced (either not stated clearly by author or
outcome not clear to reviewer). The specific form of each pollutant measured was recorded
in the outcome category for N (e.g. nitrate, ammonium, total N), P (e.g. orthophosphate,
total P, inorganic P), bacterial pathogen (e.g. E.coli) or pesticide name. Some studies were
coded with multiple outcomes (yes and no), if the outcome was dependant on the
control/comparator (e.g. both bare ground plots and inflow/outflow), flow path
(groundwater/surface), sampling location (plot or stream) or mitigation.
Study outcome slurry leakage: An overall outcome for the effectiveness of a study in
reducing slurry leakage based on reviewers interpretation of authors conclusions (no
statistical checks): slurry leakage detected, no slurry leakage detected; not clear slurry
leakage detected
Experimental factor: Experimental factors under investigation e.g. buffer width, tillage, soil
type, crop type
Heterogeneity in outcome: a clearly stated explanation by the author for variation in results
(e.g. soil type, buffer width, cover/catch crop type, pesticide type), and a summary of overall
study outcome (Mitigation-Not Successful, Mitigation-Successful, Mitigation-Outcome Not
clear, Mitigation-Outcome depends Pollutant, Mitigation-Outcome depends form Pollutant).
Due to limitations in database design it was not possible to differentiate outcomes of studies
that varied depending on mitigation, pollutant flow path, sampling point or the type of
control employed (studies with multiple controls e.g. bare ground and cropped controls)
these were flagged respectively as: Mitigation-Outcome depends Mitigation; Mitigation-
Outcome depends Flow; Mitigation-Outcome depends sampling point Mitigation-Outcome
depends control. Two notes section recorded any noteworthy comments relating to study and
outcome.
Best recorded outcomes: The authors best % reduction recorded for the following
measurements of water quality: total N, inorganic N, organic N, nitrate N, ammonium N,
26
total P, soluble P, particulate P, organic P, inorganic P, pesticide, sediment, bacteria pathogen
counts.
Summary data used for calculating measures of effectiveness: Overall effectiveness for N, P,
sediment, Pesticide, sediment, and bacterial counts with one of four possible values: Yes (all
forms of measurement reduced; Partial (at least one form of measurement reduced); No (no
form of pollutant reduced); Not clear (any other outcome).
Summary data for generating numerical counts: Soil categories, buffer types and flow path
types. A value of mixed indicated that there was more than one code for a category.
Meta-analysis: flag to indicate studies used in meta-analysis
Systematic map database
A searchable database of coded articles was created to describe the water quality research for the
topic specific mitigation measures. The searchable database is included as a Microsoft Access file
(Additional file 5). The list of references included in the database is attached as an additional file
(Additional file 6). The database can be ordered or filtered by category, and provide simple
numerical counts. There are 2 database tables included:
1.WaterQualityMapTitleAbstractFullText: This table contains all the articles that were coded,
whether at title, abstract or full text. All evidence was coded with country of study, mitigation and
water quality measurement, if the information was missing, ‘not clear’ was recorded. In addition
full text articles were coded for study design, and those articles without confounding factors were
coded for outcome. Data in this table were used to calculate the hierarchy of evidence by filtering
for studies coded at full text with no duplicates (articles reporting same study).
2. WaterQualityMapFullText: This table only includes studies without confounding factors coded at
full text. Articles reporting same study (duplicates) were also removed. Data in this table was used
to calculate measures of effectiveness.
Summary tables and graphs of study characteristics were generated from the systematic map 27
accompanied by a narrative synthesis. A mean and standard deviation were calculated from
individual article scores to give an overall score to mitigations for hierarchy of evidence and
effectiveness. Effectiveness scores, combined with the quality of evidence provided an indication of
the level of effectiveness and knowledge for each intervention.
Study quality assessment for each intervention
Every article coded in the systematic map at full text (including studies with confounding factors)
was scored according to a hierarchy of evidence adapted from systematic review guidelines used in
public health [30] and conservation [31], and using a system adapted from a method outlined by
Pullin and Knight [32]. Studies were given values for their design, based on categories applied in
the systematic map database, see Table 2. Values were calculated using standard Access queries of
specific categories in the database (see Additional file 7) (ie. whether studies were randomized, had
a control or comparator, had replicates (spatial or temporal), were conducted for longer than a year,
and whether they were manipulative, correlative, monitoring or sampling.) The values for each
category were combined for each study, and used to provide an overall indication of the type of
evidence available for each intervention. Topic-specific criteria such as sampling methodology were
not used, due to concerns from subject matter experts that this would introduce an unacceptable
level of subjectivity.
Evidence of effectiveness for each intervention
Each article coded in the systematic map at full text (excluding studies with confounding factors)
was given a value for effectiveness in reducing N, P, sediment, pesticides or bacterial pathogens in
water. A system adapted from Ramstead [33], (see Table 3), gave each study a value for
effectiveness using the following scale:
3 –Intervention fully effective for all forms of measurement
2 –Intervention partially effective for at least one form of measurement
28
1 -Intervention effectiveness not clear
0 -Mitigation not effective at all
The scores were automatically calculated from codes in the map (N effective; P effective; sediment
effective; pesticide effective; bacterial pathogen effective) using the Access queries which are
documented in an additional file (Additional file 7). The values were combined to provide an
indication of the overall effectiveness of each intervention for named outcomes.
The effectiveness of cover crops - Meta-analysis
Meta-analysis is a technique, developed in medicine, whereby the results of multiple studies are
combined and analysed together [34]. Meta-analysis is rarely carried out on raw data but on derived
statistics which are then synthesised to give an overall estimate. Study results are assigned weights
according to the sample size and the degree of error, allowing the relative value of different studies
to be compared objectively. In primary analysis, large effect sizes may be erroneously ignored
because of inherent low statistical power but in meta-analysis even small studies can usefully be
included as this approach combines effect sizes across studies, resulting in greater statistical power
[34].
The effectiveness of cover/catch crops in reducing leaching of N was selected as the focus for the
meta-analysis as a considerable body of evidence exists for cover/catch crops and N. There was also
a large amount of research available for the effectiveness of buffer strips on water quality, but 3
existing meta-analyses were found measuring the effect of the mitigation on N, P, sediment and
pesticides (Additional file 3). For the other mitigations, there were not enough experimental studies
for a meta-analysis.
29
Initial selection of studies for meta-analysis was made according to the following criteria:
The study directly compared cover/catch crop to a fallow/no vegetation treatment
The study investigated the impact of cover/catch crops on leaching of N
The study was judged to be high quality i.e. from peer-reviewed journals or scientific studies
commissioned by the government
The study reported sufficient information to be included without approaching the author
The remaining papers were inspected to determine the difficulty of data extraction and were
assigned to three categories
Simple: Data presented in either a table or text; to include a mean value with an associated
n, and either a standard deviation/ error or P value.
Medium: data appeared to be available, but some further calculation may be necessary. Data
may also be presented in simple graphs.
Complex: data are given in complex figures or complex tables, not immediately clear if data
would be extractable.
When possible data were extracted from tables and text, and DataThief [35] was used to extract data
from graphs. The mean, standard deviation and sample size (n) were extracted for each intervention
(cover/catch crop) and the corresponding control (fallow/ un-vegetated plot). In one case, a study
was split into 2 experiments or buckets[36]. The P-value was extracted if the standard
deviation/error was missing, however where the figures for insignificant values were not reported,
instead only presented as ‘not significant’, the dataset could not be used. Frequently the P-value was
reported rather generally (i.e. at 0.05, 0.01, <0.001) and in those cases the upper bound point was
used, which is not as accurate or desirable. Data were collapsed if presented over several time
points or treatments (other than cover/catch crop type) and were averaged using the arithmetic
mean. The variance (obtained by squaring the standard deviation) was averaged and weighted by
sample size.
30
Data synthesis and presentation meta- analysis
The effect of cover/catch crops as an intervention to manage N in leachate was investigated using a
standardised difference of the mean, which transforms effect sizes to a common metric so that data
are comparable between studies. For this study it was necessary as concentration of N was reported
in different ways (e.g. mgl-1; kg NO3-Nha-1 and mmol NO3-Nm-2). Hedges g is a statistical method
for estimating effect sizes, which estimates the amount of the variance within an experiment that
can be explained by the experiment. Hedges’s g was selected for this study as it gives an unbiased
estimate of δ (standard deviation of a population) suitable for small samples [34]. Meta-analysis
was used to calculate the summary effect of cover/catch crop at the study level and a test
comparison was run to measure the effects of cover/catch crop type. Data were also grouped by soil
type to investigate the impact on cover/catch crop effectiveness. The random effects model was
employed as it could not be assumed that the variance was equal between studies. Analysis was
carried out using Comprehensive MetaAnalysis, version 2.2.064.
Results
Review descriptive statistics
Online database searches identified 74,086 records after duplicate removal. Google scholar search
results identified 4168 articles after duplicate removal and oorganizational web site searches
identified 5430 records. The number of records generated for specific searches are identified in an
additional file (Additional file 1). A further 28 articles were identified from bibliographic
references. The records were screened for relevance as illustrated in Figure 1, and the included
articles recorded in a Microsoft Access TM database (Additional file 5).
The Kappa analysis showed an agreement on 45 out of the 50 articles screened on reading title or
abstract with a calculated Kappa score of 0.588 (Confidence interval: 95% ), which is considered as a
moderate agreement between reviewers [37] and is acceptable for this type of review. 31
A total of 718 records were judged to have met the inclusion criteria after all the search results were
combined and these records were used to populate the systematic map. The systematic map is
available as an electronic database and is attached as an additional file (Additional file 5). The list
of articles included in the map is listed in an additional file (Additional file 7).
Of the 718 records included in the map, 495 were coded on full text details, 147 on abstract details
and 76 on title details. Of those 718 records, 29 articles were marked as articles that reported the
same study twice resulting in 689 unique studies in total.There were potentially 94 articles (it was
unclear for 23) that were non English language texts coded on either abstract or title details. Overall
the majority of the articles were journal papers (n=494), followed by conference papers/posters
(n=118), reports (n=44), and theses (n=27). The remaining articles (n=35) were either books or the
article type was not clear.
The earliest article dated back to 1950, but after that date there were no more publications until
1971. In the early 1990s there was an increase in publications (Figure 2).
Mitigation measures
Buffer strips were the most commonly studied mitigation (n=364), followed by cover/catch crops
(n=245), slurry storage (n=93), woodland creation (n=24) and subsoiling (n=10) (Figure 3). No art-
icles were found that studied the effect on water quality of controlled trafficking on grasslands.
Country of origin
A large percentage of studies were from the northern states of the USA (n=256). The country of
study was not mentioned in many articles read at title or abstract (n=115). The UK was the domin-
ant country in Europe (n=80). The Western European countries of Germany, France, Holland, Bel-
gium, Austria, Switzerland and Ireland were also well represented (n=107) and so were the Scand-
inavian countries of Norway, Finland, Sweden and Denmark (n=85). The Eastern European coun-
tries of Latvia, Lithuania, Estonia, Poland, Romania, Belarus, Slovakia and Ukraine did not repres-32
ent a large component (n=28) of the map. Canada was better represented (n=40) than New Zealand
(n=12). No studies were found from Luxembourg, Lichtenstein, Hungary, Czech Republic or Latvia
(Figure 4).
The dominant mitigation measure studied in the Northern states of the USA was buffer strips. How-
ever, in the UK cover/catch crop were slightly more frequently studied than buffer strip (Figure 4).
Outcomes measured
The dominant water quality measurement in the map was N (n=473), followed by P (n=178) and
sediments (n=165). Less evidence was found for pesticides (n=71) and bacterial pathogen counts
(n=61). Figure 5 details the number of articles found for each mitigation measure and water quality
measurement. Measurements of N were recorded in buffer strip, cover/catch crops and slurry stor-
age studies. Most measurements of sediment were from buffer strip studies, although there were a
few cover/catch studies that measured sediments derived from soil erosion. Likewise most measure-
ments of P were from buffer strip studies, with a smaller amount of evidence found for cover/catch
crops and slurry storage. A smaller amount of evidence was found for pesticides, often recorded
from buffer strip studies. The small amount of evidence for bacterial pathogen counts came from
buffer strip and slurry storage studies.
Study description buffer strips (including tree buffers):
There were 225 buffer strip included in the database after studies with confounding factors and du-
plicate articles were removed. Studies were mainly manipulative (n=147) or correlative (n=74).
