Full-Wave Optical Modeling of Leaf Internal Light Scattering Dynamics with Potential Applications for Early Detection of Foliar Fungal Disease

TL;DR

This work addresses the limitation of traditional leaf optics models that rely on ray-tracing and radiative-transfer approximations, which fail to capture diffraction and coherent scattering when leaf microstructures are on the order of the wavelength ($D/\lambda \sim 1$). It introduces a GPU-accelerated, full-wave electromagnetic framework based on 2D FDTD to simulate light propagation through anatomically informed leaf cross-sections across $400$--$2{,}500$ nm, decomposing illumination into $S$- and $P$-polarizations and computing reflectance $R(\omega)$ and transmittance $T(\omega)$. The framework is validated against PROSPECT-PRO for healthy leaves, achieving high concordance ($CCC_{dicot}=0.8962$, $CCC_{monocot}=0.7849$) and revealing wave phenomena beyond ray-based models. It further demonstrates early-stage necrotrophic fungal infection by modeling melanized hyphae, showing reduced green reflectance and suppressed NIR reflectance plateau consistent with experimental observations, thereby establishing a physics-based link between microscale pathology and hyperspectral signatures. The approach enables physics-based pre-symptomatic disease detection from hyperspectral data and provides a foundation for future extensions to 3D geometry, dispersive materials, inverse retrieval, and multiscale canopy integration for precision agriculture.

Abstract

Light interacting with plant leaves undergoes reflection, transmission, scattering, and absorption, which together determine leaf optical properties. Changes in leaf architecture disrupt internal light scattering dynamics and consequently affect photosynthetic performance. Previous studies on internal leaf light scattering have primarily relied on ray-tracing approaches (e.g., Raytran) or radiative-transfer models (e.g., PROSPECT). However, these high-frequency approximations cannot capture diffraction and coherent multiple scattering in wavelength-scale leaf tissues, unlike full-wave electromagnetic simulations. Here, we employ GPU-accelerated Finite-Difference Time-Domain (FDTD) simulations to model internal light scattering dynamics using segmented cross-section image geometries of representative dicot and monocot leaves with wavelength-dependent complex refractive indices. The simulations accurately reproduce the reflectance and transmittance characteristics of healthy leaves, showing strong agreement with the PROSPECT model, with average Lin's concordance values of 0.8962 for dicot leaves and 0.7849 for monocot leaves. We further simulate early-stage necrotrophic fungal infection by modeling melanized hyphae penetrating the cuticle and upper epidermis. Diseased leaves exhibit a pronounced reduction in visible green reflectance and a marked suppression of the near-infrared reflectance plateau, consistent with experimental observations. Remaining discrepancies in the visible band are expected to be reduced through more advanced geometric and material modeling. This proof-of-concept study presents a full-wave FDTD optical modeling framework for plant-leaf light scattering, enabling physics-based analysis of internal scattering before and after early-stage fungal penetration and supporting the use of light scattering as an indicator for pre-symptomatic plant fungal disease detection.

Figures (22)

• Figure 1: Comparison of light scattering patterns (a) without and (b) with a wedge. The wedge introduces strong diffraction and interference effects that cannot be described by ray-based approximations.
• Figure 2: Schematic of the FDTD simulation domain showing the cross-section of a plant leaf, total-field/scattered-field (TF/SF) regions, equivalent plane-wave excitation, field collection lines for reflectance and transmittance calculations, and boundary conditions (periodic boundaries along $x$ and perfectly matched layers along $y$).
• Figure 3: Schematic of the two-dimensional Yee grid illustrating the field component arrangement for (a) TEz and (b) TMz polarizations.
• Figure 4: Schematic illustration of the leapfrog time-stepping schemes for TEz and (b) TMz polarizations.
• Figure 5: Procedure for reconstructing the internal geometry from microscope images of (a) dicot and (b) monocot leaf samples.