Search papers, labs, and topics across Lattice.
To move beyond passive scientific emulators that cannot model counterfactuals, this work casts Earth-system emulation as an action-conditioned world model capable of handling interactive, user-directed what-if interventions. The framework employs transition-action pretraining鈥攖reating naturally occurring state deltas as self-supervised action labels鈥攑aired with masked response learning to infer unobserved variables and preserve coupled system dynamics. Validated on terrestrial ecosystem simulations across six global regions, the model matches the long-horizon accuracy of traditional passive emulators while successfully generating coherent, system-wide responses to partial state manipulations.
Treating natural state trajectories as self-supervised "actions" turns static scientific emulators into interactive, counterfactual world models without requiring manually labeled intervention data.
Machine learning emulators have become essential for accelerating expensive Earth-system simulations, but most existing approaches remain passive forecasters: they reproduce simulator trajectories under prescribed forcings without an explicit interaction mechanism for user-specified interventions. This limits their use in interactive scientific workflows and Earth-system digital twins, where users often need to explore how a system would respond if selected state components were changed. We propose an action-conditioned world-modeling framework for Earth-system emulation that reformulates simulator trajectories as supervision for controllable state-transition learning. The key idea is transition-action pretraining: naturally observed state changes are treated as label-free action supervision, allowing the model to learn both prescribed dynamics and action-conditioned responses without manually annotated interventions. We further introduce masked response learning to infer unobserved variables under partial state edits and learn coupled system dependencies. We test this framework on ecosystem dynamics across six global regions and multiple stand ages. Experiments show that the model preserves competitive long-horizon emulation accuracy while enabling controllable structural interventions and coherent responses in coupled ecosystem-cycle variables. These results suggest a practical route from passive Earth-system emulators toward interactive, intervention-aware scientific surrogates.
