Content uploaded by Anna Mária Csergő
Author content
All content in this area was uploaded by Anna Mária Csergő on Feb 01, 2018
Content may be subject to copyright.
Weed species composition of small-scale farmlands
bears a strong crop-related and environmental
signature
K NAGY* , A LENGYEL†,AKOV
ACS*, D T
€
UREI‡,AMCSERG
}
O§&
G PINKE*
*Faculty of Agricultural and Food Sciences, Sz
echenyi Istv
an University, Mosonmagyar
ov
ar, Hungary, †MTA Centre for Ecological
Research, Tihany, Hungary, ‡European Molecular Biology Laboratory –European Bioinformatics Institute, Hinxton, UK, and §School of
Natural Sciences, Trinity College Dublin, Dublin, Ireland
Received 13 February 2017
Revised version accepted 2 October 2017
Subject Editor: Jonathan Storkey, Rothamsted Research, UK
Summary
Weed species loss due to intensive agricultural land
use has raised the need to understand how traditional
cropland management has sustained a diverse weed
flora. We evaluated to what extent cultivation prac-
tices and environmental conditions affect the weed
species composition of a small-scale farmland mosaic
in Central Transylvania (Romania). We recorded the
abundance of weed species and 28 environmental,
management and site context variables in 299 fields of
maize, cereal and stubble. Using redundancy analysis,
we revealed 22 variables with significant net effects,
which explained 19.2% of the total variation in spe-
cies composition. Cropland type had the most pro-
nounced effect on weed composition with a clear
distinction between cereal crops, cereal stubble and
maize crops. Beyond these differences, the
environmental context of croplands was a major dri-
ver of weed composition, with significant effects of
geographic position, altitude, soil parameters (soil pH,
texture, salt and humus content, CaCO
3
,P
2
O
5
,K
2
O,
Na and Mg), as well as plot location (edge vs. core
position) and surrounding habitat types (arable field,
road margin, meadow, fallow, ditch). Performing a
variation partitioning for the cropland types one by
one, the environmental variables explained most of
the variance compared with crop management. In con-
trast, when all sites were combined across different
cropland types, the crop-specific factors were more
important in explaining variance in weed community
composition.
Keywords: Transylvania, weed flora, arable fields,
agroecology, agro-ecosystem, altitude, field edges,
redundancy analysis.
NAGY K, LENGYEL A, KOV
ACS A, T€
UREI D, CSERG
}
OAM & PINKE G (2018). Weed species composition of small-
scale farmlands bears a strong crop-related and environmental signature. Weed Research 58 46–56.
Introduction
Changes in farming systems, mechanisation, increases
in field size, as well as the use of chemical fertilisers
and herbicides, have had a marked negative impact on
weed species diversity and abundance (Marshall et al.,
2003; Albrecht et al., 2016). Many European countries
have reported significant decrease in abundance or
even extinction of typical arable weed species (Storkey
et al., 2012).
Despite their potential importance for the health of
agricultural ecosystems, weed species may also cause
Correspondence: K Nagy, Faculty of Agricultural and Food Sciences, Sz
echenyi Istv
an University, H-9200 Mosonmagyar
ov
ar, V
ar 2, Hungary.
Tel: (+36) 70 4019399; Fax: (+36) 96 566610; E-mail: galnagykatalin@gmail.com
©2017 European Weed Research Society 58, 46–56
DOI: 10.1111/wre.12281
significant economic losses for farmers and weed con-
trol can be the most expensive agricultural practice
aimed at improving crop production (Marshall et al.,
2003). To develop efficient, sustainable and environ-
mentally friendly weed control practices, it is impor-
tant to understand the drivers of weed presence and
abundance on cultivated lands (Swanton et al., 1999).
We need to investigate how the interaction between
farming and weed management systems and the envi-
ronment affects the composition of weed vegetation in
different croplands (Py
sek et al., 2005; Pinke et al.,
2011, 2012, 2013).
Existing evidence is mixed, suggesting that the weed
composition of arable lands may primarily be deter-
mined by ecological factors (Lososov
aet al., 2004) or
by human activity (Fried et al., 2008; Andreasen &
Skovgaard, 2009; Cimalov
a & Lososov
a, 2009; Pinke
et al., 2012). However, it is sensible to expect that the
two types of factors interact, and the prevalence of one
or the other is context dependent. For instance, where
environmental conditions are less favourable to crop-
ping, the degree of agricultural intensification is also
lower and the environmental imprint on weed composi-
tion is strong (Lososov
aet al., 2004; Nowak et al.,
2015). In upland areas, the frequency of herbicide treat-
ments is usually lower than elsewhere (P
al et al., 2013),
the proportion of alien weed species is lower and weed
species richness is higher (Lososov
aet al., 2004). Never-
theless, the composition of the weed flora also depends
on the crop type, including the division between winter-
and summer-sown crops and crop-specific management
(Fried et al., 2008). Superimposed on this pattern may
be the often-reported increase in weed species richness
towards field margins, due to a lower competition pres-
sure from crops and release from chemical stressors in
border areas (Seifert et al., 2015). The role of these mar-
ginal cropland habitats in conservation is very impor-
tant and increasingly recognised (Wrzesie
n & Denisow,
2016). Rare weed species are usually restricted to the
outermost few metres of the croplands, where weed spe-
cies richness and cover are higher compared with the
field centre (Wilson & Aebischer, 1995; Fried et al.,
2009). The study fields in our area were characteristi-
cally small, potentially magnifying this affect, as the
boundary:area ratio would be increased.