Over a third of the manipulative studies were of short duration with either no temporal replication
or only time series data which was often derived from engineered runoff events. Runoff events were
generated by applying a known amount of pollutant to the start of the buffer strip and measuring
runoff at various stages along the buffer length. There were a 106 studies conducted for less than a
year. Over a third of the manipulative studies were conducted for a year or longer (n=64), whereas
33
the majority of the correlative studies were conducted for at least a year (n=52). Only 5 studies were
conducted for longer than 10 years.
There were 103 studies with a control, most frequently of bare ground or of a cropped or native ve-
getation plot and only 1 study that used a BACI experiment. In some cases the author stated that a
no buffer control was used which could be a separate buffer plot of 0m in length (where applied
runoff is collected immediately to a calibrate collection system). Studies without a control usually
had a comparator, often a comparison of water quality measurements along the width of a buffer,
starting from the inflow of water and ending at the outflow of water. Some controlled studies repor-
ted results in relation to inflow measurements rather than controls. A few studies measured changes
in water quality over time.
Most studies (n=146) were conducted at single sites/farm or in laboratories/lysimeters/mesocosms.
There were only a few multi-site studies (n=41), or larger scale studies at catchment, regional, coun-
try or international level (n=38). 74 studies had data for all 4 seasons, and more studies were con-
ducted in summer than in winter. Surface runoff water was often collected using gutters or weirs,
whereas subsurface water was frequently collected using lysimeters or ceramic cups.
Manipulation of vegetation was a common experimental factor e.g. vegetation type, age, height or
density, or cutting and harvesting of vegetation (n=98). Other factors studied were the type or
amount of fertilizer applied to plots (n=22), buffer width (n=52), soil type (n=23) and gradient and
slope of land (n=19). Buffer strips were composed of either grass (n=154), trees (n=55) or a mixture
of trees, grasses or shrubs (n=69). More studies recorded buffer strips composed of deciduous trees
than conifer species.
Study description cover/catch crops:
There were 132 cover/catch crop studies included after those with confounding factors and duplic-
ate articles were removed. Studies were mainly manipulative (n=125) with a few correlative studies 34
and one monitoring study. Most studies were conducted for at least two winter seasons (n=102) with
8 lasting for more than 10 years. Cover /catch crops were either grown alone, intercropped with a
winter crop, or drilled into the stubble left from the previous crop. Fallow, bare ground or cropped
plots were commonly used controls. A few studies did not have a control, either measuring changes
in water quality over time or between different cover/catch crop types. Volunteer weeds and winter
wheat were sometimes used as controls, but in other cases used as crop covers.
The effectiveness of cover/crops in improving water quality was mainly measured from within field
plots (n=111), there were a few lab/lysimeter studies, but only one study that sampled river water
[38]. Water quality was measured using ceramic cups, lysimeters, monitoring wells, drainage or
sometimes estimated from soil cores taken at different depths.
Commonly used experimental factors were crop type (n=62), date and amount of fertilizer applica-
tion (n=45), the date and technique for removing the cover/catch crop (n=6), type of tillage (n=27),
or soil type (n=18).
Study description slurry storage:
There were 42 slurry storage studies summarised after studies with confounding factors and duplic-
ate articles were removed. The study types could be divided into 3 categories:
Studies that measured leakage from under or nearby to slurry storage (coded in the map as
sampling location under slurry storage, near slurry storage -50m, or aquifer (n=23). These
types of studies were mainly correlative using measurements over time and distance as com-
parators. Slurry was normally sourced from swine or dairy farms. Most of the slurry stores
studied were earth lined and below ground. Only 4 of the articles in this group were less
than 10 year old and only 6 of the studies were conducted in Europe (the other studies were
from the USA and Canada). Slurry storage legislation and drinking water standards may
35
therefore not be comparable across studies. For example, one study from the USA discussed
legislation that came into effect in the state of North Carolina [39].
Studies measuring survival rates of bacterial pathogens in slurry, the comparator being time
(coded in the map as sampling method slurry, n=11).
Field-based studies that measured the effect on water quality of timing and amounts of
slurry application in winter (sampling location plot/field, n=8).
Study description woodland creation (excluding woodland buffers):
Water quality studies of buffer strips composed of trees were categorised as buffer strip studies
rather than woodland creation and so were not considered here. The woodland creation studies des-
pite their small numbers (n=12) were very diverse, making trends and comparisons difficult to es-
tablish. The studies could be divided into 3 main categories:
Studies that measured water quality under afforested former agricultural land and compared
the results to cropped or forested land or measured differences across different tree species.
Included under this category were the studies that reported findings from the AFFOREST
project which measured the effect on water quality of afforestation on former agricultural
soils in 3 different European countries [40].
Studies that measured the effect on water quality over time of trees grown for biomass.
Studies that measured the effect of water quality of trees intercropped with a cash crop.
Study description subsoiling:
There were only 5 studies coded for subsoiling, 4 of which measured soil erosion and sediment run-
off. All the studies were manipulative and used a no-subsoiling control. Controlled traffic on grass-
land had no studies coded in the map.
36
Study quality assessment
The scientific rigour of study design for each intervention was assessed by the hierarchy of evid-
ence scoring system detailed in Table 2, and was applied to all full text articles excluding duplicates
(n= 467). Individual article values were combined to and a mean used to indicate the comparative
scientific rigour of each intervention studied (or provide a hierarchy of evidence) [32]. Cover/catch
crops had a higher scientific rigour value (mean value 6.8, standard deviation (s.d) 3.1) than both
buffer strips (mean 5.9, s.d. 2.4) and slurry storage (mean 4.1, s.d. 3.1), reflecting a greater percent-
age of randomized controlled experiments of manipulative study design recorded for cover/catch
crops, but values were very variable (Figure 6 a-c). The variation was, in part, due to the awarding
of 0 to studies with confounding factors, but also reflected the variability of studies included in the
map. Slurry storage and cover/catch crops had proportionally more studies scored as 0 (mainly con-
founding factor studies), compared to buffer strips (Figure 6 a). If studies with scores of 0 were dis-
regarded then slurry storage and buffer strips studies had scores that ranged from 2-9 whilst cover/
catch crops studies had scores that ranged from 4-9. The hierarchy of evidence value of 7.3 recor-
ded for the woodland creation mitigation measure (Table 4), should be interpreted with caution as it
is based on a small sample sizes and none of the included studies had confounding factors. A score
was not calculated for subsoiling due to small sample size (n=5).
Evidence of effectiveness
There were 410 studies scored for measures of effectiveness, (all articles obtained at full text except
where an article reported the same study (duplicate) or where there were, confounding factors). The
studies were diverse for study design, comparator, sampling location and experimental factor across
and within mitigations. The records were screened as illustrated in Figure 7 and summarised in the
Access database (waterqualitymapfulltext, Additional file 5)
Most of the evidence given values for effectiveness was drawn from field or plot studies for
cover/catch crops (n=111) and buffer strips (n=187), very little evidence was drawn from stream,
37
sample measurements (Figure 8 a). This may be in part be due to difficulties in accurately
identifying within river/catchment impacts from specific mitigations. Thirty three studies were
found that did measure water quality in samples taken from river/streams, but none are shown on
Figure 8a as all had confounding factors and so were not scored for effectiveness. Both organic and
inorganic fertilizers were used in studies (Figure 8b). However, for buffer studies, the form of
fertilizer was often unclear either because runoff was derived from nearby fields rather than from
manipulated experiments, or the focus of study (e.g. changes in sediment, bacterial pathogen counts
or pesticides) was such that the form of fertilizer was not mentioned by study authors. Studies on
loam soils dominated the evidence base (Figure 8 c).
The distribution and average values recorded for measures of effectiveness for buffer strips, slurry
storage and cover/catch crops for each water quality measurement are shown in Figure 9 (a-b) and
in text form in Table 5. Values were on a scale of 0 to 3, a study coded ‘yes-reduced’ for outcome
was scored with a value of 3, a study coded ‘not clear’ for outcome was given a value of 1, a study
coded ‘not reduced’ for outcome was given a value of 0. A study value of 2 indicated a partial suc-
cess where at least one form of N, P, bacterial pathogen counts or pesticide was reduced. No studies
measuring sediments received a partial score as unlike other measurements there were not multiple
forms. No mitigation had a large amount of scores marked as 2 (partial outcome) as shown in the
distribution graph of scores (Figure 9a), suggesting that the scoring system did not disadvantage
studies measuring multiple forms of a pollutant. Mean values and standard deviations were used to
indicate the overall value for each combination of intervention and outcome (Figure 9b). The values
are rudimentary and for comparative purposes only. They depend on reviewer interpretation of
study outcomes rather than statistical analysis so should be interpreted with caution.
Six studies included in the scoring for effectiveness had at least 2 relevant interventions that would
have received an overall score across both interventions. One study was marked as having an out-
38
come dependant on the type of intervention. These could not be separated in the database so both
interventions were marked as partially successful [41].
Evidence of effectiveness buffer strips (including woodland)
Buffer strips appeared to be most effective at reducing sediment (2.7) (Figure 9), followed by pesti-
cide (2.3), N (2.2), P (2), and pathogen counts (1.8). This ranking of buffer strip effectiveness (sedi-
ment, pesticide, N,P, bacterial pathogen) was similar to the ranking reported in a pre-existing meta-
analysis (pesticide, sediment, P, N) [27]. Further analysis would be useful to investigate the impact
of sediment bound P within the value for P.
Outcome for N:
There were 139 studies that assessed the effect of buffer strips on improving water quality as meas-
ured by N (all flow paths). Loam was the most commonly studied soil type (n=71), sand (n=16) and
clay (n=6) were studied rather less, however there were a lot of studies coded with no or mixed soil
type (n=52).
Authors indicated that buffer strips are generally effective for reducing at least one type of N (72%,
n=100/139), but that this varied for different forms of N. Nitrate, total N and ammonium N were the
most commonly measured forms of N (Figure 10). Proportionally more buffer strip studies were
coded as ‘yes-reduced’ for total N (74%, n=29/39) and nitrate (67%, n=80/120) than for ammonium
N (50%, n=23/46) (Figure 10). Few studies investigated soluble or organic forms of N (n=10). The
prevalence of total N studies that were coded as ‘yes-reduced’, may reflect a greater number of
studies measuring water quality in surface flows.
Figure 11 shows the impact of buffer strips on the 3 main forms of N measured in surface, subsur-
face or ground waters. Studies with multiple flow path measurements were excluded from the fig-
ure. Subsurface was the default coding for studies that measured flow path below ground, therefore
39
this category may contain some groundwater studies. There were proportionally more studies as
‘yes-reduced’ for surface water measurements of total N (91%, n=21/23), than either nitrate 71%,
n=20/28) or ammonium N (67%, n=16/24). Proportionally more buffer strip studies were coded as
‘yes-reduced’ for subsurface/groundwater nitrate measurements (66%, n=47/71) than ammonium N
(36%, n=5/14). One study [42] measuring groundwater found that nitrate generally decreased under
buffer strips, but that ammonium could increase in groundwater, with one study suggesting [43] that
litter inputs from vegetated buffers could be creating fluxes of ammonium in groundwater. The
sample size for total N in subsurface measurements was too small (n=4) to allow any meaningful
trend to be concluded.
The outcomes coded for buffer strips of different vegetation types are shown in Figure 12. Studies
that compared differences between grass and woodland buffers were excluded from the figure.
Measurements of groundwater/subsurface flow were more common for woodland buffer studies
(75%, n=39/52), than grass buffer studies (14%, n=14/102). The number of studies coded with an
outcome of ‘yes-reduced’ for nitrate showed no apparent difference between tree buffers (66%,
n=28/42) compared to grass buffers (68%, n=27/39). Some studies that reported differences in ef-
fectiveness between vegetation types cautioned that other factors (eg differences in landscape or nu-
trient flow rates) may have influenced the results. One study [44] found that grass removed almost
double the amount of nitrate compared to a forest buffer, but the forest was experiencing higher
flow rates of nitrate than the grass buffer and had become saturated. Another author [45] suggested
that site differences in water table depth may have influenced the outcome of their comparison
between grass and tree buffer strips.
