Weed population and crop yield response to recommendations from a weed control decision aid

Selecting among the diverse options available to manage weeds requires vari ed knowledge, This study field-tested the performance of a decision aid mod el that integrates biological and economic factors to make weed control rec ommendations. The model was parameterized using data from the literature an d subjected to sensitivity analysis. Tactics generated from the resulting m odel were compared with standard-herbicide weed control in continuous corn (Zea mays L.) having a long-term history of various tillage and herbicide a pplication regimes. The model recommended preemergent and postemergent weed control based on weed seed and weed seedling densities, respectively. Usin g standard-herbicide weed control resulted in fewer weed seeds in the soil each year compared with model-based weed control. Variation among model-bas ed weed control treatments applied to plots with different management histo ries was eliminated over the course of the study as weed populations shifte d in response to weed control practices, The result was a 98% relative abun dance of foxtail (Setaria spp.) seedlings after 3 yr of model-based treatme nts, compared with 70% in standard-herbicide treatments. In 1994, the model reduced herbicide use and weed control treatment costs, but often reduced corn yields and net economic returns. In 1995. model-based treatments maint ained or increased corn yields and net returns. In 1996, however, model-bas ed treatments consistently reduced corn yields and net returns. Variability in control treatment effectiveness and weed-crop interaction greatly affec ted the performance of the model. The use of decision-aid models in weed ma nagement is still a developing technology. Interactions of the weed seed ba nk, weed emergence, treatment effectiveness, weed-crop interference, and en vironmental conditions are very complex, The model could become a valuable tool for producers if consistency of weed control were improved by better u nderstanding and modeling of threshold levels and accounting for weed seed production and by adding data on weed emergence, seed production, yield los s, and control efficacy.