Upload
mayor
View
213
Download
0
Embed Size (px)
Citation preview
In¯uence of weed management strategies on soil seedbank diversity
J.-P. MAYOR AND F. DESSAINT*
Station FeÂdeÂrale de Recherches en Production
VeÂgeÂtale de Changins, CH-1260 Nyon,Switzerland, and *INRA, Laboratoire deMalherbologie, BV 1540, F-21034 Dijon, France
Received 23 August 1996Revised version accepted 10 November 1997
Summary
The role of three weed management strategies onweed populations and community dynamics wasinvestigated from 1989 to 1994. These were
chemical weed control (CH), integrated weedcontrol (IN) and mechanical weed control (ME).Weed populations and communities were
analysed by univariate (species richness, rank:frequency diagrams) and multivariate (corre-spondence analysis) methods. Species richness ofthe soil seedbank di�ered with weed management
strategies and over time, with an observed annualnumber of species that ranged from 18 to 32. Theseedbank was dominated by the same pool of
species in all cases, but these species respondeddi�erently among years and management strate-gies; their rank (relative abundance) and densities
changed between strategy and within a strategyamong the years. In the last 2 years of the study,the soil seedbank was characterized mainly by
Capsella bursa-pastoris (L.) Medik on the MEstrategy and Chenopodium album L. on the INstrategy. The CH strategy contrasted with theother two by the presence of a high density of
Amaranthus retro¯exus L.
Introduction
The decline in crop prices and ecological pres-sures against the systematic use of herbicides arecausing farmers to consider alternative ap-
proaches to weed control (Thornton et al., 1990;Swanton & Weise, 1991; Forcella et al., 1993).Among the most e�cient of these alternative
forms of weed control is physical weed control,i.e. mechanical working of the soil (Forcellaet al., 1993).
Although the objectives of weed control are
similar in conventional and integrated produc-tion systems, as well as in organic farming, weedcontrol procedures di�er in their e�ciency, im-
pact on the environment, costs, etc. In conven-tional production systems, herbicides are usedsystematically; in integrated systems, herbicides
are used only when no other possibilities exist,whereas in organic farming herbicides are pro-hibited. In the present economic climate, costs of
weed control procedures must be taken into ac-count. Consequently, as most farms in Switzer-land are smaller than 35 ha, investments inexpensive mechanical weeders, which are neces-
sary in integrated production systems or in or-ganic farming, can be a�orded only when severalproducers get together to buy one. This invest-
ment needs to be based upon an understandingof the consequences of the di�erent weed controlmethods. Because agronomic practices are
known to in¯uence the size and composition ofweed communities, a better understanding of thepotential changes in weed population dynamicsis necessary to facilitate an e�cient transition
from conventional to alternative managementapproaches (Clements et al., 1994).
One of the major factors a�ecting annual
weed population stability is the large and po-tentially transient seedbank. Seedbanks are ofecological and evolutionary importance in the Corresponding author.
Weed Research, 1998, Volume 38, 95±105
Ó 1998 European Weed Research Society
dynamics of weed populations and communities.Seed longevity and carry-over of viable seeds in
the soil from previous years can bu�er the e�ectsof weed control and hence maintain the weedproblem. Some researchers (Lambelet-Haueter,
1985, 1986; Barralis & Chadoeuf, 1987) havefound that, with the exception of weeds withlarge seeds, the seedbank is a better indicator ofthe long-term in¯uence of agronomic practices
on weeds than the above-ground vegetation.The objective of this study was to understand
the in¯uences of three weed control strategies on
the size and composition of the weed seedbankover a 6-year period, so as to provide useful in-formation for weed managers to be able to make
decisions on the basis of balancing of risks, costsand bene®ts.
Materials and methods
Site description and ®eld procedures
The ®eld experiment was located at the researchstation of the Institut Agricole de l'Etat de
Fribourg, Switzerland. The elevation is approx-imately 655 m above sea level. Mean dailyMarch±September air temperatures and precipi-
tation were 12.5 °C and 2.4 mm respectively.The soil was a rather deep (60±90 cm) slightly
acidic (pH � 6.1), sandy loam (17% clay, 37%silt, 46% sand). A moderate humus content(1.6% organic C) and relatively high silt content
made it susceptible to compaction. Crop species,weed control procedures and dates of treatmentsare summarized in Table 1.
