Researchers from the field of environmental science and AI have investigated how uncertainty in learned transport processes affects exposure mappings and downstream spillover inference. They compared different transport models, including mechanistic and operator-learning approaches, using both simulated data and real-world pollution data from California. The study found that while these models can accurately predict pollution levels, they may produce significantly different estimates of the effects of interventions on air quality. This has implications for reliable causal inference in environmental settings where exposure mappings are learned rather than directly observed. --- Why it matters: This research matters to engineers and researchers working on AI applications in environmental science because it highlights the importance of considering uncertainty in transport processes when making predictions about the effects of interventions. Reliable causal inference is crucial for informing policy decisions that can have significant impacts on public health and the environment. Source: https://arxiv.org/abs/2608.19224