Airborne hyperspectral imaging of cover crops through radiative transfer process-guided machine learning

Authors: WANG, SHENG - University Of Illinois; GUAN, KAIYU - University Of Illinois; ZHANG, CHENHUI - University Of Illinois; JIANG, CHONGYA - University Of Illinois; ZHOU, QU - University Of Illinois; LI, KAIYUAN - University Of Illinois; QIN, ZIQI - University Of Illinois; Ainsworth, Elizabeth - Lisa; HE, JINGRUI - University Of Illinois; WU, JUN - University Of Illinois; SCHAEFER, DAN - Illinois Fertilizer And Chemical Association; GENTRY, LOWELL - University Of Illinois; MARGENOT, ANDREW - University Of Illinois; HERZBERGER, LEO - University Of Illinois

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/19/2022
Publication Date: 12/8/2022
Citation: Wang, S., Guan, K., Zhang, C., Jiang, C., Zhou, Q., Li, K., Qin, Z., Ainsworth, E.A., He, J., Wu, J., Schaefer, D., Gentry, L.E., Margenot, A.J., Herzberger, L. 2022. Airborne hyperspectral imaging of cover crops through radiative transfer process-guided machine learning. Remote Sensing of Environment. 285. Article 113386. https://doi.org/10.1016/j.rse.2022.113386.
DOI: https://doi.org/10.1016/j.rse.2022.113386

Interpretive Summary: Cover cropping is an essential conservation practice that can significantly benefit sustainable agriculture and has been highly promoted in intensive cropping regions around the world. However, the potential to leverage remote sensing data to quantify the benefits of cover cropping poorly understood. In particular, the retrieval of cover crop growth variables in the U.S. Corn Belt is challenged by mixed signals attributable to low vegetation biomass of cover crops and mixed signals from soil and vegetation. This study demonstrates airborne full-optical-range (400–2400 nm) hyperspectral imagery with state-of-the-art Process-Guided Machine Learning (PGML) to accurately quantify cover crop aboveground biomass and nitrogen content in the central U.S. Corn Belt. The new algorithms have potential for accurate monitoring of cover crop growth, remote sensing retrieval of vegetation traits, and have potential societal impacts on assessing outcomes of cover crop conservation programs.

Technical Abstract: Cover cropping between cash crop growing seasons is a multifunctional conservation practice. Timely and accurate monitoring of cover crop traits, notably aboveground biomass and nutrient content, is beneficial to agricultural stakeholders to improve management and understand outcomes. Currently, there is a scarcity of spatially and temporally resolved information for assessing cover crop growth. Remote sensing has a high potential to fill this need, but conventional empirical regression operated with coarse-resolution multispectral data has large uncertainties. Therefore, this study utilized airborne hyperspectral imaging techniques and developed new process-guided machine learning (PGML) for cover crop monitoring. Specifically, we deployed an airborne visible to shortwave infrared (400–2400 nm) hyperspectral system to acquire high spatial (0.5 m) and spectral (3–5 nm) resolution reflectance over 23 cover crop fields across Central Illinois in March and April of 2021. Airborne hyperspectral surface reflectance with high spectral and spatial resolution can be seamlessly matched with field data to quantify cover crop traits. Furthermore, we developed the PGML models pre-trained by synthetic data generated from soil-vegetation radiative transfer modeling, and then fine-tuned with field-based cover crop biomass and nutrient content data. Results show that airborne hyperspectral data with PGML can achieve high accuracy to predict cover crop aboveground biomass (R2 = 0.80, relative RMSE = 13.99%) and nitrogen content (R2 41 = 0.70, relative RMSE = 15.32%). Unlike the pure data-driven approach (e.g., partial-least-squares regression), PGML incorporated radiative transfer knowledge and obtained higher predictive performance with fewer field data. Meanwhile, with field data for model fine-tuning, PGML predicted biomass more accurately than the numerical inversion of radiative transfer models. We also found that the red edge has a high contribution in quantifying aboveground biomass and nitrogen content, followed by green and shortwave spectra. This study demonstrated the first attempt of utilizing hyperspectral remote sensing to accurately quantify cover crop traits. We highlight the strength of PGML in exploiting sensing data to advance agroecosystem monitoring for sustainable agricultural management.