The experiment was conducted from 1989 to
1994. Before the experiment, the ®eld was undera grass/clover pasture for the previous 2 years.The ®eld was divided into three unreplicated
plots, each 50 m long and 18 m wide. Each plotwas managed with one of three di�erent weedcontrol systems: (1) chemical weed control (CH),
(2) integrated weed control (IN) using mechani-cal and chemical techniques or (3) mechanicalweed control (ME).
Every year, the three plots were ploughed to adepth of 20 cm. The seedbed was prepared usinga p.t.o.-driven rotary cultivator at a speed of4 km h)1. Mineral fertilizer and organic manure
were spread each year according to commonpractices (Ryser et al., 1994). After harvest thesoil was tilled to a depth of 10±15 cm at a
speed of 5 km h)1. A rotary harrow linked tothe stubble cultivator crushed clods up to a
Table 1. Weed control strategies and dates of treatments for the di�erent ®eld plots
Culture Date Weeding
Chemical control (CH)Winter wheat 16/3/90 1386 g a.i. ha)1 isoproturon + 14 g a.i. ha)1 triasulfuronMaize 05/6/91 1500 g a.i. ha)1 atrazineWinter wheat 11/4/92 6 g a.i. ha)1 metsulfuron-methyl + 750 g a.i. ha)1 isoproturonWinter barley 15/3/93 5 g a.i. ha)1 metsulfuron-methyl + 750 g a.i. ha)1 isoproturonPasture 06/8/93 1264 g a.i. ha)1 MCPB
Integrated control (IN)Winter wheat 16/3/90 Spring-toothed weeder
26/4/90 695 g a.i. ha)1 MCPA + 112.5 g a.i. ha)1 ¯uroxypyrMaize 05/6/91 Inter-row hoeing roller + 1500 g a.i. ha)1 atrazine on rowsWinter wheat 11/4/92 6 g a.i. ha)1 metsulfuron-methyl + 750 g a.i. ha)1 isoproturonWinter barley 15/3/93 5 g a.i. ha)1 metsulfuron-methyl + 750 g a.i. ha)1 isoproturonPasture 06/8/93 1264 g a.i. ha)1 MCPB
Mechanical control (ME)Winter wheat 16/3/90 Spring-toothed weederMaize 05/6/91 Inter-row hoeing roller
01/7/91 Inter-row hoeing rollerWinter wheat 11/4/92 Spring-toothed weeder
24/4/92 Spring-toothed weederWinter barley 15/3/93 Spring-toothed weeder (2´)*
18/3/93 Spring-toothed weeder (2´)*Pasture 08/8/93 Cutting to clean (10 cm)
*Two passes in the opposite direction.Isoproturon/triasulfuron = Graminon Forte, Ciba, WG, 49.5%/O.5%; atrazine = Atrazins, Siegfried, WP, 50%; Metsulfron-methyl = Ally, Maag, WG, 20%; isoproturon = Arelons, Siegfried, SC, 50%; MCPB = MCPB, PluÈ sstaufer; SL, 31.6%; MCPA/Fluroxypyr = Apell, Maag, EC, 27.8%.
96 J.-P. Mayor and F. Dessaint
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
maximum depth of 10 cm and levelled the soil
surface. For winter wheat (Triticum aestivum L.),mechanical weed control was carried out at aspeed of 6.5 km h)1 with a spring-toothed
weeder at the main stage of tillering (DecimalCode Z23; Zadoks et al., 1974) and with herbi-cides at around the start of stem extension (Z30±31). For winter barley (Hordeum vulgare L.),
both chemical and mechanical weed control werecarried out at the end of tillering. For maize (Zeamays L., herbicide was applied at the three-leaf
growth stage (Z25) and mechanical treatmentswere made at a speed of 5 km h)1 with an inter-row hoeing roller at the three- and seven-leaf
growth stages of the crop.
Seedbank study
An area of 15 m ´ 15 m in each plot was sam-
pled each year from 1989 to 1994. A total of 100soil cores per plot (5 cm diameter, 20 cm depth)were systematically taken in autumn a few weeksafter ploughing, on the intersections of a
1.5 m ´ 1.5 m grid throughout the samplingareas. In 1993, soil cores were collected beforethe pasture was sown.
