High resolution land use maps improve understanding of carbon cycle changes
TerraNostra-Bench: Benchmarking Spatial Downscaling and Temporal Rollout for Global 1-km Land-Use Reconstruction
Machine Learning
Summary
Climate predictions struggle because of unclear land use changes on Earth's surface. The authors created a new computer method that combines big, coarse data with local features to make detailed yearly maps of land use at a 1 km scale. They also introduced a way to test how well these methods track land change over time and space. Their tests showed different model approaches have trade-offs between accuracy and errors over many years but can be adjusted without retraining. This work helps build better datasets to track land use changes continuously over time.
What this means in practice
- •For environmental monitoring teams: Create high-resolution annual land-use maps to improve carbon cycle tracking and climate impact assessments.
- •For urban planning departments: Use detailed 1 km land-use reconstructions to inform sustainable development planning and land management decisions.
Authors
Amirpasha Mozaffari, Marina Castaño, Stefano Materia, Joan Vedrí, Jordi Varela-Agrelo, Dario Garcia-Gasulla, Etienne Tourigny, Mario Acosta, Amanda Duarte
Abstract
Uncertainty in the terrestrial carbon cycle remains a major constraint in climate projections, partly due to imperfect land surface representations in Earth system models. We propose a data-driven framework for high-resolution land-use reconstruction, treated as a semantic-segmentation task, that combines coarse-resolution scenario data and climate classification with static geophysical features to produce annual land-use maps at 1 km resolution. We introduce TerraNostra-Bench, a change-aware evaluation protocol that separates spatial downscaling from autoregressive temporal rollout and scores models on the rare land-use-change pixels rather than the dominant static background. Using it, we compare convolutional (U-Net) and transformer (Swin) backbones in autoregressive and non-autoregressive schemes against persistence, linear, and shallow convolutional baselines. The autoregressive prior is the most informative input yet accumulates error over long rollouts: autoregressive models lose 5.7-7.2 accuracy points over 13 years, against 2.2-2.8 for non-autoregressive models. This trade-off is governed by the autoregressive choice rather than the backbone. It can be tuned by an inference-time prior-trust threshold that requires no retraining and by the training-time weighting of reconstruction and change objectives; the two levers trace complementary frontiers, with calibration recovering up to 10 accuracy points while keeping change sensitivity above the persistence floor. Extrapolating beyond the training window, backward to 1955-1959 and forward to 2015-2019 under scenario forcing, the models retain most of their in-window skill and show the same trade-off, indicating it is intrinsic to the design. This work provides the methodological foundation and evaluation surface for a forthcoming temporally continuous land-use product.