GFDL - Geophysical Fluid Dynamics Laboratory



The CM2.5 model (Delworth et al., 2012) is a descendant of the GFDL CM2.1 model (Delworth et al., 2006) that incorporates higher spatial resolution and a significantly improved land model (LM3). As a result of these enhancements, the CM2.5 model has a significantly improved simulation of many aspects of climate, particularly hydroclimate over continental regions (Delworth et al., 2012, Figures 5,6,7 and 9) and aspects of ocean circulation. This improvement has allowed GFDL scientists and their collaborators to use this model for innovative studies of regional hydroclimate change (Doi et al, 2012,2013; Kapnick and Delworth, 2013; Delworth and Zeng, 2014; Kapnick et al., 2014) and ocean circulation (Lee et al, 2013). The model has also proven very effective at simulating climate extremes, such as tropical cyclones (Kim et al., 2014) and drought (Delworth et al., 2015). A similar horizontal spatial resolution is being targeted for GFDL’s next-generation model, CM4.

Since the high-resolution ocean model is relatively expensive to run, a companion model (see description on this web site for CM2.5_FLOR) was developed using the same atmospheric component as CM2.5, but with a lower resolution (and therefore much faster) ocean model. This has proven very effective for seasonal to interannual prediction, especially given the need for very large sets of simulations to assess the skill of a prediction system.

Model details

The atmospheric component of CM2.5 has similar physics as CM2.1, but uses grid box cells that are 50 Km on a side, versus approximately 200 Km in CM2.1. The atmospheric component also increases the number of vertical levels from 24 to 32. The ocean component has horizontal resolution of approximately 0.25o (versus 1o in CM2.1), with grid box sizes ranging from 28 Km at the Equator to 8 Km in polar regions. The model uses increasing levels of parallelism to run efficiently on modern supercomputers. For example, the model was able to simulate 12 model years per day when using approximately 6000 processors on the NOAA Research Supercomputer (GAEA) located at the Oak Ridge National Laboratory.

CM2.5 code released March 2015

The code for this model is publicly available.


The GFDL Forecast-oriented Low Ocean Resolution version of CM2.5 (CM2.5-FLOR, or FLOR) model (Vecchi et al. 2014) is a descendent of the CM2.5 model (Delworth et al., 2012) and CM2.1 model (Delworth et al., 2006). The FLOR model incorporates the higher horizontal resolution in the atmosphere and land, higher vertical resolution in the atmosphere, and significantly improved land model (LM3; Milly et al. 2014) from CM2.5. The FLOR model also uses the relatively low resolution ocean and sea ice components of CM2.1. These choices create a coupled model that is relatively computationally efficient, but can be used to address problems of regional climate and extremes.

The enhanced resolution in the CM2.5 model has a significantly improved simulation of many aspects of climate, particularly hydroclimate over continental regions (Delworth et al., 2012, Figures 5,6,7 and 9); many of the improvements in simulation of near-surface climate in CM2.5 are recovered in FLOR (e.g., Jia et al. 2015). The FLOR model has been used extensively to understand predictability, change and mechanisms of tropical cyclones (Vecchi et al. 2014), Arctic sea ice (Msadek et al. 2014), precipitation and temperature over land (Jia et al. 2015), drought (Delworth et al., 2015), extratropical storms (Yang et al. 2015), the Great Plains Low Level Jet (Krishnamurthy et al. 2015), and the global response to increasing greenhouse gases (Winton et al. 2014).

FLOR is used in real time seasonal predictions, contributing to the North American Multi-Model Ensemble for seasonal prediction (NMME; Kirtman et al. 2014). Retrospective prediction output from FLOR is available from the GFDL data server:

Model details

The atmospheric component of CM2.5-FLOR has similar physics as CM2.1, but uses a cubed-sphere dynamical core (Putnam and Lin 2007) with grid box cells that are 50 Km on a side, versus approximately 200 Km in CM2.1. The atmospheric component also increases the number of vertical levels from 24 to 32. The ocean component has the same horizontal resolution of CM2.1, which is approximately 1(with meridional resolution of 1/3o near the Equator). The model uses increasing levels of parallelism to run efficiently on modern supercomputers. For example, the model was able to simulate 18(12) model years per day when using approximately 4000(2600) processors on the NOAA Research Supercomputer (GAEA) located at the Oak Ridge National Laboratory.

FLOR code released March 2015

The code for this model is publicly available.