Evaluating the changes of Ultisol chemical properties and fertility characteristics due to animal manure amelioration

Abstract

Amending Ultisols using organic matter encourages a paramount improvement in its chemistry and fertility characteristics. This study was aimed to evaluate the changes in soil chemical properties due to the animal manure amelioration in Ultisol in the Jasinga, Bogor, West Java, using classical and advanced statistical methods. Composite soil samples were collected then incubated with three types of animal manure (cow, chicken, and goat) and four rate levels (0, 2.5, 5, and 7.5% of dry weight). The dynamics of eleven soil variables (pH, organic C, total N, cation exchange complex/CEC, base saturation/BS, and exchangeable Al, H, Ca, Mg, K, and Na) were observed four times (0, 2, 4, and 6 weeks). Basic cation saturation ratio/BCSR and sufficiency level of available nutrients/SLAN soil fertility approaches were applied. Modeling comparison was done among multiple linear regression/MLR, machine learning/ML (tree regression/TR, random forest/RF, gradient boosting machine/GBM), and deep learning/DL (multilayer perceptron/MLP). Most of the soil chemical and fertility parameters exhibited strong relation among three applied factors. Generally, their values failed to reach the BCSR’s ideal soil and national SLAN’s sufficiency criteria; oppositely, they were categorized as sufficient based on the global SLAN approach. Multivariate analysis revealed the similarity among manure type and rate, whereas incubation time showed the opposite trend. MLR usage was convenient in modeling BS, pH H2O, and Al saturation. Meanwhile, CEC modeling requires more sophisticated methods. This study highlighted the possible improvement of Ultisol chemical properties and fertility characteristics by amending it with a higher rate and low C/N ratio of animal manure, and using ML to capture non-linear relationships in soil.