12 January 2000

Summary Report for Climate Simulation Laboratory Project

Climate Change Simulations

Warren M. Washington - Principal Investigator

Ocean Component Development

Albert Semtner (Naval Postgraduate School, NPS), Anthony Craig, Warren Washington, John Weatherly (U.S. Army Cold Regions Research Engineering Laboratory, CRREL), Gerald Meehl, and Frank Bryan, continued to evaluate, improve, and optimize a modified version of the Parallel Ocean Program (POP). The Parallel Climate Model (PCM) uses a 2/3 degree (on average) displaced pole, global configuration of  POP, while the Climate System Model (CSM) is developing a 2 degree version with a different numerical pole point. During the past year there was an effort to have closer collaboration between the CSM and PCM efforts. An attempt is being made to introduce the Gent-McWilliams diffusion scheme and K Profile Parameterization (KPP) vertical upper ocean scheme. This latter activity is being done in collaboration with scientists at Los Alamos National Laboratory. The PCM ocean model includes increased latitudinal resolution near the equator to resolve the strong tropical current systems. This version of POP was integrated for over a century with acceleration at the bottom of the ocean of a factor of 10 to bring the ocean simulation closer to equilibrium. The ocean model component of PCM continues to be in production mode on the Silicon Graphics Inc. (SGI) Origin 2000 at NCAR’s Climate Simulation Laboratory (CSL) and on the SGI machines at Los Alamos National Laboratory. Recently, the PCM components have started production on the NCAR IBM SP2. The ocean simulations revealed remarkable smaller scale structure that compared quite favorably with higher resolution versions and limited observations. Also, the narrow flows along the coastal regions were in good agreement with observations. The meridional overturning in the North Atlantic and worldwide conveyor belt circulation were well represented in the simulations. This version of POP is an integral component of the PCM, a coupled model developed as part of a multi-institutional, distributed research effort for use on current and future generations of Massively Parallel Processors (MPP). A discussion of the ocean model simulation in the coupled model can be found in papers by Semtner (2000) and Washington et al. (2000) to be published in Climate Dynamics.

Examples of the ocean simulation of the model can be found on the Parallel Climate Model web page at www.cgd.ucar.edu/pcm/.

Coupled Model Simulation

The PCM makes use of the NCAR Community Climate Model (CCM3) and Land Surface Model (LSM) for the atmospheric and land surface components, respectively, the Los Alamos National Laboratory POP, and the NPS sea-ice model. The PCM executes on several distributed and shared memory computer systems. The coupling method is similar to that used in the CSM in that a flux coupler ties the components together, with interpolations between the different grids of the component models. Flux adjustments or corrections are not used in the PCM. The ocean component has 2/3° average horizontal grid spacing with 32 vertical levels and a free surface that allows calculation of sea level changes. Near the equator, the grid spacing is approximately 1/2° in latitude to better capture the ocean equatorial dynamics. The North Pole is rotated over northern North America, thus, producing resolution smaller than 2/3 degree in the North Atlantic where the sinking part of the world conveyor circulation largely takes place. Because this ocean model component does not have a computational point at the North Pole, the Arctic Ocean circulation systems are more realistic and similar to the observed. The elastic viscous plastic sea-ice model has a grid spacing of 27 km to represent small-scale features, such as ice transport through the Canadian Archipelago and the East Greenland current region.

