Extraction of reflectance spectra features for estimation of surface, subsurface, and profile soil properties

Authors: ZHOU, PENG - China Agricultural University; Sudduth, Kenneth - Ken; Veum, Kristen; LI, MINZAN - China Agricultural University

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/2/2022
Publication Date: 3/17/2022
Citation: Zhou, P., Sudduth, K.A., Veum, K.S., Li, M. 2022. Extraction of reflectance spectra features for estimation of surface, subsurface, and profile soil properties. Computers and Electronics in Agriculture. 196. Article 106845. https://doi.org/10.1016/j.compag.2022.106845.
DOI: https://doi.org/10.1016/j.compag.2022.106845

Interpretive Summary: As farmers look for new ways to increase the efficiency of production, they often turn to the concepts of precision agriculture to optimize placement of fertilizers and other crop inputs across fields and farms. An important part of this optimization process is availability of efficient measurements of crop and soil parameters to guide input placement. Prior research has shown that diffuse reflectance spectroscopy in the visible and near-infrared wavelength ranges can provide high-resolution, nondestructive estimation of soil chemical and physical properties. However, most previous research has focused on surface soils, while subsurface soils are often quite different from surface soils within the same profile. Therefore, this study examined separately surface and subsurface soils from a large database of about 700 soil samples. Goals were to (1) evaluate the effect of a reduced number of spectral variables on the accuracy of soil property estimation, using two different variable selection methods, and (2) determine if models developed to estimate surface soil properties could also estimate subsoil properties, and vice versa. We found that selection of a reduced number of variables sometimes improved accuracy; however, in other cases fewer variables decreased accuracy. Further study using additional samples is needed, as this is an important question for development of lower-cost soil sensors based on fewer spectral variables. We also found that when a model was developed for one depth interval (surface or subsurface), applying it to data from the other interval provided less accurate results. However, those results could still be sufficiently accurate in some cases, depending on the intended use of the data. This finding is important because it shows the potential of models developed for surface soils, generally the most important depth interval, to still be useful for subsurface soils. Overall, this study provided new insight into how to best develop spectral soil property estimation models targeted to surface, subsurface, and combined datasets.

Technical Abstract: Diffuse reflectance spectroscopy in the visible and near-infrared wavelength ranges has potential to provide high-resolution, pollution-free, and nondestructive estimation of soil chemical and physical properties for use in precision agriculture. However, most previous research has focused on surface and profile soils, while subsurface soils are often quite different from surface soils within the same profile. Thus, in this study, a database of 697 soil samples was used to compare results for the three soil categories (profile, surface, and subsurface). Soil cores were obtained to approximately 1.2 m depth from ten fields, two each in Missouri, Illinois, Michigan, South Dakota, and Iowa, USA, then sieved and air-dried. Laboratory soil spectra were obtained from 350 to 2500 nm using a commercial spectrometer and soil properties (total nitrogen, soil organic carbon, total carbon, magnesium, calcium, potassium, soil texture (clay, silt, and sand) fractions, cation exchange capacity, and pH) were measured using standard laboratory analyses. The ability of ten spectral preprocessing techniques to improve analysis results was investigated. Backward interval partial least squares was used to identify those spectral regions most predictive of soil properties. Alternatively, specific characteristic wavelengths were identified by a combination genetic algorithm (GA)-back propagation neural network (BPNN) approach. Soil property estimation based on three techniques was compared: (1) partial least squares regression (PLSR) models based on the full spectrum, (2) PLSR models based on sensitive regions, and (3) BPNN models based on characteristic wavelengths. The best results for profile and subsurface soils were obtained with absorbance preprocessing, but for the surface soils, the standard normal variate transformation was best. For some soil properties, the prediction R2 of the PLSR models based on sensitive regions was better than that of the PLSR models based on the full spectrum, demonstrating that retaining only sensitive wavebands could improve estimates. However, in some cases, the reduction in wavebands decreased accuracy. Differences in prediction R2 across all calibration models over profile and subsurface soils were relatively small but were largest for surface soils. Furthermore, application of characteristic wavelength calibrations to other soil datasets resulted in a lower R2 than with the full spectrum calibration developed for that dataset. In general, this study shows that there are measurable differences in prediction R2 across all calibration models over the three soil depth categories. The experimental results of this study illustrate the potential for a set of wavelengths optimized for one depth category to still provide acceptable estimates for other depth categories. Overall, these results provide important theoretical guidance for the development of soil sensors based on discrete DRS wavebands to reduce cost and increase the speed of in-field data collection.