Problem
Research questions and friction points this paper is trying to address.
Reducing uncertainties in climate projections
Merging knowledge from models and observations
Improving regional temperature pattern accuracy
Transferring climate change physical knowledge
🤖 AI Summary
Climate prediction using Earth System Models (ESMs) suffers from substantial uncertainty due to inadequate representation of nonlinear feedback processes, hindering robust adaptation and mitigation decisions. To address this, we propose the first application of transfer learning to cross-source climate knowledge integration, jointly leveraging multi-source historical observations and ESM simulations within an observation-constrained transfer optimization framework. This approach overcomes fundamental limitations of conventional methods in modeling complex physical feedbacks. Our method achieves synergistic improvements in regional pattern fidelity and uncertainty quantification for surface air temperature predictions: global prediction uncertainty is reduced by over 50%; regional spatial patterns better align with both physical principles and observational evidence; and projection intervals are significantly narrowed while maintaining higher statistical credibility. The framework establishes a novel paradigm for high-confidence, observation-informed climate services.
📝 Abstract
Precise and reliable climate projections are required for climate adaptation and mitigation, but Earth system models still exhibit great uncertainties. Several approaches have been developed to reduce the spread of climate projections and feedbacks, yet those methods cannot capture the non-linear complexity inherent in the climate system. Using a Transfer Learning approach, we show that Machine Learning can be used to optimally leverage and merge the knowledge gained from Earth system models simulations and historical observations to reduce the spread of global surface air temperature fields projected in the 21st century. We reach an uncertainty reduction of more than 50% with respect to state-of-the-art approaches, while giving evidence that our novel method provides improved regional temperature patterns together with narrower projections uncertainty, urgently required for climate adaptation.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Transfer Learning to merge model and historical data
Machine Learning reduces climate projection spread
Uncertainty reduction over 50% for temperature patterns
🔎 Similar Papers
No similar papers found.
