[Frontpage] [Contents] [Previous] [Next]

7 Output from the oilseed rape model

7.1 Introduction

In this section we focus on the output from the oilseed rape model after 20 years under different scenarios with currently used herbicides (propyzamid and benazolin+clopyralid; only the former will be available to Danish farmers in the future), and two scenarios with transgenic glyphosate or glufosinate tolerant oilseed rape.

Crop rotation

The recommended crop rotation with oilseed rape is a six-year rotation with: oilseed rape - cereals - cereals - peas - cereals - cereals, but often oilseed rape is grown in a four-year rotation with three cereal crops in between. The main reason for this distance in time between succeeding oilseed rape crops is soil borne pests and diseases. In order to test the most simple rotation with the highest percentage of oilseed rape crops we chose to model the four-year rotation.

Thresholds

Propyzamide is normally applied to the soil during periods of cold temperatures and to simulate this, the model is restricted to spray with this herbicide during the winter period. In the cereals, the sulfonylurea (SU) herbicide: tribenuron is applied early in the season at a low threshold of weed biomass which allows for a spraying in both autumn and spring. Presently, most farmers would prefer metsulfuron in the spring, but this SU-herbicide is rather persistent in the soil which makes tribenuron the most likely herbicide choice for the future. For the scenarios with glyphosate and glufosinate, volunteer crops (crop plants acting as weeds in succeeding crops) are included in the threshold level for the oilseed rape crop, whereas the traditional oilseed rape scenarios use the same threshold based on dicotyledonous (dicot) weeds only. Elymus repens is controlled with a preharvest application with glyphosate in the cereal crops and in the traditional oilseed rape, Elymus repens and cereal volunteers can be controlled with fluazifop-P-buthyl.

Efficacy

The individual dose-response relationship of the herbicides on different weed species is (benazolin+clopyralid, glyphosat and glufosinate) based on experimental data from the greenhouse (Chapter 5). The dose-response curves for tribenuron-methyl were produced for the Danish PC-Plant Protection programme and have kindly been provided by Per Rydahl, from the Danish Agricultural Research. The dose-response relationships are well described for these herbicides, and the dose applied per spraying can therefore be changed resulting in a change of herbicide efficacy. For the remaining herbicides only efficacy at recommended dose is known, and these doses/efficacies are, therefore, fixed.

Resistance

The resistance alarm is induced when biomass of either volunteer oilseed rape, Brassica campestris or SU-resistant Stellaria media exceeds 25% of the total dicot weeds and volunteers biomass. In case of alarm, the model sprays with an additional herbicide (fluoxypyr) in the cereal crops. The frequency of resistant Stellaria media has been artificially set to initiate the SU-resistant Stellaria alarm in a hypothetical scenario with SU used in all crops.

7.2 Materials and methods

Prerequisites

The model (Fig. 7.1) is, in contrast to the ecophysiological model developed by Spitters and Kropff (Kropff, 1993), an empirical model based on alleged relationships, comparative studies between transgenic and non-transgenic crops, and practical experience from current crop rotations. The model presented here requires few growth related parameters and does not take climatic and soil conditions into account (for model parameters and equations, see Appendix 2).

