New CPO-funded research, recently published in Nature Climate Change, presents new methods for evaluating the performance of climate model predictions. The statistical methods provide a way to detect failures from a single model or set of models in order to determine whether or not the models are working and if the predictions are reliable. Detecting model failure early allows scientists to build better models, improve climate predictions, and ultimately help policy makers and managers make better decisions for the future.
Decision makers use climate models to predict outcomes of different potential management strategies or scenarios. However, climate predictions can fail when they are unable to capture the uncertainty due to unanticipated elements or behaviors of the climate system. Thus, the authors aimed to develop model evaluation methods that can more effectively determine when climate models might be failing.
The researchers applied and tested their new model evaluation methods on two data sets: northern pintail ducks (an important waterfowl for recreational hunting) and Arctic sea ice. They found that, had climate modelers applied their methods, they would have detected a shift in the pintails’ breeding distribution in 1985, 20 years before scientists discovered the change and policy makers incorporated it into hunting regulations. They also found that summer Arctic sea ice observations are now aligning more with the climate models that predict an ice-free Arctic in September by 2055.
Given the context of a changing climate, policy and management decisions that will affect the future of our planet and its people are becoming increasingly more important. These new methods could help natural resource managers and other decision makers assess climate predictions more easily and adjust or determine better management actions for a more resilient Nation.
This research was funded in part by the NOAA Climate Program Office’s Modeling, Analysis, Predictions, and Projections program.
Access the full paper: http://www.nature.com/nclimate/journal/vaop/ncurrent/full/nclimate3041.html