Evaluation of Soil Health Indicators at Long-Term Agricultural Sites in Ontario, Canada, and Indicator Prediction with Infrared Spectroscopy
Four trials across Ontario were sampled to evaluate the effects of best management practices (BMP) on six soil health indicators (SHI). As well, the use of near infrared (NIR) and mid infrared (MIR) spectroscopy to predict SHI was explored at site and regional levels. Lastly, MIR spectroscopy was utilized to identify differences in soils under differing BMPs. The study found no-tillage (NT) improved SHI values at two of three sites. Inclusion of cover crops and small grains in rotation improved SHI values at one site, while no significant differences were found at two sites. The study found that MIR models predicted SHI better than NIR models. As well, the use of competitive adaptive reweighted sampling (CARS) feature selection improved model performance compared to full spectra models. Lastly, differences were found between MIR spectra from soils under NT and moldboard plow treatments and between annual versus perennial crop rotations.