Outcome for P:
Ninety-four studies assessed the effect of buffer strips for improving water quality as measured by
phosphate. The most commonly studied soil type was loam (n=55), with a few studies coded for
clay or sand (n=7). 40
Authors indicated that buffer strips could be effective for reducing at least one type of P (65% of
studies measuring P, n=61) but that this varied for different forms of P. Total P, orthophosphate and
soluble P were the most common forms of P studied (Figure 13). There were proportionally more
buffer strip studies coded as ‘yes-reduced’ for total P (73%, n=46/63), than for orthophosphate
(55%, n=23/42), or soluble P (26%, n=5/19) as shown in Figure 13. There were 10 studies coded
with an outcome as ‘yes-reduced’ for particulate or sediment bound P out of a total of 13. Only 4
studies recorded an outcome for organic forms of P.
Figure 14 shows the outcomes coded for buffer strips and the 3 main forms of P measured in either
surface, subsurface or groundwater. Studies with multiple flow path measurements were excluded.
Proportionally more buffer strip studies that measured P in surface flows were coded as ‘yes-re-
duced’ for total P (84%, n=32/38), than for orthophosphate (71%, n=15/21). As few studies meas-
ured subsurface or groundwater flows of P no comparison between flow paths can be made (total P,
n=7; soluble P, n=2; orthophosphate, n=9). Phosphorous has a low mobility in soil therefore it is not
surprising that most evidence relates to surface flows. Thirteen studies measured P in multiple flow
paths. One of those studies [46] found that buffer strips were effective in removing sediment bound
forms of P from surface flow, but were less effective in removing total P from subsurface flows, and
were not effective at removing soluble forms of P in subsurface flow. Another study also found that
buffer strips reduced levels of P in surface water, but not from drainage water [47].
Outcome for sediment:
Ninety-eight studies assessed the effectiveness of buffer strips for reducing sediment in water (Fig-
ure 15). Studies recorded soil type as either loam (n=62), unknown (n=28) or sand/clay/mixed
(n=8). There were 4 types of measurement recorded for sediment, but the categories may reflect dif-
ferences in terminology used by article authors. Sediment was coded for 57 studies as ‘yes-reduced’
out of a total of 66 studies (86%). Total suspended sediment was coded for 22 studies as ‘yes-re-
41
duced’ out of a total 26 studies (84%). There were 5 outcomes recorded for sediment soil loss and 2
outcomes for sediment measure as turbidity in water.
Outcome for bacterial pathogen counts:
Nineteen studies assessed the effectiveness of buffer strips for influencing bacterial pathogen
counts. Most investigated surface flow (n=17). Study soil type was coded as either loam or un-
known.
Authors indicated that buffer strips can be effective for reducing at least one of the bacterial count
measurements (63% of studies measuring bacterial pathogen counts, n=12). Of 11 studies coded
with an outcome for total faecal coliform, 7 were coded as ‘yes reduced’ (Figure 16). Of 7 studies
coded with an outcome for E.coli, 3 were coded as ‘yes reduced’ (Figure 16). Two studies measured
subsurface flow and both were coded as ‘not-reduced’ for E.coli [47] [48], one study found that the
outcome depends on flow, as E.coli was reduced in surface flow, but not in drainage water [47].
There were a small number of outcomes recorded for bacterial pathogen counts measured as total
coliform, Streptococcus spp., Cryptosporidi spp. and Enterococci spp..
Outcome for pesticides:
Thirty-eight studies assessed the effect of buffer strips on improving water quality as measured by
pesticide levels. Loam was the most frequently studied soil type (n=22) although it was not possible
to code for soil type in many cases (n=13). Only 3 studies used either sand, clay or mixed soil types.
Surface flow was coded for 15 of the studies, and subsurface flow for 9 studies (A further 8 studies
measured both flow paths).
There were 35 different pesticides coded in the map, of which atrazine and metolachlor were the
most commonly studied (Table 6). Authors indicated that buffer strips are generally effective for re-
ducing at least one of the 38 pesticides measured (71% of studies measuring pesticide, n=27). Of
42
the 26 studies coded with an outcome for atrazine, 16 were coded as ‘yes reduced’ (Table 6). Of the
12 studies coded with an outcome for metolachlor, 9 were coded as ‘yes reduced’ (Table 6). How-
ever, one study [49] found that whilst levels of metolachlor and atrazine were reduced by the buffer,
the outcome was not significantly different to results from a bare ground plot. Therefore this study
was coded as outcome depends upon control/comparator. The pesticides Isoproturon, Endosulfan
and Metribuzin were measured in a few studies (n=4).
Reasons for heterogeneity in results and limitations of evidence base:
Buffer strip effectiveness may depend on experimental factors such as vegetation types, but effect-
iveness was only given as an overall value. Experimental factors such as buffer width, slope, flow
rate (of water containing nutrients coming into buffer), amount of fertilizer applied, season, vegeta-
tion type, vegetation age, vegetation height drainage, cutting harvesting biomass were cited by au-
thors as reasons for heterogeneity in results.
Buffer strip effectiveness was often assessed on either loam or unknown soil types, which may not
capture the effect of soil particle size on buffer strip performance, some authors did cite differences
between loams based on silt, sand or clay composition. A multi-site study, with silt loam, and silt
clay loam soils [50] noted that a wider buffer was needed for soils with a high clay content as soil
particles were smaller and took longer to deposit in surface flow. Buffer strip effectiveness was of-
ten assessed on either loam or unknown soil types, which may not capture the effect of soil particle
size on buffer strip performance.
Buffer strip effectiveness was often assessed at field scale, which may not capture the effects of
preferential flow paths or buffer strip placement on buffer strip performance. A Defra commissioned
buffer strip study at 3 sites representative of UK soil types [51] found no significant difference in
levels of total-N, nitrate or molybdate reactive P in river samples taken from paired catchments
(buffered and not). However, at the field site fans of sediment deposits were observed at the edge of
43
the buffer strip and ground monitoring wells recorded reductions in nitrate and total N on buffer
strip sites (not clear for P). One explanation given for the result was that phosphate could have been
stored as sediment in the river and was acting as a source for sediment bound P which, until de-
pleted, would mask any positive effects of buffer strip implementation. Another reason cited was
that water flows may have bypassed the buffer strip either through underground drainage, or vertical
movement into aquifers. Reductions in P measured at buffer strip plots not translating to reductions
in stream samples have been observed in other studies [52]. The authors suggested that the study
should have been longer than 2 years so as to observe the long term effectiveness of buffer strips.
Differences between vegetation types such as grass and trees may only become apparent over time,
as trees mature more slowly.
Variability in the hydrological landscape has been cited as an important factor for buffer strip effect-
iveness. Delivery rates of groundwater can affect a buffers ability to improve water quality. One
study found that specific regions of a river consistently received high loads of N and considered that
their identification was critical for effective catchment planning. Other studies have noted that
zones of upswelling of groundwater containing nitrates could reduce buffer strips ability to reduce
soluble pollutants, one of the areas studied supplied 4% of the streams flow, but only represented
0.006% of the riparian zone [53].
The findings of another study suggested that the implementation of buffer strips on former agricul-
tural land could increase leaching of soluble P, due to changes in plant-microbe interactions [54].
Other authors have reported that P can be leached from buffer strips over time [55, 56]. The leach-
ing of N from buffer strips has been reported once [57]. A general decline in buffer strip efficiency
under artificial rainfall was noted by another author [58].
Seasonal differences in plant growth and nutrient uptake may impact of buffer strip effectiveness.
Further analysis of the studies with data for all 4 seasons would be needed of identify any seasonal
effect.44
Evidence of effectiveness: cover/catch crops
Cover/catch crops were most effective at reducing sediment and N (both 2.3) (Figure 9), however
some of the sediment studies used a crop cover of winter wheat rather than a traditional cover/catch
crop. Cover/catch crops had a relatively low value for P (1.2).
Outcome for N:
One hundred and fourteen studies assessed the effectiveness of cover/catch crops for reducing N,
mainly from subsurface/groundwater measurements (the distinction may be artificial as subsurface
was the default when below ground measures were not specified). Loam was the most commonly
studied soil type (n=60). Twenty-nine studies were coded for sand, 9 for clay and 31 were unknown
or used an unknown/mixed soil type. Grass, cereal, brassica and legumes were the most commonly
studied cover/catch crops (Table 7). Nitrate was the most commonly measured form of N, a few
studies measured total N, ammonium N and N- inorganic, but no studies measured the organic
forms of N (Figure 17). Of the 108 studies coded for nitrate, 74 were coded as ‘yes-reduced’ (69%).
Outcome P:
Both surface and subsurface water measurements of P were taken in the 14 cover/catch crops stud-
ies, which were conducted on a range of soil types. Grass was the dominant cover/cover crop stud-
ied (Table 7). Total P, soluble P and orthophosphate P were commonly measured. Total-P was coded
as ‘yes-reduced’ for 3 studies, ‘not-clear’ for 5 studies, and ‘not-reduced’ for 1 study. Of the 7 stud-
ies that measured soluble/orthophosphate P, no studies were coded as ‘yes-reduced’.
Outcome Sediment:
Most of the 19 cover/catch crops studies measuring sediment, studied grass, winter wheat or other
cereal cover/catch crops on a loam soil type. The focus of a majority of the studies was erosion (the
45
term erosion was used in the title of 11 studies). There were 13 studies that had a coded outcome of
‘yes-reduced’, which was mainly recorded as ‘sediment-soil loss’.
Reasons for heterogeneity in results and limitations in evidence:
Authors have suggested that a number of factors can impact on the effectiveness of cover/catch
crops such as the amount of fertilizer applied, the crop rotation, crop or cover/catch crop type, cover
crop establishment or sow date, the presence or absence of crop stubble, date of tillage, date or tech-
nique used to kill the cover /crop and soil type. For further details refer to the map (filter on reason
heterogeneity results and cover/catch crops).
Climatic data was often difficult to extract from studies, however some studies reported year to year
variation in effectiveness depending upon the date when autumn rains started [16]. Only a quarter
of the studies assessed effectiveness across all 4 seasons. However, a study reported in 2 articles
cautioned that cover/catch crop effectiveness in reducing leaching of N should be assessed over the
full crop succession [59, 60]. One of the articles [59] reported that a cover/catch crop of mustard re-
duced leaching of N in winter, when compared to a fallow, but a crop planted after the cover/catch
crop did not uptake more N than a crop planted on the fallow control. The other article for the same
study [60] reported increased leaching of N after the removal of cover/catch crops in spring com-
pared to the fallow plot.
Although some studies were of long duration (up to 30 years), the effect of stopping cover/catch
cropping on effectiveness was not studied that often, one study suggested that nutrients caught by
cover catch crops can be leached in subsequent years if no cover/catch crop is planted. A study [61]
suggested that stopping cover/catch cropping could increase leaching of N in subsequent years in
comparison with treatments that had not been previously cover/catch cropped, due to a build-up of
N under cover/catch cropped soils. However, a 17 year multi-site study [62, 63] found no temporal
46
reduction in efficiency of cover/catch crops for preventing nitrate leaching, although the effect of
stopping cover/catch cropping was not assessed [62].
The only cover/catch crop study in the map that measured water quality in stream/river samples was
a long term catchment monitoring study (9-16 years) which observed no downward trend of N or
sediment, but some reduction in P which the authors noted was at odds with the outcome for sedi-
ment [38]. Cover catch crop studies were often conducted on loam or unknown soil types, which
may not capture differences between soil types and nutrient leaching (e.g. sandy soils).
Evidence of effectiveness: Slurry Storage
Evidence of effectiveness values for slurry storage are based on assessments of slurry storage leak-
age or counts of bacterial pathogen in slurry are not therefore directly comparable to other interven-
tions that directly measured water quality (Figure 8 a). Slurry storage had the highest effectiveness
value for bacterial pathogens counts (2.2), but relatively low values for N and P (Figure 9), however
these results are based on evidence that has many limitations.
Limitation of the evidence for outcome N:
Much of the evidence was from outside Europe where slurry storage construction legislation may be
different. Of the 23 studies that measured leakage of N, 17 were from the USA or Canada. Quite a
few of the studies were old and used earth lined stores which may not meet current legislation.
Much of the evidence was based on studies that measured slurry storage leakage rather than the im-
pact of timing of slurry applications to maximise plant nutrient uptake. There was a very small
amount of evidence in the map that studied the effect on water quality of varying the timing of
slurry applications although timing of slurry application was not directly searched for. At least 2 of
those studies reported that a staggered application of slurry in winter could improve water quality
compared to one large untimed application [64, 65].
47
Whilst N was often detected under or near slurry storage (Figure 18), quite a few studies were not
of the highest scientific rigour. Some authors suggested that results for leakage may have been due
to experimental error. One study found that the complete emptying of a slurry store and then re-
filling caused slurry leakage as the earth clay liner had cracked [66]. One sampling study found that
it was not possible to identify if the slurry had leaked as part of the initial sealing or much later
when the storage was operational [67].
