Characterizing and estimating rice brown spot disease severity using stepwise regression, principal component...

Area of Science:

• Agricultural Science
• Plant Pathology
• Remote Sensing

Background:

• Plant health monitoring is crucial for effective farm pest management and crop protection.
• Rice brown spot, caused by Bipolaris oryzae, significantly impacts crop yield and quality.

Purpose of the Study:

• To assess the feasibility of using hyperspectral leaf reflectance for detecting rice brown spot disease severity.
• To compare the efficacy of different statistical methods in estimating disease severity.

Main Methods:

• Hyperspectral leaf reflectance was measured for healthy and infected rice leaves (Oryza sativa L.) across a 350–2,500 nm wavelength range.
• Disease severity was quantified by estimating the percentage of leaf surface lesions.
• Statistical analyses including multiple stepwise regression, principal component analysis (PCA), and partial least-square regression (PLSR) were employed.

Main Results:

• Multiple stepwise linear regression estimated disease severity using three wavebands with root mean square errors (RMSEs) of 6.5% (training) and 5.8% (testing).
• PCA indicated the first principal component explained ~80% of spectral variance; a regression model with two components yielded RMSEs of 16.3% (training) and 13.9% (testing).
• Partial least-square regression (PLSR) with seven factors demonstrated superior performance, achieving RMSEs of 4.1% (training) and 2.0% (testing).

Conclusions:

• Hyperspectral reflectance data is a viable tool for estimating rice brown spot disease severity at the leaf level.
• Partial least-square regression offers the most effective approach for disease severity estimation compared to other statistical methods.
• This technique holds promise for advancing precision agriculture and disease management strategies in rice cultivation.