Mixed-Species Cover Crop Biomass Estimation Using Planet Imagery

Abstract

Cover crop biomass is helpful for weed and pest control, soil erosion control, nutrient recycling, and overall soil health and crop productivity improvement. These benefits may vary based on cover crop species and their biomass. There is growing interest in the agricultural sector of using remotely sensed imagery to estimate cover crop biomass. Four small plot study sites located at the United States Department of Agriculture Agricultural Research Service, Crop Production Systems Research Unit farm, Stoneville, MS with different cereals, legumes, and their mixture as fall-seeded cover crops were selected for this analysis. A randomized complete block design with four replications was used at all four study sites. Cover crop biomass and canopy-level hyperspectral data were collected at the end of April, just before cover crop termination. High-resolution (3 m) PlanetScope imagery (Dove satellite constellation with PS2.SD and PSB.SD sensors) was collected throughout the cover crop season from November to April in the 2021 and 2022 study cycles. Results showed that mixed cover crop increased biomass production up to 24% higher compared to single species rye. Reflectance bands (blue, green, red and near infrared) and vegetation indices derived from imagery collected during March were more strongly correlated with biomass (r = 0–0.74) compared to imagery from November (r = 0.01–0.41) and April (r = 0.03–0.57), suggesting that the timing of imagery acquisition is important for biomass estimation. The highest correlation was observed with the near-infrared band (r = 0.74) during March. The R2 for biomass prediction with the random forest model improved from 0.25 to 0.61 when cover crop species/mix information was added along with Planet imagery bands and vegetation indices as biomass predictors. More study with multiple timepoint biomass, hyperspectral, and imagery collection is needed to choose appropriate bands and estimate the biomass of mix cover crop species.

Assessing the capacity of high-resolution commercial satellite imagery for grapevine downy mildew detection and surveillance in New York state

“…Despite test accuracy and F1 scores both exceeding 0.78, the random forest models trained on three years of PlanetScope and SkySat data performed poorly at classifying high incidence GDM for images acquired during advanced disease stages in 2021 and 2022 (Figure 5). In a comparison of PlanetScope VIs for cover crop yield estimation, Kharel et al also found that NDVI was positively correlated with biomass across their study period, supporting the link between NDVI and amount of healthy vegetation (Kharel et al 2023). Further, they observed that correlation coefficients between biomass, VIs, and spectral bands depended on the time of image acquisition, with highest correlation observed at late vegetative growth stages.…”

“…The proliferation of the commercial high resolution Earth observation industry has reinvigorated research interest in spaceborne ecological surveillance. SkySat and PlanetScope imagery has been used for a range of ecosystem and agricultural applications, including biomass estimation, vegetation classification, plant disease detection, quantifying evapotranspiration, and soil moisture mapping (Kharel et al 2023; Guo et al 2022; Szabó et al 2021; Shi et al 2018; Baloloy et al 2018; Du et al 2022; Raza et al 2020; Aragon et al 2021). The frequent revisit times and fine spatial scale of these platforms enables monitoring diverse specialty crops, including grapevine, despite their smaller, spatially heterogeneous fields (Helman et al 2018; Meyers et al 2020).…”

Artificial intelligence and satellite‐based remote sensing can be used to predict soybean (Glycine max) yield

“…Of the many space‐based sensors to choose from, PlanetScope was selected because it has high spatial (3.12 m × 3.12 m) and temporal (daily) resolution and provides multispectral images (blue [455–515 nm], green [500–590 nm], red [590–670 nm], and near‐infrared [780–860 nm]; Jain et al., 2016; Planet Team, 2017; Yang et al., 2012). The Ortho Scene‐Analytics (Level 3B) surface reflectance imageries are orthorectified, radiometrically calibrated, and atmospherically corrected products that capture the reflectance characteristics of the lower atmosphere (Frazier & Hemingway, 2021; Houborg & McCabe, 2018; T. P. Kharel et al., 2023). Between planting and harvesting in 2019 and 2021, six cloud‐free images at six growth stages (VE/VC [June 10]; V1/V3 [July 2], R1/R2 [July 24], and R2/R3 [August 10], R4/R5 [August 25], and R6/R7 [September 10]) were acquired.…”

A comparison of traditional and machine learning corn yield models using hyperspectral UAS and Landsat imagery

“…Several studies have focused on the exploitation of remote sensing data and indices for use in constructing mathematical models for forecasting maize silage yield. [5][6][7][8][9][10][11][12] Two examples of vegetation indices (VIs) that have shown promise as predictors are the normalized difference vegetative index (NDVI) and enhanced vegetative index (EVI). Both indices are regularly used in linear and exponential regression models:…”

“…The National Aeronautics and Space Administration's (NASA) Landsat and Moderate Resolution Imaging Spectroradiometer (MODIS) imagery systems, Planet Labs' Dove and Superdove constellations, and the European Space Agency's (ESA) Sentinel satellites provide calibrated reflectance imagery which have been used for such yield predictions. [5][6][7][8] Recent advances in machine learning have created new opportunities for spectral data exploitation, using neural networks to perform regression analysis on imagery instead of traditional least squares regressions. However, machine learning regression performance benefits from more information than is available in the multispectral systems previously mentioned.…”