New CarbonBench benchmark evaluates global carbon flux upscaling with zero-shot learning

A new benchmark called CarbonBench has been introduced to evaluate the performance of zero-shot learning models in upscaling carbon fluxes globally. This benchmark includes over 1.3 million daily observations from 567 flux tower sites worldwide, covering the period from 2000 to 2024. CarbonBench is designed to test model generalization across different vegetation types and climate regimes, providing a standardized method for comparing transfer learning approaches and advancing climate modeling efforts. AI

IMPACT Enables more rigorous evaluation of AI models for climate change monitoring and policy.

RANK_REASON The item is a research paper introducing a new benchmark for machine learning applications in Earth system science. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CarbonBench
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →