Explicit depth representation improves the 3D modelling of soil carbon to nitrogen ratio in croplands

Abstract

The soil carbon-to-nitrogen ratio (C:N ratio) is a fundamental indicator of organic matter quality and nutrient cycling, yet its three-dimensional (3D) variability in croplands remains poorly understood. Accurate mapping of the C:N ratio is essential for evaluating sustainable agricultural practices, as it reflects the balance between soil organic carbon (SOC) and total nitrogen (TN) stabilization. Using the Hangjiahu Plain in the Yangtze River Basin of China, as a case study, we evaluated the predictive performance of 2.5D and 3D digital soil mapping (DSM) frameworks for soil C:N ratio to a depth of 1 m. A key methodological focus was comparing the direct prediction of the C:N ratio against indirect calculation from separate SOC and TN models. The results demonstrate that C:N ratio declined with soil depth, decreasing from 16.63±2.12 in the 0-20 cm to 14.58±3.87 in the 60–100 cm. Models based on a 3D assumption where depth is treated as an explicit spatial dimension consistently outperformed 2.5D models in estimating C:N ratio, achieving an R2 of 0.56 and an RMSE of 2.25. Crucially, direct prediction of the C:N ratio significantly improved model stability and accuracy, particularly in deeper layers. By reducing uncertainty in the 60–100 cm, the 3D direct-prediction approach provides a more robust representation of vertical stoichiometric gradients. These findings suggest that vertically informed DSM frameworks are vital for the reliable mapping of soil C:N ratio in high-resolution, profile-resolved agricultural assessments.