Soil organic matter prediction with benchtop and implement-mounted optical reflectance sensing approaches

Authors: CONWAY, LANCE - University Of Missouri; Sudduth, Kenneth - Ken; Kitchen, Newell; ANDERSON, STEPHEN - University Of Missouri; Veum, Kristen; MYERS, BRENTON - Corteva Agriscience

Submitted to: Soil Science Society of America Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/5/2022
Publication Date: 10/31/2022
Citation: Conway, L.S., Sudduth, K.A., Kitchen, N.R., Anderson, S.H., Veum, K.S., Myers, B.D. 2022. Soil organic matter prediction with benchtop and implement-mounted optical reflectance sensing approaches. Soil Science Society of America Journal. 86(6):1652-1664. https://doi.org/10.1002/saj2.20475.
DOI: https://doi.org/10.1002/saj2.20475

Interpretive Summary: Collecting soil information from sensors mounted on row-crop equipment allows for the opportunity to improve machinery management and crop performance. Commercially available technologies currently allow for the integration of optical sensors that measure soil properties, such as organic matter, into traditional row-crop equipment. Evaluation and improvement of these sensors is needed to maximize their utility and the systems they control. Therefore, research was conducted to determine organic matter prediction capability across selected soils and soil water contents with a commercially available sensor, as well as through advanced analytical techniques and multiple combinations of soil reflectance bands in the visible and near-infrared spectrum. Soils from Missouri and Illinois, USA were utilized for the study, and were evaluated at three different soil water contents. Results found that the commercial system was able to detect relative differences between low and high soil organic matter, but that the estimates were impacted by the amount of water content present in the soil. Implementing advanced analytic techniques in combination with additional spectral information resulted in an improved organic matter estimation across soil water contents. This information will help producers interpret measurements from the commercially available sensor and contribute to development of improved sensor-based control systems for precision agriculture applications.

Technical Abstract: Proximal sensing technologies can densely quantify soil organic matter (OM) variability utilizing visible and near infrared (VNIR) reflectance spectroscopy. However, issues such as soil moisture and sensor-to-soil engagement can influence predictions. Therefore, research was conducted to determine OM prediction accuracy across selected soils and soil volumetric water contents (VWC) with (i) a commercially-available, planter-mounted sensor and (ii) machine learning techniques applied to multiple combinations of soil reflectance bands within the VNIR spectrum. Ninety soils collected across Missouri and Illinois, USA were used in the study. Data were collected at three VWC with a commercially available, planter-mounted sensor and a benchtop spectrometer. Spectral pre-processing and machine learning techniques were utilized for prediction of OM in all modeling approaches. Results found that the commercial sensor predictions were affected by soil VWC, with OM predictions decreasing with increasing VWC. Findings from three modeling approaches showed that a continuous spectrum (i.e., 400-1500 nm) improved performance (RMSE = 5.25 g kg-1) over the targeted waveband approach. Furthermore, including the entire VNIR region (400-2500 nm) resulted in the best predictive capability (RMSE = 1.42 g kg-1). However, because a full-spectrum approach may not be practical due to economic and computational expense, utilizing continuous reflectance from 400 to 1500 nm along with spectral pre-processing and machine learning may be an acceptable method for estimating OM. These findings contribute to the development and improvement of commercially available proximal sensors that may be used to monitor soil carbon stocks, assess changes in soil health, or for other precision agriculture applications.