online proximal hyperspectral detection of fusarium head blight in winter wheat
An innovative online (tractor-driven) visible and near-infrared (vis-NIR) hyperspectral imager that collects geo-located plant measurements at high sampling resolution is available at Ghent University. The sensor provides a wide range of valuable crop health information for the farmer, based on a single data collection exercise. The spectra of plants is captured at a height of around 50 cm above the canopy. Once modelled the online measurements are capable of detecting fusarium head blight (FHB) disease within a crop stand, detecting disease variability and severity through the field. This proximally sensed information can be used to map and classify FHB infection areas to generate zones of variable fungicide treatments and selective harvest. Selective harvest is appropriate for FHB infection as the fungal disease is linked to mycotoxin presence in the grain, which is toxic to human health. Variable fungicide applications help reduce the quantity of fungicide required by applying the fungicides only where there is disease risk. Variable fungicide treatments can lead to greater profits for the farmer, and reduce environmental damage, helping to improve sustainable farm production, and ensuring maximum yield for human consumption.
The use of hyperspectral imagery for crop disease detection is gaining popularity in glasshouse and laboratory research (Nagasubramanian et al., 2019; Thomas et al., 2018). The methodologies we have been developing move the sensors into the field to an online approach, allowing detection of crop diseases in-situ. This helps increase the resolution of disease detection in the field and provides a non-destructive approach to crop disease sensing (Munnaf et al., 2021; Pane et al., 2021). The hyperspectral imager is mounted on either the side of the tractor or on the boom of a sprayer, the camera is positioned at least 1 meter into the crop stand, away from the tramlines, this allows a good representation of the crop. A differential global positioning system (Trimble AG., Trimble Navigation Ltd., Sunnyvale, Canada) recorded the corresponding latitudes and longitudes during online crop scanning and sampling. The GPS and speed of the camera are recorded, along with time stamps of the hyperspectral image, this allows the spectral data to be geo-located across the field. The crop spectra are recorded through (Lumion, Planmeca, Helsinki, Finland) software, installed on a semi-rugged laptop computer (Toughbook, Panasonic UK Ltd., Bracknell, UK). The hyperspectral imager is driven at 3 km/h forward speed along each tramline.
Crop field assessment of disease severity is necessary to appropriately apply fungicide treatments. Fungicide treatments are usually broadcast through a field and are traditionally based on visual evaluations of crop health by walking and inspecting the field. This process requires expert knowledge and is time-consuming. Lab-based measurements are an advancement on the traditional approach, however, the resolution depends on destructive sampling, which can be time-consuming, and has little ability to move into real-time detection. To overcome the bottleneck of data acquisition by traditional and lab measurements, online measurement through proximal sensing is required. Inherited spatial variations in a crop stand due to both environmental and soil properties lead to different disease pressures within a crop stand, and broadcast treatment is not always necessary (Schisler et al., 2007; Xia et al., 2021). Broadcast treatments lead to excess fungicide being applied, which is costly for the farmer and damaging to the environment (Pazdiora et al., 2021; Wegulo et al., 2015). Mycotoxin contaminations from FHB infection can reduce grain quality. The severity of mycotoxin contamination causes the grain to be downgraded at the market, which means grain originally intended for human consumption can be downgraded to cattle feed or in severe cases even biofuels (McMullen et al., 2012; Wegulo et al., 2015). This reduces both the farmers' profit and the amount of food suitable for human consumption. High-resolution information on crop disease severity can permit a precision approach for fungicide applications and selective harvest. Considering both the negative environmental impacts from fungicides and the growing demand for food to meet the population increase. It is essential that methods are developed to maximize the amount of high-quality grain reaching the human food market through a sustainable approach.
Through high-resolution data collection, via online hyperspectral imaging, accurate maps can be produced of FHB disease spread and severity. Improving the resolution of information on the health of a field, compared to the traditional “walking the field” method. Furthermore, the proximal camera attached to a tractor can reduce data collection time, improve data resolution, and is non-destructive compared to the laboratory hyperspectral methods. Although digital agricultural tools exist commercially, the majority of farmers apply different forms of management processes homogeneously over the entire field area. The high-resolution FHB maps can be used for variable fungicide applications and selective harvest methods. These two precision approaches to farm management decisions can help improve the farmers' income, reduce the amount of fungicide needed, and maximize the high-quality grain available for human consumption. This will also help reduce the risk of mycotoxin-contaminated grain entering the market for human consumption. Variable rate applications allow fungicides to be applied in target areas, where high-risk areas can receive the full fungicide dose, and the lower-risk areas can then receive a reduced rate. This helps reduce the quantity of fungicide applied whilst maintaining disease control (Dammer, 2010; Whetton et al., 2018). Selective harvest methods in response to disease severity maps ensure high-quality grain can be harvested separately from mycotoxin-contaminated grain. Allowing the high-quality grain to be sold for human consumption, maximizing the farmers' financial return, along with ensuring a high yield reaches the growing population's food demand. The downgrading of a harvest can occur when small areas of the field are of poor quality/high in mycotoxins. Selective harvest removes the risk of downgrading a field's entire yield from human consumption to cattle feed or biofuels when sold.
Currently, we are in the development stage, we have a suitable method, and we are trialling the approach which was successful in the U.K. and Belgium in further winter wheat fields in Spain, Belgium and Lithuania, and for barley in Lithuania.