Most studies were not of the highest scientific rigour without baseline pre and post slurry storage
water quality data. A manipulative study with baseline data found that after building the slurry stor-
age nitrate levels rose in groundwater for the first 6 months then afterwards returned to pre slurry
store levels [68].
Most studies were conducted for less than 2 years therefore the effect over time e.g. age of slurry
storage may not have been accurately assessed. Soil type has also been given as a reason for differ-
ences in slurry storage leakage.
Limitation of the evidence for P:
There was only a small amount of evidence for P spread across the different study types therefore
no major conclusions can be drawn.
Outcome bacterial pathogen counts:
Studies showed that when no fresh additions of slurry were made to a slurry store pathogen counts
could reduce over time (Figure 18). Some studies found that bacterial pathogen die off rates could
be species dependant. One study [69] reported a 90% reductions in bacterial counts of E.coli in
slurry stored for 26 days, whereas there was not a considerable reduction in counts of Y. enterocolit-
ica after 73 days. Some studies found that temperature could affect the bacterial pathogen die off
rate and one study found that the die off rate of a Salmonella spp. increased at higher temperatures
[70].48
Evidence of effectiveness: woodland creation
Woodland creation studies most frequently measured N (n=11), whereas P, sediment and bacterial
pathogen counts were only once measured. The variety of controls/comparators employed in wood-
land creation studies made it difficult to code outcomes. Some afforestation studies did not have a
non woodland control, but instead measured changes in water quality over different aged woodlands
making it difficult to certain if woodland had improved water quality compared to agricultural land
[71, 72]. Some biomass studies did not have a non woodland control, but instead used a non fertil-
ized treatment as a control [73].
Modelling studies were excluded from the review, however they are useful for woodland studies
which experimentally can take years to. Furthermore the role of trees in pesticide reduction drift
was not included as pesticide was measured a deposit rather than within water. Forest Research has
recently reviewed the role of trees on water quality combining both woodland creation and buffer
strip studies and provides a comprehensive review in this area [18].
Evidence of effectiveness: subsoiling/controlled traffic on grasslands
Four out of the 5 subsoiling studies measured soil erosion and sediment loss from plots, but none
were coded as ‘yes-reduced’.
Review statistics meta-analysis
There were 114 cover/catch crop studies coded in the systematic map that measured nitrate leach-
ing. Of those studies, 48 directly compared the effect of cover/catch crops to a fallow or no vegeta-
tion control. The application of exclusion criteria immediately rejected 16 studies in a first pass and
a further 8 studies were rejected in a second pass. The remaining 24 papers were placed into one of
3 categories determined by the perceived difficulty of data extraction. There were 6 studies categor-
ised as easy for data extraction, 5 as medium for data extraction and 13 as difficult for data extrac-
49
tion. The screening system used to identify the records for inclusion in the meta-analysis is illus-
trated in Figure 19.
Study quality assessment meta-analyses
The 10 studies included in the meta-analysis were also scored for hierarchy of evidence. There were
3 studies that scored 9, 5 that scored 8 and 3 that scored 7. The studies with a score of 9 were ran-
domized, controlled, replicated and conducted for longer than a year.
Meta-analysis overall effect of cover/catch crop at the study level
Overall the meta-analysis suggests a consistent positive effect of cover/catch crop in reducing
leaching of N. It was disappointing that it was possible to include so few studies in the meta-
analysis. Largely this was due to poor reporting. Frequently there was no clear statement of what
had been used to calculate means, graphs either present no error bars or mean error bars which lack
precision and cover multiple comparisons. Many of the studies compare cover/catch crop with a
second set of treatments such as additional N or ploughing date or depth over multiple time points.
It is essential to partition the variance correctly and to do this a good understanding of how
summary data has been calculated is necessary. To produce this analysis, data were collapsed over
time and treatment (but not cover/catch crop type), by calculating means and averaging standard
deviation. The benefit is that these studies are comparable but it does not allow an inspection of the
variation associated with the various study designs and this is reflected in the large differences
observed between the studies. The precision of each study is influenced by the data we were able to
glean. The dataset showed significant differences between studies but also demonstrated relatively
little error within many of the studies. This is not surprising given the data that was included. 1)
studies with diverse aims often addressing more than just cover/catch crop which could lead to
differences in the effect size 2) well planned and well executed replicated studies, therefore the
within study variance (represented by the whiskers in the Forest plot) tended to be relatively small.
50
Using the study as Unit of Analysis, the results suggest that cover/catch crop consistently reduced
nitrate leaching (Z = 7.869, P = <0.001) but that there was significant variation between the studies
(Q = 131.31, df =10, P = <.001).
Almost all of the variation is due to difference between the studies rather than within study error (or
noise) as represented by I2 (92.). This analysis is based on combined data (across crop type) where
studies included more than one crop type [63, 74-76]. The data for the various comparisons
included a common comparison group and the assumption of independence is not true,
consequently the crop types cannot be treated as independent. Very few studies included legumes
(3) and grass (2). These were excluded from the analysis and a comparison was made between
cereal and brassica only.
Effect of cover/catch crop type (cereal v brassica)
Based on the mixed effects model, both brassica (Z = 3.18, P = <.001) and cereal (Z = 6.57, P = <
0.001) cover/catch crops are effective, and that there was no significant difference in the extent to
which they are so, based on this data set (Q = 0.83, P = 0.362). The variance, as given by T2, is
larger for brassica (1.774) than Cereals (0.979) indicating that there was more variation between the
brassica studies (a larger observed dispersion in effects in brassica studies). Again there is
significant variation between studies and this is associated with between-study differences rather
than within study error (93% and 96% for brassica and cereals respectively) (Additional file 8, Page
2). The analysis is illustrated in the Forest plot shown in Figure 20.
Effect of soil type
Three soil types were identified
Sandy and light soils
Medium soils
Chalk and limestone soils
51
(as categorised by Defra [77]).
No soil types were identified for heavy and peat soils.
Analysis was carried out at the study level, again combining across crop types where necessary but
there was no difference between soil types (Q = 2.5, P =0.4). However these data reveal very little
as for two of the studies it was not possible to determine soil type [78] [79] and there was only one
study on medium soil [75] and two on chalky soils [80, 81].
Discussion
General Trends
The most commonly studied interventions were buffer strips (including woodland buffers)
and cover/catch crops. Some evidence was found for slurry storage, but it was sometimes at
least 10 years old and conducted in North America where legislation may be different from
that of the UK. Buffer strips composed of trees were only categorized under buffer strips
therefore only a small number of woodland creation studies were found. These woodland
creation studies either measured changes in water quality after afforestation on former agri-
cultural land or planting of trees for biomass. Very little evidence was found for subsoiling
(break up of compacted soil) or controlled traffic on grassland.
Many studies included in the systematic map database were not randomized. About two
thirds of the studies were conducted for less than 2 years. Over a half of the studies used a
control, but measurements of water quality pre and post intervention implementation were
rarely recorded (BACI). Nearly three quarters of the studies were manipulative and the
remaining studies were predominantly correlative. Cover/catch crops studies when assessed
for scientific rigour were slightly more likely to score higher for these factors than buffer
strips studies. Slurry storage studies were often not randomized or controlled and a relatively
high number of studies had confounding factors compared to other interventions.
52
Water quality was mostly sampled in fields or plots rather than within river systems. Loam
was the most common soil type studied, although sometimes the soil type was not reported.
Therefore, given the current evidence base, it would be difficult to assess intervention effect-
iveness at a catchment scale and to generalize results across all soil types.
Average effectiveness values suggested that buffer strips were most effective for reducing
sediments, followed by pesticides, N, P, and bacterial pathogens in decreasing order. Buffer
strips were also found to be effective in reducing N, P, sediments and pesticides by a pre-ex-
isting meta-analysis. However, that meta-analysis found that buffer strips were slightly more
effective for P than N. Some research in the database suggested that saturated buffer strips
could leach P, which may explain this difference.
Evidence in the map could also suggest that the form of N or P can impact upon mitigation
effectiveness, as proportionally more buffer strip studies were scored as effective in
reducing levels of nitrate, total N, total-P than ammonium-N or soluble forms of P.
Average effectiveness values suggested that cover/catch crops were most effective at
reducing N and sediments, whereas values for P were much lower. Cover/catch crops were
not assessed for measurements of effectiveness for pesticides or bacterial pathogen counts
due to small sample sizes.
A meta-analysis found that cover/catch crops consistently reduced leaching of N when
compared with fallow, although there was significant variation between the studies. No
significant difference was found between the effectiveness of brassica and cereal cover/catch
crops for reducing N. A dominance of loam soil types in the studies meant that it was not
possible to carry out any soil comparisons. Poor reporting in primary studies, meant that
only 10 studies could be analysed in the time available, so the meta-analysis is likely to be
subject to bias.
Most of the evidence for N and P was assessed from studies measuring leakage from slurry
storage, rather than studies that investigated the timing of slurry to maximise plant nutrient
53
uptake. This was a result of the search strategy not focusing on plant uptake so the evidence
in this area will be underrepresented. Slurry storage was on average at least partially
effective at reducing bacterial counts but the outcome was unclear for N and P
Studies were often designed to address questions that differed from those posed in this
review which made it difficult to assess the effectiveness of some interventions. For
example, some woodland creation compared water quality across different aged trees or
types and lacked a non woodland control. Subsoiling is a primarily a tools for improving
soil infiltration rather than water quality which may explain the small number of studies
found this intervention.
Improvements in water quality measured from within plots did not always translate to
improved river water quality as found by a few studies [53, 82]. Some studies suggested that
preferential flow paths or upswellings of groundwater could result in water bypassing buffer
strips and flowing directly into river systems therefore reducing mitigation effectiveness if
assessed from river water measurements. One study suggested that certain regions of rivers
systems can deliver a disproportionate amount of water to river flows and that these should
be targeted with buffer strips otherwise improvements may not be observed at a catchment
level [53].
Gaps in the research
Some of the following research gaps have been identified:
The evidence base for slurry storage and effect on surrounding water quality is dated and
may not relate to current/regional legislation.
There is little evidence for the direct impact of subsoiling or controlled traffic on grasslands,
on water quality, however studies to measure improvements in soil water infiltration were
not included in this review.
54
The amount of evidence for woodland creation (excluding tree buffer strips, which were
considered separately) was quite small being composed of studies measuring water quality
after afforestation on former agricultural soil or planting of biomass studies. However,
woodland creation studies often need many years to complete therefore modelling studies
which were excluded in the review can provide important insight when longer term data is
needed.
Buffer strip studies that measured pesticides and bacterial pathogen were less common than
studies measuring N, P or sediment.
Most pesticide studies were performed on loam or an unknown soil type and used a wide
variety of pesticides. No grouping of pesticides based on chemical properties was attempted
within this review which could highlight further research gaps. There were 22 buffer studies
that measured changes in pesticide levels, which were not coded at full text and could
contain valuable information.
There were only 3 studies that measured the effect of cover/catch crops on pesticide levels.
There was some evidence for P and sediment, but it was not sufficiently well reported to be
usable in a meta-analysis.
There were only a small number of studies conducted at catchment scale in the map. Some
of the studies measured the effect of multiple mitigations (including non-topic mitigations)
and could therefore not be used to assess individual mitigation effectiveness.
Few studies measured organic forms of N or P, which are much more dependent on soil
conditions e.g. temperature, aeration and structure.
Loam soils dominated the evidence base; however some studies soil type were marked as
unknown, therefore research gaps for soil type may be artificial.
55
Potential systematic review topics
Evidence in the map often had a general inconsistency in approach that makes combining
information for meta-analysis a challenge. However, there was sufficient enough evidence for a
meta-analysis for buffer strip and catch/cover crops.
Cover/catch crops
When a meta-analysis was attempted for cover/catch crops and N it was found that authors did not
always report all the statistics necessary for meta-analysis which greatly impacted sample size.
However, some further topics could be investigated for feasibility:
The effect of time on the effectiveness of cover/catch crops
The interaction between cover/catch crops and applications of nitrogen and tillage
The effect of cover/catch crops compared to a cropped control (winter crop)
Buffer strips
There are some pre-existing meta-analyses which measured changes in levels of sediments, N, P
and pesticides [21, 27, 83] as measured along the length of a buffer strip (comparing
inflow/outflow). However, some further topics could be investigated for feasibility:
The effect of time on the effectiveness of buffer strips
The effect of pollutant solubility on mitigation effectiveness e.g. P
The effect of buffer strips compared to a cropped or bare ground control
Limitations during searching
Non English language search terms were excluded. However, over 100 articles in the map
were assumed to be foreign language texts and only included on titles/abstracts. Their
translation would extend the evidence base. For example, some woodland creation reports,
56
written in French or German, were not coded on full text [84, 85].
