An assessment of the use of site-specific weed control for improving prediction-based management decisions and automating on-farm research

Prediction of the competitive effect of a weed on a crop has generally been accomplished by extrapolating the results of small-plot experiments. I investigated the potential of site-specific technologies to improve the predictability of competitive response to chemical inputs by combining the concepts of management and experiment. The spatial distribution of wild oat (Avena fatua L.) was mapped in four north-central Montana study sites in spring wheat production. Two herbicide treatment strategies, site-specific and broadcast, were applied in a large-scale block design and contrasted using site-specific yield data. I found that the proportion of the field requiring treatment determined the short-term profitability of site-specific weed management. I also determined that sitespecific on-farm research was feasible but experiments could not distinguish treatment effects unless the effects were relatively large, primarily because the resolution of the current generation of yield monitors is poor. Using supplemental quadrat count data, I parameterized a relationship describing the competitive effect of wild oat on wheat using the site-specific geld data. I suggest that creating locally parameterized stochastic functions may improve the accuracy of input response prediction over regional small-plot experiments because of field-specific effects, which are not calibrated by regional small plot experiments.; I examined the potential of improving small-plot study predictions by adding precipitation to weedcrop impact model structure. When using simulated small-plot competition study data and March through May precipitation, I found that the threshold (value of weed control) could not be predicted accurately, even when we had knowledge of the correct structural form of the weed-crop competitive relationship.; I also investigated the importance of weed demography in weed control decisions and developed an analytical approximation of the long-term economic threshold based on the growth rate of the weed with and without control. I suggest that site-specific weed management and on-farm studies may offer a means to estimate demographic model parameter distributions.