Soil cores were stored at 4 °C in darkness,before being washed separately through sieves of4 mm and 0.25 mm. The residue was spread on a
TergalÒ cloth which was placed on a sterilizedmould contained in 12.5 cm ´ 10 cm trays.These were then placed for 4±5 weeks in a
growth chamber (14 h light at 25 °C and 10 hdark at 19 °C). During this period, each seedlingwas identi®ed and counted. At the end of theperiod, all trays were placed for 1 month at 4 °Cin darkness for vernalization. Finally, germina-tion was again monitored for a period of 4 weeksin the conditions described previously. Nomen-
clature of species follows Kergue len (1993).
Data analysis
Species richness for the soil seedbank commu-nities was calculated for each plot and year. The
species richness (S) was estimated by the jack-knife procedure (Heltshe & Forrester, 1983)using the formula:
S � S0 � nÿ 1
n
� �k
where n is the number of samples, k the numberof species present in only one sample (``unique''
species) and S0 the observed total species num-
ber. The variance of S was calculated as:
var�S� � nÿ 1
n
� � XS0j�0
j2fj ÿ k2
n
!
where fj is the number of samples containing j`unique' species (Heltshe & Forrester, 1983).Palmer (1990) showed that this index is more
accurate and less biased than S0. Statisticalcomparisons of species richness SÃ employedjack-knife con®dence intervals (CIs) calculated
for each year and plot. Two estimates of S wereregarded as statistically di�erent at the 95%signi®cance level if their CIs did not overlap.
Seed density data were used to compare total
community densities, to calculate diversity indi-ces and to perform multivariate analysis. Shan-non's diversity index (H¢) was calculated for each
year and each plot using the formula:
H 0 � ÿX�ni=N� log2�ni=N�
where ni is the number of seeds of species i and Nis the total number of seeds.
Rank:frequency diagrams are another visualway to assess community structure (Frontier,1985). Relative abundance values of species
within each plot were plotted against speciesrank to obtain the rank:frequency diagrams;both axes were log-transformed. The purpose of
such curves is to extract information on thedominance pattern within a plot, without re-ducing that information to a single summarystatistic, such as a diversity index. They form a
class of techniques which can regarded as inter-mediate between univariate summaries and a fullmultivariate analysis of the species/plots matrix.
Correspondence analysis (CA) was used toanalyse variations in species composition betweenplots and years. CA (Greenacre, 1984) is an ordi-
nation technique that has been found to be usefulfor describing changes in species composition ofcommunities over time (Dessaint et al., 1990).
Most of the measures used in this study aredescriptive and, because there was no replica-tion, the analyses are hypothesis-generating only.Our objective was to discover possible aspects of
the relationship between weed control strategiesand weed seedbank composition and to deter-mine whether important characteristics are af-
fected by the treatments.
Weed seedbank diversity 97
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
Results
Species richness
During the 6 years of the study, seeds of 56 weedspecies were found over the whole experiment.Among these, 30 species were common to the
three plots whereas 17 species were restricted toonly one plot. Thus, 39 species were recorded inthe CH plot, 42 appeared in the IN plot and 44
were found in the ME plot (Table 2).Species richness of the seedbank varied as a
function of the plot and year, with an observed
annual number of species (S0) that ranged from18 to 32, and an estimated species richness (S)that ranged from 28 to 38.9 (Table 3). Shannon's
diversity index (H¢) ranged from 1.63 to 2.94(Table 4).
In 1989, at the beginning of the study, 25species were found in the CH plot, 27 in the IN
plot and 21 in the ME plot, but there were nosigni®cant di�erences of species richness (S)among the plots (Table 3). Over the time, the
CH plot showed a more or less constant numberof species with no signi®cant change in speciesrichness. There was more variability in species
number on the IN and ME plots, with signi®cantchanges in species richness. On the IN plot, thespecies richness (S) was less in 1992 than in 1989,1993 and 1994. On the ME plot, species richness
(S) di�ered with year with no temporal trend(Table 3). Signi®cant di�erences occurred in1990 between the ME plot and the two other
plots, in both 1992 and 1993 between IN and CHplots and in 1992 between ME and CH plots.
