Monitoring defoliation by Chrysodeixis includens in soybean using proximal hyperspectral imaging

Chrysodeixis includens is the most important defoliator in Brazilian soybean crops. It is a laborious task to monitor the fields, searching for this pest. Therefore, we propose the use of proximal hyperspectral imaging to establish reflectance profiles that correspond to this injury. We used 60 potted soybean plants, grown until phenological stage V3-V4. Then, each pot was kept in a cage, where artificial infestation took place. We tested four infestation levels: 0, 2, 5 and 10 caterpillars per cage, for 10 days. At day 0, 5 and 10 after infestation, one leaflet from each plant was taken to the lab to be imaged using a Resonon Pika-L hyperspectral camera, that covers the 400 - 1000 nm spectral range. Also, the leaflets were measured for their foliar area, to compare the percentage of defoliation between the infestation levels. The images were processed and reflectance data were analyzed using a multivariate approach, including a Principal Component Analysis. Our results showed that at day 0, no difference was observed between the plants' reflectance nor the foliar area. At day 5, although the visual diference between the plants was not too intense, the reflectance data showed that the plants could be separated considering their infestation levels. At day 10, there was a great difference between the plants' reflectance from each infestation level. Also, the foliar area was different because of the intense defoliation by the caterpillars. Hence, we conclude that proximal remote sensing can be used to monitor defoliation in soybean plants.