Actual data

The model was based the following actual data: the crop rotation was constructed from Danish agricultural statistics (Danmarks Statistik, 1994; Madsen et al., 1997). Obtainable yields for crops were based on attainable yield levels. Selection of model weeds was based upon a recent Danish weed survey (Andreasen, 1990). Initial levels of different volunteer crops were estimated. Seed decay in the soil was assumed to be exponential. Crossing frequencies between oilseed rape and B. campestris were based on experimental data (Jørgensen & Andersen, 1994), from which inverse linear curves were extrapolated with the presence of one species relative to the other as the predictor. Data of a naturally occurring sulfonylurea-resistant Stellaria media (L.) Vill. in Denmark was from Kudsk et al. (1995) and the initial frequency was set at a high level (10^-4) corresponding to a worst case scenario which in the model caused resistance problems within 20 years in a hypothetical crop rotation with consistent sprayings with tribenuron-methyl in all crops. Herbicide resistant biotypes have similar fitness as the non-resistant biotypes of both crops and weeds (Madsen, 1994; Fredshavn et al., 1995; Jensen, 1993), however, biomass and seed production appear to be reduced in second generation hybrids (F₂ or BC₁) between oilseed rape and B. campestris (T. Hauser, personal communication). To account for the observed yield depression in later generations (which are not modelled separately), the harvest index from the hybrid was reduced in the model. The resistance trait segregated according to Mendelian principles. Recommended dose used for calculation of treatment frequency was 406 g a.i. ha^-1 for the mixture of benazolin and clopyralid and 375 g a.i. ha^-1 for fluazifop-P-butyl (Anonymous, 1995). As no recommended dose is yet available for the transgenic oilseed rape, we chose to use the Canadian recommendations for spring varieties of glyphosate tolerant ‘Roundup Ready Canola’ and the glufosinate tolerant ‘Liberty Link Canola’ for which the recommended dose was up to 445 g a.i. ha^-1 for glyphosate and up to 600 g a.i. ha^-1 for glufosinate (Anonymous, 1997).

Model development

The model (Fig. 7.1), programmed in STELLA II (Peterson & Richmond, 1994), simulated the growth of crops and weeds in a rotation with oilseed rape. It included six ‘model’ weed species.

Seed bank

The weed species and volunteers regenerate in succeeding crops from a seed bank (or for Elymus repens (L.) Gould from a bank of propagules). The number of seeds per species that yearly enter the seed bank was quantified as a linear function of the biomass present at harvest time:

N_input : Number of seeds m^-2 that enter seed bank

TCW: Thousand corn weight according to Korsmo et al. (1981)

HI: Harvest index

s: Proportion of seeds that survive from mature seed to seed bank

Y_t=h: Dry matter production g m^-2 (100 [t ha^-1]) at harvest time

Seed number that leaves the seed bank was determined by exponential seed decay and the number of seeds that germinate when the crop is sown:

N_output : Number of seeds m^-2 day^-1 that leave seed bank

N_total: Total number of seeds m^-2 in the seed bank of the species

d: Annual seed decay

e_t=s: Proportion of seeds that germinate when the crop is seeded

Growth and competition

Sigmoid growth curves described growth of the different species which competed for a yield potential determined by a maximum potential biomass of the crop:

Y_t+1: Dry matter production from day t to day t+1.

g: The relative growth rate day^-1

Y_t : Dry matter production up to time t

Y_Max: The maximum attainable biomass production ha^-1 for the species

Y_Total : The accumulated biomass produced t ha^-1 of all species in the model.

The simple competition model (eqn. 3) allowed the individual weed species to compete with each other as well as with the crop in order to obtain a share of the maximum available yield potential.

Weed control

At a fixed threshold of biomass of weeds, each crop was sprayed with herbicides. The efficacies depended upon the weed species, herbicide and dose. A realistic timing of application and numbers of sprayings was determined by the threshold levels, which were: propyzamide: 0.5 t ha^-1 of biomass of weeds and volunteers during winter; benazolin+clopyralid: 1 t ha^-1 of dicot weeds followed by fluazifop-P-butyl at 0.5 t ha^-1 of monocotyledonous weeds; glyphosate and glufosinate: 1.4 t weeds and volunteers ha^-1. Efficacy of glyphosate, glufosinate or the mixture of benazolin+clopyralid in the oilseed rape crop was determined by dose response curves (based on greenhouse experiments) for for S. media, Chenopodium album L., Capsella bursa-pastoris, Myosotis arvensis (L.) Hill, and oilseed rape (data are not shown); details on comparison of herbicide efficacy with the log-logistic dose-response model are found elsewhere (Streibig et al., 1993; Madsen & Jensen, 1995). The dose-response curves for B. campestris were assumed to be identical to the curve for oilseed rape. Dicot weeds in the barley and wheat crops were sprayed with a sulfonylurea herbicide, tribenuron-methyl (threshold: 0.1 t biomass ha^-1) with efficacy levels based on dose-response curves (P. Rydahl, personal communication). If the fraction of sulfonylurea-resistant/sensitive weeds exceeded 0.25 in the cereal crops, then fluroxypyr (144 g a.i. ha^-1) was added to the treatment. Where no dose-response relationships were available, the efficacy according to agronomic practice was used (propyzamide: 500 g a.i. ha^-1 and fluazifop-P-butyl: 188 g a.i. ha^-1). E. repens growing in cereal crops was controlled with a pre-harvest application of glyphosate (800 g ha^-1) when the development of this grass weed exceeded 0.3 t biomass ha^-1 at the end of the growing season.

