Is there a Weed Shift in Roundup Ready Maize? (Poster Presentation at the 4^th International Workshop on PMEM of Genetically Modified Plants, Quedlinburg, Germany, 2010)

Gibt es eine Verschiebung im Unkrautspektrum im Roundup Ready Mais-Anbau? (Poster-Beitrag beim 4. internationalen Workshop zum Anbaubegleitenden Monitoring von gentechnisch veränderten Pflanzen, Quedlinburg, Deutschland 2010)

Material and Methods

The soil was prepared before and during the test period by plowing in Vesbeck and Werne, although the Gerbitz site had been in non-conversion-tillage production for many years. The Roundup Ready maize was planted as usual between April 29 and May 18.

With a total test area of 81 m × 93 m (including a buffer of non-transgenic maize), the plot size was 18 m × 18 m. In our study we compared eight different herbicide treatments (see below), each treatment in one plot. In each of these treatment plots, 4 square sub-plots in a fixed position with an area of 9 m² were systematically installed for the weeds assessments. This test design with relatively large plots, but no real replicates, was designed to ensure accurate positioning and to prevent disruptive peripheral effects, e.g. by the accidental carry-over of soil.

Treatments 6–8 were treated similar to treatment No. 4, but in every 3rd year (2005 and 2008) there was a change with herbicides with different mode of action (Table 2).

According to the specific weed situation, the standard treatment contained the following herbicides in combinations of 2–3 herbicides in tank mixtures: Callisto (100 g L^–1 mesotrione), Cato (250 g kg^–1 rimsulfuron), Gardo Gold (312.4 g L^–1 s-metolachlor and 187.5 g L^–1 terbuthylazine) and Motivell (40 g L^–1 nicosulfuron). The herbicides were applied between May 10 and June 16, i.e. 3–38 days after the maize was planted. The first Roundup Ready treatments were applied simultaneously with the standard treatment; the second application was made after 14 to 40 days, depending on the site and year.

The seedbank, e.g. the number of weed seeds per species, was determined by germination tests in the glasshouse: Each year, shortly after the maize was planted, soil samples were taken from the sides of the above referenced sub-plots. For this purpose, undercuts from a depth of 30 cm respectively 590 cm³ soil volume were combined into a mixed sample. These soil samples were spread in a 2–3 cm thick layer in flat pans over an area of approximately 550 cm². In order to assess as many viable seeds as possible by breaking the dormancy, the soil samples were mixed and spread out multiple times, cooled and dried in between germination timings.

On account of the extremely high spatial variability, the weed dynamic was calculated primarily on the basis of indi­vidual annual differences within the same sub-plot. No standard distribution of the data and no homogeneity of the variances were found, so if possible non-parametric statistical tests like the Kruskal-Wallis test were used to analyse treatment effects. Also descriptive methods like box-whisker-plots and frequency distributions were applied. Statistical calculations and graphic presentations were done using the Statgraphics Plus program (Version 5.1). We put the main focus on the comparison of the local standard (LOCSTD) with the Roundup Ready treatments (1.5 + 1.5, 2 + 2, 3 + 3).

Furthermore treatment effects were assessed by plotting the relative abundance of the weed species against the species list, ranked in order of descending abundance (relative species abundance). This descriptive approach is especially applicable to detect possible shift of rare weed species (Preston, 1948).

Due to the different site conditions mostly all evaluations were done site by site. The evaluation of weed diver­sity was based on the following indicators:

Results

During the complete test period we found 94 weed species in the seedbank at the three sites, but 15 species occurred only once. The most frequent weed species were Chenopodium album, Solanum nigrum and Stellaria media, but only Chenopodium album showed reasonably similar densities at all 3 sites (Table 4). Compared to the extensive weed survey by Mehrtens et al. (2005), the weed communities can be described as typical for maize fields in Germany. However, compared to this study Echinocloa crus-galli and Matricaria chamomilla were found less whereas others species like Gnaphalium uliginosum and Digitaria ischaemum appeared more frequently.

Starting from the beginning in 2003 the number of weed species increased from 37 to 53 at the end of the study in 2009. Of those, 33 species occurred with a frequency of less than 25% and 18 of those were found only on one site in 1–2 years. Alternatively, 14 weed species were found in at least 75% of the study.

Both the spectrum of weed species and the seedbank size were highly location-dependent. Averaged across the sites and treatments, the relative frequency of the following weeds increased within the six years: Amaranthus retro­flexus, Solanum nigrum and Chenopodium album. In some years the soil seedbank increased by a factor of up to 5. There was a significant decrease in the seed number of Capsella bursa-pastoris, Juncus bufonius and Solanum nigrum. Other weeds such as Echinocloa crus-galli, Gnaphalium uliginosum and Galinsoga ciliata reacted weakly and indifferently.

Looking at the complete data set, there was a clear inter­action between site and treatment effects: Especially at Gerbitz, the site with the lowest initial weed diversity both seedbank and species richness increased independently from the herbicide treatment. In contrast to Gerbitz, the weed diversity at Vesbeck declined during the study period (Fig. 1).

Fig. 1. Effects of herbicide treatments on the ratio of (a) seedbank, (b) species richness, (c) Shannon’s diversity (a ratio of > 1 indicates an increase of the parameter from 2003 to 2009), mean values and SEM.