In many parts of Eastern Europe, the traditional
management practices have been preserved for longer
compared with Western Europe, conserving important
arable biodiversity in small-scale mosaic landscapes
(Loos et al., 2015). Although significant land use
changes are currently underway (Ny
ar
adi & B
alint,
2013; Loos et al., 2015), due to the high number of
small farmlands and a high variety of cropping prac-
tices, these landscapes still provide ideal ground for
gauging the imprints of environment on weed composi-
tion in agricultural lands.
In this study, we investigated the relative effect of
agricultural management and environmental factors on
weed species composition of arable fields in small-scale
farmlands. Our study system was a mosaic of small
farmlands in Central Transylvania (Romania), charac-
terised by a high diversity of cropping practices.
Detailed surveys of weed vegetation of arable lands in
the area have been scarce and the existing studies pro-
vided little mechanistic understanding of the persis-
tence of weed species in traditional landscapes (Chiril
a,
2001; Cioc^
arlan et al., 2004; Loos et al., 2015).
We performed a comprehensive survey of weed veg-
etation in this area and examined the effects of 14
management, 12 environment and two site context
variables on species composition of weed communities.
We hypothesised that, due to the persistence of tradi-
tional management practices and the small-scale farms,
the weed composition of arable lands would carry a
strong imprint of environmental factors, in addition to
the effect of management techniques.
Materials and methods
Site description
We carried out our survey in 2013 in Central Transyl-
vania, Romania (23°590260″–26°110992″North,
46°080520″–46°540597″East), covering nearly the total
area of 6714 km
2
of Muresßcounty in this region
(Fig. 1). The proportion of agricultural land in this
county is 61%, of which 54% is classified as arable
land. The most widely cultivated crops are cereals and
maize (INS, 2016). Our study covered an elevational
gradient ranging between 260 and 543 m (Table 1).
The lower elevations included the Transylvanian Pla-
teau, more suitable for agriculture due to wide valleys
and a milder climate. The higher elevation north-east-
ern corner of the county consisted of the C
alimani and
Gurghiu Mountain foothills, where arable fields were
rarer. Here, the temperature and precipitation regimes
have been less suitable for crop production and there-
fore agricultural intensification has been lower, for
example 4–6 times lower doses of chemical fertilisers
and herbicides on average compared with France or
Germany (Storkey et al., 2012).
Data collection
We selected a total of 299 arable fields for the survey
in a broadly random pattern, but also depending on
farmer’s co-operation (Fig. 1). Within each field, we
sampled weed vegetation in six randomly selected,
©2017 European Weed Research Society 58, 46–56
Environmental signatures of small-scale farming weeds 47
4-m
2
plots (2 92 m), totalling 1794 plots. Three plots
were located on the field edge (within 2 m from the
outermost seed drill line), and three were in the field
centre; 101 fields were cereal crops (74 Triticum aes-
tivum L., 11 Triticosecale 9rimpaui Wittm., eight Hor-
deum vulgare L., five Hordeum distichon L., three
Avena sativa L.) and 97 maize (Zea mays L.). The
remaining 101 sites were stubbles of cereal fields. While
cereal stubbles are not crops, we analysed them as a
separate cropland type due to their unique weed vege-
tation (Pinke et al., 2010). We surveyed the cereal
fields between May 10 and June 6, and the maize and
the cereal stubble fields between July 31 and August 20
to ensure that we captured the most comprehensive set
of weed species within each cropland type.
Within each 4-m
2
plot, we estimated visually the
percentage ground cover of all species, including crop
species, and the vegetation data recorded were subse-
quently digitised and stored in TURBOVEG format
(Hennekens & Schamin
ee, 2001). In addition, we inter-
viewed landowners for information on crop manage-
ment of each investigated field. We recorded the
cropping history (indicating the preceding crop as
either cereal or a hoed crop, such as sunflower, potato
and maize), the amount of organic manure applied,
whether farmers used chemical fertilisers (N, P
2
O
5
,
K
2
O), as well as crop sowing season (previous autumn
or spring) and field size. Information on weed manage-
ment (type of herbicides used and number of times
mechanical weed control treatments were applied) were
also recorded. Herbicides applied on less than 10 fields
of the total of 299 were subsequently dropped from
the analyses. To reduce the number of management
categories, the ‘cropland type’ variable was coded as
cereal crop, maize crop or cereal stubble.
We used soil chemical and physical properties as
local environmental variables. From each field, we col-
lected one soil sample of 1000 cm
3
from the top 10-cm
layer. Soil samples were air-dried and stored at room
temperature until further analyses were performed at
UIS Ungarn GmbH (Mosonmagyar
ov
ar, Hungary).
Soil variables included: soil pH, texture, salt and
humus content, CaCO
3
,P
2
O
5
,K
2
O, Na and Mg. In
addition, we used three proxies of regional environ-
mental conditions quantified as the geographic lati-
tude, longitude and elevation above sea level of each
field, as recorded by a GPS device.
Finally, we considered two site variables: plot loca-
tion (edge or field core) and neighbouring habitat (ara-
ble field, road margin, meadow, fallow or ditch) to
represent composite management and environmental
Fig. 1 The distribution of the surveyed arable fields across the
study area (Mureșcounty, Central Transylvania, Romania). At
this scale, individual points may represent a number of fields with
different cropland types. [Colour figure can be viewed at wileyon-
linelibrary.com].
