Search papers, labs, and topics across Lattice.
The paper introduces AI4Land, a scalable framework utilizing a U-Net architecture to reconstruct high-resolution historical and future land use data by integrating coarse-resolution scenario data with static geophysical features. This approach addresses significant uncertainties in the terrestrial carbon cycle that hinder climate projections by providing spatially explicit and physically consistent land surface patterns, even extending to periods lacking direct observations. Trained on advanced GPU-accelerated infrastructure, AI4Land produces open-source emulators that enhance real-time coupling with digital twin platforms, thereby improving the predictive capabilities of climate simulations.
AI4Land reduces critical uncertainties in climate models by delivering high-resolution land use reconstructions that enhance predictive accuracy.
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.