The data reveal that reducing the glyphosate dosage from 2 × 3 L/ha^–1 by 50% will result in a heavy weed buildup. Also the full glyphosate dosage as well as the standard treatments could not avoid additional weed seed production in each year, but finally these treatments obtained the least weed seed densities.

Fig. 4. Proportion of weed species according to their sensivity class (1 = high sensitivity to glyphosate, 6 = tolerant to glyphosate).

Based on the above mentioned general indicators, we could not find clear evidence that the Roundup Ready system resulted in a stronger weed shift than the tested standard herbicide treatments did.

In a next step the ranking of weed species according to their relative abundance was analysed. Especially we are interested in a possible response of very rare as well as very dominant weed species. Fig. 2 shows the relative species abundance for the 3 + 3 L ha^–1 Roundup Ready treatments: By comparing initial and final year there is no evidence that the 3 + 3 treatment resulted in a loss of less frequent species or in a stronger dominance of other weed species.

Fig. 2. Effect of the 3 + 3 L ha^–1 Roundup Ready application on the relative species abundance.

From an agronomical point of view it is of interest if the continuous application of glyphosate will create situations with less but more dominant weed species. Fig. 3 shows the frequency distribution of the dominance indicator (D) across all sites and years (except 2003). Based on this parameter, there was no shift towards to dominant weed species in the 3 + 3 treatment. We actually found that by the full glyphosate dosage the frequency of dominant weed species (D > 0.6) was lower than in all other treatments.

Fig. 3. Frequency of the most predominant weed species (dominance (D), data of all sites, 2004–2009).

Regarding the class of weed species characterised by a low sensivity against glyphosate, there was no shift until 2008, but a different situation in 2009 (Fig. 4). Since the reduced data set in the final year (no data at Werne) this result should not be overestimated. However, at Gerbitz reduced glyphosate rates resulted clearly in a serious build-up of Chenopodium album and Urtica urens. Also the seedbank of Polygonum convolvulus increased to some extend (80 seeds m^–2 in 2009). Other less sensitive species like Menthis arvensis, Convolvulus arvensis and Trifolium repens occurred in low densities without showing any trend.

Based on the initial occurrence (Fig. 5) weed species which where less frequent at the beginning of the study increased by trend in all Roundup Ready plots. Depending on site and year we found 10–23 weed species which were not abundant in 2003 and 2004 (single plant samples not considered). Some of these species achieved rather high seedbank densities at the end, e.g. Echinocloa crus-galli (333 seeds m^–2), Erophila verna (1403 seeds m^–2), Lamium spp. (489 seeds m^–2) and Veronica spp. (223 seeds m^–2). However, due to the high spatial variation it was not possible to distinguish treatment effects on any of these weed species.

Fig. 5. Proportion of weed species according to their initial frequency in 2003 (1 = 0–1.9%, 2 = 2–9.9%, 3 = 10–79%, 4 = 80–89%, 5 = 90–100%).

In a next step we analysed possible treatment effects on those weed species for which herbicide resistance is known. All species were considered with at least three general records by Heap (2010) or at least one record for the HRAC group G (Glycine, including glyphosate). Out of these 24 relevant weed species, Chenopodium album was the only one which increased extra-proportionally. Averaged across the 3 different Roundup Ready treatments and all sites, the density of Chenopodium album increa­sed from 206 to 920 seeds m^–2 during the study period. Even the high rate of 3 + 3 L ha^–1 Roundup Ready could not avoid spreading out of this weed.

Altogether the results of this study are in line with recent studies which have been conducted in North America where glyphosate-tolerant maize is widely grown since more than 10 years. According to a survey by Culpepper (2006) weed shifts have not been observed in maize, but in other GMHT crops in the United States. However, simi­lar to our findings it is also reported that low glyphosate dosages of less than 0.8 kg ha^–1 may cause serious weed problems, especially for Chenopodium album (Wilson et al., 2007). Furthermore, especially in fully glyphosate-tolerant crop rotations spreading of volunteer crops like oilseed rape, soybean or winter wheat have to be con­sidered. Because of the missing residual effect of glyphosate, a higher seed production of late emerging weeds is more likely in GMHT maize rather than in conventional cultivars (Heard et al., 2003). For these reasons American long term experiences have clearly demonstrated that the exclusive use of glyphosate raises the agronomical risks in terms of weed shifts.

Focussing on the ecological aspect we could not find a clear proof, that the Roundup Ready system reduces the weed diversity in maize. There are similar findings of the Field Scale Evaluation in UK. According to Dewar (2009) this is mainly due to the fact that growing of maize itself as well as any measure of weed removal are of higher rele­vance than the type of herbicide. As far as selective herbicides are applied at optimal conditions, an almost complete weed kill can be achieved. Because of a very low competitiveness of maize and thus a high need for an effi­cient weed control, there are only very few options for enhancing the biodiversity in this crop.

The planting and conducting of the field tests were performed by employees of the Landwirtschaftskammer Nordrhein-Westfalen, the Landwirtschaftskammer Nieder­sachsen and the Landesanstalt für Landwirtschaft und Gartenbau in Sachsen-Anhalt.