Supported by growing observational data and computational power, as well as success stories from other research fields, data-driven methods are increasingly used in soil and biogeosciences. New artificial intelligence (AI) and machine learning (ML) methods enhance our ability to describe soils and model soil variables. However, their introduction has been accompanied by overpromises and challenges, such as the relative data sparsity in soil science compared to other biogeoscience domains. In this session, we would like to address the following questions:
• How can AI support soil and land resource inventories, leading to better decision-making in land management? What role can AI play in capturing climate and vegetation dynamics' impacts on soil properties?
• To what extent can data-driven soil research complement traditional hypothesis-driven approaches by generating new hypotheses for observed patterns?
• How do data-driven, ML-based approaches compare with mechanistic models in predicting soil properties and understanding soil processes? What are the benefits and challenges of hybrid models that combine data-driven insights with mechanistic knowledge? How can we better communicate challenges of data-driven research, such as uncertainty quantification, especially when extrapolating beyond training data?
• What new remote sensing data or covariates are emerging as valuable tools in predictive soil modeling? For instance, can Synthetic Aperture Radar (SAR) data effectively predict soil organic carbon (SOC) and other soil properties, and how do bare soil observations enhance these predictions?
• In what ways can we use foundational Earth models to improve soil predictions and simulate scenarios like land use changes and climate impacts? Can these models help scale local research to regional or global levels, enhancing predictions and supporting sustainable land management?
• How can we leverage large language models (LLMs) to enhance data analysis and improve research efficiency in soil science? Can LLMs synthesize complex soil data to provide insights for sustainable land management?
This session encourages contributions from various research fields, especially case studies where soil scientists, computer scientists, agronomists, and data scientists collaborate to solve soil-related problems. We invite abstracts on the practical use of AI in soil research, including successes, failures, and future ideas.
• How can AI support soil and land resource inventories, leading to better decision-making in land management? What role can AI play in capturing climate and vegetation dynamics' impacts on soil properties?
• To what extent can data-driven soil research complement traditional hypothesis-driven approaches by generating new hypotheses for observed patterns?
• How do data-driven, ML-based approaches compare with mechanistic models in predicting soil properties and understanding soil processes? What are the benefits and challenges of hybrid models that combine data-driven insights with mechanistic knowledge? How can we better communicate challenges of data-driven research, such as uncertainty quantification, especially when extrapolating beyond training data?
• What new remote sensing data or covariates are emerging as valuable tools in predictive soil modeling? For instance, can Synthetic Aperture Radar (SAR) data effectively predict soil organic carbon (SOC) and other soil properties, and how do bare soil observations enhance these predictions?
• In what ways can we use foundational Earth models to improve soil predictions and simulate scenarios like land use changes and climate impacts? Can these models help scale local research to regional or global levels, enhancing predictions and supporting sustainable land management?
• How can we leverage large language models (LLMs) to enhance data analysis and improve research efficiency in soil science? Can LLMs synthesize complex soil data to provide insights for sustainable land management?
This session encourages contributions from various research fields, especially case studies where soil scientists, computer scientists, agronomists, and data scientists collaborate to solve soil-related problems. We invite abstracts on the practical use of AI in soil research, including successes, failures, and future ideas.