Formats
Format
BibTeX
MARC
MARCXML
DublinCore
EndNote
NLM
RefWorks
RIS

### Abstract

A user-friendly computerized decision model has been developed for selecting profitable site-specific herbicide applications in winter wheat. The model is based on six years of field research in southeastern Washington State, USA. The model calibrates herbicide applications to weed densities, soil properties, and preceding management, as well as to expected input and output prices. The model increased broadleaf herbicide rates by an average of 0.65 label rates compared to the recommendations by farmers and weed science professionals, but cut the more expensive grass herbicides by an average of 0.56 label rates. The model increased average projected profitability, excluding model application costs, by 65 percent. Both the model and the cooperating farmers properly chose no grass herbicides for the study sites, but weed science experts chose up to 1.0 label rates. The estimated payoff from using the model substantially exceeded the cost of weed scouting and other information collection. Determining economically optimal sampling and management units is an important challenge for precision agriculture models like this one. The computerized site-specific herbicide decision model for winter wheat reported here (Kwon et al. 1998) was based on six years of large-plot experimental data in the Palouse region of eastern Washington State, USA. The model proved easy to use and showed potential to substantially increase profit while reducing postemergence grass, but not broadleaf, herbicides in the study region. The model increased broadleaf herbicide rates by an average of 0.45 to 0.91 label rates compared to competing recommendations, but reduced the more expensive grass herbicides by an average of 0 to 1.0 label rates. The projected costs of weed control using the model were slightly higher than for the farmer and extension recommendations, but much lower than the weed scientist and label rate recommendations. On average, the model recommendations boosted projected profitability (which accounted for yield and revenue increases as well as cost changes) by 65% compared to the farmer, extension consultant, weed scientist and label rate recommendations. The estimated $6 ha-1 cost for using the weed decision model could be easily absorbed by the model's projected profitability advantages which ranged from$39 to \$185 ha-1, but the costs of weed monitoring and adjusting herbicide application to irregular subfields might be higher in real world conditions. More research is needed on cost effective monitoring of weed densities and other site characteristics and for adjusting herbicides to subfield management units. The authors believe the weed decision model described in this paper represents a substantial improvement over an earlier version (Kwon et al., 1995; Kwon et al., 1998). The early model performed well in field validations and we expect the revised model to also perform well in further field testing. Future research will also examine cost effective procedures for defining management and sampling units. Successive field validation and development of affordable implementation procedures will remain important steps to promote adoption of precision agriculture tools.