Table 1 Units and ranges of continuous variables and values of
categorical variables recorded on each cropland for this study
Variable (unit) Range/Values
Site
Plot location Edge, core
Neighbouring habitat Arable field, ditch,
fallow, meadow, road
margin
Environmental
Longitude (E) 46°080520″–46°540597″
Latitude (N) 23°590260″–26°110992″
Altitude (m) 260–543
Soil pH (KCl)* 5.02–7.60
Soil texture (KA) 29–57
Soil properties (% as g g
1
)
Humus 1.58–7.57
CaCO
3
0.1–18.5
Soil salt* 0.02–0.17
Soil properties (mg kg
1
)
P
2
O
5
20–4460
K
2
O 83.3–1030
Na* 14.2–148
Mg 72.1–803
Management
Field size (ha) 0.06–32
Cropland type Cereal crop, maize crop,
cereal stubble
Sowing season Autumn, spring
Preceding crop Cereal, hoed crop
Organic manure (t ha
1
)0–45
Chemical fertiliser Yes, no
Mechanical weeding (times)* 0–6
Herbicides
2,4 D* Yes, no
Bromoxynil* Yes, no
Dicamba Yes, no
Isoxaflutol +cyprosulfamide Yes, no
Florasulam Yes, no
Fluroxypyr Yes, no
Thiencarbazone-methyl Yes, no
*Variables dropped during the backward selection process.
©2017 European Weed Research Society 58, 46–56
48 K Nagy et al.
effects. Overall, we recorded 28 parameters: two site
variables, 12 environmental variables and 14 manage-
ment variables (Table 1).
Statistical analyses
Prior to analyses, we averaged the abundance of
species across field edge and field core plots respec-
tively, which we subsequently transformed following
the Hellinger approach (Legendre & Gallagher,
2001). We also transformed the categorical variables
(the amount of chemical fertilisers and herbicides)
into ‘dummy’ indicator variables. To analyse the
relationship between the composition of weed vege-
tation and site, environmental and management
variables, we performed a redundancy analysis
(RDA). RDA links species abundance data to
explanatory variables more accurately than the com-
monly used canonical correspondence analysis
(CCA), even when species responses to environmen-
tal gradients are unimodal (Legendre & Gallagher,
2001). Only species with >10 occurrences were
involved in the analyses. We reduced the number of
explanatory variables using stepwise backward selec-
tion with a P<0.05 threshold. With this procedure,
six variables were eliminated: soil pH, Na and salt
content, mechanical weeding and herbicides 2,4 D
and bromoxynil, resulting in a reduced RDA model
with 22 terms with significant effects. The gener-
alised variance inflation factor GVIF (Fox & Mon-
ette, 1992) ranged between 1.0 and 5.51, indicating
no serious collinearity between explanatory
variables.
We then compared the gross and net effects of each
explanatory variable, following the methodology
described in Lososov
aet al. (2004). The gross effects
represented the variation explained by a ‘univariate’
RDA containing the predictor of interest as the only
explanatory variable. The net effect was calculated
using a partial RDA (pRDA), which included the vari-
able of interest as explanatory variable and the other
21 variables as conditional variables (‘covariables’).
We extracted the explained variance and the adjusted
R-squared (R2
adj) for models of both gross and net
effects of each variable. In models of net effects, model
fit was also assessed by the F-value for which a type I
error rate was estimated using 999 permutation tests of
the constrained axis. The importance of each explana-
tory variable was ‘ranked’ using the R2
adj values of the
pRDA (i.e. net effect) models. Subsequently, we identi-
fied the 10 species with the highest fit for each explana-
tory variable.
We report only the RDA ordination diagrams of
the reduced model with the finally selected 22
variables. In these diagrams, continuous variables were
represented by their linear constraints, while positions
of categorical variables were calculated by weighted
averaging of co-ordinates of plots representing each
level.
In addition, we performed a variation partitioning
analysis to assess the relative effects of site, environ-
mental and management variables on weed species
composition, either within each cropland type sepa-
rately or across all the fields, and separated by edge vs.
centre position (Borcard et al., 2011). This procedure
identified unique and shared contributions of groups
of variables using adjusted R-squared values. Statisti-
cal analyses were performed using the vegan (version
2.3-3) and car (version 2.0-25) packages in R 3.1.2 (R
Development Core Team). Species fit on the con-
strained ordination axes was calculated using the ‘in-
ertcomp’ function of vegan package.
Results
Across the 1794 plots sampled from 299 arable fields,
we found a total of 141 weed species, 110 in cereals, 88
in stubble fields and 76 in maize crops. From the top
most threatened 48 arable weeds in Europe (Storkey
et al., 2012), only four occurred in our data set, all in
cereal fields. Their frequency of occurrence ranged
between 1.0 and 9.7% (Adonis aestivalis L. 9.7%, Cen-
taurea cyanus L. 6.1%, Ranunculus arvensis L. 5.9%,
Lathyrus aphaca L. 1.0%).
