Remote sensing-based estimation of advanced perennial grass biomass yields for bioenergy

Authors: HAMADA, YUKI - Argonne National Laboratory; ZUMPF, COLLEEN - Argonne National Laboratory; CACHO, JULES - Argonne National Laboratory; LEE, DOKYOUNG - University Of Illinois; LIN, CHENG-SHIEN - University Of Illinois; HEATON, EMILY - University Of Illinois; BOE, ARVID - South Dakota State University; BOERSMA, NICK - Iowa State University; Mitchell, Robert - Rob; NEGRI, CRISTINA - Argonne National Laboratory

Submitted to: Land
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/3/2021
Publication Date: 11/10/2021
Citation: Hamada, Y., Zumpf, C., Cacho, J., Lee, D., Lin, C., Heaton, E., Boe, A., Boersma, N., Mitchell, R., Negri, C. 2021. Remote sensing-based estimation of advanced perennial grass biomass yields for bioenergy. Land. https://doi.org/10.3390/land10111221.
DOI: https://doi.org/10.3390/land10111221

Interpretive Summary: The bioeconomy promotes sustainable economic growth under changing climate conditions by reducing our dependence on fossil fuels and fossil-derived materials. A sustainable bioeconomy integrates food, feed, fiber and bioenergy crops to protect our natural resources. A general premise of the bioeconomy is to grow advanced, high-yielding bioenergy crops that require fewer inputs on marginal areas while growing commodity and food crops on the productive areas of the agricultural landscape. Because marginal areas are often small and spread across the agricultural landscape, this proposed cropping system requires rapid, accurate, and cost-effective estimation of biomass yield prior to harvest. Our objective was to determine if remotely sensed imagery could accurately and rapidly estimate biomass yield for dedicated energy crops during the growing season across 5 locations in the U.S. Midwest. A linear regression model using mid-summer remote sensing data accurately predicted end of season switchgrass yields, except for the establishment year. The model also accurately predicted end of season switchgrass yields as early as 152 days before harvest, except for the establishment year. This initial investigation indicates remote sensing can predict perennial bioenergy grass yields across multiple locations up to 5 months prior to harvest. This yield forecasting would support the important economic and logistical decisions of farmers, breeders, policy makers, and future biomass refineries.

Technical Abstract: Bioenergy and bio-based materials are critical for sustainable economic growth under changing climate conditions by reducing our dependence on fossil fuels and fossil-derived materials. The key to the successful development of a sustainable biomass-based economy is an integrated production of food, feed, and fiber and bioenergy crops on a finite land resource while protecting wildlife and natural resources. A viable solution would be growing advanced or high-yielding bioenergy crop cultivars that require lower production inputs on marginal areas while growing commodity and food crops on inherently productive parts of the agricultural landscape. Because marginal areas are often small and spread across the agricultural landscape, this proposed cropping system requires rapid, accurate, and cost-effective estimation of biomass yield at harvest time at a plot or sub-field scale. Remotely sensed imagery collected during the growing season has been successfully applied for estimating yields across scales for various crop types. This paper demonstrates (1) the initial investigation of multispectral optical remote sensing for predicting warm-season perennial grass yields at harvest using a linear regression model with a spectral vegetation index, more specifically the Green Normalized Difference Vegetation Index (GDNVI), and (2) the model’s predictive power for at-harvest dry biomass yields evaluated using five study locations in the U.S. Midwest. The results show that the linear regression model using mid-summer GNDVI predicted at-harvest switchgrass yields with R2 as high as 0.879 and a mean absolute error and root mean squared error as low as 0.592 Mg/ha and 0.539 Mg/ha, respectively, except for the establishment year. The model also indicated at-harvest switchgrass yields may be predicted as early as 152 days before the date of harvest on average, except for the establishment year. Additionally, the study showed that the green band appeared to have a greater contribution for predicting at-harvest switchgrass dry biomass yields than the red band, which is consistent with an increase in chlorophyll content during the early growing season. Although additional testing is warranted, this initial investigation showed great promise for a remote sensing approach utilizing a spectral vegetation index for forecasting perennial bioenergy grass yields in advance to support the important economic and logistical decisions of farmers, breeders, policy makers, and future biomass refineries.