AI Insight
CarbonBench is a new standardized benchmark dataset designed to evaluate machine learning models that estimate terrestrial carbon exchange across different geographic locations and ecosystem types. The dataset contains over 1.3 million daily observations from 567 eddy covariance flux tower sites worldwide collected between 2000-2024, and specifically tests whether models can accurately predict carbon fluxes in ecosystems they have not been trained on (zero-shot spatial transfer learning). The benchmark provides harmonized remote sensing and meteorological data, stratified evaluation protocols that separate spatial generalization from temporal patterns, and baseline model comparisons to enable rigorous testing of transfer learning approaches.
Why it matters
Accurate carbon flux estimation is critical for climate policy and carbon accounting, but current models struggle with ecosystems that lack observational data. This benchmark enables systematic development and comparison of machine learning methods that can generalize to underrepresented regions, potentially improving global carbon cycle predictions and supporting more effective climate mitigation strategies.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
-cross
Abstract: 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.
Source: CarbonBench: A Global Benchmark for Upscaling of Carbon Fluxes Using Zero-Shot Learning