Estimating cover crop biomass nitrogen credits with Sentinel‐2 imagery and sites covariates

Abstract

Cover crops can positively impact productivity and the environment. While improved estimates of cover crop N benefits could help promote their adoption, little information is currently available at the broad scale. We conducted a multi-site study to determine whether use of satellite images and site factors could fill this gap. Six spectral indices were extracted from Sentinel-2 satellite imagery and used with modeling covariates to estimate cover crop properties and biomass N credits (biomass × N contents). The partial least squares regression (PLSR) models were calibrated and validated with samples from 42 cover crop fields located in the midwestern and northeastern United States collected in 2017-2018. Remote sensing (RS)-derived spectral indices strongly correlated (r > .7) with red clover (Trifolium pratense L.) but not with rye (Secale cereale L.) biomass. Growing degree days (GDDs), cover height, ground cover percentage, and temperature often had high importance (variable importance in projection >1) in PLSR models. Model predictive power was limited for estimates of biomass N credits when data from all validation sites and cover types were used (adjusted [adj] R 2 = .52). However, models for both biomass (adj R 2 = .81) and biomass N credits (adj R 2 = .89) were successful for red clover fields. This suggests N benefits could be more effectively modeled for specific cover crop types. We also found RS-based estimation of C/N ratios performed moderately well when applied to the complete dataset (adj R 2 = .54), suggesting a way to differentiate grass and legume cover crops that can potentially inform biogeochemical models. INTRODUCTIONInterest in cover crops has grown rapidly (Dunn et al.

Process-based modeling of soil nitrous oxide emissions from United States corn fields under different management and climate scenarios coupled with evaluation using regional estimates

“…However, if cover crops were managed properly, metaanalysis found that soil N 2 O emissions would be reduced or not affected compared to non-cover crop treatments considering the balance between increased SOC and reduced nitrate contents (Basche et al, 2014;Kaye and Quemada, 2017). The key to N 2 O mitigation in a diversified cropping system is, therefore, to avoid the overapplication of fertilizers by better estimating N balance that takes into account N credits from cover crops, which can be achieved through the use of remote sensing (Xia et al, 2020) and N calculators (Gaskin et al, 2019).…”

Integrating APSIM model with machine learning to predict wheat yield spatial distribution

“…This phenomenon, also known as equifinality, prevents the estimation of unique and recognizable parameter values (D. He et al., 2017). However, preparing crop growth and phenology for a large number of fields using standard ground‐based sample methods can be inconveniently time‐consuming, opening the way for the use of remote sensing datasets as a quick and low‐cost method (Ko et al., 2006; Maas, 1993; Xia et al., 2021). The integration of the developed APSIMr package with ANN acts as a scientific modeling workflow that provides fast, accurate, and low‐cost spatiotemporal simulation, which is very important for food security and nutritional analysis.…”

Quadrat misuse: Confessions of a cover crop researcher's biomass sampling journey

“…In the past two decades that the author has been involved in cover crop research, the tools available for this work have expanded to include new high‐tech tools such as unmanned arial vehicles (Roth & Streit, 2018; Yuan et al., 2019), remote sensing (Marcillo et al., 2020; Xia et al., 2021), hand‐held sensors (White et al., 2019), and mobile phone apps (Patrignani & Ochsner, 2015), but regardless of how high‐tech cover crop research becomes there will always be a need for accurate biomass sampling methods. Accurate sampling is critical to estimate field‐scale cover crop biomass production and associated metrics (nitrogen uptake, weed suppression, soil erosion control) and to “ground truth” and calibrate high‐tech tools that may provide new insights and more time efficient data collection.…”