Although web searches were conducted for a variety of organisations, grey literature may
be under-represented, where it is not available online.
Some included studies contained forms of the interventions that were not specifically
searched for (e.g. winter wheat to provide a crop cover, winter slurry applications in split
over multiple dates, or trees intercropped with crops). These topics may be less
comprehensively covered in the database.
Limitations of the systematic map
Articles lacking full text were coded on title and abstract which may result in the inclusion
of some non-relevant studies.
Only studies that demonstrated a direct effect of the intervention on water quality were
included in the map, thereby excluding studies that measured indirect (but important) effects
such as soil water infiltration, crop yields, crop biomass, soil mineralization rates, and
herbicide degradation. Studies that assessed the effect of buffer strips on reducing pesticide
drift or trapping of aerial pollutants were excluded in this review but these subjects have
been reviewed recently by the Forestry Commission Woodland report
Only overall outcomes were recorded for a study therefore differences in sampling location,
mitigation, study site, and flow path were not captured. The map could be designed to
capture this information, but it would then become more unwieldy. Data extraction for meta-
analysis can address this shortcoming.
The terms used in the map are not standardized due to a lack of topic ontologies.
There are missing soil types for some studies as no mapping was performed for soil series.
Climatic data proved difficult to extract.
57
Limitations in hierarchy of evidence assessment
The standard scoring that was applied to all studies may have excluded important water
quality specific factors, or experimental design factors that were not considered.
Limitations in mitigation effectiveness assessment
The effectiveness scores are not based on rigorous data analysis, but rather are based on
categories applied to a study by the reviewer on reading a studies outcome [86]. Despite
those limitations the ranking of buffer strip effectiveness scores from this review (sediment,
pesticide, N,P, bacterial pathogen) was not dissimilar to that reported for a pre-existing a
meta-analysis (pesticide, sediment, P, N) [27].
No differentiation between the effectiveness of trees, grass and other vegetation was made
for buffer strips (although a comparison was made using a subset of the data that measured
either grass or tree buffers, which did not show any difference). An existing meta-analysis
for buffer strips suggested that there was no difference in vegetation effectiveness as regards
reducing N [21].
Many of the buffer strip studies are short term and would not address vegetation
management and the overall effect of time on buffer strip performance.
Modelling studies were excluded from the review, however they are useful for woodland
studies which experimentally can take years to assess.
In some cases, studies addressed different questions to the review, making it difficult to
assess the overall effectiveness of interventions. For example, some woodland creation
studies compared water quality under different aged trees, or to plots lacking additions of
fertilizer (biomass studies), rather than to a control without trees.
Scores were too rudimentary to be used to assess correlations between measurements such
as sediment and sediment bound forms of P or pesticide.
58
Many related factors (such as the potential for pollution swapping) have not been considered
by this work.
Data extraction for meta-analysis was very difficult. A number of studies presented data in
graphical form for data collected over several time points, which gave no indication of
standard deviation (SD) or standard error for each point. Initially, data were calculated for
all SD using all of the data points so that SD represented dispersion over the sampling
period. A more complex model which takes into account time and a wider range of
covariates is desirable but although time has limited the development of such a model, it
must be emphasised that better reporting would have greatly enhanced the analysis.
The final meta-analysis analysis is based on few studies and so presents limited information
and may be subject to bias. It may be possible to build a more complex and more
informative model but it would preferable to invest time in contacting authors to improve
the precision and breadth of the study before doing so.
Conclusion
Studies conducted at predominantly field/plot scale suggested that cover/catch crops and
buffer strips can improve water quality, although there was not enough evidence recorded in
the database to assess mitigation effectiveness at a catchment scale. A recent COST action
knowledge exchange programme for buffer strips also observed that most evidence for
buffer strips was from plot based studies [87]. A lot of the evidence was from short duration
studies which did not always have seasonal data, therefore the impact of rainfall events and
mitigation effectiveness over time may not have been fully captured. Most evidence was
from loam or unknown soil types. The evidence base as a whole was not of the highest
scientific rigour; although on average cover/catch crops studies were slightly more
rigorously executed than those of buffer strips.
59
Evidence in the map suggests that at a field scale buffer strips composed of either grass
and/or trees can on average be partially effective at reducing levels of sediments, pesticides,
N, but slightly lesser effective at reducing levels of P and not so effective at reducing levels
of bacterial pathogen counts. Evidence in the map suggests that cover/catch crops at a field
scale can be effective at reducing levels of N and sediment, but not levels of P (although
these were quite diver studies). There was not enough evidence found for cover/catch crops
and measurements of pesticides or bacterial pathogen counts to draw any conclusions on
mitigation effectiveness. The conclusions on mitigation effectiveness were based on
standard categories using reviewer interpretation of studies rather than rigorous data
analysis. However, pre-existing meta-analyses for buffer strips and a meta-analysis
conducted as part of this review on cover/catch crops did support some of these findings.
A small amount of research suggested that, over time, the storage of slurry could reduce
bacterial pathogen counts. A very few studies were found that investigated the impact, on
water quality, of altering the timing of slurry applications to crops, but this was identified
as an topic that would benefit from future synthesis and has been funded as a separate
project since the completion of this systematic review [88]
The woodland creation evidence that was not buffer strip studies was diverse and often
lacked a non-woodland comparator making it difficult to assess effectiveness. There were
too few subsoiling and controlled trafficking on grasslands studies to give any assessment of
mitigation effectiveness.
Further work could start looking at the evidence in more detail to understand under which
conditions mitigations perform best.
Implications for policy and management
Most evidence was drawn from journal articles, despite the search strategy being designed to
capture unpublished evidence. Although several projects were found on websites, little
60
information could be used in the map. The allocation of resources to improving project
documentation and archiving can be invaluable for improving the evidence base for a given
topic [33].
The review covered a wide topic area which could be broken down into 25 different
questions as there were 5 interventions and 5 different water quality measurements e.g. One
of the 25 questions was ‘the effect of buffer strips on water quality as measured by changes
in N’. The review could only consider the direct effect of mitigations on water quality, as the
topic was so large therefore future work should aim to ask a more focused question.
Evidence can be collated as a systematic review, rapid evidence assessment or systematic
map care needs need to ensure that the question is suitable for each tool.
The systematic map provides a large database of research on the primary topic that can be
used to filter information by mitigation or water quality measurement, which should help
enable decision makers and delivery agencies to better facilitate catchment planning as
required under the Water Framework Directive [89-91].
The systematic map can be used as a tool to find research for a particular experimental
factor such as buffer width, slope, or tillage. As an example, the map contains 3 buffer strip
studies that investigated the effect on buffer strip performance of harvesting plant biomass.
A review published as part of a recent COST action knowledge exchange programme for
buffer strips [87] suggested that cutting and removal of vegetation could alleviate P
saturation of buffer strips, the studies in the map could be used to investigate this further.
However, the review also commented that management needs to be adapted to the local area
and buffer strip access may be limited if it fenced making it difficult to pass a mower .
Implications for water quality research
Studies designed with controls, and pre and post water quality measurements would improve
the quality of the evidence base.
61
Multiple sampling points from both within field and rivers would provide greater insight
into the impact of preferential flow paths, upswellings of groundwater and critical points in
river systems.
Long term studies with seasonal data would allow the effects of full crop rotations and
degradation of mitigation effectiveness over time to be assessed.
The evidence base would be enhanced if statistics were reported more comprehensively as
standard in primary research papers. For example, reporting of summary data with intuitive
metrics, associated sample sizes and measures of dispersion such as confidence intervals or
standard deviations would increase the value of reported data. Submission of data with
journal papers would ensure that water quality data is not lost to science [92].
.
Competing interests
Financial competing interests – The authors have been commissioned and funded by the UK
Department of Environment Food and Rural Affairs (Defra), and by the UK Natural Environment
Research Council (NERC) to carry out this research.
Authors Contributions
All authors involved in drafting/revising the manuscript
NPR – Conception and design of review, involved in drafting and revision of review, final approval.
PJL - Conception and design of review, guidance on environmental quality and protection and
subject expert for buffer strips and slurry storage.
LMD – Conception and design of review, database searches, extracted data for map and meta-
analysis. Involved in drafting and revision of review
BS -extracted data for meta-analysis and data analysis.
62
Acknowledgements
This systematic review is funded by the UK Natural Environment Research Council and the UK
Department for Environment Food and Rural Affairs under work order WT0965. The authors are
grateful to the following subject experts from Harper Adams University for their comments and
suggestions in the drafting of the protocol: Jim Waterson (Woodland creation), Nigel Hall
(cover/catch crops) and Dick Godwin (loosening compacted soils, controlled trafficking and slurry
storage). The authors would like to thank the librarians at Harper Adams University, and in
particular Mathew Bryan for his help in ordering articles. Thanks are due to Laura Kor and Amy-
Jane Smith at the Game and Wildlife Conservation Trust for help in data extraction for the meta-
analysis. The authors are also grateful to stakeholders Defra, NERC, the Environment Agency and
the Forestry Commission for their input at review meetings
63
Figures
Figure 1 Literature included and excluded at each stage of the systematic mapping process
64
19501972
19761978
19801983
19851987
19891991
19931995
19971999
20012003
20052007
20092011
0
5
10
15
20
25Buffer StripsCover/Catch CropSlurry StorageWoodland Creation
Year of publication
Num
ber o
f arti
cies
Figure 2 Number of articles published each year per mitigation (all texts)
The totals reported on the graph are greater than the number of records held in the systematic map
as a publication can investigate multiple mitigations. Numbers are for all texts read (title, abstract
and full text). There are some studies duplicated in the article totals.
65
Not clear
Subsoiling/Controlled Traffic
WoodlandCreation
Slurry Storage
Cover/Catch Crop
Buffer Strips
0 50 100 150 200 250 300 350 400
6
10
24
93245
364
Full Text Abstract Title
Number of articles
Miti
gatio
n
Figure 3 Number of articles included in the database per mitigation (all texts)
The total numbers of publications per mitigation is shown at the end of the column. The totals
reported on the graph are greater than the number of records held in the systematic map as a
publication can investigate multiple mitigations. Numbers are for all texts read (title, abstract and
full text). There were 6 articles read at title where the mitigation was not clear. There are some
studies duplicated in the article totals.
66
BelarusRomania
UkraineAustria
LithuaniaPoland
New ZealandHollandFrance
DenmarkUK
USA
0 50 100 150 200 250 300
1
1
2
4
4
5
6
7
Buffer StripsCover/Catch CropSlurry Storage
Number of articles
Coun
try of
stud
y
Figure 4 Number of articles for each country of study per mitigation (all texts).
The totals reported on the graph are less than the sum of the values contained in each coloured
section of the bar, as they represent the total not broken down by mitigation (one study can have
multiple mitigations). More than one country can occur in a publication; therefore the total of the
numbers reported on the graph is greater than the number of records held in the systematic map.
Numbers are for all texts read (title, abstract and full text). There are some studies duplicated in the
article totals.
67
N P Sediment Pesticides Pathogens Unknown0
50
100
150
200
250
209
136128
63
32 26
203
24 28
8 111
5842
4 0
34
1119
4 1 0 1 4
Buffer StripsCover/Catch CropSlurry StorageWoodland CreationSubsoiling/ Controlled Traffic
Water quality measurement
Num
ber o
f arti
cles
Figure 5 Number of articles per mitigation for each water quality measurement (all texts).
The totals reported on the graph are greater than the number of records held in the systematic map
as more than one measurement or mitigation can occur in a publication. Numbers are for all texts
read (title, abstract and full text). There are some studies duplicated in the article totals.
68
0 1 2 3 4 5 6 7 8 9 100
10
20
30
40
50
60
70
80(a) Buffer Strips
Cover/catch cropSlurry Storage
Hierachy of evidence value
Nu
mb
er o
f st
ud
ies
Buffer Strips Cover/Catch Crop Slurry Storage0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
5.96.8
4.1
(b)
Mitigation
Avera
ge hie
rachy
value
(c) Average rounded value
Mitigation number of studies in brackets
1098
7Cover/catch crops (n=156)
6 Buffer strips (n=252)
54 Slurry storage (n=61)3210
Figure 6 Hierarchy of evidence values (a) distribution of values (b) average values, error bars
are standard deviations (c) averages scaled.
