Introduction

Maize is grown in the Netherlands as silage feed for cattle and occupies the largest area of any arable crop. covering 210,000 ha in 2000. Pesticide use in maize per ha is relatively low compared to other arable crops, but because the maize crop covers such a large area it accounts for a considerable proportion of pesticide use on a national scale. Pesticides used on maize in the Netherlands are almost exclusively herbicides.

In view of the importance of maize on a global scale, the improvement of cropping methods is steadily attracting more effort and interest. The improvements include the success achieved by the agro-chemical industry in genetically modifying maize in order to make the crop tolerant to herbicides such as glyphosate and glufosinate-ammonium. The essence of this improvement is that this type of herbicides can be used on maize to suppress most weeds while leaving the maize unaffected. This study is concerned with glufosinate-ammonium (Liberty) developed by Bayer Crop Science and the genetically modified maize known as "Liberty Link maize".

Aim

As regards the Netherlands, concern about the effects of glufosinate-ammonium on non-target vegetation exists. Therefore a field study is being carried out in the Netherlands. Furthermore guidelines for this type of studies are lacking.

The present study has two aims:

• To acquire information on the effects of glufosinate-ammonium on adjacent vegetation at the community level in the field situation in the longer term (one year or more).
• To assemble knowledge about how to conduct field tests with vascular plants at the vegetation level.

The proposed study design and interim results were presented and discussed in three workshops with representatives of chemical industry, authorities for the approval of pesticides and researchers from the Netherlands, Germany and the United Kingdom.

Set up

Since it was impossible to conduct a field study in a nature reserve or on natural vegetation, we therefore looked for habitats (ditch banks and road verges) in appropriate areas (potential exposure) with relatively natural vegetation. (no herbicides used on or adjacent to the vegetation in the last 10 years, relatively high number of species, diversity ) representative for the Netherlands and for Mid-Europe. It was decided to study the effects of glufosinate-ammonium in a “realistic” worst case situation. Hereto the dosages were applied twice per season (highest rate was 64% of the registered rate with 2 x 800 g ai/ha per season in two consecutive years), as can be the (worst) case in practice and the vegetation is sprayed directly with the different dosages, thus simulating drift. The main question concerned the effects at the vegetation level.

The central parameters are therefore vegetation surveys: records of species presence and abundance. The parameters were recorded (2000-2002) in May (before spraying) and in August (after spraying). The effects at the vegetation level were expressed in the two following research questions:

• Are there any effects in August of the herbicide applications in May/June within the same year? (short / midterm effects)
• Are there any effects in May before the herbicide applications due to the application in the year before ? (longer term effects)

Also some other parameters have been recorded:

• Exposure of the vegetation to the herbicide during application of the herbicide
• Phytotoxic symptoms of the vegetation shortly after spraying (short-term effects)
• Biomass: assessment of the effects of spraying on the above ground biomass (mid- and longer term effects).

Sites & pesticide treatments
The study was carried out at four sites, two sites were road verges on sandy soil and the other two are ditch banks, on clay and peat respectively. Plots measuring 25 m2 (1 m x 25 m) were arranged in blocks to minimise within site variation. The study was carried out with 5 different treatment levels and a control (0%):

• Low dosages: 2% and 4% of the maximum registered field dosage of 2 x 800 g a.i. glufosinate-ammonium/ha per season. These low deposition levels (4%: 2 x 32 g a.i./ha) , occur at approximately 2.5 m and 1 m resp. from sprayed fields
• Intermediate dosage: 16% (2 x 128 g a.i./ha).
• Higher dosages (positive controls): 64% (2 x 512 g a.i./ha) and 32% (2 x 256 g a.i./ha). At these dosages effects would be expected.

The number of replications was 5 per site per treatment. The herbicide was applied with an advanced handheld knapsack sprayer. The interval between the first and second spraying (May-June) was 15-20 days. Spray deposition was measured with water-sensitive paper at different heights above ground level and calculated with an image processing and analysis program.

Parameters
In order to quantify the short-term effects of spraying, the observable external damage to the plants was recorded 8-10 days after each spraying in each year by eye in intervals of 5% ( phytotox). In 2001 the phytotox was also determined at about 33 and 62 days after the second application. The effects of the spraying on the biomass in 2000 at two sites and in 2001 and 2002 at all sites was studied. For this purpose three subsamples of 30 cm x 30 cm per plot (of 1m x 25m) were harvested, taking the entire above-ground biomass. Braun-Blanquet relevés were made in each plot in order to study the effects at the vegetation level. At all four sites data were gathered before spraying (May/June) and after spraying (August). From the relevé data the following parameters per plot were derived: total number of species, number of dicotyledonous species, number of pioneer species, total coverage, coverage of dicotyledon species, coverage of pioneer species, diversity (Shannon-Weaver index), evenness and plant species composition.

