Great brome and barley grass – modelling the long-term value of harvest weed seed control

Key messages

  • Great brome had high seed production in Western Australia and South Australia. Barley grass had much higher seed production in South Australia than Western Australia.
  • Modelling indicated that harvest weed seed control is a valuable tool in the management of these species, even in years where most weed seed is shed before harvest.

Aims

Brome grass and barley grass are ranked as the fourth and ninth most detrimental weeds nationally, with an annual revenue loss of $22.5 and $1.7 million respectively (Llewellyn et al, 2016). The results of a five-year GRDC project to determine the ecology of great brome and barley grass in Western Australia and South Australia (‘Seed bank ecology of emerging weeds’ UA00156) were used to update a decision support tool; the Weed Seed Wizard (WSW). Data used in the model included plant density, seed production, competitive ability with the crop and time of seed shedding. The proportion of seed shed prior to harvest varies widely between ecotype and season for both these weed species (Borger & Hashem, 2019). The updated model was used to investigate harvest weed seed control (HWSC) as a potential management technique for great brome and barley grass. We tested the hypothesis that HWSC is a valuable weed management technique even if you can only capture a low proportion of seed at harvest.

Method

Changes to the Weed Seed Wizard

Weed Seed Wizard is an agricultural decision support tool that is fully described in Borger et al (2018) and can be freely downloaded from Agriculture and Food Western Australia (2016). Changes to ‘Model’ and ‘Edit Species’ were used to update ‘Barley Grass’ and ‘Brome (diandrus)’ within the species list, based on earlier field trials on emergence, persistence, seed production and competitive ability (Borger & Hashem, 2018b, a; Borger & Hashem, 2019; Hashem et al, 2019). Species parameters include characteristics of ‘Germination’, ‘Monthly Germination’ rates, ‘Yield’, ‘Persistence’, ‘Monthly Death’, ‘Interaction’ with other species and ‘Acquired Dormancy’. The updated parameters for ‘Bromus (diandrus)’ included changes to: ‘Monthly Germination’ (germination in Jan to Dec changed from 0, 0, 0.8, 1, 1, 0, 0, 0, 0, 0, 0, 0 to 0, 0, 0.8, 0.8, 0.8, 1, 0.2, 0.2, 0, 0, 0, 0), ‘Max seed production per m2’ (changed from 10 000 to 60 000), ‘Plants per m2 for half of max seed production’ (changed from 20 to 30), ‘Mass of 1000 seeds’ (changed from 9 to 11.7) and ‘Standard antagonism’ (changed from 1.4 to 2.0). The updated parameters for ‘Barley grass’ included changes to ‘Max seed production per m2’ (changed from 40 000 to 6000) and ‘Interaction’ of lentils (‘Antagonism’ and ‘Penalty’ changed from 1.5 and 0.8 to 1.9 and 0.4).

Field trials used to validate changes to the Weed Seed Wizard

Trials were run at Department of Primary Industries and Regional Development Wongan Hills Research Station WA (-30.8463, 116.7442) and Roseworthy Agricultural College South Australia (-34.5195, 138.6929). Trials were established as a split-plot design, with weed species including great brome and barley grass as the main-plot factor and weed density as the sub-plot factor, with four replications (Table 1). In all sites, non-selective herbicide was applied where necessary at the start of the year, but no other herbicides were applied. Hand weeding was used to clear additional weeds. In each trial, data included plant density, weed seed production and crop yield. Full details of the trials can be found in Hashem et al (2019) and in the 2019 annual report for GRDC project UA00156.

It was not the intent of the current paper to present all details and results from these six field trials, but rather to take mean data on great brome and barley grass plant density and seed production, and crop yield, to compare to predicted values from the model. The agronomic details of the field trials used to parametrise the model are listed in Table 1. A scenario for each field trial was constructed in the model. Under ‘Scenario Setup’ the ‘Scenario Timeframe’, ‘Soil Type’ and ‘Weather’ were specified (Table 1). ‘Initial Seed Bank’ values were selected to ensure that plant density in the model matched the values of actual plant density in the field, on the specified date of weed assessment in the field (Table 3 and 4). Under ‘Event Management’ a ‘Spray event’ was used to add the non-selective herbicide, a ‘Sow event’ was specified using sowing date, crop species, seeding rate in kg/ha and ‘Till type’, and a ‘Harvest event’ was specified using ‘Date’, ‘Crop’, ‘Yield based on average season’ and ‘Harvest type’ of ‘All chaff spread/normal harvest’. After selecting ‘Scenario’ and ‘Simulate’, results were examined in ‘Simulation’. ‘Export’ was selected to download results on weed density, weed seed production and crop yield. Each field trial scenario was run in the Weed Seed Wizard model using both the original species parameters for great bromeand barley grass and the altered species parameters detailed in Changes to the Weed Seed Wizard.

