Matt Osman
Matt Osman is Assistant Professor of Climate Science in the Department of Geography at the Universtiy of Cambridge. His interests are in exploring new ways of comparing model outputs with large observational datasets; the goal is to better our understanding of long-term climate changes and their drivers. His group is developing several reanalysis-based estimates of global Quaternary change by compiling, redating, and assimilating globally distributed proxies into models, accounting for uncertainties in both datasets, in collaboration with modellers, statisticians, and proxy experts. Matt also maintains a strong fascination with the cryosphere, particularly on past ice–climate interactions, Arctic sea-ice sensitivity, and marine productivity.
Towards uncertainty reduction of future climate projections using paleoclimate inference
Assessing climate model performance under vastly different paleoclimate states offers the chance to refine future projections. But few, if any, frameworks exist that can systematically account for changing uncertainties and data availability across geologic intervals, models, and proxy systems. Here, I outline a new probabilistic (Bayesian) method that pairs together available PMIP3/4 models (n = 30) with >1300 chronologically consistent geochemical temperature proxies spanning the Mid-Pliocene, Last Interglacial, Last Glacial Maximum, and mid-Holocene. Established and newly developed proxy forward models are used to capture changing seasonality and multi-environmental effects, to provide more physically grounded model-data comparisons. The method then iteratively “learns” which models outperform others across intervals and boundary conditions, thereafter weighting future CMIP5/6 projections accordingly. This data-driven approach strives toward a longstanding ambition of paleoclimatology: using past climates to (probabilistically) inform our future.
