Using proximal vis-NIR soil spectroscopy to assess soil health

Authors: Sudduth, Kenneth - Ken; Veum, Kristen; Ransom, Curtis

Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 9/14/2024
Publication Date: 10/14/2024
Citation: Sudduth, K.A., Veum, K.S., Ransom, C.J. 2024. Using proximal vis-NIR soil spectroscopy to assess soil health. In: Proceedings of the 6th Global Proximal Soil Sensing Workshop, October 14-17, 2024, Ghent, Belgium. p. 74-79.

Interpretive Summary: Assessment of soil health involves determining how well a soil is performing its biological, chemical, and physical functions by measuring soil properties referred to as soil health indicators. Applied soil health research is focused on the needs of on-farm soil health assessment and interpretation and requires tools and strategies that are affordable and accessible. The high cost and labour requirements of most currently available soil health assessments renders them impractical for large-scale efforts. Ideally, an on-farm soil health assessment would be easy to measure, applicable to field conditions, affordable, and capture a wide range of soil functions that vary spatially and temporally. Soil spectroscopy techniques offer high-density spatial and temporal soil information and may help achieve agronomic and sustainability goals related to soil health. In this research, we evaluated soil spectroscopy data, both with and without auxiliary sensor variables, for estimating common soil health indicators. Spectral data were collected both in the laboratory on prepared soil samples, and in-situ in the field using a commercial sensing probe. Models developed with lab-collected data were generally more accurate, but models estimating five indicator variables using field data were of sufficient accuracy to be useful. In-situ estimation of soil health indicators was promising, showing a potential for providing efficient and cost-effective on-farm soil health assessment data.

Technical Abstract: Soil health is a topic of much interest, but efficient and cost-effective on-farm measurement methods are lacking. This research investigated the use of visible and near infrared spectroscopy (vis-NIRS), with and without auxiliary sensor variables of soil bulk electrical conductivity (ECa) and soil strength (as cone index, CI), for the measurement of common soil health indicators. In laboratory studies, a spectrometer (400-2500 nm) collected data on dried and sieved soil samples from multiple landscape positions and cropping systems, where ECa and CI data were collected using field sensors. Seven soil health indicators were quantified using standard methods. Best estimates were for the biological indicators of soil organic carbon (SOC; R2 = 0.82) and ß-glucosidase activity. Chemical and physical indicators were not well estimated (R2 < 0.6). Adding the auxiliary sensor variables to the model improved estimates of some indicators and of overall model scores, illustrating the advantage of multi-sensor fusion. In-situ profile spectroscopy data, along with CI and ECa, were obtained with a commercial probe at 77 locations in 7 fields in Missouri, USA. Data from the surface horizon were used to estimate soil health indicators measured on 0-15 cm samples. Cross-validated model estimates were less accurate than in the laboratory study (e.g., SOC R2 = 0.59 vs. 0.82); however, estimates of several indicators were in a useful (R2 > 0.5) range: SOC, total nitrogen, aggregate stability, active carbon, and soil respiration. Only for SOC and aggregate stability was accuracy improved by including auxiliary variables. Overall, in-situ estimation of soil health indicators was promising, showing a potential for providing efficient and cost-effective on-farm soil health assessment.