Crop stress analysis based on hyperspectral vegetation index

Vegetation attributes are typically measured by converting reflectance spectra into a single numerical value or vegetation index. Hyperspectral or narrow-band vegetation indices contain narrow-band vegetation characteristics and wavelength information that can only be captured by hyperspectral instruments.

Vegetation properties determined by hyperspectral vegetation indices can be divided into three main categories: 1. Structure; 2. Biochemical properties; 3. Plant physiology/stress.

The structural properties measured include vegetation cover, leaf biomass, leaf area index (LAI), biomass of senescent leaves, and photosynthetically active radiation absorptive ratio (FPAR). Most indices currently used for structural analysis are constructed in broadband systems, while narrow-band hyperspectral indices also exist.

Biochemical properties include water content, pigments (chlorophyll, carotenoids, and anthocyanins), other nitrogen-rich mixtures (such as proteins), and plant structural materials (lignin and cellulose). Physiological and stress indices are used to measure minute changes caused by factors such as stress-induced changes in xanthophyll status, chlorophyll content, fluorescence effects, or leaf humidity. Generally, biochemical properties and physiological/stress indices are mostly constructed through laboratory and field instrument measurements (spectral sampling intervals ≤10 nm) and are highly specific spectral characteristics. Therefore, these indices are strictly hyperspectral indices.

Compared to broadband vegetation indices, hyperspectral vegetation indices have bands not sampled in broadband systems and specific band centers, enabling them to identify and extract more subtle absorption features, thereby better estimating crop attributes.

Compared to point spectral measurements, hyperspectral imagery adds spatial dimension information, which is crucial for precision agriculture. It can support commonly used analytical methods in precision agriculture, such as soil and crop clustering, hotspot and outlier analysis, and can introduce spatial information at different levels to help estimate crop attributes. It can reveal the spatial distribution patterns and characteristics of crop attributes from point to surface, providing a more powerful tool for the engineering practice of precision agriculture.

Methods and Principles

We used hyperspectral images of the Xi'an area taken by Xiguang-1 05 satellite on April 6, 2025, to calculate various hyperspectral vegetation indices. Based on the structural/biochemical/physiological indicators corresponding to the indices, we conducted crop health and stress analysis of winter wheat in Xi'an during the jointing stage in the 2024-2025 season, and identified potentially stressed areas.

Figure 1. Data Introduction

process:

Radiation calibration:The original DN value data is radiometrically corrected using radiometric calibration coefficients to obtain the apparent radiance data.

Atmospheric correction:Atmospheric correction was performed using the 6S model to obtain surface reflectance data.

Spectral smoothing:The Savitzky-Golay (SG) algorithm was used to smooth the surface reflectance data, which improved the reliability of subsequent red edge correlation index calculation while reducing spectral noise.

Image registration:The processed data is then georeferenced based on the base map.

Index calculation:The hyperspectral vegetation index was calculated from the registered data.

Vegetation area masking/plot extraction:Using the calculated partial indices, vegetation area masking was extracted using a fixed threshold method. The previously completed vector dataset of wheat-growing plots in the Yellow River Basin was also utilized.

Regional statistics:Based on the land parcel vector, each index is statistically analyzed by region, and the statistical method is the mean.

Multi-index comprehensive analysis:Based on statistical vectors, and referring to the physical meaning represented by different indices, the distribution of different indices is analyzed to explain some phenomena.

The table below lists hyperspectral vegetation indices, categorized into structural, biochemical (pigment, chlorophyll, anthocyanin, carotenoid, water, lignin and cellulose, nitrogen) and physiological (light use efficiency, stress). Each index includes its Chinese name, English abbreviation, and formula, which is based on narrow-band reflectance (Rλ represents reflectance at wavelength λ, in nm).

