Despite the cogency of addressing risks from climate change and the availability of terabytes of data, robust quantitative estimates of these risks still elude us. Translating our climate science studies into actionable information remains a challenge. Ironically the difficulties of uncertainty quantification fueled my career, bginning with the very first time I handled climate model output, 25 years ago. Those difficulties continue to provide study material and motivation for many of us. Climate models’ structural uncertainties arguably increase with every CMIP phase; the noise of internal variability and its interaction with climate change signals pose difficult trade-offs to our simulation capabilities and resources; non-stationary extreme events, and their tendency now to occur more frequently and jointly across space, time and variables, stretch the capabilities of our statistical modeling; and last but not least, the human system components of risk - exposure, vulnerability, responses – are even harder to model, especially when we try to characterize feedbacks and loops between Earth and humans. Grateful for this opportunity to give a brief overview of the themes of my research, addressed together with my invaluable collaborators over the years, I want to leave the audience with the message that truly interdisciplinary work, from which my career greatly benefited, is key to our progress in tackling these questions.