Modeling temporal variation of soil acidity after the application of liming materials

Soil acidification is a natural phenomenon that human activity can accelerate or manage. Acid soils limit plant growth by reducing nutrient availability and causing aluminum toxicity, significantly reducing crop production. Liming has proven efficient for increasing crop yield on acidic soils. Knowledge is required on the effect of liming on the soil pH dynamics to find the suitable material, application rate, timing, and method. The objective of this study was to develop a prediction model of soil pH temporal variations after lime application using data from the literature. A database was built from research results extracted from 16 scientific papers that provided data on soil acidity changes over time under different liming treatments. Machine learning (ML) was used to predict soil pH dynamics from eight predictive parameters: rate of application, time since application, neutralizing value (CCE), grind fineness (D50), pH measurement depth, soil acidity (pH prior to liming), the type of solution used for pH measurement (water or CaCl2) and the soil:solution ratio used to measure pH. Since soil pH fluctuates with seasons, all pH values were expressed as difference between the actual pH value and that of an unlimed control (ΔpH). The Random Forest (RF) algorithm was tested to predict ΔpH over time. On testing, we obtained an R2 between measured and predicted ΔpH values of 0.881 and an RMSE of 0.230. These results are excellent, considering the heterogeneity in soils, liming materials, and pedoclimatic conditions found in the 16 papers. The primary factors influencing ΔpH, ranked by their impact are application rate, time elapsed post-application, and lime characteristics including its grind fineness and neutralizing value.