The full RDA model comprising all 28 explanatory
variables explained 20.25% of the variance, while the
reduced model with 22 explanatory variables still
explained 19.15% of the total variation in species com-
position. All 22 variables (cropland type, geographic
position, altitude, soil parameters, plot location and
neighbouring habitat) had significant net effects at a
P<0.05 level (Table 2). Weed species with the stron-
gest responses to these factors are listed in Tables S1,
S2 and S3.
In the reduced RDA ordination (Fig. 2), the first
two axes explained 7.65% and 2.51% of the total vari-
ation respectively. Cropland type (cereal crop, maize
crop and cereal stubble) resulted in the largest distinc-
tion in weed species composition, followed by the sow-
ing season (autumn and spring) (9.46 and 3.84% of
explained variation, respectively; Table 2). Species pos-
itively associated with the first axis were typical of
maize crops (e.g. Amaranthus retroflexus L., Chenopo-
dium album L., Hibiscus trionum L.), while species
characteristic of cereal crops were negatively associated
with the first axis (e.g. Galium aparine L., Papaver
rhoeas L., A. aestivalis). Species found in cereal stub-
bles had a positive weight on the second axis (e.g.
©2017 European Weed Research Society 58, 46–56
Environmental signatures of small-scale farming weeds 49
Stachys annua L., Anagallis arvensis L. and Setaria vir-
idis (L.) P. Beauv) (Fig. 2).
Neighbouring habitat (a site variable) was the next
best important predictor of variation in weed composi-
tion (net effect: 0.76% and gross effect: 1.42%
explained variation; Table 2). Arable fields were posi-
tively, and road margins and meadows were negatively
associated with the first axis, while ditches weighted
positively on the second axis. Further variables with a
strong weight on the first axis were organic manure
and soil properties (calcium, potassium and humus
content), while variables with strong weight on the sec-
ond axis were soil texture, chemical fertilisers and lati-
tude (Table 2, Fig. 2).
The variation partitioning within each cropland
type revealed that environmental variables outper-
formed the management and site variables, with nearly
equal values in stubbles and maize, and slightly lower
in cereals (6.6%, 6.5% and 4.8%, respectively Fig. 3).
The management variables had the highest relative
effect in maize and equally lower in cereals and stub-
bles. The relative effects of site and management vari-
ables were similar in cereals (2.5% vs. 2.6%,
respectively), but in maize and stubbles site explained
only a tiny fraction of the variance (0.9–0.2%) (Fig. 3).
Variation partitioning over all the 299 fields resulted
the highest influence of management variables, being
largely driven by crop type, explaining three times
more of the total variance compared with the
environmental variables (10.9% vs. 3.4%; Fig. 4). The
variation partitioning of the RDA according to the
plot location revealed that the effect of environmental
variables is only slightly higher in field edges than in
the cores (3.2% vs. 2.6%, respectively), while the influ-
ence of management was nearly equal (10.4% vs.
10.5%; Fig. 5).
Discussion
Farmland management practices such as cropland
type, fertilisation and sowing season were the major
drivers of weed composition in the studied system.
However, environment and site effects were also
important contributors to the revealed patterns. Our
report represents the most exhaustive assessment to
date of the weed vegetation of arable lands in Central
Transylvania, showcasing factors that structure weed
composition under agronomical practices currently
typical of Eastern Europe.
Management effect
We found that 11 of the 22 significant predictors of
weed composition were elements of the management
system. From all management variables involved in the
study, only three (two herbicides and frequency of
mechanical weeding) were dropped during the back-
ward selection process, and the effect of all of the
Table 2 Gross and net effects of the explanatory variables on the weed species composition identified using (p)RDA analyses with single
explanatory variables
Factors d.f.