The model simulated four scenarios for weed management in a rotation with winter varieties of oilseed rape - wheat (Triticum aestivum L.) - wheat - barley (Hordeum vulgare L.):

1. A non-transgenic oilseed rape sprayed only with propyzamide and fluazifop-P-butyl against E. repens and cereal volunteers.
2. A non-transgenic oilseed rape sprayed with a mixture of benazolin+clopyralid against dicot weeds and fluazifop-P-butyl against E. repens and cereal volunteers.
3. A glyphosate tolerant oilseed rape sprayed only with glyphosate.
4. A glufosinate resistant oilseed rape sprayed only with glufosinate.

7.3 Results

Simulating practice

The simulated growth of crops and weeds in Fig. 7.2 was used to evaluate different scenarios with herbicides in oilseed rape and a common herbicide for the cereal crops. The simple competition model allowed the individual weed species to compete with each other as well as with the crop in order to obtain a share of the maximum available yield potential. The thresholds were adjusted to simulate common practice with herbicide sprayings and were unchanged at different herbicide rates. When the threshold induced a spraying, it controlled the biomass of the individual weed species with a certain efficacy level, which was determined by the dose-response curve for the weed and herbicide combination. The controlled biomass then became available to the remaining crop and weeds. However, the spraying had favoured growth of the crop relative to growth of the weeds, as the relative weed biomass had diminished.

By including the greenhouse dose-response curves (Chapter 4) it was possible to test the effect of dose on accumulated herbicide use over 20 years which makes it possible to compare herbicide scenarios even though recommended dose and number of applications are not yet known. This can be used as part of a sensitivity analysis, but it also illustrates that sensitivity of herbicide-dose relationship varies with dose, because the closer the dose is to the steep part of the dose-response curves the more sensitive it becomes to adjustment. Furthermore, these runs showed that the minimum combination of dose and numbers of application, which will result in a satisfactory weed control, is not necessarily the lowest dose tested in the model.

Herbicide use

Herbicide use over 20 years was low for all scenarios compared to accumulated use in sugarbeets (Chapter 6), because current weed control in oilseed rape is conducted with relatively low herbicide doses. At the assumed recommended rate (based on Canadian recommendations) herbicide use over 20 years, measured in kg a.i. ha^-1, was similar for the scenarios with traditional and glyphosate tolerant oilseed rape, whereas it was higher in the scenario with glufosinate tolerant oilseed rape (Figs 7.3 and 7.4), however, if herbicide use was measured as treatment frequency, herbicide use was lower in the transgenic crops (Fig. 7.5). The difference in accumulated treatment frequency was rather consistent with changing rates per application, and the rotation with glyphosate or glufosinate tolerant oilseed rape had significantly lower treatment frequencies at different fractions of recommended rates (Fig 7.5). When rates were reduced below approx. 0.3 - 0.5 times recommended dose, the number of sprayings increased radically, thus making such scenarios unpractical (Fig. 7.6).

Crop yields

At the recommended rate (based on the Canadian recommendations) crop yield in the different scenarios was similar around 96-97% of the maximum yield level.

