1. Surveys have shown that there has been a dramatic decrease in the weed flora of fields under rotational cultivation during the last 30 years. This trend has been particularly noticeable in winter cereals, a crop of increasing importance in the landscape.
2. The weed flora of spring and winter cereals was compared in 19 unsprayed fields during a 5-year study to test the hypothesis that cereal type exerts no effect on the flora or on the absolute and relative abundance of single species.
3. Plant and species densities, and accumulated species richness, were lower in winter than in spring cereals.
4. Floristic similarity was greater among spring cereal fields and between spring and winter cereals within the same fields than among winter cereal fields.
5. Species that occurred with unequal density in spring and winter cereals occurred at higher densities in the spring cereals; these species germinated mainly in the spring. However, for a few species the relative plant abundance was highest in winter cereals; these species were able to germinate both in the spring and autumn.
6. Some species – on the relative scale – occurred indifferently of season of sowing; all but one of these species were able to germinate both in the spring and autumn.
7. Plant species and taxa that are important food resources for arthropod herbivores occurred at greater densities in spring than in winter cereals and, in addition, occurred with the highest relative abundance in spring cereals.
8. Change in land use from spring to winter cereals involves not only an immediate reduction of more than 25% in the density of plants and species, but also a change and increased uncertainty in the composition of the weed flora. These findings may have serious implications for the ecology of wildlife in the agricultural landscape.
During the last half century agriculture has intensified, especially in northern European countries, and there has consequently been a change in the weed flora of arable fields (Fryer & Chancellor 1970; Reuss 1981; Erviö & Salonen 1987). This has indirectly affected the herbivores associated with the plants (Green 1984; Rands 1985). Furthermore, the area and types of crop grown are changing more quickly now than ever before, in response to fluctuations in market prices.
In Denmark, surveys of the potential weed flora, i.e. pre-spraying (Andreasen, Stryhn & Streibig 1996) and seed bank (Jensen & Kjellsson 1995), in rotational arable fields have shown that both plant and species densities have declined and that the frequency of some species has declined more than others. Of the 67 species most common in 1967–70, the number of species per 0·1 m2 has reduced by about 60% within 20 years and the reduction has been greater in winter wheat than in spring barley (Andreasen, Stryhn & Streibig 1996). Jensen & Kjellsson (1995) found that the seed bank had halved from 1964 to 1989, and out of the 27 most frequent species in the seed bank 10 species had decreased in abundance. A comparison of the flora in conventionally farmed cereal fields, pre-spraying, and cereal fields under organic management for more than 10 years has shown that the species and plant densities of the latter were approximately twice those of conventionally farmed fields (Hald & Reddersen 1990).
Hitherto, the results from comparisons of the weed flora in spring and winter cereals have been based on surveying the flora at different sites, and could therefore be biased by different soil types (Andreasen, Streibig & Haas 1991), different climates and other confounding factors such as distribution of farm types. The frequency of weed species may also change from year to year (Andreasen 1990). Unbiased data are therefore needed to test the suggested trends.
The aim of this study was to test the hypothesis that there are no differences in the weed flora of unsprayed spring and winter cereals, and to make an unbiased quantification of any differences. The weed flora of spring and winter cereals was compared in a 5-year study in unsprayed, permanent plots distributed throughout Denmark, which ensured a wide range of soil types, climates and farm types. Comparisons were made of the floristic composition, plant and species density, and occurrence of species. The results are discussed in relation to season of germination of the species, and the impact of an increase in area of winter cereals relative to spring cereals on the ecology of wildlife in the agricultural landscape.
