The butterfly’s shadow: how subtle surface forcing elicits remote atmospheric responses

Land-atmosphere coupling influences critical climate processes from regional drought to seasonal predictability, yet existing methodologies struggle to capture the full spectrum of teleconnection mechanisms. Traditional approaches—coordinated multi-model experiments, such as GLACE comparing interactive versus fixed land surface variables, forecast initialization studies, and idealized extreme event simulations—each have distinct drawbacks in signal detection capability, often requiring unrealistic perturbations that likely change model climatology. We present an enhanced system identification approach that overcomes such limitations through optimized signal design and improved response analysis. Our methodology injects artificial signals with zero-mean characteristics that preserve climatological states while enabling unprecedented detection of atmospheric responses via sliding regression. Unlike conventional step-response experiments, this approach enables detection of atmospheric responses while minimizing system perturbation through small neutral forcing. Results demonstrate successful capture of dynamic atmospheric responses to surface forcing. Results demonstrate successful capture of dynamic atmospheric responses to surface forcing. The technique traces propagation pathways from localized surface heating to upper-tropospheric circulation patterns. This approach enables climate model evaluation, seasonal prediction enhancement, and mechanistic understanding of climate feedbacks, with future applications extending to assess teleconnection capabilities in AI-based climate models such as NeuralGCM and Ai2 Climate Emulator.