Values are given for buffer strips, cover/catch crops and slurry storage read at full text including
studies with confounding factors. Values are automatically calculated from the map using
randomized, control/comparator, replicates and study length codes. A score of 10 would represent a
randomized, fully replicated study with a BACI conducted for longer than a year, 0 would indicate
the converse or a study with confounding factors. 69
Figure 7 Literature included and excluded at each stage of the hierarchy of evidence and
measures of effectiveness.
70
Stream/river
Plot/field Lab/lysimeter/mesocosm
In/under/near slurry storage
0
20
40
60
80
100
120
140
160
180
200
23
187
27
01
111
21
01 8 3
30
(a)Buffer StripsCover/Catch cropSlurry Storage
Sampling location
Nu
mb
er
of
stu
die
s
Inorganic fertilizer Organic fertilizer Not clear0
20
40
60
80
100
120
140
160
180
200
3441
159
70
3345
(b)Buffer StripsCover/catch crops
Type of fertilizer
Nu
mb
er
of
stu
die
s
Loam Sand Clay Not clear/Not in category
0
20
40
60
80
100
120
140
160
180
200
121
1810
7871
30
10
31
(c) Buffer Strips Cover/catch crops
Soil type
Nu
mb
er
of
stu
die
s
Figure 8 Variation in studies that were scored for effectiveness (a) sampling location (b)
fertilizer (c) soil type. Slurry storage was not plotted for fertilizer as it is organic or for soil type an
due to small sample size. Woodland creation studies that were not buffer strips were not plotted due
to small sample size and likewise for subsoiling. Not clear/not in category was recorded when the
soil type used in the study was not clear or could not be placed in one of the 3 main categories. Less
frequent sampling locations were excluded (e.g.river bank). Numbers are from full text studies
without confounding factors (studies used to calculate measures of effectiveness).
71
0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3N P Sediment Pesticide Pathogen
0
10
20
30
40
50
60
70
80
90(a) Buffer StripsCover/catch cropSlurry Storage
Measures of effectiveness value per water quality meas-urement
Num
ber o
f stu
dies
N P Sediment Pathogen Pesticide0.0
0.5
1.0
1.5
2.0
2.5
3.0
2.2 2.0
2.7
1.82.32.3
1.2
2.3
1.0 1.0
2.2
(b) Buffer StripsCover/Catch CropSlurry Storage
Water quality measurement
Ave
rag
e e
ffe
ctiv
ne
ss v
alu
e
Figure 9 Measures of effectiveness values (a) distribution of values (b) average values, error
bars are standard deviations Values are given for buffer strips, cover/catch crops and slurry
storage read at full text excluding studies with confounding factors. Values are calculated from
reviewer’s interpretation of an author’s conclusion on study outcome. Values are automatically
calculated from the map using the scores for each study scored on a scale of 0-3. A score of 3 was
all forms of a measurement were reduced, 2 some form of measurement was reduced, 1 not a clear
outcome, 0 no form of measurement reduced. The scale of 4 on the graph is to accommodate
standard deviation bars.
72
N-Nitrate N-Total N-Am-monium
N-Inorganic N-Organic N-Nitrate-
Nitrite
N-Soluble/ N-Organic
soluble
0
10
20
30
40
50
60
70
80
9080
2923
04
1 0
19
710
0 0 2 2
21
3
13
0 2 0 2
Buffer stripsYes reducedNot reducedOutcome not clear
Form of N
Nu
mb
er o
f stu
die
s
Figure 10 Number of buffer strip studies per outcome for each form of N measured
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Numbers are from full text studies without confounding factors (studies used to
calculate measures of effectiveness).
73
Surfa
ce
Gro
undw
ater
Subs
urfa
ce
Surfa
ce
Gro
undw
ater
Subs
urfa
ce
Surfa
ce
Gro
undw
ater
Subs
urfa
ce
N-Nitrate N-Total N-Ammonium
05
101520253035404550
20
38
9
21
03
16
50
6 62 1 1 0
3 402
13
3 1 0 05 3 2
Buffer strips N per flow path Yes reducedNot reducedOutcome not clear
Num
ber o
f stu
dies
Figure 11 Number of buffer strip studies per outcome for each form of N measured divided by
flow path
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Values for the 3 main forms of N are divided by either surface, subsurface or
groundwater flows. These results should be interpreted with caution as studies with multiple flow
paths or studies where flow paths were not clear were excluded. When flow path was not stated and
measurements were taken below ground a default coding of subsurface was used, therefore the
distinction between groundwater and subsurface may not be valid.
74
Nitrate Total N Ammonium0
5
10
15
20
25
30 28
3 45
2
5
9
0
4
(a) Tree buffer strips (tree, tree-grass, tree grass shrub, grass tree) excluding studies where tree buffers compared to grass
buffers
Yes reducedNot reducedOutcome not clear
Nu
mb
er
of
stu
die
s
Form of N
Nitrate Total N Ammonium0
5
10
15
20
25
3027
18
14
9
433
1
4
(b) Grass buffer strips (grass) excluding studies where tree buf-fers compared to grass buffers
Yes reducedNot reducedOutcome not clear
Num
ber o
f stu
dies
Form of N
Figure 12 Number of buffer strip studies per outcome for each form of N measured (a) tree
buffers (b) grass buffers. Each individual water quality measurement was coded with one of 3
values for study outcome (yes successful, not successful, not clear outcome) based on reviewer’s
interpretation of authors conclusions. These results should be interpreted with caution as studies
with multiple buffer types were excluded. Grass-shrub buffers were excluded.
75
P-To
tal
P-O
rthop
...
P-So
lubl
e
P-O
rgan
i...
P-Pa
rticu
...
P-Se
dim
en...
P-R
eact
i...
05
101520253035404550 46
23
50
8
2 0
8 9 9
1 1 1 3
9 105 3
0 1 0
Buffer strips
Yes reducedNot reducedOutcome not clear
Num
ber o
f stu
dies
Form of P
Figure 13 Number of buffer strip studies per outcome for each form of P measured
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Numbers are from full text studies without confounding factors (studies used to
calculate measures of effectiveness).
76
Surfa
ce
Grou
ndw
ater
Subs
urfa
ce
Surfa
ce
Grou
ndwa
ter
Subs
urfa
ce
Surfa
ce
Grou
ndw
ater
Subs
urfa
ce
P-Total P-Orthophosphate P-Soluble
05
101520253035404550
32
2 2
15
2 2 41 00
30
3 1 1 3 1 0
6
0 03 3
0 2 0 0
Buffer strips P flow pathYes reducedNot reducedOutcome not clear
Num
ber o
f stu
dies
Figure 14 Number of buffer strip studies per outcome for each form of P measured divided by
flow path
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Values for the 3 main forms of P are divided by either surface, subsurface or
groundwater flows. These results should be interpreted with caution as studies with multiple flow
paths or studies where flow paths were not clear were excluded. When flow path was not stated and
measurements were taken below ground a default coding of subsurface was used, therefore the
distinction between groundwater and subsurface may not be valid.
77
Sediment Sediment -Total Suspended Solid
Sediment-Soil Loss
Sediment-Water turbity
0
10
20
30
40
50
60 57
22
512 3
0 0
7
1 0 1
Buffer Strips Yes reducedNot reducedOutcome not clear
Term used to record sediment
Num
ber o
f stu
ides
Figure 15 Number of buffer strip studies per outcome for each term used to record sediment
measurement
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Numbers are from full text studies without confounding factors (studies used to
calculate measures of effectiveness).
78
Pat
hoge
n-To
ta...
Pat
hoge
n -T
ot...
Pat
hoge
n-S
tr...
Pat
hoge
n-E
.col
i
Pat
hoge
n-C
ryp.
..
0
1
2
3
4
5
6
7
87
2 2
3
1
4
0
1
4
00
1
0
1
0
Buffer Strips Yes reducedNot reducedOutcome not clear
Bacterial pathogen count
Num
ber o
f stu
ides
Figure 16 Number of buffer strip studies per outcome for each form of bacterial pathogen
measured
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Numbers are from full text studies without confounding factors (studies used to
calculate measures of effectiveness).
79
N-N
itrat
e
N-T
otal
N-A
mm
oniu
m
N-In
orga
nic
N-O
rgan
ic
N-N
itrat
e-
...
N-S
olub
le/ N
-...
0
10
20
30
40
50
60
70
80
90
74
2 06
0 0 08
1 3 0 0 0 0
26
3 0 0 0 0 1
Cover crops NYes reducedNot reducedOutcome not clear
Form of N
Num
ber o
f stu
dies
Figure 17 Number of cover/catch studies per outcome for each form of N measured
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Numbers are from full text studies without confounding factors (studies used to
calculate measures of effectiveness).
80
N P Pathogen N P Pathogen N P PathogenSlurry storage leakage Survival rate pathogens
slurrySlurry applications winter
0
2
4
6
8
10
12
14 Effectivity scores separated out by study type
3 2 1 0
Nu
mb
er o
f stu
die
s
Figure 18 Number of slurry storage studies per study type for N,P or pathogen bacterial
counts per study type
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. The distribution of values is plotted. Numbers are from full text studies without
confounding factors (studies used to calculate measures of effectiveness). Pathogen refers to
bacterial pathogen counts. Slurry storage studies were divided into 3 main types those that
measured leakage from slurry stores, those that measured bacterial pathogen survival in slurry and
those that measured water quality during variable applications of slurry during winter.
81
Figure 19 Literature included and excluded at each stage of data extraction for meta-analysis.
82
Selection of sources from database - 48 identified
48 papers scan read - 16 clearly unsuitable papers rejected. Some papers had no data or incomplete
were data presented.
The remaining papers reviewed again. 8 unsuitable papers rejected at second pass
24 papers read carefully-papers with available data selected
Data extraction. Data extracted from tables and graphs. Data from graphs extracted with
Datatheif.
Data reviewed. In 12 cases data were unsuitable.
Papers read again to confirm correct data extracted. Much of the detail frequently buried in text.
10 studies included in meta-analysis
Group byComparison
Study name Comparison Statistics for each study Std diff in means and 95% CI
Std diff Standard Lower Upper in means error Variance limit limit Z-Value p-Value
Brassica 144Bontemps Brassica 5.730 0.922 0.851 3.922 7.537 6.212 0.000Brassica 591Justes Brassica 2.419 0.284 0.081 1.862 2.975 8.516 0.000Brassica 226Constatin Brassica 0.808 0.219 0.048 0.378 1.238 3.686 0.000Brassica 783Merbach Brassica 1.072 0.535 0.286 0.024 2.120 2.005 0.045Brassica 2.277 0.715 0.512 0.875 3.679 3.184 0.001Cereal 159Brandi-Dohrn Cereal 1.211 0.071 0.005 1.071 1.351 16.954 0.000Cereal 260Davies Cereal 2.209 0.124 0.015 1.966 2.451 17.837 0.000Cereal 281Defra Cereal 5.981 0.827 0.684 4.360 7.602 7.232 0.000Cereal 757McCracken Cereal 1.414 0.456 0.208 0.520 2.309 3.099 0.002Cereal 895Parkinexp1 Cereal 6.749 1.056 1.115 4.679 8.819 6.390 0.000Cereal 895Parkinexp2 Cereal 3.917 0.697 0.486 2.550 5.284 5.617 0.000Cereal 3.056 0.465 0.216 2.145 3.967 6.573 0.000Overall 2.825 0.390 0.152 2.061 3.589 7.246 0.000
-10.00 -5.00 0.00 5.00 10.00Fallow reduces N Cover crop reducing N
Meta Analysis
Figure 20 Forest plot illustrating the relative impact of Brassica and Cereals on N in leachate.
In this diagram the size of the squares shows the impact of that study in the analysis i.e. studies with
large squares have a greater influence than studies with small squares. The whiskers represent
confidence intervals. The diamonds represent summary data. Grey diamonds are summaries of
crop type. The red diamond represents the overall summary.
83
Tables
Table 1 Keywords and qualifiers to be used in literature search.
