Hyperspectral measurements of yellow rust and fusarium head blight in cereal crops: Part 2: On-line field measurement

Abstract

Yellow rust and fusarium head blight cause significant losses in wheat and barley yields. Mapping the spatial distribution of these two fungal diseases at high sampling resolution is essential for variable rate fungicide application (in case of yellow rust) and selective harvest (in case of fusarium head blight). This study implemented a hyperspectral line imager (spectrograph) for on-line measurement of these diseases in wheat and barley in four fields in Bedfordshire, the UK. The % coverage was assessed based on two methods, namely, infield visual assessment (IVA) and photo interpretation assessment (PIA) based on 100-point grid overlaid RGB images. The spectral data and disease assessments were subjected to partial least squares regression (PLSR) analyses with leave-one-out cross-validation. Results showed that both diseases can be measured with similar accuracy, and that the performance is better in wheat, as compared to barley. For fusarium, it was found that PIA analysis was more accurate than IVA. The prediction accuracy obtained with PIA was classified as good to moderately accurate, since residual prediction deviation (RPD) values were 2.27 for wheat and 1.56 for barley, and R2 values were 0.82 and 0.61, respectively. Similar results were obtained for yellow rust but with IVA, where model performance was classified as moderately accurate in barley (RPD = 1.67, R2 = 0.72) and good in wheat (RPD = 2.19, R2 = 0.78). It is recommended to adopt the proposed approach to map yellow rust and fusarium head blight in wheat and barley.

Plant phenotyping relevance

ハイパースペクトルラインイメージャと画像評価・PLSRを用いて、作物上の病害被覆率を測定・検証する方法が研究の中心であり、植物病害状態の表現型計測に該当する。

Code and data availability

The supplied article blocks describe hyperspectral on-line field measurement of yellow rust and fusarium head blight with PLSR analysis, but contain no data availability statement, public dataset deposit, author code release, or repository URL. The spectral data, RGB images, disease assessments, and PLSR models are not

No evidence-backed public reproduction asset is currently recorded.