Browsing by Subject "Weed mapping"

Now showing 1 - 1 of 1

Investigations on site-specific weed management for a decision support system for patch spraying
(2012) Gutjahr, Christoph; Gerhards, Roland
During the past five years, powerful sensor technologies have been developed which are capable of classifying weed species in digital images based on shape features and which allow assessing weed seedling distributions automatically in arable crops. Classification algorithms have been computed based on shape features to differentiate between the most abundant weed species in winter wheat, winter barley, maize and sugar beet. Those cameras were used in combination with GPS and GIS-technologies to create weed distribution maps or they can be mounted in front of a sprayer to detect and spray weed patches in real-time. It has been shown in previous studies that patch spraying, based on weed distribution maps and simple decision rules for herbicide application significantly reduces the amount of herbicides needed. Therefore, site-specific weed management practices have economic and ecological benefits by reducing the amount of herbicides applied. It has further been shown that populations of Galium aparine and Alopecurus myosuroides did not significantly change in location and size when site-specific weed control methods were applied over a period of 8 years. However, precise decision rules for site-specific weed management are still lacking. The objectives of this study were to derive and verify decision rules for site-specific weed management in winter annuals grains and maize. This study includes three work packages: In the first work package, weed species were grouped into three classes based on their competitive ability and sensitivity to herbicides. The first group contained annual grasses, the second group annual dicotyledons and the third group perennial weed species. Weed distribution maps were created for all groups of weed species in winter wheat, winter barley, maize and sugar beets. It was then analysed at which locations in the field weed control measures were warranted and which herbicides and combinations of herbicides were required. Weed control measures were realized with a multiple tank sprayer and spatial and temporal stability of weed patches was assessed in the following year. In the second work package, a so-called Precision Experimental Design using Precision Farming technologies and Geographic Information System, was applied in maize, winter barley and winter wheat to determine the effects of each weed species group, soil variability and herbicide application on grain yield separately. Data of these experiments were used to calculate yield loss functions for individual weed species. In the third work package, the structure of a decision support system for site-specific weed control was created including yield loss function and dose-response curves for the most relevant weed species in winter wheat and maize. The results of the three work package can be summarized as followed: All weed species and weed classes were distributed heterogeneously within the fields with densities ranging from 0 to more than 200 plants m-2. Patch spraying resulted in 30-40% herbicide saving when a tank mixtures of all herbicides needed was applied. Savings of 77% were achieved when a three tank sprayer was used to apply each herbicide at different locations. For the Precision Experimental Design, a linear mixed model with spatial correlation structure has been modified and fitted to the data. It was found that competition of E. crus-galli resulted in significant yield losses of 0.027 t ha-1 plant m-2 in maize and G. aparine in 0.034 t ha-1 plant m-2 yield loss in winter wheat. However, herbicides against grasses and annual dicotyledons also reduced grain yield by approximately 0.3 t ha-1, which again underlines the necessity to save herbicides at location where no or only few weed species are present. ?HPS Online? describes a possible structure of a decision support system for patch spraying. The combination of yield loss functions for the most abundant weed species/group of species with dose response curves for the most relevant herbicides to control these species allowed determining the most economic weed control strategy at each location in the field. It is recommended to include weather conditions or historical data of the fields, if available, such as maps of perennial weed species to optimize weed control decisions. In conclusion of the results, precision weed management offers a great potential for herbicide savings in arable crops. It requires the combination of automatic sensor technology for weed detection, a decision support system for weed control and application technology to vary the herbicide mixture in real-time.