Remote-sensing-based sampling design and prescription mapping for soil acidity

Authors: Casanova, Joaquin; CARLSON, JENNY - Washington State University; Letourneau, Melissa

Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/11/2023
Publication Date: 6/14/2023
Citation: Casanova, J.J., Carlson, J.L., LeTourneau, M. 2023. Remote-sensing-based sampling design and prescription mapping for soil acidity. Remote Sensing. 15(12). Article 3105. https://doi.org/10.3390/rs15123105.
DOI: https://doi.org/10.3390/rs15123105

Interpretive Summary: Farmers in the Inland Pacific Norwthwest face increasing soil acidity, which limits crop productivity. This is from a number of factors, including fertilization, base cation leaching, and intensive crop production. One solution is applying lime to neutralize acidity, but this is expensive. Precision application of lime saves money, but requires detailed maps of soil acidity. This paper demonstrates a method to reduce the sampling requirements to generate prescription maps for lime application, decreasing labor and monetary costs compared to conventional methods.

Technical Abstract: Soil acidification is a major problem in the inland Pacific Northwest. A potential solution is application of lime to neutralize acidity and raise pH. As lime is an expensive input, precision variable-rate application is necessary. However, high resolution mapping of pH and buffer pH for lime prescription requires costly sampling and analysis. To reduce the amount of sampling needed, remote sensing correlates with soil pH and buffer pH can help decide optimal sampling locations and allow optimal interpolation. This paper uses soil and crop data from a USDA research farm to develop an optimal sampling plan on a farmer’s property, then following that sampling design, uses the measured pH and buffer pH to fit a Bayesian Hierarchical Model using remote sensing variables specific to that farmer’s land. Following this, a new model is developed for the research farm with similar covariates. Ultimately, on the farmer’s field, we see a root mean square error (RMSE) in 0-10 cm soil pH of 0.2256 and of 0-10 cm Modified Mehlich buffer pH of 0.1240. For the research farm, where buffer pH has not been measured, we see a RMSE in 0-10 cm soil pH of 0.2956 and 10-20 cm soil pH of 0.3563. The ability of this technique to make predictions of soil acidity with uncertainty allows for prescription lime application while optimizing soil sampling and testing. Further, this paper serves as a case study of on-farm research.