Developing a weed patch spraying system for use in arable crops


Cereals & Oilseeds
Project code:
01 June 1998 - 31 March 2001
AHDB Cereals & Oilseeds.
AHDB sector cost:
£165,499 from HGCA (project 1982)
Project leader:
P.J.W. Lutman1, N.H. Perry1, R.I.C. Hull1, P.C.H. Miller2, H.C. Wheeler2, R.O. Hale2 1Rothamsted Research, West Common, Harpenden, Hertfordshire AL5 2JQ 2Silsoe Research Institute, Wrest Park, Silsoe, Bedford, Bedfordshire MK45 4HS



About this project


The overall aim of the project was to develop weed patch spraying (GPS location, weed mapping, creation of a treatment map and application of the herbicide) into a commercially viable package that could be used by farmers and contractors. It was concluded that the most appropriate weeds to be targeted were black-grass, wild-oats, bromes, cleavers, thistles and common couch.

Weed mapping. As the project aimed to develop a system for 2002, it was decided that weed detection should be based on visual mapping. Over 60 geo-referenced presence/absence weed maps were created from tractors, ATVs and combine harvesters. The optimum time for mapping varied between the test species. Practical recording systems based on voice recognition and toggle switches were developed. Some weed maps were validated by quadrat counts on a 5m grid. Correlations were acceptable, though mapping small grass weeds in cereals was difficult. Weeds could be mapped at 3-4ha/h, depending on the vehicle used. The potential of automated detection was also explored, but video images failed to detect weeds in cereals, reliably. However, semi-automated, reflectance based 'vegetation anomaly' detection was quite successful. Both these techniques need further research.

Behaviour of weed patches. Knowledge of the spatial behaviour of weeds is essential if patch spraying is to be based on historic maps. Research showed that patches were fairly stable, as the core areas only moved 3- 4m over several years, and so patches of most species only need re-mapping every 3 or so years. However, this did not apply to wild-oat patches, which were potentially unstable, because small numbers of seeds, which were moved some distance by the combine, had the potential to create new patches. Studies of the causes of patches showed that soil characteristics could be linked to the patch distribution in some fields.

Creation of the treatment map. The ‘raw’ weed map has to be ‘processed’ before it can be used by the computer control system (such as Fieldstar) to control the sprayer. Software was developed in association with Patchwork Technology that will interpolate data from the weed maps and add buffers to account for possible patch movement and location variation. Finally, appropriate products and doses can be selected and the application map prepared.

Herbicide application and system testing. The project used the dual boom, variable volume rate, Micron Patchspray sprayer for most herbicide applications, though the AGCO Spra-Coupe was also studied. Several fields were sprayed, as controlled by the treatment map. In the majority of treated fields, full rates of product were applied to patches and reduced rates to the rest of the field. Applications were generally successful. As well as demonstrating the system at ‘Cereals’ and ‘Sprays & Sprayers’ a specific demonstration of the project was held in February 2002 at SRI.

Economics and other aspects relevant to farmers. It was estimated that average herbicide savings of £6/ha were possible from patch spraying. In some situations, this could increase to £20/ha. Cost of patch spraying equipment ranges from £8000 down to less than £2000, depending on whether the user already using ‘precision agriculture’ equipment. Weed mapping would cost c.£1/ha. Thus, overall the equipment would have to used on approximately 2000 ha, to recoup the costs. The technology also provides printed evidence of herbicide application, of relevance to crop assurance schemes and to regulatory authorities.