Impact of Consistency between the Climate Model and its Initial Conditions on Decadal Climate Prediction

• Main Author: Liu, Xueyuan
• Other Authors: Stammer, Detlef (Prof. Dr.)
• Format: Doctoral or Postdoctoral Thesis
• Language: English
• Published: Staats- und Universitätsbibliothek Hamburg Carl von Ossietzky 2014
• Subjects: decadal prediction; initialization; El Niño; momentum balance; equatorial Pacific; 550 Geowissenschaften; 38.90 Ozeanologie; Ozeanographie; ddc:550; Pacific; North Atlantic
• Online Access: http://nbn-resolving.de/urn:nbn:de:gbv:18-72533; https://ediss.sub.uni-hamburg.de/handle/ediss/5815

In my thesis I investigated the influence of dynamical consistency of initial conditions with the model used to perform forecasts starting from this initial condition, on the predictive skill of climate predictions on decadal time scales. The investigation builds on the coupled global model “Coupled GCM for Earth Simulator” (CFES) developed by Japan Agency for Marine-Earth Science and Technology (JAMSTEC). It is initialized in the ocean using a full state initialization strategy with two different initial fields: (1) oceanic initial conditions obtained through the same CFES coupled model using a 4-DVAR assimilation scheme (CDA: Coupled Data Assimilation) as provided by JAMSTEC; and (2) interpolated oceanic initial conditions obtained from the ocean-only GECCO2 Synthesis, while in the atmosphere initialized with bulk parameters controlling air-sea fluxes assimilated through CFES. Estimations on the performances of two differently initialized forecasts, i.e. CDA initialized hindcasts (CIH) and GECCO2 initialized hindcasts (GIH) through anomaly correlation coefficient (ACC) reveals that the forecast skill is better in the former than latter, with the most significant improvement of skill observed at the first lead year. For this specific model, improvements in SST predictive skill are especially obvious over the tropical Pacific, suggesting that the dynamical consistency of initial conditions can indeed improve the predictive skill of climate predictions at least in the first lead year. For longer lead times of 4-yr averages, a large decrease of SST predictive skill is observed almost everywhere for both CIH and GIH. This holds especially over the North Atlantic (NA) where previous studies indicate that predictive skill of SST is associated with the predictability of Atlantic Meridional Overturning Circulation (AMOC), with the latter leading the former several years. Our results indicate that significant predictive skill in terms of AMOC is obtained at the first lead year for both CIH and GIH. However, significant ...