This is one of 16 Rapid Response (RAPID) projects funded as the result of a Dear Colleague Letter (NSF 11-006) encouraging diagnostic analyses of climate model simulations prepared for the Intergovernmental Panel on Climate Change Fifth Assessment Report (IPCC AR5). Research conducted in these projects is expected to lead to more detailed model intercomparisons, better understanding of robust model behaviors, and better understanding and quantification of uncertainty in future climate simulations.
This project considers the potential predictability of decadal climate variabiality in hindcast simulations prepared for the AR5. These 10-year hindcast experiments are performed with models in which ocean properties, principally sea surface temperature (SST), are initialized using observations every five years starting in 1960, with additional 30-year hindcasts starting in 1960, 1980, and 2005 (actually a forecast in the later years). The hindcast experiments are a novel feature in AR5 with no counterpart in the Fourth Assessment Report, and they are designed to determine the extent to which decadal-scale climate changes of relevance to policy makers are predictable given the state of the ocean and the beginning of the forecast period.
Research conducted in this project assess the skill of atmospheric teleconnections forced by SST anomalies in the Pacific and Indian Oceans that are predictable on decadal timescales in the hindcast ensemble. Skill is assessed by comparisons between observed and hindcasted patterns of rainfall, air temperature, and circulation anomalies. Additional comparisons are performed with uninitialized simulations of the 20th century forced by greenhouse gas increases and other external forcings (e.g. volcanos, aerosols, and solar variability), so that the extent to which predictability depends on external forcing rather than initialized internal variability can be assessed.
The broader impact of the project lies in its support of the IPCC AR5, which is intended to provide information on climate change and its consequences to decision makers worldwide. Successful decadal climate forecasts could be of value to to resource managers developing informed adaptation strategies, but first the degree to which such forecasts are achievable must be determined.
In this work I have developed, in collaboration with Matt Newman at the University of Colorado, a technique to distinguish ENSO variability. We take advantage of the fact that observed tropical SST variability is well described as multivariate rednoise, a stochastically forced linear dynamical system where all evolving perturbations are stable to exponential growth but some can, and for ENSO do, experience substantial transient growth and decay over finite time intervals. Because multivariate red noise provides a baseline for the statistics of observed tropical seasonal anomaly evolution, it also serves as a useful null hypothesis against which possible changes in SST can be tested. Motivated by these results, we developed a new optimal perturbation filter that uses a LIM to remove space and time-varying ENSO anomalies from the ocean temperature data record. We demonstrate that uncertainty in tropical Indo-Pacific SST trends in different reconstruction scan be explained by disparate estimates of ENSO variability and that removing this variability results in consistent centennial trends. This work was presented as an invited contribution at the Third Santa Fe Conference on Global and Regional Climate Change and at the American Geophysical Union 2011 Fall Meeting. This study was published in Nature Climate Change online 8 July 2012. In a companion study, using ensembles of initialized and uninitialized decadal hindcasts with yearly start dates from one coupled climate model, I have demonstrated that these experiments are useful in identifying model biases that contribute to uncertainties in simulations of the response of the tropical Indo-Pacific to an increase in greenhouse gases. In this analysis I have focused on diagnosing the processes that cause the prominent cooling trend in the equatorial thermocline in ocean reanalyses and what causes this cooling trend to dissipate in decadal hindcasts for forecasts beyond one year. This methodology can be extended to identifying additional biases (for example,why the eastern equatorial surface temperature trends increase for longer thanone year forecasts even though the overturning circulation strengthens) and other modelst o identify potentialmodel uncertainties in climate change projections. This work was presented at the CMIP5 workshop on Model Analysis in Honolulu, Hawaii 5-9 March 2012. Journal Publications: Amy Solomon and Matthew Newman, "Reconciling disparate twentieth-century Indo-Pacific ocean temperature trends in the instrumental record", Nature Climate Change, vol. , (2012), doi:10.1038/nclimate1591. Amy Solomon, "Using initialized decadal forecasts to identify model uncertainties in the response to external forcing in the tropical Indo-Pacific Ocean", Journal of Climate, vol. , (2012), p. ., Submitted.