Capability of high-resolution RGB imagery to accurately document residue in row-crop fields

Abstract

Sustaining agriculture and civilizations requires protecting topsoil from erosion. Residue on the soil surface dissipates energy in raindrops, runoff, and wind, reducing detachment and transport of soil particles. Our objective was to contrast transect residue estimates based on (1) in-field evaluation of 100 evenly spaced points along a 15 m tape; (2) evaluation of imagery of the same tapes obtained at two ground sampling distances: 0.014 and 0.06 cm pixel^–1; and (3) a 100 point grid analysis of 50 contiguous 0.3 × 0.2 m images (0.014 cm pixel^–1) adjacent to the tape. Data were collected from May through early July of 2018 in row-crop fields in four Missouri counties: Audrain, Boone, Callaway, and Cooper. We used data from 60 15.2 m transects in 20 fields. Residue estimates based on field readings of the transect tape ranged from 6% to 99%. Reading images of tapes resulted in a mean decrease of 2.5 percentage points of residue. The bullseye grid point method documented tremendous variability along the transect (mean transect sd = 12.4) and a substantially lower estimate of residue cover compared to field tape readings (mean difference –10.3 percentage points). Consistent with a systematic bias to overestimate residue with the tape transect method, the difference increased with increasing residue for residue &lt;50%, was more pronounced with soybean (<i>Glycine max</i> [L.] Merr.) than corn (<i>Zea mays</i> L.), and the difference could be eliminated by rereading tapes with scrupulous attention to the details of transect methodology. Image-based methods have the potential to improve the accuracy of residue estimates by facilitating methods less prone to reader bias such as the bullseye grid approach. Additionally, to improve accuracy, image-based protocols should be integrated into training and quality control assurance protocols of USDA Natural Resources Conservation Service (NRCS) assessments of residue using transect methods.