Leveraging long-term field experiments and other datasets to develop AI-ready decision support systems for sustainable soil management

There is a need for standardized methods in collecting and reporting soil data to ensure consistency and comparability across different studies and regions.

Opportunities exist to integrate soil data from long-term field experiments with other data sources (including the Mission Soil projects results, but not only) to provide a more comprehensive understanding of soil health dynamics and trend, as well as response to policies and management strategies.

Enhancing the accessibility and interoperability of soil data across platforms and sources can facilitate collaborative research and accelerate advancements in soil health management.

More high-resolution temporal and spatial data are needed to capture short-term soil dynamics and site-specific variations that can influence broader interpretations of soil health trends.

design and implement standardised protocols and procedures for harmonised soil data collection, ensuring consistency and comparability, from different LTEs and regions across the EU and Horizon Europe Associated Countries;

develop robust frameworks for integrating LTE data with other relevant soil health datasets, including outputs from Mission Soil projects, to create comprehensive soil health databases;

develop open-access and user-friendly interoperable systems and platforms to improve data sharing and accessibility, allowing researchers, advisors, land managers and other stakeholders to easily access and utilize comprehensive soil health information;

build a network of at least 50 LTEs covering most representative pedo-climatic regions in the EU and Associated Countries involving at least 7 owning institutions, to test and validate the developed infrastructure;

promote the use of the developed infrastructure for widespread collection and integration of as many as possible soil-health relevant databases (LTEs and others) by, for example, developing intuitive interfaces and user-friendly platforms, partnering with relevant organizations managing LTEs and/or generating datasets, demonstration projects, feedback and improvement loops or training and support services;

develop AI-driven tools to analyze integrated datasets (including publicly available such as CORDIS, SoilWise repository or repositories like Zenodo), extracting meaningful patterns, and generating predictive models that inform soil health dynamics and management strategies;

examine potentially correlated explanatory covariates and their relative contribution to the outcome to facilitate spatial downscaling and forecasting in data poor regions and areas by using pre-trained deep learning models;

develop and train open-source and/or modular AI components, providing comprehensive documentation and tutorials, and establish and nurture open-source communities by, for example, hosting hackathons, workshops, or online platforms to encourage the development, sharing, and integration of the developed modular AI components into commercial applications for land managers and advisors, with a focus on small-scale producers;

mine large data from publicly available databases (e.g. CORDIS, SoilWise repository or repositories like Zenodo) to pre-train deep learning models and artificial intelligence mobile apps that will facilitate real-time soil status assessments.