Multi-sensor proximal remote sensing for cover crop biomass estimation at high and moderate spatial resolutions

Authors: Jennewein, Jyoti; Davis, Brian; SEEHAVER EAGEN, SARAH - North Carolina State University; Nicolette, Jordan; PITTMAN, JOSH - Bayer Cropscience; HIVELY, DEAN - Us Geological Survey (USGS); GOLDSMITH, AVI - North Carolina State University; HIDALGO, CHRIS - North Carolina State University; REBERG-HORTON, CHRIS - North Carolina State University; Mirsky, Steven

Submitted to: Smart Agricultural Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/13/2025
Publication Date: 7/18/2025
Citation: Jennewein, J.S., Davis, B.W., Seehaver Eagen, S., Nicolette, J.A., Pittman, J., Hively, D.W., Goldsmith, A., Hidalgo, C., Reberg-Horton, C., Mirsky, S.B. 2025. Multi-sensor proximal remote sensing for cover crop biomass estimation at high and moderate spatial resolutions. Smart Agricultural Technology. 12:101201. https://doi.org/10.1016/j.atech.2025.101201.
DOI: https://doi.org/10.1016/j.atech.2025.101201

Interpretive Summary: Farmers use cover crops to improve soil health and provide other benefits to their fields, but accurately measuring their growth and effectiveness remains a challenge. To address this, we used a handheld or tractor mountable sensor called the Active Canopy Sensor (ACS-214) to track cover crop growth and biomass accumulation over four winters, in 13 states, and with 11 different types of cover crops. The ACS-214 uses light and distance measures to estimate cover crop biomass. Results showed that the ACS-214 worked well for certain cover crop types, such as grasses and legumes, showing a good ability to predict their biomass if models were trained on local data. However, the ACS-214 struggled with brassica crops, which had uneven growth and flowered early. Additionally, this study showed that the ACS-214 could be used to train publicly available satellite imagery, which could help farmers estimate cover crop growth over larger areas. Overall, we found that ACS-214 can be a valuable tool for biomass predictions in grasses and legume cover crops but require region-specific data to provide the most accurate predictions. This advancement could make it easier for farmers to monitor cover crop growth, which has a direct impact on soil health.

Technical Abstract: Planting cover crops is a climate smart, conservation agriculture practice that has important implications for precision agriculture. Cover crop performance (biomass and quality) influence the magnitude of agroecological services they provide necessitating the need for accurate performance estimation. The Active Canopy Sensor (ACS) 214 – a proximal sensing device designed for crop biomass assessments – includes active red and near-infrared sensors, a time-of-flight laser, and an ultrasonic sensor. We deployed the ACS-214 over four winter cover crop seasons (2020-24) in 13 states and 11 cover crop species (797 grasses, 264 legumes, 181 brassicas). We conducted four forms of cross validation: 1) all region train-test splits, 2) leave-one-region out, 3) region-specific, and 4) out-of-bag predictions on withheld years. Finally, we assessed the use of ACS-214 to train multispectral Sentinel-2 imagery. Random forest (RF) models showed good performance for grasses (R2 = 0.51 – 0.64) and legumes (R2 = 0.44 – 0.76) for all models except leave-one-region out results indicating poor spatial generalizability without local training data. Importantly, slope changes were detected at ~3,000 kg ha-1 for legumes and 4,000 kg ha-1 for grasses suggesting an upper threshold for biomass predictions. All brassica models performed poorly due to flowering and patchy growth (R2 < 0.30). Sentinel-2 RF models showed good agreement with ACS-214 estimated biomass (R2 = 0.70) and destructively sampled biomass (R2 = 0.58 – 0.61), with a similar 4,000 kg ha-1 breakpoint in prediction accuracy. These findings demonstrate the capability of the ACS-214 for estimating cover crop biomass, while emphasizing the need for locally-trained models.