After modelling each field trial scenario, the values for weed seed production and crop yield measured in the field trials were compared to the predicted values from both the initial model (original parameters for barley grass and great brome) and the updated model. The mean absolute error (MAE) was used to compare observed (field trial) values with predicted means from the initial and updated model.

Table 1. Trial location (note that location, crop and year are used to name each scenario), year, scenario timeframe used in the model, actual and modelled soil type, weather record used in the model, weed assessment date, actual and modelled non-selective herbicide, sowing details, actual and modelled seeding system, and harvest date.

image of trial location

*Herbicide formulations include Spray.Seed® (paraquat/diquat 135/115 g a.i. L-1, SL, Syngenta Pty. Ltd.) and Weedmaster DST® (glyphosate 475 g a.i. L-1, SL, Nufarm) + Hammer (carfentrazone-ethyl 400 g a.i. L-1, EC, ®, FMC).

Long-term harvest weed seed control in the Weed Seed Wizard

The long-term impact of HWSC (destroying 1, 20, 40, 60, 80 and 100% of great brome or barley grass seed) was modelled in the Weed Seed Wizard. The model used the altered weed species parameters detailed in Changes to the Weed Seed Wizard. A scenario was established with a ‘Scenario Timeframe’ of 11/12/2010 to 31/12/2016, and 100 seeds of either species in the ‘Initial Seed Bank’. The ‘Soil Type’ was sand and the ‘Weather’ was WA_Wongan Hills. The six-year scenario was a rotation of wheat-wheat-lentil-wheat-wheat-lentil. All agronomic events in year one to three of the rotation were repeated in year four to six (i.e. a chemical sprayed in year one was sprayed on the same day of the year in year four, detailed in Table 2). Harvest was changed from ‘All Chaff Spread/Normal Harvest’ with 1% of seeds caught in the grain to ‘Seed Destructor’, and a specified 0%, 20%, 40%, 60% and 80% of seed dropped to consider varying levels of HWSC.

Both species in the model finish dropping seeds and die prior to harvest. To explore the option of HWSC, we selected ‘Model’, ‘Edit Species’, ‘Yield’ and changed ‘DD finish dropping seeds and die’ (degree days required for the plant to finish shedding seed and die) to zero. This disables the parameter and causes all seed to be retained until a ‘Harvest Event’ in ‘Event Management’ kills all plants and drops or destroys the specified proportion of weed seed. The total number of great brome and barley grass seeds in the seed bank on 1 December in each year were recorded, to determine the value of varying levels of seed destruction at harvest.

Table 2. The date, event and description for each event in ‘Event Management’ of the model scenario to consider harvest weed seed control. The scenario is a six-year rotation of wheat-wheat-lentil-wheat-wheat-lentil, where each event in year 1 to 3 was repeated in year 4 to 6.

image of Event management

Note that the herbicides listed here do not refer to specific products; they are options selected from the list of available ‘herbicides’ in the model.

Field trials

For both WA and SA, the initial model underestimated great brome seed production (Table 3). The MAE of great brome seed production was reduced for the updated model, indicating that predicted values from the updated model were closer to the actual field values (MAE of 4250.3 for the updated model compared to 6551.4 for the initial model). However, seed production predicted by the updated model was still usually lower than seed produced in the field, particularly for SA 2017. The initial model always overestimated yield, except for SA wheat 2017, and the updated model slightly improved yield predictions (MAE of 1.052 for the initial model and 0.933 for the updated model).

Table 3. The great brome density, seed production and crop yield from six field trials in WA and SA (growing wheat or lentils in 2017 and 2018), as well as estimated great brome seed production and crop yield from the initial model and updated model parameters for great brome. The mean absolute error compares seed production or yield data from the field trials with estimated values from the initial or updated model.

image of great brome density

The initial model overestimated barley grass seed production in WA (Table 4). In SA, both the initial and updated model underestimated seed production from a very low density of barley grass, especially in 2017. Both the initial and updated model overestimated yield, but the updated model gave a slightly improved estimation of yield in lentil crops in SA. The MAE of barley grass seed production and crop yield were both reduced for the updated model, indicating that predicted values from the updated model were closer to the actual field values.

Table 4. The barley grass density, seed production and crop yield from six field trials in WA and SA (growing wheat or lentils in 2017 and 2018), as well as estimated barley grass seed production and crop yield from the initial model and updated model parameters for barley grass. The mean absolute error compares seed production or yield data from the field trials with estimated values from the initial or updated model.

image of barley grass density

Long-term harvest weed seed destruction in the Weed Seed Wizard

Great brome was highly competitive in the updated model. In the scenario where great brome was controlled by herbicide alone and a normal harvest was used (i.e. 1% of weed seed captured in the grain), the total weed seed in the soil seed bank on 1 December each year increased from 319 seeds in 2011 to 10954 seeds in 2016 (Table 5). However, HWSC of 20% of great brome seed reduced the number of seeds in the soil seed bank to 5925 in 2016. When 60% of seed was captured at harvest, total seeds in the soil seed bank in 2016 were lower than the initial starting seed bank of 100 seeds/m2.

