Mapping Multiple Soil Properties Using Artificial Neural Networks: Creating Coherent Soil Dataset

Stockholm

2024

1 S.

Bandaufführung

9087; 13354S

40005066

Mapping multiple soil properties is essential for soil-related ecological modeling, including the prediction of tree species suitability. In this study, we propose an approach that utilizes artificial neural networks (ANN) to simultaneously map soil properties, ensuring the generation of a coherent and plausible dataset. Our aim is to enhance the compatibility and plausibility of the predicted dataset by considering the interdependencies among different soil properties. To achieve this, we employed ANN to model individual soil properties while incorporating independent covariates such as climatic, morphometric, geological, and remote-sensing information. Our approach aimed to maintain the accuracy of the predictions by considering the relationships among soil properties and their interplay with the covariates.