Materials and methods
Nomenclature follows Hansen (1981). Vegetation surveys were carried out between 1988 and 1992 by Hald et al. (1994) in 6-m wide crop margins in 22 experimental fields distributed all over Denmark. Within each experimental field there were permanently unsprayed and sprayed plots laid out in four blocks; the farmer determined the crop rotation. In this paper only results from the unsprayed plots are included. The weed flora in each 6 × 20-m experimental plot was surveyed every year in spring in 10 permanently located circles of 0·1 m2. The species were identified (Hansen 1981; Haas & Laursen 1986) and the number of individuals of each species was counted. Some of the plants could only be identified to genus. In the present paper the accumulated species and plant density in the circles of the four permanently unsprayed plots (in total 4 m2) in one cereal field in one year is equal to one sample. The number of experimental fields included in this paper was reduced to 19, as each field was required to support at least one spring and one winter cereal sample during the 5-year period, resulting in a total of 72 samples. A sample in spring cereal (barley) and a sample in winter cereal (winter wheat or winter barley) in another year within the same field was designated a fixed sample pair.
The analyses were performed using two approaches: species density (floristic) and plant density (abundance of single species and higher taxa).
To test the hypothesis that the species density and floristic composition did not differ between the two crop types, fixed sample pairs were analysed using species presence–absence information. Analyses involved a paired test of the number of species and plants in samples, calculation of the Sørensen Similarity Index (MVSP 1993), and an indirect gradient analysis with non-linear rescaling of axes (DCA) (canoco; ter Braak 1991). Sørensen Similarity Indices were compared between crop types (between and within fields) and within crop types (between fields). The detrended correspondence analysis was performed using the default options, and the scores of axis 1 and axis 2 were interpreted in relation to species density of the samples. Only species that occurred with two or more specimens in the data were included in the presence–absence data, i.e. 114 species.
The fixed sample pairs were selected among the 72 samples with the following restrictions. (i) Each field should be represented with n sample pairs, where n = the number of samples of the crop type with the fewest samples. The partner within a pair was selected randomly within each field. (ii) Samples should be included in one sample pair only. (iii) The number of samples from each crop type within a year should not differ by more than two samples. As a result, a total of 22 sample pairs could be included.
Abundance of single species and taxa
The hypothesis that each species occurred independently of crop type was analysed with a van Elteren test (stratified Wilcoxon Mann–Whitney U-test; Lehmann 1975), with field as stratum. The empirical distribution for the test-value was found by performing 999 randomizations. This method allows inclusion of unequal numbers of each crop type within a stratum and thereby extension of the number of samples in the data set compared to the sample numbers of the fixed pairs data set. Only species that occurred in two or more years in the same field were included in the test, in total 55 species. As some of the wild plant genera and families are more important to arthropod herbivores than others (Hald et al. 1994), occurrence of the plant genera and families mentioned by Hald et al. (1994) was also tested. The analyses were performed using the plant density of the species as a variable. However, to reduce the influence from different plant densities among the crop types, the importance value of a species was also represented by its relative abundance, calculated as the species’ share of the total number of plants in the sample.
WP language English (U.S.)A non-significant van Elteren P-value (P > 0·05) might imply that the species occurs independently of crop type. Therefore, the power of the van Elteren test was calculated for each species to investigate which species might have a relative abundance independent of crop type. This was done by calculating the empirical power function based on 1000 randomizations given that the ratio of the mean relative abundance of the species in the two crop types is 5 [F(5x)] (for empirical power functions, see WP language English (U.K.)WP language English (U.S.)Hirsch & Slack 1984). The calculation was based on the observed numbers of occurrence of the species in the two crop-type samples and on the observed mean relative abundance of the species in the crop with the lowest abundance. Thus, F(5x) is the probability of rejecting the hypothesis of equal relative abundances in the two crop types from a simulated experiment with the given ratio of 5 between the relative abundance in spring and winter cereals. A high power value [F(5x) > 0·80] in combination with a non-significant van Elteren-test (P > 0·05) indicates that a difference in observed abundance in winter and spring cereals is not present, i.e. the species occurs independently of crop type.WP language English (U.K.)
