CALIPSO

Using explainable AI to diagnose the representation of environmental drivers in process-based soil organic carbon models

Posted by Design Studio

16 December 2025

Challenge 2: Soil Carbon


Using explainable AI, we quantified the poor performance of two process-based soil carbon models, including the model developed in Project 2. The poor performance is primarily associated with the omission of key variables, such as soil cation exchange capacity, and with substantial errors in the modelled relationship between soil carbon and carbon inputs. These findings will guide future model improvements.