Gross effect Net effect
Explained variation (%) R2
adj Explained variation (%) R2
adj FP-value
Cropland type 2 9.459 0.0915 5.619 0.0556 19.8414 0.001
Longitude 1 1.469 0.0130 0.696 0.0058 4.9130 0.001
Altitude 1 0.819 0.0065 0.619 0.0050 4.3698 0.001
Organic manure 1 0.818 0.0065 0.507 0.0038 3.5807 0.001
Soil Ca content 1 0.612 0.0045 0.477 0.0035 3.3716 0.001
Plot location 1 0.459 0.0029 0.459 0.0033 3.2407 0.001
Soil texture 1 0.568 0.0040 0.455 0.0033 3.2122 0.001
Soil K content 1 0.787 0.0062 0.442 0.0031 3.1188 0.001
Chemical fertiliser 1 0.568 0.0040 0.383 0.0025 2.7073 0.002
Soil Mg content 1 0.443 0.0028 0.367 0.0024 2.5945 0.001
Fluroxypyr 1 0.735 0.0057 0.359 0.0023 2.5351 0.001
Field size 1 0.511 0.0034 0.346 0.0021 2.4463 0.003
Latitude 1 0.414 0.0025 0.341 0.0021 2.4085 0.001
Neighbouring habitat 4 1.416 0.0075 0.763 0.0020 1.3480 0.017
Preceding crop 1 0.480 0.0031 0.329 0.0020 2.3231 0.002
Florasulam 1 0.576 0.0041 0.317 0.0018 2.2359 0.003
Soil P content 1 0.328 0.0016 0.290 0.0015 2.0469 0.006
Isoxaflutol +cyprosulfamide 1 0.917 0.0075 0.269 0.0013 1.8981 0.014
Sowing season 1 3.843 0.0368 0.262 0.0013 1.8535 0.018
Soil humus content 1 0.598 0.0043 0.260 0.0012 1.8360 0.012
Thiencarbazone-methyl 1 0.852 0.0069 0.260 0.0012 1.8340 0.013
Dicamba 1 0.222 0.0005 0.235 0.0010 1.6610 0.030
©2017 European Weed Research Society 58, 46–56
50 K Nagy et al.
remaining management variables was significant. Of
these, cropland type had the most pronounced effect,
reinforcing the view that crop type is a primary driver
of weed vegetation (Cimalov
a & Lososov
a, 2009). This
can be explained by major differences in cultivation
practices between cereals and hoed crops, such as sun-
flower, potato and maize (Andreasen & Skovgaard,
2009; Nowak et al., 2015). Cereal fields are exposed to
mechanical disturbance (and stresses caused by herbi-
cides) only at the beginning of the season and after
harvesting, ensuring a longer undisturbed growing per-
iod for weeds in comparison with hoed crops. Most
rare and endangered species (such as A. aestivalis,
C. cyanus,L. aphaca,R. arvensis in our data set) have
been associated with cereals, because they germinate
mainly in autumn and have their life cycle adapted to
that of cereals rather than to that of hoed spring-sown
crops (Kol
a
rov
aet al., 2013). Following cereal harvest,
stubbles are left undisturbed until late autumn, leaving
open sunny habitats suitable for the establishment of
species that are able to germinate at high temperatures
and tolerate summer drought, for example summer
therophytes (S. annua,A. arvensis,Kickxia elatine (L.)
Dumort.). In contrast, species identified as typical of
maize crops have their germination associated with
later crop sowing date (Gunton et al., 2011) and are
able to tolerate continuous disturbance regimes (Echi-
nochloa crus-galli (L.) P. Beauv, Setaria pumila (Poir.)
Schult., H. trionum,C. album) (Fig. 2). A typical dis-
turbance-tolerance strategy is the steady germination
ability of seeds throughout the cultivation period
(Fried et al., 2012).
It would have been interesting to distinguish
between the effect of the season (using the date of
observation) and the effect of the management. How-
ever, these two factors are confounded in the one vari-
able, cropland type, making their separate analysis
impossible. It is likely that season and management
interacted to shape the characteristics we associated
with stubble in our analysis. Despite similar sowing
dates of cereals, subsequent germination later in the
season would have contributed to the different floras
recorded in their stubble. Preceding management
regimes, that is cropping technologies applied in cere-
als and maize, also have their impact on weed floras.
Furthermore, environmental conditions in the stubble
are different, for example free from the shading.
Accordingly, not only the flora of cereals and that of
their stubbles differ remarkably, but stubble and maize
also have different weed flora, even though the fact
that they were surveyed in the same season. Conse-
quently, stubble is not a homogenous category among
cropland types; its subdivision and introduction of sea-
son as a new variable would have made it possible to
Fig. 2 Ordination diagrams of the reduced RDA model contain-
ing the 22 significant explanatory variables and the species. Only
the species with the highest weight on the first two RDA axes are
presented.
©2017 European Weed Research Society 58, 46–56
Environmental signatures of small-scale farming weeds 51
further dissect the causalities behind the patterns of
weed composition.
Fertilisation was an important filter of weed species
and a selective driver of weed abundance (for similar
results, see Lososov
aet al., 2006; Pinke et al., 2012;
Seifert et al., 2015). Several species responded to
organic manure with increased abundances (e.g.
Convolvulus arvensis L., S. pumila,E. crus-galli), while
chemical fertilisers could be linked to higher
abundances of only three species (Rubus caesius L.,
H. trionum,Elymus repens (L.) Gould). Almost all
weed species that responded positively to higher
organic manure were associated with maize fields (e.g.
E. crus-galli,C. album, A. retroflexus), due to higher
doses applied in hoed crops (Lososov
aet al., 2006).
A strong negative relationship between field size
and weed diversity at the landscape level has often
been reported due to a higher associated heterogeneity
of cultivated areas and a larger edge:area ratio in smal-
ler field sizes (Marshall et al., 2003; Gaba et al., 2010;
Fahrig et al., 2015). Some mechanical operations are
less efficient in smaller fields, and farmers cultivating
small fields tend to have limited access to weed man-
agement technology or expertise (Pinke et al., 2013).
In our study, this effect, albeit significant, was less pro-
nounced (field size ranked only 12th among the
explanatory variables), as our data cover only a nar-
row range of field sizes (most fields in our survey were
small, 59% had ≤1 ha).
The sowing season was an important driver of weed
composition in our survey, where we investigated winter-
and spring-sown cereals and spring-sown maize. Winter
annual weed species (Veronica persica Poir., Consolida
orientalis (J. Gay) Schr€
odinger, G. aparine,P. rhoeas)
were strongly associated with autumn-sown cereals,
while summer annual weed species (A. retroflexus,C. al-
bum,H. trionum,S. pumila,E. crus-galli) preferred
Fig. 3 Percentage contributions of groups
of explanatory variables to the variation
in weed species composition in the three
investigated cropland types, identified by
variation partitioning (only non-negative
adjusted R-squared values are shown).
Fig. 4 Percentage contributions of groups of explanatory vari-
ables to the variation in weed species composition using all the
299 fields, identified by variation partitioning (only non-negative
adjusted R-squared values are shown).