The samples included in the van Elteren test were selected with the following restrictions: (i) within years, the number of samples of each crop type should be ≥ the number of samples available for the crop type with the fewest samples; and (ii) samples from the other crop type were sampled randomly within year and the number of samples within year should not differ by more than two samples. As a result, 32 and 30 samples from spring and winter cereals, respectively, were included in the extended data set (Table 1). A comparison of plant and species density in the extended data set used in the van Elteren analysis with the data set of fixed sample pairs showed that an extension had no significant impact on the values of the two variables (Table 2).
Table 1. Number of samples from spring (s) and winter (w) cereals included in the two data sets in total and by year
Table 2. Plant and species density (number per 4 m2) in spring and winter cereals and P-value of comparison between the two data sets (Wilcoxon Mann–Whitney U-test and t-test)
Plant density (median)
Species density (mean)
As plant density in the permanently unsprayed plots generally increased during the experimental period (Hald et al. 1994), the time sequence of spring and winter crop types in the analyses should be balanced. Comparison of the number of experimental years for the samples within each crop type for the two data sets confirmed that the criteria of being balanced was met. Whenever the data were not normally distributed or the variances were not of equal size a non-parametric test was used, and consequently the median was used as a parameter instead of the mean. All tests were two-sided and calculated by means of SAS (SAS 1990) and ASTUTE (ASTUTE 1995).
The mean species density was more than 25% lower in winter than in spring cereals (Table 2; paired t = 5·2, n = 22 pairs, P < 0·0001), and the species density was positively correlated with plant density (Spearman rs = 0·37, n = 44, P = 0·014). Correspondingly, a lower plant density was found in winter cereals than in spring cereals (Table 2; Wilcoxon matched-pairs signed-ranks test z = −3·0, n = 22 pairs, P = 0·002).
The samples from winter cereals showed higher variation in ordination space than the samples from spring cereals, especially along the first axis (6 SD units) (Fig. 1). The samples from spring cereals showed greatest variation along the second axis (4 SD units). The eigenvalues of the first two axes were 0·24 and 0·21, respectively. In all, the two axes explained 19·3% of the total variation. The scores of axis 2 were negatively correlated with species density as a result of the spring cereals (Table 3). Within sample pairs, the median of the pair difference score (spring–winter) was −0·61 for axis 1 and 0·74 for axis 2, i.e. in the opposite direction of the species density vector.
Table 3. Correlation coefficient (Pearson product-moment) and P-values between species density of the samples and sample scores on DCA axis 1 and axis 2, respectively, for the samples in total and within crop type
A total of 114 species was found. In all, 97 were found in spring cereals and 87 species in winter cereals. The lower accumulated species richness in winter compared to spring cereals was also reflected in the distribution of common species. For example, the fifth most widespread species within crop type occurred in 20 spring samples, but only in 15 winter samples. Similarly, the tenth most widespread species occurred in 17 spring and 12 winter samples. In most cases, two species contributed ≥50% of the plants (dominant species) within a sample and consequently the species abundance curves were l-shaped. In most cases one dominant species was shared in sample pairs while the other was different. The 19 species that were registered as dominating (see the Appendix) also varied among fields. The Sørensen Similarity Index between crop types was higher within fields than between fields, and it was higher between spring cereal fields than between winter cereal fields (Table 4).
Table 4. Mean and standard error of the mean (SE) of Sørensen Similarity Index between crop types (between and within fields) and within crop type (between fields). n: number of comparisons
W and S
W and S
W and W
S and S
Abundance of single species and taxa
A list of the species and taxa included in the van Elteren test, the season of germination of the species, the results for each species/taxon of the van Elteren analysis and the power value are listed in the Appendix and summarized in Fig. 2. Nineteen taxa were found significantly (P≤ 0·05) or to have a tendency (0·05 < P≤ 0·10) to occur with different absolute plant density in the two crop types, and all of them occurred with highest density in spring cereals. Similarly, 15 taxa occurred with different relative abundance in the two crop types, including five taxa that occurred with the highest relative abundance in the vegetation in winter cereals. Ten taxa occurred preferentially in spring cereals irrespective of the importance value used. Nine taxa that occurred with highest plant density in spring cereals did not show differences in relative abundance in the two crop types. However, five taxa that did not occur at different absolute plant densities in the two crop types were relatively more important in the vegetation in winter cereals.
