Does drone remote sensing accurately estimate soil pH in a spring wheat field in southwest Montana

Soil acidification is a growing problem in semi-arid agroecosystems. In the state of Montana, USA, soil pH levels below 5.5 have been documented in nearly half of the counties. Acidic soils have the potential to reduce crop yield, but methods to identify and remediate acidic soils are costly and time-intensive. This study tests a relatively new approach for identifying areas of acidic soils using imagery derived from UAS (unmanned aerial systems). UAS provide a means to collect fine-scale, multi-spectral imagery at user-defined intervals for an area of interest—in this case, a 22 ha spring wheat field in southwestern Montana. In addition to 12 dates of spectral observations across a growing season, field measurements of soil pH and other soil attributes were collected to analyze their relationship with the normalized difference vegetation index (NDVI) using linear regression models, and to spatially predict soil pH across the field using a random forest model. The linear regression models indicated that most of the variation in early-season NDVI was attributed to differences in soil pH and soil organic matter, whereas variation in later-season NDVI was less related to soil pH. The random forest model predicted soil pH with reasonable accuracy (RMSE = 0.72). This study helps to fill a knowledge gap by bridging UAS-derived observations of NDVI with field-derived measurements of soil pH to identify areas of soil acidity. The methodology put forth by this study would enable land managers to easily identify and hence, remediate acidic soils in a more cost-efficient and timely manner.