Exact keyword and qualifier combinations varied in order to optimise searching efficiency and have
been informed by a scoping search
Mitigation Keyword AND Qualifier
1 Slurry storage Slurr* stor*
Animal waste lagoon*
Animal waste stor*
Slurr* lagoon*
Slurr* tank*
Dairy lagoon*
Water qualit*
Water pollut*
Control of pollut*
Nitrat* OR Nitrogen
Phosph*
Nutrient loss*
Bacter*
Fecal OR faecal
Pesticid*
Sediment*
River* OR Stream*
OR Catchment*
Leak* OR Seap* OR Spill*
Ground* water*
Run off OR runoff
Directive* OR Europe*
Infiltrat*
Leach*
Water AND (Erosion OR
Erod*)
Eutrophication
Water
2 Woodland Afforest*
(Wooded OR woodland*) AND
(agricult* OR arable OR grass*)
(Shelterbelt* OR windbreak* OR
hedge*)
Spray drift and tree*
3 Buffer Riparian AND (buffer* OR zone* OR
filter* Or strip*
Filter strip*
Vegetat* AND( buffer* OR barrier*)
4 Loosening
Compacted Soil/
Controlled trafficking
“Subsoiling”
Loosen* Compact*
Deep ripping
Wheel* AND compact* AND grass*
Traffic* AND compact* AND grass*
Soil compact* AND grass*
Controlled traffic* AND grass*
5 Cover Crop
/Catch Crop
“Cover crop” OR “Cover crops” OR
“Covercrop” OR “Covercrops”
“Catch crop” OR “Catch crops” OR
“Catchcrop” OR “Catchcrops”
84
Table 2 Scoring system used to assess hierarchy of evidence calculated from values in
map Adapted from: Pullin and Knight [32].
Category Score Hierarchy of evidence
Randomized 1
0
Yes - Randomized (includes partial)
Not Randomized
Control 3
2
1
0
Controlled BACI
Control
Comparator
None
Study length 1
0
Study length greater than or equal to a year
Study length less than a year
Replicates 2
1
0
Replicate temporal (includes time series) and spatial
Replicate temporal or spatial
No replicates
Study type 3
2
1
0
Manipulative study
Correlative study
Monitoring study
Sampling study
85
Table 3 Scoring system used to assess mitigation effectiveness calculated from values in
map: Adapted from: Ramstead et al. [33]
Category Measure of effectiveness
3 Yes reduced -All forms of a measurement were reduced by the
mitigation.
OR
Slurry leakage not detected for any forms of measurement
2 Partial - At least one form of a measurement was reduced by the
mitigation regardless of the outcome of other measurements
OR
Slurry leakage not detected for one form of measurement
1 Not clear – Outcome not clear as stated by authors, or not clear as
mixed outcome for forms of measurement (No and not clear)
OR
Slurry leakage outcome not clear.
0 No – No forms of a measurement were reduced by the mitigation.
OR
Slurry leakage detected for all forms of measurement
86
Table 4 Average values for hierarchy of evidence calculated for each mitigation, standard
deviations are given in brackets and number of studies is n. Studies with confounding factors
are included, but subsoiling was excluded due to low sample size (n=5).
Mitigation Average(standard deviation)Number of studies including confounding factor studies
Buffer Strips 5.9(2.4)n=252
Cover/Catch Crop 6.8(3.1)n=156
Slurry Storage 4.2(3.1)n=61
Woodland Creation 7.3(1.2)n=12
87
Table 5 Average values for effectiveness calculated for each mitigation, standard deviations
are given in brackets and number of studies is n.
Studies with confounding factors were excluded and mitigation water measurement combinations
with less than 10 studies.
N P Sediment Bacterial Pathogen Pesticide
Buffer Strips 2.2 2.0 2.7 1.8 2.3(1.1) (1.2) (0.8) (1.3) (1.1)n=139 n=94 n=98 n=19 n=38
Cover/catch crops 2.3 1.2 2.3(1.0) (0.9) (1.1) - -n=114 n=14 n=19
Slurry Storage 1.0 1.0 2.2(1.1) (1.2) - (1.1)n=30 n=10 n=18
Woodland creation 2.0(1.0) - - - -n=11
Subsoiling- - - - -
88
Table 6 Outcomes for buffer strips and pesticides
Each individual water quality measurement was coded with one of 3 values for study outcome (yes
successful, not successful, not clear outcome) based on reviewer’s interpretation of authors
conclusions. Numbers are from full text studies without confounding factors.
Pesticide Yes Reduced Outcome Not Clear Not ReducedAtrazine 16 7 3Metolachlor 9 2 1Isoproturon 3 2 0Endosulfan 2 1 1Metribuzin 2 2 0Acetochlor 1 2 0Alachlor 0 2 1Cyanzine 3 0 0Chlorothalonil 1 1 0Chlorpyrifos 0 2 0DIA 0 1 0Fenpropimorph 2 0 0Glyphosate 2 0 0Propiconazole 2 0 0Terbuthylazine 2 0 0Ametryn 0 1 0Carbofuran 1 0 0Dacthal 1 0 0DEA 0 1 0Dicloroprop 1 0 0Diflufencian 1 0 0Diuron 1 0 0Isoxaben 1 0 0Lindane 1 0 0Linuron 1 0 0mancozeb 1 0 0Metalaxyl 1 0 0Oryzalin 1 0 0Pendimethalin 1 0 0Proprymidone 1 0 0Simazine 0 1 0Tebuconazole 1 0 0Triadimenol 0 1 0Trifluralin 1 0 0Isoxaflutole 0 0 0
89
Table 7 Types of cover/catch crops used in studies
Numbers are from full text studies without confounding factors.
Cover/catch crop type N P SedimentGrass 55 9 8Cereal 44 2 6Crucifer 30 1 1Legume 28 3 2Other 3 0 1Volunteer weeds 7 2 2Winter wheat 12 2 5Not clear 5 3 3
Additional files
Additional file 1 –SearchTerms.xls
Spread sheet contains the exact search terms used to search each database.
90
Additional file 2– OtherReferences.doc
Sample of articles that were excluded from systematic map, but showed indirect effects
Additional file 3– ReviewReferences.doc
Systematic reviews and meta-analysis of relevance
Additional file 4– CategoriesCodings.doc
Coding categories used in the systematic map
Additional file 5 – SystematicMap.accdb
Access database of coded review evidence searchable by category
Additional file 6 – SystematicMapReferences.doc
References included in the systematic map database.
Additional file 7 – AccessQueries.doc
Queries that can be run on access databases to calculate scores.
Additional file 8 – MetaAnalysis.xls
Spread sheet containing meta-analysis details
References
1. Herzog F, Steiner B, Bailey D, Baudry J, Billeter R, Bukácek R, De Blust G, De Cock R, Dirksen J, Dormann C: Assessing the intensity of temperate European agriculture at the landscape scale. European Journal of Agronomy 2006, 24:165-181.
2. Department for Environment, Food and Rural Affairs (UK): Safeguarding our soils. UK: Department for Environment, Food and Rural Affairs; 2009.
3. The European Commission: Report from the commission to the council and the European parliament: On implementation of Council Directive 91/676/EEC concerning the protection of waters against pollution caused by nitrates from agricultural sources based on Member
91
State reports for the period 2004-2007. Brussels: The European Commission; 2011.4. von der Ohe PC, Dulio V, Slobodnik J, De Deckere E, Kühne R, Ebert RU, Ginebreda A, De
Cooman W, Schüürmann G, Brack W: A new risk assessment approach for the prioritization of 500 classical and emerging organic microcontaminants as potential river basin specific pollutants under the European Water Framework Directive. Science of the Total Environment 2011, 409:2064-2077.
5. Salomons W, Brils JT: Contaminated Sediments in European River Basins. The Netherlands: European Sediment Research Network; 2004.
6. Collins AL, Anthony SG, Hawley J, Turner T: Predicting potential change in agricultural sediment inputs to rivers across England and Wales by 2015. Marine and Freshwater Research 2009, 60:626-637.
7. Edwards A, Withers P: Transport and delivery of suspended solids, nitrogen and phosphorus from various sources to freshwaters in the UK. Journal of Hydrology 2008, 350:144-153.
8. Kay D, Crowther J, Fewtrell L, Francis C, Hopkins M, Kay C, McDonald A, Stapleton C, Watkins J, Wilkinson J: Quantification and control of microbial pollution from agriculture: a new policy challenge? Environmental Science & Policy 2008, 11:171-184.
9. Environment Agency: The State of Soils in England and Wales. UK: Environment Agency; 2004.
10. Department for Environment, Food and Rural Affairs (UK): Water for life White Paper. UK: Department for Environment, Food and Rural Affairs; 2011.
11. Burke A: Synthesis of Diffuse Pollution Research in England and Wales funded by Department for Environment, Food and Rural Affairs and Environment Agency. UK: Demonstrating Catchment Management; 2011.
12. Newell Price JP, Harris D, Taylor M, Williams JR, Anthony SG, Duethmann D, Gooday RD, Lord EI, Chambers BJ, Chadwick DR, Misselbrook TH: An Inventory of Mitigation Methods and Guide to their effects on Diffuse Water Pollution, Greenhouse Gas Emissions and Ammonia Emissions from Agriculture. UK: Department for Environment, Food and Rural Affairs; 2011.
13. Department for Environment, Food and Rural Affairs (UK): Competition details and project specification: Project WT0965 Independent systematic review of impact of on-farm mitigation measures for delivering an improved water environment. UK: Department for Environment, Food and Rural Affairs; 2011.
14. Johnson D, Hodgkinson R, Lord E, Silgram M, Cottrill B, Gooday R, Morrow K, Smith S, Adele H: Nitrates Directive Consultation Document Environmental Impact of the 2009 Nitrates Action Programme and of potential further measures. UK: Department for Environment, Food and Rural Affairs; 2011.
15. Dabney S, Delgado J, Reeves D: Using winter cover crops to improve soil and water quality. Communications in Soil Science and Plant Analysis 2001, 32:1221-1250.
16. Shepherd M: The effectiveness of cover crops during eight years of a UK sandland rotation. Soil Use and Management 1999, 15:41-48.
17. Townsend LG: What evidence is there that the utilization of catch crops can reduce nitrate leaching in temperate Europe? . MSc thesis. Harper Adams University, 2010.
18. Nisbet T, Silgram M, Shah N, Morrow K, Broadmeadow S: Woodland for Water: Woodland measures for meeting Water Framework Directive objectives. Forest Research Monograph, 4, Forest Research, Surrey, 156pp 2011.
19. Donnison LM: Review of the effects of farmland trees on erosion and pollution in the local farmed landscape. Report to the Woodland Trust by Harper Adams University. UK: The Woodland Trust; 2011.
20. Hamza M, Anderson W: Soil compaction in cropping systems: A review of the nature, causes and possible solutions. Soil and tillage research 2005, 82:121-145.
21. Mayer PM, Reynolds SK, McCutchen MD, Canfield TJ: Meta-analysis of nitrogen 92
removal in riparian buffers. Journal of Environmental Quality 2007, 36:1172-1180.22. Muscutt A, Harris G, Bailey S, Davies D: Buffer zones to improve water quality: a review
of their potential use in UK agriculture. Agriculture, ecosystems & environment 1993, 45:59-77.
23. Collaboration for Environmental Evidence: Guidelines for Systematic Review in Environmental Management. Version 4.0.: Envionmental evidence; 2010.
24. Randall NP, James KL: The effectiveness of integrated farm management, organic farming and agri-environment schemes for conserving biodiversity in temperate Europe-A Systematic Map. Environmental Evidence 2011, 1:4.
25. Randall NP, Donnison LM, Lewis PJ: How effective are slurry storage, cover or catch crops, woodland creation, controlled trafficking or break-up of compacted layers, and buffer strips as on-farm mitigation measures for delivering an improved water environment? Environmental Evidence 2012, 1:12.
26. Tonitto C, David M, Drinkwater L: Replacing bare fallows with cover crops in fertilizer-intensive cropping systems: A meta-analysis of crop yield and N dynamics. Agriculture, ecosystems & environment 2006, 112:58-72.
27. Zhang X, Liu X, Zhang M, Dahlgren RA, Eitzel M: A review of vegetated buffers and a meta-analysis of their mitigation efficacy in reducing nonpoint source pollution. Journal of Environmental Quality 2010, 39:76-84.
28. Hansen K. (eds): Literature review for AFFOREST: Planning afforestation on previously managed arable land - influence on deposition, nitrate leaching, and carbon sequestration. 2002.
29. Correll DL: Principles of planning and establishment of buffer zones. Ecological Engineering 2005, 24:433-439.
30. Stevens A, Milne R: The effectiveness revolution and public health. Progress in public health Royal Society of Medicine Press, London 1997:197-225.
31. Pullin AS, Knight TM: Effectiveness in conservation practice: pointers from medicine and public health. Conservation Biology 2001, 15:50-54.
