AI4Land: Scalable Deep Learning for Global High-Resolution Land Use Reconstruction

2026-06-10

AI Summary

This study addresses the critical limitation in climate prediction accuracy imposed by uncertainties in land surface process representation, primarily due to the lack of high spatiotemporal resolution historical and future land use data. To overcome this, we propose a two-stage deep learning framework based on the U-Net architecture that fuses coarse-resolution scenario data with static geophysical features to achieve the first scalable, global, annual, high-resolution reconstruction of land use and land cover. Trained on Earth observation data and accelerated via GPU on the MareNostrum5 supercomputing platform, the model supports real-time coupling with digital twin systems. The resulting product exhibits physical consistency, spatial explicitness, and temporal continuity, substantially reducing land surface representation uncertainty and providing essential inputs for next-generation climate simulations.

Abstract

Uncertainty in the terrestrial carbon cycle remains a major constraint in climate projections, partly driven by the uncertainties affecting the land surface representation and variability in Earth system models. To address this limitation, we present a data-driven framework AI4Land, for generating high-resolution historical reconstructions and future projections of key land surface variables. The framework follows a two-phase approach using a U-Net architecture. In the first phase, which is the focus of this work, it reconstructs annual land use and land cover by integrating coarse-resolution scenario data with static geophysical features. In a planned second phase, the resulting high-resolution maps will be used to predict dynamic biophysical variables, particularly leaf area index, at finer temporal scales. Trained on Earth observation data, the models learn to reproduce spatially explicit and physically consistent land surface patterns, extending temporal coverage to periods lacking direct observations. AI4Land was developed and trained on MareNostrum5, demonstrating how GPU-accelerated HPC infrastructure enables global-scale climate AI pipelines. The final product is a suite of open-source emulators designed for real-time coupling with digital twin platforms, such as those developed under the Destination Earth initiative. By delivering realistic and evolving land surface conditions on demand, this work aims to reduce critical uncertainties and improve the predictive power of next-generation climate simulations.

Problem

• land use reconstruction
• terrestrial carbon cycle
• Earth system models
• high-resolution
• climate projections

Innovation

• U-Net architecture
• high-resolution land use reconstruction
• GPU-accelerated HPC
• Earth system modeling
• digital twin integration