A high power value [F(5x) > 0·80] on 10 occasions in which the van Elteren P-value of relative abundance was >0·05 indicated that the taxon occurred independently of crop type. Differences in occurrence of the species as well as indifference towards crop type was found both within more widespread species and within the more sparsely distributed species, and within all taxonomic levels.
Among the tested species with a known season of germination, most (94%) are able to germinate in spring cereals (Sg + SgWg) (see the Appendix), but only 62% are able to germinate in winter cereals (Wg + SgWg). Among the 10 taxa found to occur preferentially in spring cereals irrespective of importance value, nine are known to be or consist of species that only germinate in the spring (Sg). Of the two species occurring with the highest relative abundance in the vegetation in winter cereals and the species occurring independently of crop type, all but one are able to germinate in both spring and autumn (SgWg).
Among the tested plant genera and families, Chenopodiaceae (mostly Chenopodium album), Fabaceae and Polygonaceae were found to occur with higher density in spring cereals, while Veronica spp. and Caryophyllaceae (mostly Stellaria media) occurred with highest relative abundance in the vegetation in winter cereals.
The experimental fields included in the analyses were fairly representative of Danish agriculture in terms of the crops grown (Hald et al. 1994). As the flora was surveyed in the spring, the results for winter cereals included plants that had germinated in the spring as well as those that had germinated in the autumn, while plants that died during winter could not be included. The results thus give an estimate of the effects of cropping spring and winter cereals on the spring and summer flora, which is likely to have most consequence for the abundance of arthropods.
The van Elteren tests based on absolute plant density and on relative abundance were complementary. Thus the two analyses together differentiated between species that occurred in spring cereals, because they only germinate in the spring (10 taxa), and those that occurred in spring cereals due to the generally higher densities in spring cereals (nine taxa). The analysis based on density may be of ecological interest in relation to food chains. However, the analysis unbiased by the discrepancies in total plant density in the two crop types gives most information when the species’ share of the germinated vegetation is the focus of interest.
Unsprayed spring cereals had higher plant densities and species densities than similar winter cereals, confirming the findings of Andreasen, Stryhn & Streibig (1996). They found, among 67 common weed species with ≥50% difference in occurrence between spring barley and winter wheat in 1967–70, that 68% occurred with the highest frequency in spring barley – recalculated from Andreasen, Stryhn & Streibig (1996). In this study, all the taxa with unequal density between the two crop types occurred at highest densities in spring cereals. One of the reasons for the effect of crop type may be that the seed and species pool is much higher for spring than for winter cereals, i.e. over many years the flora has been selected towards the ecological conditions prevailing in a spring-sown crop. Only when relative abundance was considered was it found that a few taxa preferred winter cereals. Furthermore, some of the species generally found to dominate the samples (in most cases two species) showed crop preference. This was also evident from the fact that in a fixed sample pair, i.e. samples from the same physical area, only one common species dominated in both crop types. The relative abundance of some species was independent of cereal type, and all but P. aviculare are able to germinate in both spring and autumn. Thus changes in the relative abundance of these species are unlikely to be found purely in response to the type of cereal grown.
The weed flora associated with spring cereals across the range of experimental fields was similar, as indicated by the high similarity among spring cereal fields and the dispersion of sample points within 4 SD units in ordination space. As the sample area was 4 m2 and similarity was based on the presence–absence of species, similarities were mainly based on non-dominant species. The greater similarity between winter and spring cereals within the same field, than between winter cereals, indicates that the flora in the winter cereals were subsets of the vegetation in the spring field, but a different subset from field to field.