32. Pullin AS, Knight TM: Support for decision making in conservation practice: an evidence-based approach. Journal for Nature Conservation 2003, 11:83-90.
33. Ramstead KM, Allen JA, Springer AE: Have wet meadow restoration projects in the Southwestern US been effective in restoring geomorphology, hydrology, soils, and plant species composition? Environmental Evidence 2012, 1:11.
34. Borenstein M, Hedges LV, Higgins JPT, Rothstein HR: Introduction to Meta-Analysis. John Wiley & Sons, Ltd; 2009.
35. Tummers B: Shareware Software DataThief III [http://datathieforg] 2006.36. Parkin TB, Kaspar TC, Singer JW: Cover crop effects on the fate of N following soil
application of swine manure. Plant and Soil 2006, 289:141-152.37. Landis JR, Koch GG: The measurement of observer agreement for categorical data.
Biometrics 1977:159-174.38. Bechmann M, Deelstra J, Stalnacke P, Eggestad HO, Oygarden L, Pengerud A: Monitoring
catchment scale agricultural pollution in Norway: policy instruments, implementation of mitigation methods and trends in nutrient and sediment losses. Environmental Science & Policy 2008, 11:102-114.
39. Huffman RL: Seepage evaluation of older swine lagoons in North Carolina. Transactions of the Asae 2004, 47:1507-1512.
40. Rosenqvist L: Afforestation of former arable land in North-Western Europe. PhD thesis. University Of Uppsala, 2007.
41. Kaspar TC, Jaynes DB, Parkin TB, Moorman TB: Rye cover crop and gamagrass strip effects on NO3concentration and load in tile drainage. Journal of Environmental Quality 2007, 36:p1503-1511.
42. Young EO, Briggs RD: Nitrogen dynamics among cropland and riparian buffers: Soil-landscape influences. Journal of Environmental Quality 2007, 36:801-814.
93
43. Szajdak L, Zyczynska-Baloniak I, Jaskulska R: Impact of afforestation on the limitation of the spread of the pollutions in ground water and in soils. Polish Journal of Environmental Studies 2003, 12:453-459.
44. Hefting M, Beltman B, Karssenberg D, Rebel K, van Riessen M, Spijker M: Water quality dynamics and hydrology in nitrate loaded riparian zones in the Netherlands. Environmental Pollution 2006, 139:143-156.
45. Blattel CR, Williard KWJ, Baer SG, Schoonover JE, Zaczek JI: Ground Water Nitrogen Dynamics in Giant Cane and Forest Riparian Buffers. Castanea 2009, 74:259-270.
46. Dillaha T, Sherrard J, Lee D, Mostaghimi S, Shanholtz V: Evaluation of vegetative filter strips as a best management practice for feed lots. Journal (Water Pollution Control Federation) 1988:1231-1238.
47. Duchemin M, Hogue R: Reduction in agricultural non-point source pollution in the first year following establishment of an integrated grass/tree filter strip system in southern Quebec(Canada). Agriculture Ecosystems & Environment 2009, 131:85-97.
48. Schellinger G, Clausen J: Vegetative filter treatment of dairy barnyard runoff in cold regions. Journal of Environmental Quality 1992, 21:40-45.
49. Seybold C, Mersie W, Delorem D: Removal and degradation of atrazine and metolachlor by vegetative filter strips on clay loam soil. Communications in Soil Science and Plant Analysis 2001, 32:723-737.
50. Syversen N, Borch H: Retention of soil particle fractions and phosphorus in cold-climate buffer zones. Ecological Engineering 2005, 25:382-394.
51. Leeds-Harrison PB, Quinton JN, M.J.Walker, K.S.Harrison, D.J.Gowing, S.F.Tyrrel, J.Morris, J.Mills, T.Harrod: Report NT1101 Buffer zones. UK: Department for Environment, Food and Rural Affairs; 1996.
52. Newbold JD, Herbert S, Sweeney BW, Kiry P, Alberts SJ: Water Quality Functions of a 15-Year-Old Riparian Forest Buffer System. Journal of the American Water Resources Association 2010, 46:299-310.
53. Angier JT, McCarty GW, Prestegaard KL: Hydrology of a first-order riparian zone and stream, mid-Atlantic coastal plain, Maryland. Journal of Hydrology 2005, 309:149-166.
54. Stutter MI, Langan SJ, Lumsdon DG: Vegetated Buffer Strips Can Lead to Increased Release of Phosphorus to Waters: A Biogeochemical Assessment of the Mechanisms. Environmental Science & Technology 2009, 43:1858-1863.
55. Barfield BJ, Blevins RL, Fogle AW, Madison CE, Inamdar S, Carey DI, Evangelou VP: Water quality impacts of natural filter strips in karst areas. Transactions of the Asae 1998, 41:371-381.
56. Dillaha TA, Reneau RB, Mostaghimi S, Lee D: Vegetative filter strips for agricultural nonpoint source pollution control. Transactions of the Asae 1989, 32:513-519.
57. Barone VA, Christopher KI, Coffey ME, Gehring ME, Kirkpatrick AP, Vincent AA, Buck SP, Mostaghimi S, Dillaha TA, Wolfe ML, Vaughan DH: Effectiveness of vegetative filter strips in reducing NPS pollutant losses from agricultural lands: sediment, nutrients, bacteria, and pesticides. In American Society of Agricultural Engineers Annual International Meeting; Orlando, Florida, USA. 1998
58. Magette WL, Brinsfield RB, Palmer RE, Wood JD: Nutrient and Sediment Removal by Vegetated Filter Strips Transactions of the Asae 1989, 32:663-667.
59. Herrera JM, Liedgens M: Leaching and Utilization of Nitrogen During a Spring Wheat Catch Crop Succession. Journal of Environmental Quality 2009, 38:1410-1419.
60. Herrera JM, Stamp P, Liedgens M: Root development of catch crops and nitrate losses by leaching after spring wheat. Aspects of Applied Biology 2005:35-40.
61. Hansen EM, Djurhuus J, Kristensen K: Nitrate leaching as affected by introduction or discontinuation of cover crop use. Journal of Environmental Quality 2000, 29:1110-1116.
62. Constantin J, Beaudoin N, Laurent F, Cohan J-P, Duyme F, Mary B: Cumulative effects of catch crops on nitrogen uptake, leaching and net mineralization. Plant and Soil 2011,
94
341:137-154.63. Constantin J, Mary B, Laurent F, Aubrion G, Fontaine A, Kerveillant P, Beaudoin N: Effects
of catch crops, no till and reduced nitrogen fertilization on nitrogen leaching and balance in three long-term experiments. Agriculture Ecosystems & Environment 2010, 135:268-278.
64. Withers PJA, Bailey GA: Sediment and phosphorus transfer in overland flow from a maize field receiving manure. Soil Use and Management 2003, 19:28-35.
65. Parkes ME, Campbell J, Vinten AJA: Practice to avoid contamination of drainflow and runoff from slurry spreading in spring. Soil Use and Management 1997, 13:36-42.
66. Thomas H, Collins E, Hallock D, Kornegay E, Martens D, Ciravolo T: Pollutant movement to shallow ground water tables from anaerobic swine waste lagoons. Journal of environmental quality 1979, 8:126-130.
67. Gooddy D, Hughes A, Williams A, Armstrong A, Nicholson R, Williams J: Field and modelling studies to assess the risk to UK groundwater from earth based stores for livestock manure. Soil Use and Management 2001, 17:128-137.
68. Miller MH, Robinson JB, Gillham RW: Self-Sealing of Earthen Liquid Manure Storage Ponds: I. A Case Study. Journal of Environmental Quality 1985, 14:533-538.
69. Cote C, Villeneuve A, Lessard L, Quessy S: Fate of pathogenic and nonpathogenic microorganisms during storage of liquid hog manure in Quebec. Livestock Science 2006, 102:204-210.
70. Arrus KM, Holley RA, Ominski KH, Tenuta M, Blank G: Influence of temperature on Salmonella survival in hog manure slurry and seasonal temperature profiles in farm manure storage reservoirs. Livestock Science 2006, 102:226-236.
71. Jussy J-H, Colin-Belgrand M, Ranger J: Production and root uptake of mineral nitrogen in a chronosequence of Douglas-fir (Pseudotsuga menziesii) in the Beaujolais Mounts. Forest Ecology and Management 2000, 128:197-209.
72. Salm Cvd, Gon HDvd, Wieggers R, Bleeker A, Toorn Avd: The effect of afforestation on water recharge and nitrogen leaching in The Netherlands. Forest Ecology and Management 2006, 221, issue 1-3:170-182.
73. Mortensen Jr, Hauge Nielsen K, JØrgensen U: Nitrate leaching during establishment of willow (Salix viminalis) on two soil types and at two fertilization levels. Biomass and Bioenergy 1998, 15:457-466.
74. Bontemps PY, Lambert R, Devillers C, Peeters A: Comparison of the efficiency of different catch crops onpotentially leachable nitrate. In Controlling nitrogen flows and losses 12th Nitrogen Workshop; University of Exeter, UK. 2003
75. Defra: OF0118T Optimisation of nitrogen mineralisation from winter cover crops and utilisation by subsequent crops. 2000.
76. McCracken DV, Smith MS, Grove JH, MacKown CT, Blevins RL: Nitrate leaching as influenced by cover cropping and nitrogen source. Soil Science Society of America Journal 1994, 58:1476-1483.
77. Department for Environment, Food and Rural Affairs (UK): Single Payment Scheme Cross compiance Guidance for Soil Mangement. UK: Department for Environment Food and Rural Affairs; 2010.
78. Oelmann Y, Kreutziger Y, Bol R, Wilcke W: Nitrate leaching in soil: Tracing the NO3 sources with the help of stable N and O isotopes. Soil Biology and Biochemistry 2007, 39:3024-3033.
79. Merbach W, Wurbs A, Pfeffer H, Latus C: The effect of catch crops on the N-15 nitrogen percolation and leaching during the early winter period. Isotopenpraxis 1993, 29:93-97.
80. Davies DB, Rochford ADH: Effect of husbandry on nitrate leaching from a chalk soil. Aspects of Applied Biology 1992:271-274.
81. Justes E, Mary B, Nicolardot B: Comparing the effectiveness of radish cover crop, oilseed rape volunteers and oilseed rape residues incorporation for reducing nitrate
95
leaching. Nutrient Cycling in Agroecosystems 1999, 55:207-220.82. Angier JT, McCarty GW: Variations in base-flow nitrate flux in a first-order stream and
riparian zone. Journal of the American Water Resources Association 2008, 44:367-380.83. Liu X, Zhang X, Zhang M: Major factors influencing the efficacy of vegetated buffers on
sediment trapping: A review and analysis. Journal of Environmental Quality 2008, 37:1667-1674.
84. Aureau F: Afforestation, a benefit for water in Brittany (in french). RenDez Vous Techniques 2008, 22:44-46.
85. Anton K: Afforestation of agricultural lands as a means of conserving drinking water in the water conservation zone of Holdorf (in german). Forst und Holz 1999, 54:404-408.
86. Stewart G: Meta-analysis in applied ecology. Biology letters 2010, 6:78-81.87. Stutter MI, Chardon WJ, Kronvang B: Riparian Buffer Strips as a Multifunctional
Management Tool in Agricultural Landscapes: Introduction. Journal of Environmental Quality 2012, 41:297-303.
88. Waterson A and Randall NP What impact does the alteration of timing to slurry applications have on leaching of nitrate, phosphate and bacterial pathogens? A Rapid Evidence Assessment Department for Environment, Food and Rural Affairs 2013
89. Fish R, Potschin M: Catchment Planning and the Ecosystems Approach: Progress towards Application LWEC Pilot Review. Contract Number: R8/H12/107. UK: Nottingham University; 2009.
90. Brown LE, Mitchell G, Holden J, Folkard A, Wright N, Beharry-Borg N, Berry G, Brierley B, Chapman P, Clarke SJ, et al: Priority water research questions as determined by UK practitioners and policy makers. Science of the Total Environment 2012, 409:256-266.
91. McGonigle D, Harris R, McCamphill C, Kirk S, Dils R, Macdonald J, Bailey S: Towards a more strategic approach to research to support catchment-based policy approaches to mitigate agricultural water pollution: A UK case-study. Environmental Science & Policy 2012.
92. Whitlock MC: Data archiving in ecology and evolution: best practices. Trends in ecology & evolution 2011, 26:61-65.
96
