Search papers, labs, and topics across Lattice.
CarbonBench is introduced as the first benchmark for zero-shot spatial transfer learning in carbon flux upscaling, addressing the challenge of generalizing carbon exchange models to underrepresented ecosystems. It comprises 1.3 million daily observations from 567 flux tower sites globally and provides stratified evaluation protocols to test generalization across unseen vegetation types and climate regimes. The benchmark also includes harmonized remote sensing and meteorological features and baseline models for comparison.
Finally, a standardized benchmark to rigorously evaluate how well models generalize carbon flux predictions to geographically distinct ecosystems they've never seen before.
Accurately quantifying terrestrial carbon exchange is essential for climate policy and carbon accounting, yet models must generalize to ecosystems underrepresented in sparse eddy covariance observations. Despite this challenge being a natural instance of zero-shot spatial transfer learning for time series regression, no standardized benchmark exists to rigorously evaluate model performance across geographically distinct locations with different climate regimes and vegetation types. We introduce CarbonBench, the first benchmark for zero-shot spatial transfer in carbon flux upscaling. CarbonBench comprises over 1.3 million daily observations from 567 flux tower sites globally (2000-2024). It provides: (1) stratified evaluation protocols that explicitly test generalization across unseen vegetation types and climate regimes, separating spatial transfer from temporal autocorrelation; (2) a harmonized set of remote sensing and meteorological features to enable flexible architecture design; and (3) baselines ranging from tree-based methods to domain-generalization architectures. By bridging machine learning methodologies and Earth system science, CarbonBench aims to enable systematic comparison of transfer learning methods, serves as a testbed for regression under distribution shift, and contributes to the next-generation climate modeling efforts.
