We present ongoing research using weather foundation models to generate tropical cyclone counterfactuals for experimental design and causal discovery to better understand the underlying causes of extreme events in the Earth's atmosphere. Predicting atmospheric extreme events is critical for preparation and damage mitigation. Accurate prediction of extreme events is challenging because they are the result of multiple conditions converging through complex spatiotemporal structures. Each natural event is unique, so research is often conducted with Earth system models (ESMs) to simulate different conditions and causal factors. However, ESMs are computationally expensive, significantly limiting counterfactual studies. The Aurora Earth system foundation model [1] enables the rapid generation of significantly more counterfactual studies of tropical cyclone formation compared to fully-coupled Earth system models. Rapid counterfactual generation enables more rigorous causal investigations, with methods such as randomized controlled trials and active causal discovery of the drivers of cyclone formation and intensification. In contrast to typical post-hoc causal discovery of non-interventional data [2], active causal discovery iteratively intervenes on the system to produce a more robust causal graph [3]. In this talk, we will present preliminary results of generating targeted interventions to model inputs in the Aurora model for causal analysis of tropical cyclone drivers. New methods for more rigorous counterfactual analysis of extreme events can help identify their causes, improving prediction accuracy and enhancing preparedness for these destructive natural forces. [1] Bodnar, C., Bruinsma, W. P., Lucic, A., Stanley, M., Allen, A., Brandstetter, J., Garvan, P., Riechert, M., Weyn, J. A., Dong, H., Gupta, J. K., Thambiratnam, K., Archibald, A. T., Wu, C.-C., Heider, E., Welling, M., Turner, R. E., Perdikaris, P. (2025). A foundation model for the Earth system. Nature, 641, 1180-1187. [2] Glymour, C., Zhang, K., Spirtes, P. (2019). Review of Causal Discovery Methods Based on Graphical Models. Frontiers in Genetics, 10, 524. [3] Wang, Y., Liu, M., Sun, X., Wang, W., Wang, Y. (2025). Bayesian Active Learning for Bivariate Causal Discovery. Forty-second International Conference on Machine Learning.