Agriculture Reference
In-Depth Information
and Heisel
1998
). The variation in weed seedling population has often been ignored
for weed management decisions since techniques to assess the weed seedling distri-
bution in acceptable time were not available. Many studies were conducted to apply
post-emergence herbicides in winter wheat and maize based on georeferenced maps
of the weed seedling distribution (Nordmeyer and Niemann
1992
; Tian et al.
1999
;
Gerhards and Christensen
2003
). Herbicide use with this map-based approach was
reduced some 40-50 %. With a large within-fi eld variation in weed occurrence,
patch spraying that is based on the need for weed control reduces costs, herbicidal
pollution of the environment and the risk of herbicide residues in the food chain
(Dammer et al.
2003
; Timmermann et al.
2003
; Gerhards and Oebel
2006
).
In many studies, weed species were grouped into grass weeds, annual broadleafs
and perennial weeds. Perennials such as bindweed (
Convolvulus arvensis
) and thistle
(
Cirsium arvense
) were found to be highly aggregated in arable crops with less than
20 % of the fi eld being infested. Grass weeds covered on average 30-40 % of the
fi elds at infestation levels higher than the economic thresholds and annual broad-
leaves between 20 and 90 % (Timmermann et al.
2003
; Gerhards and Oebel
2006
).
Site-specifi c weed management needs patch sprayers as well as automatic and
real-time sensors for weed detection. The objective of this study is to describe the
state-of-the-art and evaluate current
patch spraying systems
.
10.2
Weed Mapping
Weed seedling distribution in the fi eld was usually assessed using discrete weed
mapping or continuous-area sampling (Rew and Cousens
2001
). In most studies,
discrete
weed mapping
was applied in a regular sampling grid that was established
in the fi eld. The side length of the squared grid varied from a few meters up to
approximately 50 m and depended on the width of the spray boom used for site-
specifi c herbicide application. Density and/or coverage of emerged weed seedlings
were counted and measured prior to and after post-emergence herbicide application
in a sampling frame placed at all grid intersection points.
Effi cacy of weed control
was determined relating weed density after post-
emergence herbicide application to prior herbicide application. Different mapping
programs have been applied to characterize spatial distribution of weeds within
fi elds. Maps differed based on the interpolation method that was applied, the area
sampled and the distance between sampling points (Isaaks and Srivastava
1989
;
Johnson et al.
1995
; Rew and Cousens
2001
; Gerhards et al.
1997
). Geostatistics
and interpolation methods were applied to overcome the problem of discontinuities
between adjacent sampling points that result from grid sampling. Interpolated weed
maps were reclassifi ed based on weed infestation levels (Gerhards et al.
1997
).
Most weed patches will be detected when sampling grids are not wider than 6 × 6 m
(Gerhards and Oebel
2006
).
A weed treatment map was derived from the weed distribution maps using weed
control thresholds to provide a decision rule for the patch sprayer (Fig.
10.1
).
