Our climate is currently changing in a dangerous and potentially unprecedented manner due to human interference. The best information we have on our future climate comes from the past as past climate provides a window into the speed at which climate can change, and an indication of how drastic these changes can be. Unfortunately, we cannot measure past climate using thermometers or weather stations, instead we must rely on the imperfect records left by plants, trees, and chemical signatures in rocks and ice. Transforming these imperfect records into estimates of past climate requires complex mathematics, statistics, and knowledge of the chemistry and biology of how climate is recorded in the natural environment. The models we create to combine these various sources of information then pose computational challenges that are daunting in scale.
One of the key developments in the last 10 years has been the idea that, whilst an individual record might provide an incomplete and highly uncertain message about past climate, combining different types of records from different places can massively reduce uncertainty. So, rather than using a single past measure, e.g. just tree rings from a single forest, we can use tree rings from multiple forests together with pollen, stalactites, midges, ice from the Arctic, and hundreds of other imperfect records, all combined and tested against our knowledge about how climate changes in space and time.
Most importantly, the goal of combining all this information together coherently has not yet been realized. Some groups have taken small chunks of the data and averaged it together to try to provide some sort of climate signal, but no-one has been able to do it yet with properly quantified uncertainties to estimate past climate change over space and time.
In WIREs Computational Statistics, a paper provides a review of both the original and latest methods in this field. The authors found that, remarkably, the most promising approach involves an equation known as Bayes’ theorem. The Bayes’ theorem is a method for combining uncertain data invented by a Presbyterian minister Thomas Bayes (and his executor Richard Price) over 250 years ago. James Sweeney and colleagues, the authors of this paper, hope that those who are building the next generation of climate forecasts can use the identified methods to better inform the likelihood of rapid changes or tipping points in the climate system.
Kindly contributed by the Authors.
