Soil health plays a crucial role in crop production, both in terms of quality and quantity, highlighting the importance of effective methods for preserving soil quality to ensure global food security. Soil quality indices (SQIs) have been widely utilized as comprehensive measures of soil function by integrating multiple physical, chemical, and biological soil properties. Traditional SQI analysis involves laborious and costly laboratory analyses, which limits its practicality. To overcome this limitation, our study explores the use of visible near-infrared (vis-NIR) spectroscopy as a rapid and non-destructive alternative for predicting soil properties and SQIs. This study specifically focused on seven soil indicators that contribute to soil fertility, including pH, organic matter (OM), potassium (K), calcium (Ca), magnesium (Mg), available phosphorous (P), and total nitrogen (TN). These properties play key roles in nutrient availability, pH regulation, and soil structure, influencing soil fertility and overall soil health. By utilizing vis-NIR spectroscopy, we were able to accurately predict the soil indicators with good accuracy using the Cubist model (R² = 0.35–0.93), offering a cost-effective and environmentally friendly alternative to traditional laboratory analyses. Using the seven soil indicators, we looked at three different approaches for calculating and predicting the SQI, including: (1) measured SQI (SQI_m), which is derived from laboratory-measured soil properties; (2) predicted SQI (SQI_p), which is calculated using predicted soil properties from spectral data; and (3) direct prediction of SQI (SQI_dp), The findings demonstrated that SQI_dp exhibited a higher accuracy (R² = 0.90) in predicting soil quality compared to SQI_p (R² = 0.23).

Large organic deposits in the southwestern plain of Montreal have been converted to agricultural land for vegetable production. In addition to the variable depth of the organic deposits, these soils commonly have an impermeable coprogenous layer between the peat and the underlying mineral substratum. Estimations of the depth and thickness of these materials are critical for soil management. Therefore, five drained and cultivated peatlands were studied to estimate their maximum peat thickness (MPT)—a potential key soil property that can help identify management zones for their conservation. MPT can be defined as the depth to the mineral layer (DML) minus the coprogenous layer thickness (CLT). The objective of this study was to estimate DML, CLT, and MPT at a regional scale using environmental covariates derived from remote sensing. Three machine-learning models (Cubist, Random Forest, and k-Nearest Neighbor) were compared to produce maps of DML and CLT, which were combined to generate MPT at a spatial resolution of 10 m. The Cubist model performed the best for predicting both features of interest, yielding Lin’s concordance correlation coefficients of 0.43 and 0.07 for DML and CLT, respectively, using a spatial cross-validation procedure. Interpretation of the drivers of CLT was limited by the poor predictive power of the final model. More precise data on MPT are needed to support soil conservation practices, and more CLT field observations are required to obtain a higher prediction accuracy. Nonetheless, digital soil mapping using open-access geospatial data shows promise for understanding and managing cultivated peatlands.

Since soil health is impacted by inherent soil properties, it is, therefore, challenging to apply the same soil health frameworks across multiple regions and soil types. Here, we examined the effect of soil textural group (coarse, medium, and fine) on four soil health indicators of soils sampled from diverse agricultural systems across Ontario. Scoring functions were developed by calculating cumulative normal distributions, using the mean and standard deviation of each soil health indicator, for three or five soil textural groups. For each soil health indicator, soil health scoring values were provided using the “more is better” approach, where greater soil health scores implied better soil health. Soil health indicators were significantly affected by three but not all five soil textural groups. Evolved NH₃ and CO₂, and potentially mineralizable N had stronger associations with each other as revealed by correlation and principal component analysis. Our results also suggested that mean separation of the tested soil health indicators was more consistent with three soil textural groups (coarse, medium, and fine) than five soil textural groups (clays, clay loams, loams, sandy loams, and sand); therefore, we recommend using three soil textural groups to develop soil health scoring functions. The findings of this study lay a groundwork for future soil health assessment involving a larger number of samples across Ontario and more soil indicators, which will facilitate the regional interpretation of soil health.