The soil seedbank density ranged from around
2800 to around 43 300 seeds m)2 (Table 4). In1989, the three plots had a seed density that ran-ged from 5700 to 8700 seeds m)2. Between 1989
and 1993, the seed density decreased steadily onthe CH plot by a factor of 2 (from 5700 to 2800seeds m)2) and then increased to 12 000 seeds m)2
in 1994 (Table 4). The large increase in seed
density was mainly because of Capsella bursa-pastoris (L.)Medik., which representedmore than50% of seeds found. On the IN plot the seedbank
density increased more or less with a peak in 1994(from 8700 to 14 900 seeds m)2) whereas on theME plot the density of seeds increased strongly,
by a factor of 6 (from 6900 to 43 300 seeds m)2).Increases in seed density were always of greatermagnitude (greater than 50%) than decreases(less than 25%) on the three plots (Table 4).
Table 2. Occurrence of weeds species (number of years inwhich species was present)
Weed CH IN ME Total
Common speciesArabidopsis thaliana (L.) Heynh. 6 6 6 18Capsella bursa-pastoris (L.) Medik. 6 6 6 18Cerastium glomeratum Thuill. 6 6 6 18Chenopodium album L. 6 6 6 18Chenopodium polyspermum L. 6 6 6 18Plantago major L. 6 6 6 18Poa annua L. 6 6 6 18Rumex obtusifolius L. 6 6 6 18Stellaria media (L.) Vill. 6 6 6 18Trifolium repens L. 6 6 6 18Apera spica-venti (L.) P. Beauv. 5 6 6 17Lamium purpureum L. 4 6 6 16Juncus bufonius L. 6 5 4 15Viola arvensis Murray 4 4 6 14Amaranthus retro¯exus L. 5 4 4 13Veronica persica Poiret 3 6 4 13Veronica serpyllifolia L. 4 6 3 13Alopecurus myosuroides Hudson 5 3 4 12Sagina procumbens L. 3 3 5 11Epilobium tetragonum L. 3 4 4 11Sonchus spp. 3 3 3 9Kickxia spuria (L.) Dumort. 4 2 3 9Hypericum perforatum L. 3 2 3 8Polygonum aviculare L. 1 4 3 8Poa trivialis L. 3 2 2 7Matricaria recutita L. 1 3 2 6Senecio vulgaris L. 1 2 2 5Taraxacum o�cinale Weber 2 1 2 5Trifolium alexandrinum L. 2 1 1 4Poa pratensis L. 1 1 1 3
Uncommon speciesConyza canadensis (L.) Cronq. 4 2 ± 6Geranium molle L. 3 ± 2 5Anagallis arvensis L. 1 3 ± 4Sonchus arvensis L. 2 2 ± 4Papaver rhoeas L. ± 2 1 3Polygonum persicaria L. ± 1 2 3Cardamine hirsuta L. 1 ± 1 2Centaurium erythraea Rafn 1 1 ± 2Veronica arvensis L. ± 1 1 2*Chaenorrhinum minus (L.) Lange 1 ± ± 1*Lepidium ruderale L. 1 ± ± 1*Myosotis arvensis Hill 1 ± ± 1Veronica hederifolia L. ± 2 ± 2*Atriplex patula L. ± 1 ± 1*Silene pratensis L. ± 1 ± 1*Solanum nigrum L. ± 1 ± 1*Urtica dioica L. ± 1 ± 1Arenaria serpyllifolia L. ± ± 3 3*Achillea millefolium L. ± ± 1 1Aethusa cynapium L. ± ± 1 1Digitaria sanguinalis (L.) Scop. ± ± 1 1Echinochloa crus-galli (L.) P. Beauv. ± ± 1 1Euphorbia helioscopia L. ± ± 1 1*Filago spp. ± ± 1 1*Mercurialis annua L. ± ± 1 1Oxalis fontana Bunge ± ± 1 1
*Species found in only one sample. CH, chemical control; INintegrated control; ME, mechanical control.
98 J.-P. Mayor and F. Dessaint
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
Structure of weed community
The structure of the weed community was
dominated by a few species that had high relativeabundance values, whereas most of the specieswere of low abundance. Depending on the year
and the plot, between two and six species rep-resented more than 90% of the total seedbank.
Dominance:diversity curves based on seed
density are shown in Fig. 1. All curves weremore or less convex and parallel in the threeplots, with the same group of abundant species.