By comparison, barley grass was not highly competitive. The seed bank reached less than 1 seed/m2 by 2014 with herbicide use alone. Increasing levels of HWSC still reduced the weed seed bank faster than the scenario with 1% HWSC.

Table 5. The total number of great brome or barley grass seeds/m2 in the soil seed bank on 1 December of each year (i.e. the day after crop harvest in each year of the six year scenario), following harvest weed seed control of 1%, 20%, 40%, 60%, 80% and 100% of the seed. Note that the starting density of the weed seed bank for each species was 100 seeds/m2.

image of brome and barley grass seed

Conclusion

Even if most great brome seed has shed at harvest, HWSC is still worthwhile. An annual seed destruction of just 20% of great brome grass made a large difference to the long-term seed bank of this species (i.e. reducing the weed seed bank from 10954 seeds to 5925 seeds at the end of the six-year rotation). As Borger and Hashem (2019) demonstrated, great brome seed retention by the end of November ranges from 0% to 47%, but seed retention may be as high as 72% when the crop first reaches maturity. In the current simulation, a HWSC of 60% or more was necessary to drive the seed bank down over six years to lower levels than the starting seed bank of 100 seeds. However, in the field, the actual percentage of HWSC required to eradicate this weed will depend on the agronomic rotation and efficiency of the herbicides used. HWSC alone will not eradicate great brome but it is still a vital part of the IWM system.

Long-term barley grass control also benefited from HWSC, but barley grass was much easier to control in-crop than great brome. In the model, herbicide alone (without HWSC) was sufficient to eradicate barley grass. It is notable that the model overestimated barley grass seed production in WA, while underestimating barley grass seed production in SA. By comparison, the initial model underestimated seed production of great brome grass in both WA and SA. This suggests that barley grass in WA may be less competitive than the populations found in eastern Australia. Further research is required to determine if HWSC is feasible for barley grass in the field, given that the plants can be much shorter than great brome (Borger & Hashem, 2019).

Note that ® indicates Registered Trademark.

Varieties displaying this symbol beside them are protected under the Plant Breeders Rights Act 1994.

Acknowledgments

The research undertaken as part of this project is made possible by the significant contributions of growers through both trial cooperation and GRDC investment, the authors would like to thank them for their continued support. We would also like to thank Nerys Wilkins, Peter Gray, Shari Dougall and Bruce Thorpe at the Wongan Hills Research station. Thanks are due to Alex Douglas for reviewing this paper.

Contact details

Dr Catherine Borger, Department of Primary Industries and Regional Development
75 York Rd (PO Box 483) Northam WA 6401
0467 816 082
08 9622 1902
catherine.borger@dpird.wa.gov.au

References

Agriculture and Food Western Australia (2016) Weed Seed Wizard. Version 7.1.7. INTERNET, available at: https://www.agric.wa.gov.au/weed-seed-wizard-0 (last accessed 29 April 2016).

Borger C & Hashem A (2019) Harvest weed seed destruction – does it work for great brome and barley grass? In: GRDC Grains Research Updates, Perth, Western Australia. Grain Industry Association of Western Australia.

Borger CPD & Hashem A (2018a) Emergence of great brome grass and barley grass. In: Proceedings Weed Biosecurity - Protecting our Future. 21st Australasian Weeds Conference. (eds S Johnson, L Weston, H Wu & B Auld). 9-13 September, Manly, Australia. 341-344. The Weed Society of NSW, Sydney, Australia.

Borger CPD & Hashem A (2018b) Recruitment and fecundity of annual ryegrass, great brome grass, barley grass, doublegee and sowthistle. In: Proceedings 21st Australasian Weeds Conference. 9-13 September 2018, Manly, Sydney. The Weed Society of NSW.

Borger CPD, Riethmuller G & Renton M (2018) Weed Seed Wizard: a tool that demonstrates the value of integrated weed management tactics such as harvest weed seed destruction. Computers and Electronics in Agriculture 127, 27-33. https://doi.org/10.1016/j.compag.2018.02.011

Hashem A, Riethmuller G & Borger C (2019) Competitiveness of emerging weed species in a wheat crop. In: Proceedings Grains Research Updates. 25-26 February 2019, Perth, Western Australia. Grains Industry Association of Western Australia, Grains Research & Development Corporation.

Llewellyn R, Ronning D, Clarke M, Mayfield A, Walker S & Ouzman J (2016) Impact of weeds on Australian grain production: the cost of weeds to Australian grain growers and the adoption of weed management and tillage practices. Grains Research and Development Corporation, Commonwealth Scientific and Industrial Research Orgainsation, Canberra.

GRDC Project code: UA00156