Using Precision Agriculture to Control Herbicide-Resistant Weeds in Brazil

Using Precision Agriculture to Control Herbicide-Resistant Weeds in Brazil

Region, in red, infested with resistant weeds in Brazil (Embrapa, 2015).


The occurrence of weeds in agricultural fields of grain and fiber production has caused losses to farmers for a long time. With the introduction of resistance genes in cultivated species such as soybean and cotton, many believed that this problem would be solved. In Brazil, glyphosate-resistant soybeans have been cultivated since 2003. But after almost 15 years, the problem with weeds has not been solved and has regained importance, mainly due to the occurrence of glyphosate-resistant weeds. As can be seen in the map, the problem is spreading in the main producing regions of Brazil, mainly in the areas where glyphosate-resistant crops, such as soybean, corn and cotton have been cultivated.

The photo below illustrates a common scene in Brazilian fields, the presence of three species of glyphosate-resistant weeds in the same area (Digitaria insularis, Conyza canadensis, and Eleusine indica), plus a dozen other species, but with a spatial occupation of less than 50% of the total area. Naturally, weeds are not evenly distributed throughout the fields. Most of the time they are aggregated in reeds mainly due to the dispersion form of the seeds. In others, the outbreaks may be very sparse or isolated plants.


A common scene in Brazilian fields: the presence of three species of glyphosate-resistant weeds in the same area (Digitaria insularis, Conyza canadensis and Eleusine indica), plus a dozen other species.

Due to this spatial variability, there is a great potential for localized herbicide application. Real-time localized weed spraying is based on plant identification by sensor and instant application of herbicide only on the target. The process of identifying a plant is performed by recognizing a reflectance pattern when subjected to a radiation source. In this case, the sensors are generally “active” because they have their own source of radiation, allowing them to work both day and night.

Currently, there are two commercial equipment in Brazil that perform the identification and spraying of weeds in real time: WEEDit and WeedSeeker. For the detection of plants, the first system is based on the technique of detection of chlorophyll fluorescence that is created by the action of a powerful light source, while the second one uses reflectance in two spectral bands. The technology of these tools is not just about the sensors but the extremely fast speed of the valves responsible for opening and closing the nozzles. In the case of WEEDit, Pulse Width Modulation (PWM) technology at a frequency of 60 Hz allows the right rate of herbicide to be accurately sprayed over the weed independent of variations in the speed of the machine. This type of technology can bring potential savings varying from 20% to 90% and is directly proportional to the level of weed infestation in the area.


Another benefit of these types of systems is the possibility of carrying out the applications with a higher frequency. Because it will only be sprayed on the weeds, it is not necessary to wait for all the weeds to germinate and to take the risks of low control due to the presence of weeds out of the right control stage. In addition, waiting may give these plants the chance to produce new seeds, which aggravates the problem for future crops. More frequently applications will naturally reduce the seed bank in the area over the years. There are reports from Australian producers with 98% herbicide savings after 7 years using the technology.

One of the limitations of using these tools lies in the limited ability to differentiate plant species. Both of the aforementioned equipment have adjustable detection limits that allow the targeting of larger or smaller targets, yet without the full capacity to differentiate species. Aiming to solve this limitation, there are technologies being developed and tested at the research level, mainly using pattern recognition in RGB images (leaves format) and hyperspectral cameras (reflectance intensity in specific regions of the spectrum). An example of this type of technology is being developed by the American startup Blue River Technology. Their concept of intelligent machine for viewing and spraying makes use of computer vision and artificial intelligence for the differentiation of plant species and application in real time.

The presence of weeds at different infestation levels within a single field makes it possible to use precision farming tools as a more efficient way of weed control through localized application using the right rates required for each part of an agricultural area. The increase in the occurrence of herbicide-resistant weeds generates a greater demand for these technologies, since the traditional forms of control have high costs and low efficiency. The risk of introduction or selection of new resistant species, associated with the slow development of new herbicide molecules, makes correct weed management increasingly important for the maintenance of a sustainable production system. Localized weed control with sensor identification and herbicide application in real time enables large product savings, as well as reducing impacts on the environment and contributing to reducing the problem in the long term. New technologies must be developed to meet these demands in the coming years, as there will certainly be a huge market to be explored.