Six species were regularly the most abundant:Capsella bursa-pastoris, Poa annua L., Stellariamedia (L.) Vill., Chenopodium album L., Cheno-podium polyspermum L. and Apera spica-venti
(L.) P. Beauv. Their combined abundance ran-ged from 78% to 98% of the seedbank (Table 5).
Although the most abundant species were thesame on the three plots, the ranking of thesespecies changed over the time. On the IN plot,P. annua was replaced as the most abundant
successively by C. album and by C. bursa-pas-toris. On the ME plot, P. annua was replaced asthe most abundant by C. bursa-pastoris.
Species composition
Correspondence analysis was performed on seeddensity for species that occurred in more than 15cores at any time during the study. The ®rst three
axes accounted for about 74.1% of the totalvariation of the data (35.3%, 27.2% and 11.6%).Figures 2 and 3 show the changes in the threeplots over time with respect to the ®rst three
principal axes.
Table 4. Estimated densities of germinated seed (m)2), yearly change in density (%) and diversity (H¢) in the three experimentalplots between 1989 and 1994
Year
1989 1990 1991 1992 1993 1994
Chemical control (CH)Density 5663 5653 4242 3183 2765 12 029Yearly changes (%) ± 0 )25 )25 )13 335Diversity H¢ 2.55 2.68 2.86 2.94 2.49 1.85
Integrated control (IN)Density 8678 8520 14 087 12 151 9696 14 917Yearly changes (%) ± )2 65 )14 )20 54Diversity H¢ 2.47 2.70 2.52 2.42 2.78 2.40
Mechanical control (ME)Density 6895 11 530 21 757 16 348 28 596 43 274Yearly changes (%) ± 67 89 )25 75 51Diversity H¢ 1.83 2.30 2.47 2.20 1.93 1.63
Table 3. Species richness of the seedbank over time in the three experimental plots
Year
1989 1990 1991 1992 1993 1994
Chemical control (CH)S0 25 23 22 24 20 24S 31.9 30.9 27.0 34.9 25.0 30.995% CI 26.9±37.0 25.6±36.3 22.6±31.3 28.7±41.1 20.6±29.3 25.9±36.0
Integrated control (IN)S0 27 24 23 18 29 25S 34.9 29.9 28.0 23.0 36.9 35.995% CI 28.9±41.0 25.3±34.6 23.6±32.3 18.6±27.3 31.6±42.3 29.1±42.7
Mechanical control (ME)S0 21 20 27 20 26 32S 26.9 22.0 34.9 23.0 32.9 38.995% CI 22.3±31.6 19.2±24.7 28.9±41.0 19.6±26.3 27.9±38.0 33.9±44.0
S0, Number of observed species; S, jack-knife estimator of the species richness; 95% CI, 95% con®dence interval.
Weed seedbank diversity 99
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
100 J.-P. Mayor and F. Dessaint
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
There was a good separation between the
three plots over time. The ®rst axis revealed amarked separation between the early years(1989, 1990) on the right, and the later years
(1991±94) on the left for the IN plot and the MEplot (Fig. 2). On the second axis, the IN plot waswell separated from the ME plot (Fig. 2) in the
later years (1991±94) of the study. On the thirdaxis, both the IN and ME plots were separatedfrom the CH plot (Fig. 3) during the same
period.The species most closely associated with the
early years was P. annua (Fig. 2). The soil seed-bank of the later years was characterized mainly
by C. bursa-pastoris on the ME plot and C. al-bum on the IN plot. The CH plot di�ered fromthe other plots primarily as a result of Amaran-
thus retro¯exus L. density (Fig. 3). There werefewer changes in species composition in the CHplot than in the other plots.
Discussion
In small grain crops grown in temperate coun-tries, up to 30 species of weeds are generally
present, but only about six are well represented(Barralis & Chadoeuf, 1987; Mayor & Maillard,1995). In this study, two classes of weeds were
recognized on a basis of species occurrence(Table 2): species common to all plots (30 specieswith an occurrence of at least one in each plot)
and uncommon species (26 species with a lack ofoccurrence in at least one plot). Among thecommon species, six species are clearly the mostabundant (Table 5). They represent 88% of the
total amount of seeds in the CH plot, 90% in theIN plot and 96% in the ME plot throughout the6 years of the study. A. retro¯exus had an oc-
currence of only 13 (Table 2), but in 1991, 1992and 1993 represented 5.4%, 12.9% and 6.1%,respectively, of the total amount of seeds in the
CH plot. In the IN and ME plots its percentagenever exceeded 0.5% of the total amount ofseeds.