©2017 European Weed Research Society 58, 46–56
52 K Nagy et al.
spring-sown cultures, many of the latter being typical
weeds of hoed crops, such as sunflower, potato and
maize (Fig. 2). These results concur with earlier evi-
dence, confirming that the presence of multiple crops
and cropping times may considerably increase the regio-
nal weed species pool (Marshall et al., 2003; Pinke et al.,
2011; Fried et al., 2012; Vidotto et al., 2016).
Among preceding crops, winter cereals usually
favour winter annuals, while hoed crops favour sum-
mer annuals. In our analysis, preceding crop ranked
only the 15th among the predictors, not independently
from the common practice in the surveyed area to
alternate winter cereals with hoed crops. The rotation
of cereals and hoed crops aims to interrupt the build-
up of weed populations associated with particular crop
types (de Mol et al., 2015).
We found that the use of herbicides significantly
affected the occurrence and abundance of weed species.
The active ingredients of the herbicides with significant
effect were fluroxypyr, florasulam, isoxaflutol with
cyprosulfamide, thiencarbazone-methyl and dicamba
(Table 2). All of these were used for post-emergence
control. Florasulam, fluroxypyr and dicamba can be
used against dicotyledonous weeds, and isoxaflu-
tol +cyprosulfamide and thiencarbazone-methyl are
broad-spectrum herbicides for the control of both
monocotyledons and dicotyledonous weeds. Although
we identified several weed species that were correlated
with herbicides according to their explained variation
in the constrained axes, without a survey before and
after herbicide treatment, we cannot draw firm
conclusions on the effect of herbicides. Accordingly,
these correlations are not shown in the supporting
information.
Environmental effect
We found nine environmental variables with significant
net effects on weed composition, including both regio-
nal and local factors (Table 2). Longitude ranked the
2nd, altitude the 3rd and latitude the 13th among all
predictors. These variables have been used as proxies
of regional climate conditions, such as precipitation
and mean temperature (Lososov
aet al., 2004, 2006;
Hanzlik & Gerowitt, 2011; de Mol et al., 2015). Spe-
cies strongly associated with lower altitudes were trou-
blesome weeds such as Solanum nigrum L., Xanthium
italicum Moretti, Polygonum aviculare L. and R. cae-
sius, while species correlated with higher altitudes were
cereal weeds typical of traditional farming, e.g. C. ori-
entalis,C. cyanus. This pattern has often been reported
from agricultural landscapes situated in heterogeneous
geographic conditions (Lososov
aet al., 2004; P
al
et al., 2013; Nowak et al., 2015). The north-eastern
higher altitude part of our study area is less favour-
able; especially for maize but also for other crops, and
as a consequence the cultivation is less intense (Fig. 1).
We interpret the change in weed composition along
this geographic gradient because of both environmen-
tal effects and differences in farming methods between
lowland and upland areas.
As expected, soil physical and chemical properties,
such as texture, Ca, K, Mg, P and humus content,
exerted significant effects on the occurrence of certain
weed species (Pinke et al., 2012, 2016). For example,
we found that Cirsium arvense (L.) Scop., a species
common in all crop types, preferred soils with high
humus and Mg content, but avoided alkaline soils.
Although in many studies pH was a crucial determi-
nant of weed species presence (e.g. Py
sek et al., 2005;
Fried et al., 2008; Vidotto et al., 2016), other investiga-
tions, including ours, found this factor to be non-sig-
nificant (see also Nowak et al., 2015), likely because
neutral soils were prevalent in our study area.
Site effect
The plot location (edge vs. core position) and the
neighbouring habitat type had moderate effects on
weed composition (the 6th and the 14th most impor-
tant predictors, respectively). Most weeds preferred
field edges and only one species, C. arvensis, had
Fig. 5 Percentage contributions of groups
of explanatory variables to the variation
in weed species composition in field edges
and field cores, identified by variation
partitioning (only non-negative adjusted
R-squared values are shown).
©2017 European Weed Research Society 58, 46–56
Environmental signatures of small-scale farming weeds 53
higher abundance towards field interiors. It is well
known from other agricultural ecosystems that crop
margins support higher species richness and the princi-
ple is applied in weed conservation (e.g. Pinke et al.,
2012; Kol
a
rov
aet al., 2013; Seifert et al., 2015; Wrze-
sie
n & Denisow, 2016). Mechanisms behind these pat-
terns include the crop’s lower competition ability,
dilution or lack of chemical stressors in the border
areas (Seifert et al., 2015), release from competition for
light exerted by crop species (Pinke et al., 2012) and a
higher external propagule supply from adjacent habi-
tats (Gaba et al., 2010; Concepti
on et al., 2012; Pinke
et al., 2012; Wrzesie
n & Denisow, 2016).
In our mosaic of small farmlands, neighbouring
habitats were diverse (arable field, ditch, fallow, mea-
dow, road margin) and were linked to the presence/
abundance of specific weeds in the crop fields. Main-
taining a diversity of non-farmed habitats adjacent to
farmlands may therefore result in an enriched weed
flora in crop fields. Here, we have shown that this
externally driven enrichment diminishes substantially
towards field interiors (see also Gaba et al., 2010;
Pinke et al., 2012).