7.4 Sensitivity of selected parameters

The output from the different parameter settings used in the model varies with the chosen scenario. However, there are some general trends:

Growth rate

Growth rates of the crops and weeds are very sensitive to adjustments, because growth of a species is defined by only two parameters, growth rate and maximum biomass. However, growth is not the main focus of this model.

Dose

Sensitivity of the dose varies according to locations on the dose response curve for the individual herbicide/weed species. If doses vary around a rather flat curve at high efficacy levels, then the parameter is not sensitive, but if the dose varies around the ED₅₀ of the model (Chapter 4) then dose is a sensitive parameter.

Other parameters

The parameters that determine the loss of seed and seed decay for oilseed rape both seem sensitive at the tested range. Seed numbers and decay of Brassica campestris and Stellaria media seem less sensitive towards changes.

7.5 Discussion

Lack of data

The literature survey revealed lack of knowledge about growth characteristics of the individual weed species, and several parameters on harvest index, seed bank, growth rate and maximum biomass were estimated based on the expected behaviour in the system. Satisfactory data concerning survival and growth characteristics of weeds under field conditions would be indispensable when developing this kind of model and would not only have practical implications for predicting the herbicide use in genetically modified herbicide resistant crops, but also in any simulation model that deals with crop-weed competition.

Low dose herbicides

Compared with glyphosate and glufosinate, benazolin+clopyralid and propyzamide have low efficacy on several weed species which causes a shift toward more sprayings in the succeeding cereal crops with the SU-herbicide. Use of sulfonylureas adds only little to the amount of herbicide used, but adds equally to the other herbicides in terms of treatment frequency. This explains the contradicting conclusions about herbicide use.

SU-resistance

In the scenario where oilseed rape was sprayed with benazolin+clopyralid, the high number of SU applications, the cereal crops posed a high selection pressure on the weeds which favoured growth of SU-resistant Stellaria media, and at high rates the model induced fluoxypyr to be mixed with the SU-herbicide after 18 years of crop rotation, whereas no such problems occurred in the scenarios with glufosinate or glyphosate tolerant oilseed rape during the 20 years. It should, however, be emphasized that the initial frequency of SU-resistant Stellaria media was set at a high level in order to initiate resistance problem if SU was used in all crops.

Volunteers and hybrids Resistant volunteers and hybrids between oilseed rape and B. campestris were present but did not cause problems in any of the four scenarios. However, it is common knowledge that a SU herbicide, which was used in the cereal crops generally controls Brassica species effectively. This emphasises the importance of studying consequences for the complete rotation system, when a transgenic herbicide tolerant crop is introduced.

Number of treatment

Figures 7.4 and 7.5 leave the impression that total herbicide use can be decreased by lowering the dose indefinitely, but this strategy increases the number of treatments dramatically with a consequent unpractical increase in application costs (Fig. 7.6).

Treatment frequency 1996

In practical agriculture, average treatment frequency in 1996 was 0.95 in winter oilseed rape and 1.33 in winter cereals, which ‘all other things being equal’ means that the accumulated treatment frequency over 20 years is 25 standard recommended rates for this particular crop rotation (Miljøstyrelsen, 1996). The model predicts a treatment frequency of 25 for the traditional herbicide strategy with propyzamide and 27 for the strategy with benazolin+clopyralid. This indicates that the model simulated the crop rotation in accordance with current agricultural practices.

Predictions

The model described here is a first attempt to integrate knowledge about herbicide-tolerant crops with known agricultural practices. It should be emphasized that the model is preliminary and needs validation before any reliable predictions can be made about long-term consequences. In order to validate and adjust the model, short-term simulation results should later be compared with field data from crop rotations with transgenic herbicide tolerant crops after their release on the market. Baring these precautions in mind, we believe that the qualified prediction from the model is an improvement to just guessing, and the approach in our model could be used in revealing potential risks of growing herbicide-tolerant crops and thus prevent problems with unwise herbicide use and resistance in weeds and volunteers.

7.6 Conclusions