Soil spectral library of Piauí State using machine learning for laboratory analysis in Northeastern Brazil

Abstract

Soil chemical and physical analyses are the major sources of data for agriculture. However, traditional soil analyses are time-consuming, not cost-efficient, and not environmentally friendly. An alternative to traditional soil analyses is soil spectroscopy. This technique is a low-cost and quick analytical method, which can be implemented in a laboratory and/or in-situ. Nevertheless, some spectrometers are expensive and do not contemplate the entire spectrum. Despite this limitation, the main objective of the study was to create a soil spectral library of the Piauí State using only the 1000-2500 nm range. In this sense, it was evaluated and standardized the soil spectral library by accessing the combination of smoothing, standard normal variate, continuum removal, and Savitzky-Golay derivative spectral preprocessing procedures with partial least squares, random forest, and cubist machine learning algorithms. It was collected 262 geo-referenced soil samples at the layer of 0.00-0.20 m across the entire Piauí State, representing most of its soil variability. The soil properties evaluated were pH(H 2 O), sand, clay, and soil organic carbon (SOC) contents. This study demonstrated that the Standard Normal Variate was one of the most promising preprocessing procedures to improve model predictions for pH(H 2 O), sand, and clay. For SOC and pH, the best overall results were without preprocessing the soil spectra. Moreover, the cubist model was the most accurate in predicting soil properties. Finally, our study showed evidence of the potential and feasibility of using this soil spectral library to estimate soil properties such as pH(H 2 O), sand, clay, and SOC.

The Brazilian soils from a spectral library perspective: from fundamental to applications

“…Similar descriptions between 350-2500 nm spectra and soil attributes (Figures 3-5) were reported by the Brazilian Soil Spectral Library (Dematte et al, 2019) for soils from Cerrado and Atlantic Forest biomes (Figure 2). The Soil spectral library of Piauí State in Brazil (Mendes et al, 2021) also reported similar relationships, where the soil clay and organic carbon contents had negative correlations and sand had positive correlation values with 350-2500 nm spectra. Moura-Bueno et al (2020) found that the first principal component from 350-2500 nm soil spectra had inverse correlation with organic carbon (r = −0.75) and clay (r = −0.63) contents, and positive correlation with sand (r = 0.45) content from subtropical Brazilian soils.…”

“…Similar influential wavelengths were also reported by Moura-Bueno et al (2020), who reduced 350-2500 nm spectra to three principal components and found eigenvectors related to iron oxides (near 450, 600-800 nm), OM (near 800-900 and 1850 nm), and clay minerals (near 1400 and 1900 nm). Mendes et al (2021) reported that the 1000-2500 nm spectral range had high importance to estimate soil attributes such as pH, sand, clay, and organic carbon. The first the principal components from the Brazilian Soil Spectral Library (Dematte et al, 2019) were related to iron oxides and OM (near 500 and 900 nm), kaolinite, gibbsite, quartz and 2:1 clay mineral (near 1000-2100 and 2300 nm).…”

“…Our soil predictions from raw spectra yielded performance values similar to those obtained from soil spectral libraries (Dematte et al, 2019; for sand, clay, organic carbon, CEC, and V% using Cubist and several spectral modelling strategies-e.g., preprocessing, stratification, splitting and auxiliary predictors. We also had a high number of laboratories that outperformed the results from the soil spectral library of Piauí in Brazil (Mendes et al, 2021) and other studies on spectral modelling (Moura-Bueno et al, 2020;Santana et al, 2018;. The prediction of clay and sand content outperformed estimation of other soil attributes because they reflect the direct influence of soil mineralogy on soil spectral patterns within the 350-2500 nm range (Lacerda et al, 2016).…”