Environment vs. management factors
In the variation partitioning within each cropland type,
the environmental variables explained the largest frac-
tions of the variance, which is in accordance with the
results of previous studies (Lososov
aet al., 2004; Pinke
et al., 2012, 2016; de Mol et al., 2015). The effect of
environmental variables reached the highest proportion
in cereal stubbles, explaining two and a half time more
variance than the effect of management variables. This
may be due to the lack of particular cropping practices
on stubbles. In maize crops, the relative influence of
environmental variables was similarly high. Both maize
and stubble represented the late summer weed flora,
and the higher contributions of environment could be
due to the longer period following weed management
practices, which allows the weed vegetation to recover
from the seed banks primarily under the influence of
soil and climatic conditions. Furthermore, in maize, the
management variables explained a higher proportion of
variance in weed communities when compared with
cereals and stubbles, possibly due to the frequently
repeated cultivation tasks typical of maize crops.
In contrast with the crop-specific analyses, the vari-
ation partitioning carried out over all sites highlighted
the importance of the management variables. This
shows that the involvement of crop type can increase
the contribution of management remarkably, highlight-
ing the generally powerful impact of crop-related
factors on the weed flora (Fried et al., 2008; Gunton
et al., 2011).
Splitting up the variance allocated to the plot loca-
tion, the management factors account for approxi-
mately three times more variance compared with the
environmental variables both in field cores and edges.
We found no difference between field edges and cores
in the importance of management variables, contrary
to the findings of Pinke et al. (2012). This could be
explained by the generally small field sizes in this
study, where the cultural and ecological conditions
between edge and core are likely to be more similar
than in large fields (Wilson & Aebischer, 1995).
Acknowledgements
The publication is supported by the EFOP-3.6.3-
VEKOP-16-2017-00008 project. The project is cofi-
nanced by the European Union and the European
Social Fund.
References
ALBRECHT H, CAMBEC
EDES J, LANG M&WAGNER M (2016)
Management options for rare arable plants in Europe.
Botany Letters 164, 389–415.
ANDREASEN C&SKOVGAARD IM (2009) Crop and soil factors
of importance for the distribution of plant species on
arable fields in Denmark. Agriculture, Ecosystems and
Environment 133,61–67.
BORCARD D, GILLET F&LEGENDRE P (2011) Numerical
Ecology with R. Springer, New York Dordrecht London
Heidelberg.
CHIRIL
AC (2001) Biologia Buruienilor. Ceres, Bucuresßti,
Rom^
ania.
CIMALOV
A
S&L
OSOSOV
AZ (2009) Arable weed vegetation of
the northeastern part of the Czech Republic: effects of
environmental factors on species composition. Plant
Ecology 203,45–57.
CIOC
^
ARLAN V, BERCA M, CHIRIL
AC, COSTE I&POPESCU G
(2004) Flora segetal
a a Rom^
aniei. Ceres, Bucuresßti,
Rom^
ania.
CONCEPTI
ON ED, FERN
ANDEZ-GONZ
ALEZ F&D
IAZ M (2012)
Plant diversity partitioning in Mediterranean croplands:
effects of farming intensity, field edge, and landscape
context. Ecological Applications 22, 972–981.
FAHRIG l, GIRARD J, DURO Det al. (2015) Farmlands with
smaller crop fields have higher within-field biodiversity.
Agriculture, Ecosystems and Environment 200, 219–234.
FOX J&MONETTE G (1992) Generalized collinearity
diagnostics. Journal of the American Statistical Association
87, 178–183.
FRIED G, NORTON LR & REBOUD X (2008) Environmental
and management factors determining weed species
composition and diversity in France. Agriculture,
Ecosystems and Environment 128,68–76.
©2017 European Weed Research Society 58, 46–56
54 K Nagy et al.
FRIED G, PETIT S, DESSAINT F&REBOUD X (2009) Arable
weed decline in Northern France: crop edges as refugia for
weed conservation? Biological Conservation 142, 238–243.
FRIED G, KAZAKOU E&GABA S (2012) Trajectories of weed
communities explained by traits associated with species
response to management practices. Agriculture, Ecosystems
and Environment 158, 147–155.
GABA S, CHAUVEL B, DESSAINT F, BRETAGNOLLE V&PETIT S
(2010) Weed species richness in winter wheat increases with
landscape heterogeneity. Agriculture, Ecosystems and
Environment 138, 318–323.
GUNTON RM, PETIT S&GABA S (2011) Functional traits
relating arable weed communities to crop characteristics.
Journal of Vegetation Science 22, 541–550.
HANZLIK K&GEROWITT B (2011) The importance of climate,
site and management on weed vegetation in oilseed rape in
Germany. Agriculture, Ecosystems and Environment 141,
323–331.
HENNEKENS SM & SCHAMIN
EE JH (2001) TURBOVEG, a
comprehensive database management system for vegetation
data. Journal of Vegetation Science 12, 587–591.
Institutul Natßional de Statistic
a–National Institute of
Statistics (INS) (2016) Available at: http://www.mures.
insse.ro/main.php?lang=fr&pageid=420 (last accessed 9
March 2016).
KOL
A
ROV
AM, TY
SER L&SOUKUP J (2013) Impact of site
conditions and farming practices on the occurrence of rare
and endangered weeds on arable land in the Czech
Republic. Weed Research 53, 489–498.
LEGENDRE P&GALLAGHER ED (2001) Ecologically
meaningful transformations for ordination of species data.
Oecologia 129, 271–280.