C. bursa-pastoris was susceptible to most of
the herbicides, as is clearly shown in CH and INplots. This weed was the most abundant in theseedbank of the ME plot (without herbicide) and
Fig. 1. Rank:frequency diagram of the weed seedbank for thethree-weed control strategies: ÐsÐ, Chemical weed control(CH); ÐdÐ, integrated weed control (IN); ÐhÐ, mechani-cal weed control (ME).
Table 5. Relative abundance of the dominant species in the three experimental plots between 1989 and 1994
Year
Species 1989 1990 1991 1992 1993 1994
Chemical control (CH)Poa annua 35.3 37.5 29.9 24.5 46.4 13.7Capsella bursa-pastoris 31.9 24.1 25.6 27.5 22.3 53.7Stellaria media 14.1 11.0 16.7 13.9 8.5 26.3Chenopodium album 4.9 8.5 4.6 4.3 4.8 0.9Chenopodium polyspermum 3.0 2.2 3.2 3.2 4.2 0.8Apera spica-venti 2.0 7.0 6.1 5.4 0 0.3Total 91.2 90.3 86.2 78.9 86.2 95.7
Integrated control (IN)Poa annua 35.9 36.3 10.3 7.8 17.6 12.6Capsella bursa-pastoris 33.3 25.0 24.2 29.0 23.2 45.9Stellaria media 10.9 9.1 17.9 4.9 5.4 8.0Chenopodium polyspermum 3.5 2.6 4.8 8.5 9.3 7.3Chenopodium album 1.9 1.1 34.6 39.0 31.7 18.8Apera spica-venti 0.5 7.9 1.6 5.6 4.6 1.9Total 85.9 82.1 93.5 94.6 91.8 94.6
Mechanical control (ME)Poa annua 63.8 43.1 8.9 6.7 5.9 11.6Capsella bursa-pastoris 16.4 8.0 33.4 51.9 60.6 71.5Stellaria media 9.9 24.1 25.8 10.0 17.0 4.2Chenopodium polyspermum 1.7 0.9 0.6 14.2 6.6 5.6Chenopodium album 0.5 0.6 16.6 10.7 4.4 2.8Apera spica-venti 0.4 17.0 1.0 4.8 3.3 0.5Total 92.8 93.7 96.3 98.2 97.8 96.2
b
Weed seedbank diversity 101
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
increased from 1990 to 1994. This increase wascorrelated with a decrease in the relative abun-dance of P. annua. As C. bursa-pastoris emergesthroughout the year, it can rapidly take advan-
tage of poorly competitive crops in the absenceof herbicide pressure and thus increase its seed-bank. The large increase in seed density of all
three plots in 1994, when only MCPB was ap-plied to the CH and IN plots to control Rumexobtusifolius L. in the pasture, was mainly caused
by C. bursa-pastoris, illustrating the poor e�cacyof MCPB on this weed.
A chemical strategy alone was insu�cient to
control P. annua. However, a chemical combinedwith mechanical weed control had a better im-pact on this weed. Mayor et al. (1994) showedthat about 15% of the seedbank of the mono-
cotyledonous weeds (A. spica-venti + P. annua)emerge in winter barley. Thus, the presence ofP. annua in a small-grain-dominant rotation wasnot surprising. This has also been observed by
Roberts & Neilson (1981).C. album and C. polyspermum were better
controlled in the CH than in the IN and ME
plots, which allowed these species to increase thesize of their seedbank in these plots, particularlyin maize (1991). As shown by Ogg & Dawson
(1984), C. album begins to emerge towards theend of March and continues to emerge untilSeptember. Disturbance of the soil increases the
overall emergence of C. album, as germination ispossible only in the presence of light and whenthe seeds are buried less than 2 cm deep. In ad-dition, each time the soil is disturbed, its nitrate
Fig. 2. Plot of the two ®rst axes of the correspondence analysis of weed seedbank data: chemical (CH), integrated (IN) andmechanical (ME) weed control plots. Only species and plots well represented by axes are labelled.
