Abstract
Data from decadal hindcast experiments conducted under CMIP5 were used to assess the ability of CM2.1, HadCM3, MIROC5, and CCSM4 Earth System Models (ESMs) to hindcast sea-surface temperature (SST) indices of the Pacific Decadal Oscillation (PDO), the tropical Atlantic SST gradient (TAG) variability, and the West Pacific Warm Pool (WPWP) SST variability from 1961 to 2010. The ESMs were initialized at specific times with observed data to make 10- and 30-year hindcasts/forecasts. Deterministic and probabilistic skill estimates show predictability of detrended WPWP index to 5 years’ lead time and of non-detrended WPWP index to 10 years’ lead time. These estimates also show atypical skill dependence of PDO and TAG indices on lead times, with increasing skill in the middle to end of 10-year hindcasts. The skill of ESMs to hindcast an observed DCV index (signal skill) is surprisingly greater than the skill to hindcast their own DCV index (noise skill) at some lead times. All ESMs hindcast occurrence frequencies of positive and negative phases of the indices, and probabilities of same-phase transitions from one year to the next reasonably well. Four, major, low-latitude volcanic eruptions are associated with phase transitions of all observed and some of the ensemble-average hindcast indices. All ESMs’ WPWP index hindcasts respond correctly to all eruptions as do three observed PDO phase transitions. No one of the ESMs’ hindcasts of the TAG index responds correctly to these eruptions. Some of the ESMs hindcast correct phase transitions in the absence of eruptions also, implying that initializations with observed data are beneficial in predicting phase transitions. The skills of DCV indices’ phase prediction up to at least two years in advance can be used to inform societal impacts adaptation decisions in water resources management and agriculture. The Atlantic region’s responses in these ESMs appear to be fundamentally incorrect.
Highlights
Among natural decadal climate variability (DCV) phenomena, the Pacific climate variability generally known as the Pacific Decadal Oscillation (PDO; Mantua et al 1997) or the Inter-decadal Pacific Oscillation (IPO; Power et al 1999), the tropical Atlantic sea surface temperature (SST) gradient
CCSM4, CM2.1, HadCM3, and MIROC5 earth system models (ESMs)—and the model ensemble (MME) formed from these ESM hindcasts—from 1961 to 2010
We conducted case studies of individual, sustained phase transitions in the ensembles of decadal hindcasts in order to attribute the transitions to volcanic eruptions or internal ocean–atmosphere variability
Summary
Among natural decadal climate variability (DCV) phenomena, the Pacific climate variability generally known as the Pacific Decadal Oscillation (PDO; Mantua et al 1997) or the Inter-decadal Pacific Oscillation (IPO; Power et al 1999), the tropical Atlantic sea surface temperature (SST) gradientPresent Address: NOAA/Climate Prediction Center, College Park, MD, USA (TAG; Hastenrath 1990; Houghton and Tourre 1992; Mehta and Delworth 1995; Mehta 1998; Rajagopalan et al 1998), variability of the West Pacific Warm Pool (WPWP) SST (Wang and Mehta 2008), and their impacts on global climate are attracting increasing attention in predictability and prediction studies because of their impacts on water resources, agriculture, hydro-electricity generation, inland water-borne commerce, and fish and crustacean stocks and captures (Mehta 2017). Some fundamental and substantial problems of decadal climate prediction are (Meehl et al 2009, 2014; Mehta et al 2011a): (1) relatively short time series of instrument-based global ocean observations, especially sub-surface observations, for understanding, model initialization, and comparison with prediction; (2) an insufficient understanding of fundamental physics of DCV; (3) an insufficient theoretical understanding of possible behaviors of geographicallyvarying, complex and non-linear dynamical systems with mixed initial and boundary values; (4) global climate models displaying less than satisfactory skill in simulating climate in general and DCV in particular; and (5) insufficient guidance from stakeholders and policymakers as to which DCVrelated climate, weather, and impacts information would be useful for applications to societal impacts of DCV if predicted In spite of these problems, there have been many encouraging decadal prediction studies with global earth system models (ESMs). Meehl et al (2014) have described results from hitherto published CMIP5 and other decadal hindcasting experiments, so only major results pertaining to predictability of indices of decadal SST variability are reviewed here
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