A change in land use from spring to winter cereals is occurring in Denmark and the flora is also changing to include new species (Andreasen, Stryhn & Streibig 1996). The results of the similarity analyses and the greater variation among winter cereal samples in the DCA ordination space may be interpreted as lack of adjustment of the flora to the winter crop, i.e. a new equilibrium with the increased frequency of winter cereal growing in Denmark had not yet been established in 1988–92. It is likely that the flora of the fields will not attain equilibrium in the future as farmers adjust their cropping programmes quickly in response to fluctuations in market prices of the crops.
The results of this study show that the dramatic reduction of the potential weeds in cereals is accompanied by a serious reduction in species diversity (measured as species density) with the switch from spring to winter cropping. In contemporary agriculture most of the cereal fields are sprayed with herbicide, impoverishing species diversity in both crop types. However, in the future we may expect a reduction in pesticide use in conventional farming and an increase in pesticide-free farming, as in organic farming. Therefore, the cropping of relatively more winter cereals may have considerable implications for species diversity and the ecology of the agricultural landscape, and may counteract the beneficial effect on biodiversity of reduced herbicide application. The density of weed species will be reduced by more than 25% in comparison with spring cereals and this reduction will have a particular impact on spring-germinating dicotyledonous species. Not only the present flora, but also the flora in the future will be affected, because the floristic composition of the input into the seed bank will change as well. Species that are very widely distributed in spring cereals and which support the arthropod herbivores may also be reduced in density.
In the analysis of unsprayed crop margins in Denmark, Hald et al. (1994) found certain plant species, especially within the genus Urtica and the families Asteraceae, Brassicaceae, Fabaceae and Polygonaceae, to be most associated with arthropod herbivores, many of which are important food items for birds living and nesting within cereal crops. The above taxa made up 31% of the dicotyledonous plants in their survey. Plant species within the genera Lamium, Myosotis, Veronica and Viola, and the families Caryophyllaceae and Chenopodiaceae, made up 66% of the individuals but had few associated arthropods (Hald et al. 1994). In the present analysis two plant taxa among the first group (Fabaceae, Polygonaceae) occurred at highest densities in spring cereals. In contrast with the vegetation of spring cereals, two of the less important taxa for the arthropod herbivores (Veronica, Caryophyllaceae) were of relative importance in the vegetation of winter cereals. Therefore, as a consequence, the arthropod herbivores in cereal fields – although unsprayed – may be influenced to a highly negative degree by a change from spring to winter cereal growing, not only through lower plant and species densities but also through changes in the association of weeds in the future.
This work was carried out with financial support from The Danish Environmental Protection Agency. I thank H. Pontoppidan for help in the field, C.T. Agger and J. Carstensen for help with the van Elteren test and the power function, G. Cracknell, C. Topping and A.R. Watkinson for linguistic help, B. GaÅrdmand for help with the figures, and C. Andreasen, J. Reddersen, two anonymous referees and A.R. Watkinson for useful comments on a previous version of the paper.
Received 14 January 1997; revision received 27 October 1998
Table 5. Taxa included in the test of equal occurrence in the two cereal crop types. A: All species (main species underlined within pseudo-species). B: Species gathered in genera and families that are discussed in relation to arthropod herbivores (after Hald et al. 1994). The columns show: dominant species; season of germination; test of significance for the hypothesis that plant density of a given taxon is equal in the two crop types (van Elteren P-value); test of significance for the hypothesis that the relative abundance of a given taxon is equal in the two crop types (van Elteren P-value), and the matching power of the van Elteren test given a ratio of 5 between the relative abundances in a randomization experiment; and number of samples within crop type in which the taxon occurred. D: Species that were registered as dominant species in at least one sample, i.e. were among the species that contributed with ≥50% of the plants in the sample. Sg: Spring germination. Wg: Autumn germination. SgWg: Spring and autumn germination. Nomenclature follows Hansen (1981).