LOOS J, TURTUREANU PD, VON WEHRDEN Het al. (2015) Plant
diversity in a changing agricultural landscape mosaic in
Southern Transylvania (Romania). Agriculture, Ecosystems
and Environment 199, 350–357.
LOSOSOV
AZ, CHYTR
YM, CIMALOV
A
Set al. (2004) Weed
vegetation of arable land in Central Europe: gradients of
diversity and species composition. Journal of Vegetation
Science 15, 415–422.
LOSOSOV
AZ, CHYTR
YM, K€
UHN Iet al. (2006) Patterns of
plant traits in annual vegetation of man-made habitats in
central Europe. Perspectives in Plant Ecology, Evolution
and Systematics 8,69–81.
MARSHALL EJP, BROWN VK, BOATMAN ND, LUTMAN PJW,
SQUIRE GR & WARD LK (2003) The role of weeds in
supporting biological diversity within crop fields. Weed
Research 43,77–89.
DE MOL F, VON RC & GEROWITT B(2015)Weed
species composition of maize fields in Germany is influenced
by site and crop sequence. Weed Research 55,574–585.
NOWAK A, NOWAK S, NOBIS M&NOBIS A (2015) Crop type
and altitude are the main drivers of species composition of
arable weed vegetation in Tajikistan. Weed Research 55,
525–536.
NY
AR
ADI I&B
ALINT J (2013) Erd
ely gyomn€
ov
enyzete,
gyomprobl
em
ak, v
edekez
esi lehet}
os
egek. Gyomn€
ov
enyek,
Gyomirt
as 14,25–34.
P
AL RW, PINKE G, BOTTA-DUK
AT Zet al. (2013) Can
management intensity be more important than
environmental factors? A case study along an extreme
elevation gradient from central Italian cereal fields Plant
Biosystems 147, 343–353.
PINKE G, P
AL R&BOTTA-DUK
AT Z (2010) Effects of
environmental factors on weed species composition of
cereal and stubble fields in western Hungary. Central
European Journal of Biology 5, 283–292.
PINKE G, P
AL RW, T
OTH K, KAR
ACSONY P, CZ
UCZ B&
BOTTA-DUK
AT Z (2011) Weed vegetation of poppy
(Papaver somniferum) fields in Hungary: effects of
management and environmental factors on species
composition. Weed Research 51, 621–630.
PINKE G, KAR
ACSONY P, CZ
UCZ B, BOTTA-DUK
AT Z&
LENGYEL A (2012) The influence of environment,
management and site context on species composition of
summer arable weed vegetation in Hungary. Applied
Vegetation Science 15, 136–144.
PINKE G, KAR
ACSONY P, BOTTA-DUK
AT Z&CZ
UCZ B (2013)
Relating Ambrosia artemisiifolia and other weeds to the
management of Hungarian sunflower crops. Journal of
Pest Science 86, 621–631.
PINKE G, BLAZSEK K, MAGYAR Let al. (2016) Weed species
composition of conventional soyabean crops in Hungary is
determined by environmental, cultural, weed management
and site variables. Weed Research 56, 470–481.
PY
SEK P, JARO
S
IK V, KROP
A
CZ, CHYTR
YM, WILD J&TICH
Y
L (2005) Effects of abiotic factors on species richness and
cover in Central European weed communities. Agriculture,
Ecosystems and Environment 109,1–8.
SEIFERT C, LEUSCHNER C&CULMSEE H (2015) Arable plant
diversity on conventional croplands –The role of crop
species, management and environment. Agriculture,
Ecosystems and Environment 213, 151–163.
STORKEY J, MEYER S, STILL KS & LEUSCHNER C (2012) The
impact of agricultural intensification and land-use change
on the European arable flora. Proceedings of the Royal
Society B: Biological Sciences 279, 1421–1429.
SWANTON CJ, SHRESTHA A, ROY RC, BALL-COELHO BR &
KNEZEVIC SZ (1999) Effect of tillage systems, N, and cover
crop on the composition of weed flora. Weed Science 47,
454–461.
VIDOTTO F, FOGLIATTO S, MILAN M&FERRERO A (2016)
Weed communities in Italian maize fields as affected by
pedo-climatic traits and sowing time. European Journal of
Agronomy 74,38–46.
WILSON P&AEBISCHER N (1995) The distribution of
dicotyledonous arable weeds in relation to distance from
the field edge. Journal of Applied Ecology 32, 295–310.
WRZESIE
NM&DENISOW B (2016) The effect of agricultural
landscape type on field margin flora in south eastern
Poland. Acta Botanica Croatica 75, 217–225.
Supporting Information
Additional Supporting Information may be found in
the online version of this article:
Table S1. Names, fit and score values of species giv-
ing the highest fit along the first constrained axis in the
partial-RDA models of the significant environmental
variables specified in Table 2. (Only the most abundant
ten weed species are shown)
©2017 European Weed Research Society 58, 46–56
Environmental signatures of small-scale farming weeds 55
Table S2. Names, fit and score values of species giv-
ing the highest fit along the first constrained axis in the
partial-RDA models of the significant management
variables specified in Table 2. (Only the most abundant
ten weed species are shown)
Table S3. Names, fit and score values of species giv-
ing the highest fit along the first constrained axis in the
partial-RDA models of the significant site variables
specified in Table 2. (Only the most abundant ten weed
species are shown)
©2017 European Weed Research Society 58, 46–56
56 K Nagy et al.