On the Development of GFDL's decadal prediction system: initialization approaches and retrospective prediction assessment
Abstract
Initializing climate models for decadal prediction is a major challenge, in part due to the lack of long-term subsurface ocean observations and the changing nature of observing systems. In order to overcome these limitations, we have developed two approaches for initializing GFDL's SPEAR decadal prediction system that do not use subsurface ocean observations. In the first initialization approach (called I1), the atmospheric winds and temperature were restored towards atmospheric data from the JRA reanalysis, and sea surface temperature (SST) was restored towards the time-varying observations. In the second approach (called I2), the observed surface pressure data was assimilated in the atmospheric component of the coupled SPEAR model using the ensemble data assimilation technique, while the SST was restored towards the same time-varying observations as in I1. Physically, by constraining the air-sea interface in the coupled model using observed SST, air temperature and winds, the ocean component of the coupled model is forced by a similar sequence of surface heat fluxes and wind stresses as observations, thus reproducing similar oceanic decadal variations as in the observed system.
Two coupled reanalyses have been completed over the period 1960 to 2019 using the I1 and I2 techniques. Two sets of retrospective decadal forecasts were then conducted using initial conditions from the I1 and I2 reanalyses. The two reanalyses show very similar multi-decadal variations of Atlantic meridional overturning circulation (AMOC), e.g., a weakened state during the 1960s and 1970s, a strengthened state in the late 1980s and 1990s, and a weakened state after 2000. In comparison with the uninitialized large ensemble simulations, the SPEAR decadal predictions using both initialization techniques show skillful decadal prediction of surface and subsurface temperature over the North Atlantic subpolar gyre and the Southern Ocean. The decadal prediction skill in the SPEAR decadal prediction system is comparable to that in the GFDL's previous system, while the model bias is substantially improved in the North Atlantic and the Southern Ocean. These results suggest that surface pressure observations over the last century may provide sufficient atmospheric observational information for the initialization of decadal predictions.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMA110...12Y
- Keywords:
-
- 3305 Climate change and variability;
- ATMOSPHERIC PROCESSES;
- 3322 Land/atmosphere interactions;
- ATMOSPHERIC PROCESSES;
- 3337 Global climate models;
- ATMOSPHERIC PROCESSES;
- 3373 Tropical dynamics;
- ATMOSPHERIC PROCESSES