Statistical procedure
All analyses were conducted with the data packages per site and with the pooled data, but for each year separately. The statistical analysis is divided into two parts: univariate analysis of separate variables, e.g. phytotoxicity, biomass and number of species, on the one hand, and on the other hand the multivariate analysis of plant species composition (the Principle Response Curves: PRC). In the univariate analysis of the relevé, biomass and phytotox data comprised of three steps. First a standard analysis of variance was carried out. If an effect had been found with a significance value of P<0.10 in the first step, the second step followed with a Williams’ test on trend (is there a monotonically increasing or decreasing relation between dosage and effect parameter). If there was a significant trend (P<0.05) in the second step, then in the third step the Williams’ test was used again, to calculate the so called no observed effect dosages (NOED). The highest dosage that was not significantly different from the control was taken as the NOED. Since in most cases there was a significant autocorrelation between August and May relevé data of the same year and between May values of the different years, in the analysis of the effects the contrasts were used, instead of the data itself. The significance level ? applied, is 0.05. If the statistical analysis resulted in a P-value of 0.05 or less than the dosage (etc.) had a significant effect. Because several variables were sampled or derived from the relevés, the significance level a for multiple comparisons was corrected by the improved Bonferroni procedure.

Three different multivariate tests were performed. First in the CANOCO computer program, Redundancy Analysis is accompanied by Monte Carlo permutation tests to assess the statistical significance of the effects of the explanatory variables on the species composition of the samples. The significance of the PRC diagram in terms of displayed treatment variance was tested by Monte Carlo permutation of the plots following the PRC analysis. Second it was tested if the treatment regime has a significant effect on the composition of the vegetation community on a particular sampling date, and third it was tested if a particular treatment had a significant effect on the composition of the vegetation community at a particular sampling date. Besides the overall significance of the treatment regime, we also wanted to know which treatment levels differed significantly from the controls. From the results at each site a NOECcommunity could be deduced for each sampling date, with the NOECcommunity being the highest treatment level for which no statistical significance of effects could be demonstrated.

Results

Pesticide deposition
Deposition on the off crop vegetation was measured during the first and second spraying. The results show that the dosages in the field were applied with great precision. During the first spraying in both years deposition significantly increased with increasing height. Differences in deposition during the first spraying are smaller in 2001 compared to 2000. During the second spraying deposition is dependent on height and dosage.

Short term phytotoxic effects
Significant short term phytotoxic effects (8-10 days after application) of glufosinate-ammonium were present in both years, after each treatment, at all sites and even at low dosages. At the low dosages 2% and 4% phytotox values ranged between 3 and 14% and phytotox values after the first application were higher then after the second application. NOED is in almost all cases lower than the lowest dosage of 2% used. In August of the same year of spraying, no significant effects on the vegetation could be detected anymore, even at the highest dosage (64%).

Mid- term ecological and community effects: within the years of application
Within the year of spraying at low concentrations (2% and 4% of the field dosage) there were no significant effects on the vegetation parameters studied. In the second year (2001) of the study no effect could be detected at low concentrations (or even high concentrations) on this type of parameters at all. At higher concentrations (32% or higher) in both years of the study significant effects were found on the vegetation community. In the community analysis the ditch bank locations (Lexmond and Zegveld) showed the most pronounced effects. In the first year of the study also significant effects on the vegetation biomass were found within the year of spraying (at 32%, all sites combined). In the second year at one site (Lexmond) a significant effect was found on the biomass (32%).

Longer term ecological and community effects: one or two years after application
In the experiment even at the highest dosages no significant effects have been found on the vegetation in May (2001 and 2002) of the herbicide applications of the years before (2000, 2001) for all relevant criteria / parameters. Thus, no longer term effects on the plant community were found one or two years after application.

Conclusions and discussion

The aims of the study were to acquire information on the effects of glufosinate-ammonium and recovery on adjacent vegetation at the community level in the field situation under “realistic” worst case conditions and to assemble knowledge about how to conduct field tests with vascular plants at the vegetation level regarding herbicides and non-target vegetation.

Concerning the effects of glufosinate-ammonium on off crop vegetation it can be concluded as follows:

• Short term effects: About 10 days after spraying in spring, small, significant phytotoxic effects could be detected on non target vegetation even at low concentrations of 2 and 4% of the field dosage. The effects are not visible any more later in the season (August).
• Mid term effects: Within the year of spraying at low concentrations (2% and 4% of the field dosage) there were no significant effects found on the vegetation parameters studied. At higher concentrations (32% or higher) in both years of the study significant effects were found on the vegetation community. At 32% also significant effects on the vegetation biomass were found within the year of spraying (in 2000 for all sites combined and in 2001 at one site).
• Longer term effects: In the pair wise comparison between years (2000/2001, 2001/2002 and 2000/2002) no significant effects of the sprayings could be detected in spring on all parameters.

Concerning the mid-term effects of this study the results should be handled with care in some instances there were nearly significant effects and clear negative trends between dosage and effect parameters.

From the study it is clear that large scale field studies can be conducted in practice. Moreover, sound and generally reproducible results could be generated. From the study some general methodological aspects could be derived regarding experimental set up (set up with drift simulation and not-adapted vegetation), field assessments (phytotox of plots incl. studying recovery, biomass and vegetation relevé’s) and statistical procedures (multivariate and stepwise univariate approaches).

• Snoo, G.R. de , F.M.W. de Jong, R.J. van der Poll & M.G.A.M. van der Linden (2001) Effects of Glufosinate-ammonium on off crop vegetation.Interim results.Med. Fac. Landbouww. Univ. Gent 66/2b: 731-742
• Snoo, G.R. de, W.L.M. Tamis & P.J. van de Brink (m.m.w.v. R.J. van der Poll, F.M.W. de Jong, M.G.A.M. van der Linden, P. van Hooft & N. van der Hoeven) (2005) Non target plant field study: Effects of glufosinate-ammonium on off crop vegetation. CML report 161, Department of Environmental Biology, 114 p.