102 J.-P. Mayor and F. Dessaint
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
content is reactivated, which exhibits an increasein C. album emergence (Roberts & Potter, 1980;Ogg & Dawson, 1984; Yenish et al., 1992). Withmore than 1000 seeds m)2 of C. album, non-
chemical control measures alone could not pre-vent small grain crop yield losses. Forcella et al.(1993) have also shown that against C. album in
maize and in soyabean, non-chemical controlmeasures alone cannot prevent large crop yieldlosses.
S. media became the most abundant speciesafter C. bursa-pastoris in the CH plot in the lastyear. Both species can be very competitive when
other dicotyledonous species are poorly repre-sented. The seedbank of A. retro¯exus was par-ticularly abundant in the CH plot in 1992 (13%of the total seedbank). It was poorly controlled
by atrazine in the maize in 1991 (more than 5%),and probably not e�ectively controlled by thechemicals in barley in 1992. Considering that thisweed emerges from mid-April to September (Ogg
& Dawson, 1984), it is not surprising that itsseedbank size increased in a signi®cant way afteronly one, quite late, atrazine treatment. This
herbicide does not control A. retro¯exus at agrowth stage later than six leaves. Two monthsafter the atrazine treatment, this herbicide would
have lost its e�ectiveness. Thus, mechanicalprocedures appear to be more able to controlA. retro¯exus than a chemical treatment alone.
Even though the weed control strategy was thesame in the CH plot as in the IN plot in 1992, thehigh percentage of A. retro¯exus in the seedbankin the CH plot may be due to the low competitive
Fig. 3. Plot of axes 2 and 3 of the correspondence analysis of weed seedbank data: chemical (CH), integrated (IN) and mechanical(ME) weed control plots. Only species and plots well represented by axes are labelled.
Weed seedbank diversity 103
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
pressure of the other weed species. In the IN plot
there were more than four times as many seedsm)2 as in the CH plot.
The type of weed management not only in-
¯uences the size of the seedbank, but also itscomposition and the relative abundance of thedi�erent species. In an integrated productionsystem and in organic farming crop yields may
decrease dramatically, which may a�ect thepro®tability of the enterprise. Moreover, pro-ducers may not expect to get higher prices for
small grain crops and maize grown withoutchemicals, as is the case in vegetables and fruits.For example, in 1993 the yield of winter barley
was 7% lower in the IN plot and 17% lower in theME plot than the CH plot (Mayor et al., 1994).
Cereal crops are very susceptible to weedcompetition during the growth stage of tillering
(Ammon & Irla, 1996). Consequently, early weedcontrol is essential. After the tillering growthstages of small grain crops, any further weed
control is used to prevent weeds from ¯oweringand hence producing seeds that will enrich theseedbank. Perennial weeds such as Cirsium ar-
vense (L.) Scop., R. obtusifolius and Convolvulusarvensis L. will also be controlled. For an earlyweed control in winter cereals in Switzerland,
farmers must act in October, March or April.During these months, the ground generally driesslowly after rainfall. Thus, the e�ectiveness ofmechanical weeding procedures may be uncer-
tain. For instance, in 1992 and 1993 we wereunable to use spring-toothed weeders. Moreover,in March most of the weeds are already too well
developed to be controlled e�ciently with aspring-toothed weeder in crops that have beensown at the beginning of October. Considering
that the e�ciency of a spring-toothed weeder isaround 50%, it is necessary to reinforce me-chanical weed control with a herbicide treat-ment. When about 30 weeds m)2 (all species) are
still present after weeding, yield may not bea�ected (Ammon & Irla, 1996). However, theincrease in the seedbank size can be such that in
the following crop it becomes absolutely neces-sary to use a herbicide. For these reasons, we donot suggest that farmers should buy expensive
mechanical tools as a ®rst priority, even if theywant to decrease the amount of herbicide ap-plied. A more appropriate chemical reduction
strategy may be to control weeds with a lowdose of herbicide applied at the three-leaf totillering growth stages of the cereals when
weather conditions are suitable. In maize, it is
possible to control weeds mechanically in be-tween the rows at the three- to four-leaf growthstages of the crop, coupled with a herbicide
treatment in the rows. An inter-row hoeingroller is well-suited for such weed control inmaize. Weeds that emerge after the six- to eight-leaf growth stages of maize do not a�ect its yield
(Hall et al., 1992).
