Retrieval of crop biophysical and biochemical traits using airborne hyperspectral observations and inversion of radiative transfer models over diverse cropping regions

Accurate estimation of vegetation biophysical and biochemical parameters is essential for effective crop monitoring, agricultural management and yield assessment. Among these parameters, Leaf Area Index (LAI) and Leaf Chlorophyll Content (LCC) are key indicators of crop canopy structure and physiological status. Hyperspectral remote sensing offers significant potential for retrieving such parameters due to its ability to capture detailed spectral signatures associated with vegetation biochemical and structural properties. In this study, high-resolution airborne hyperspectral imagery acquired from the Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) sensor was used to retrieve LAI and LCC across four different agro-ecological regions of Gujarat, India viz. Anand, Bharuch, Surendranagar and Talala. The study integrated the PROSAIL radiative transfer model with machine learning inversion techniques viz. Canonical Correlation Forest, Random Forest and Support Vector Regression to estimate vegetation parameters from hyperspectral reflectance data. A Gaussian Process Regression (GPR) based band selection approach was employed to identify the most informative spectral wavelengths for parameter retrieval. The results indicated that twenty-two and six spectral bands provided optimal performance for LAI (NRMSE = 6.96) and LCC (NRMSE = 4.72) estimation, respectively. Among the evaluated machine learning models, canonical correlation forest achieved the highest predictive accuracy, with R2 values ranging from 0.84 to 0.98 for LAI and exceeding 0.92 for LCC. The retrieved parameter maps revealed significant spatial variability in crop canopy structure and chlorophyll distribution across the study regions. The findings highlighted strong potential of integrating airborne hyperspectral data, radiative transfer modelling and machine learning approaches for reliable retrieval of crop biophysical and biochemical parameters and improved agricultural monitoring.