Table 1. Some Hyperspectral Vegetation Indices and Calculation Formulas

Results Evaluation

True-color composite image - Xiguang-1 05 satellite (celestial match) - 20250406

Reflectance curves of different ground features - reflectance magnified 10,000 times - Xiguang-1 05 satellite (Tianxianpei) - 20250406

Normalized Difference Vegetation Index (NDVI) Distribution Map - Xiguang No. 1 05 Satellite (Tianxianpeihao) - 20250406

Photochemical Reflectance Index (PRI) Distribution Map - Xiguang-1 05 Satellite (Tianxianpei) - 20250406

Improved Normalized Red-Edge Index Distribution Map - Xiguang-1 05 Satellite (Tianxianpei) - 20250406

Distribution map of structure-insensitive chlorophyll index (SIPI) - Xiguang No. 1 05 satellite (Tianxianpeihao) - 20250406

Vogelmann Red Edge Index 1VOG1 Distribution Map - Xiguang No.1 05 Satellite (Tianxianpeihao) - 20250406

Atmospheric Impedance Vegetation Index (ARVI) Distribution Map - Xiguang-1 05 Satellite (Tianxianpeihao) - 20250406

Anthocyanin reflectance 2ARI2 distribution map - Xiguang-1 05 satellite (Tianxianpei) - 20250406

Carotenoid reflectance index 2CAI2 distribution map - Xiguang No.1 05 satellite (Tianxianpeihao) - 20250406

Most of the aforementioned hyperspectral vegetation indices are also based on the differences in absorption/reflectance between different bands. However, compared with traditional broadband indices (such as NDVI), they can more sensitively reflect the concentration and distribution of specific pigments, increasing the accuracy of crop classification, phenological stage determination, crop physiological and biochemical index inversion, and crop stress assessment. For example, for winter wheat, anthocyanins gradually accumulate from the tillering stage to maturity, so the current phenological stage of the crop can be determined by the anthocyanin index. When subjected to environmental stress, the expression of anthocyanin biosynthesis-related genes (such as CHS, CHI, and F3H) is induced, leading to an increase in anthocyanin content. At the same time, the red edge position also shifts towards shorter wavelengths. Therefore, a comprehensive stress assessment can be carried out through multiple photochemical indices and red edge morphology.

Distribution map of anthocyanin reflectance index 2ARI2 at the plot level - Xiguang No.1 05 satellite (Tianxianpeihao) - 20250406

Land Parcel-Level Red Edge Location Index (REPI) Distribution Map - Xiguang No. 1 05 Star (Heavenly Match Number) - 20250406

By combining the land parcel boundary vectors, more accurate parcel-level analysis can be performed, thereby identifying potential problem areas and corresponding parcels, and then conducting more detailed statistical analysis.

Localized Red Edge Vegetation Index (RENDVI) Distribution Map - Xiguang No. 1 05 (Tianxianpeihao) - 20250406

Localized distribution map of carotenoid reflectance index (CRI2) - Xiguang-1-05 satellite (Tianxianpeihao) - 20250406

Local anthocyanin reflectance 2ARI2 distribution map - Xiguang-1 05 satellite (Tianxianpei) - 20250406

Distribution map of local photochemical reflectance index (PRI) - Xiguang-1 05 satellite (Tianxianpei) - 20250406

The above shows a local plot-level index distribution map. In the central region of the map, the red-edge vegetation index (RENDVI), carotenoid index (2CRI2), and photochemical reflectance index (PRI) are relatively high, while they are lower in the western region. This indicates that the pigment content in the western region is also relatively low. However, in the anthocyanin index (2ARI2) distribution map, the western region is similar to the central region and is higher, indicating that this region may be under the same stress, but the stress has a relatively smaller impact on the central region, possibly related to the increased carotenoid concentration in the central region. Carotenoids are photosynthetic co-pigments and, as antioxidants, can scavenge reactive oxygen species (ROS, such as superoxide anion and hydrogen peroxide), protecting cell membranes and photosynthetic structures from oxidative stress, thus not leading to a decrease in chlorophyll concentration. For the western region, however, no significant increase in the carotenoid reflectance index was observed, indicating that this type of stress has a greater impact.

By comprehensively analyzing multiple hyperspectral vegetation indices and combining them with crop canopy and soil moisture distribution maps, we can further determine the type of stress and its impact, enabling agricultural interventions at an earlier stage to improve the environmental resilience of large-scale planting systems and ensure the stability of grain yield and quality.