Acknowledgements
The authors wish to thank Mr J. Emmenegger
from the Agricultural Institute of Grangeneuve(Fribourg) for his help in the management of thetrial, Mr G. Mermillod for excellent technicalassistance and two anonymous reviewers for
constructive comments on earlier versions of themanuscript.
References
AMMON HU & IRLA E (1996) UnkrautbekaÈ mpfung in Get-reide. In: UnkrautbekaÈmpfung im Acker und Futterbau.Landwitscha¯iche Lehrmittelzentrale, Zollikofen, MerkurDruck.
BARRALIS G & CHADOEUF R (1987) Potentiel semencier desterres arables. Weed Research 27, 417±24.
CLEMENTS DR, WEISE SF & SWANTON CJ (1994) Integratedweed management and weed species diversity. Phytopro-tection 75, 1±18.
DESSAINT F, CHADOEUF R & BARRALIS G (1990) Etude de ladynamique d'une communaute adventice. II. In¯uence aÁlong terme des techniques culturales sur le potentiel se-mencier. Weed Research 30, 297±306.
FORCELLA F, ERADAT-OSKOUI K & WAGNER SW (1993) Ap-plication of weed seedbank ecology to low-input cropmanagement. Ecological Applications 3, 74±83.
FRONTIER S (1985) Diversity and structure in aquatic ecosys-tems. Oceanography and Marine Biological Annual Review23, 253±312.
GREENACRE MJ (1984) Theory and Applications of Corre-spondence Analysis. Academic Press, London.
HALL MR, SWANTON CJ & ANDERSON GW (1992) The criticalperiod of weed control in grain corn (Zea mays). WeedScience 40, 441±7.
HELTSHE JF & FORRESTER NE (1983) Estimating species rich-ness using the jackknife procedure. Biometrics 39, 1±11.
KERGUE LEN M (1993) Index synonymique de la ¯ore de France.Muse um National d'Histoire Naturelle, Paris.
LAMBELET-HAUETER C (1985) Comparaisons entre ¯ore re elleet ¯ore potentielle en grandes cultures de la re gion gene-voise. Candollea 40, 99±107.
LAMBELET-HAUETER C (1986) Analyse de la ¯ore potentielle,en relation avec la ¯ore re elle, en grandes cultures de lare gion genevoise. Candollea 41, 299±323.
MAYOR JP & MAILLARD A (1995) Re sultats d'un essai deculture sans labour depuis plus de 20 ans aÁ Changins.Revue Suisse d'Agriculture 27, 229±36.
MAYOR JP, MERMILLOD G & EMMENEGGER J (1994) E�et desme thodes de de sherbage sur le stock semencier du sol etsur la ¯ore avdventice. In: Proceedings 5th EWRS Medi-
104 J.-P. Mayor and F. Dessaint
Ó 1998 European Weed Research Society, Weed Research 38, 95±105
terranean Symposium, Weed control in sustainable agricul-ture in the Mediterranean area, Perugia, 35±42.
OGG AGJ & DAWSON JH (1984) Time of emergence of eightweed species. Weed Science 32, 327±35.
PALMER MW (1990) The estimation of species richness byextrapolation. Ecology 71, 1195±8.
ROBERTS HA & POTTER ME (1980) Emergence patterns ofweed seedlings in relation to cultivation and rainfall. WeedResearch 20, 377±86.
ROBERTS HA & NEILSON JE (1981) Changes in the soil seedbank of four long-term crop/herbicide experiments. Jour-nal of Applied Ecology 18, 661±8.
RYSER JP, WALTER U & MENZI H (1994) Donne es de basepour la fumure des grandes cultures et des herbages. RevueSuisse d'Agriculture 26, 193±242.
SWANTON CJ & WEISE SF (1991) Integrated weed manage-ment: the rationale and approach. Weed Technology 5,657±63.
THORNTON PK, FAWCETT RH, DENT JB & PERKINS TJ (1990)Spatial weed distribution and economic thresholds forweed control. Crop Protection 9, 337±42.
YENISH JP, DOLL JD & BUHLER DD (1992) E�ects of tillage onvertical distribution and viability of weed seed in soil.Weed Science 40, 429±33.
ZADOKS JC, CHANG TT & KONZAK CF (1974) A decimal codefor the growth stages of cereals. Weed Research 14,415±21.
Weed seedbank diversity 105
Ó 1998 European Weed Research Society, Weed Research 38, 95±105