Results from a 300-year present-day coupled climate control simulation by Washington, Weatherly, Meehl, Semtner, Thomas Bettge, Craig, Gary Strand, Julie Arblaster, Wayland, Rodney James (SCD) and Yuxia Zhang (NPS) showed that the PCM gave a very stable simulation with approximately the observed interannual and decadal variability.  Five transient 1% per year CO2 increase experiments have been carried out that showed a global warming of about 1.26°C for a 10 year average at the doubling point of CO2.One of the experiments was allowed to go to the quadrupling point and the global average warming was 2.89°C. There was a gradual warming beyond the doubling and quadrupling points. A 0.5% per year CO2 increase experiment also was performed showing a global warming of 1.49°C and a similar geographic warming pattern to the 1% per year doubling experiment. Globally averaged sea level rise at the time of CO2 doubling was approximately 7cm and at the time of quadrupling it was 23 cm. The regional sea level changes were much larger and reflect the adjustments in the temperature, salinity, internal ocean dynamics, surface heat flux, and wind stress on the ocean. El Niņo and La Niņa events in the tropical Pacific Ocean show approximately the observed frequency distribution and amplitude, which leads to realistic variability on interannual timescales of tropical and extratropical planetary wave patterns. Washington et al (2000) and Weatherly and Zhang (2000) show the early results from these simulations in two articles accepted for publication and a paper by Meehl et al., (2000) is in preparation and shows more details on the factors that affect El Niņo amplitude.

Presently, ensemble historical and future climate model simulations are being conducted.  The simulations make use of the same forcing as the CSM (the business as usual (BAU) and policy limited ACACIA scenarios). They simulate from 1870 to year 2100.  Additional simulations with the added effect of solar forcing on the climate system are also being conducted.

The following are scientists and programming staff involved in the PCM effort or its components in alphabetical order: J. Arblaster (NCAR), T. Bettge (NCAR), A. Craig (NCAR), J. Dennis (NCAR), J. Dukowicz (LANL), J. Hack (NCAR), S. Hammond (NCAR), E. Hunke (LANL), R. James (NCAR), P. Jones (LANL), R. Loft (NCAR), R. Malone (LANL), M. Maltrud (LANL), W. Maslowski (NPS), G. Meehl (NCAR), A. Middleton (NCAR) A. Semtner (NPS), R. Smith (LANL), G. Strand (NCAR), W. Washington (NCAR), J. Weatherly (CRREL), V. Wayland (NCAR), D. Williamson (NCAR), and Y. Zhang (NPS).

CSL Usage from June 1998 to the Present

The following is the breakdown by case configuration (type of simulation), years of simulation, and which CSL computer was used.

Because of different computational time required for G configurations (ocean and sea ice only), E case (atmosphere – CCM3 only) and B configurations (full-coupled version of the PCM) configurations, it is difficult to obtain total wallclock-pe time involved.

We have been substantial users of the IBM and we have used our CSL allocations on the Origin. It should be noted that over the last few months we have moved our work off of the Origin on to the IBM; thus, making time available to other CSL users who do not have their codes converted to the IBM.

Configuration Years Simulated Machine Comments
G04 197 Origin Active ocean and sea ice only
G05 115 Origin Active ocean and sea ice only
E04 32 Origin Active atmosphere only
E05 109 Origin Active atmosphere only
B04 399 Origin PCM Original Control simulation
B05 777 Origin PCM (B05) Control and Historical Simulations
B06 702 IBM PCM (B06) Control and Historical Simulations

Total number of years simulated is 2331.

Machines: Origin: 1629     IBM: 702

In terms of:        

Ocean Equilibrium: 453
Fully Coupled Control: 985
Fully Coupled Historical: 380
Fully Coupled Future BAU: 100
Fully Coupled Stabilization: 100
Fully Coupled Historical + Solar: 260
Residual (testing, etc.):   53

Note: This is not a total of wall clock hours or processor hours, it is the simulation years.

CSL Machine Performance Using PCM

The following table shows the number of simulated years per wallclock day achieved by the PCM on near-dedicated CSL machines. Scalability is shown down processor columns. For comparison, the IBM SP WinterHawk2 performance, benchmarked at Oak Ridge National Laboratory, and the performance achieved on the NCAR Compaq ES40 system is included.

pes     SGI Origin IBM SP (NCAR WH1) IBM SP (ORNL WH2) Compaq ES40
8 0.8 1.1 1.4
16 1.7 1.9 2.7
32 3.2 3.2 3.5
64 5.1 4.8 6.3

Estimates of Computer Time Used at Other Centers

We estimate that we have conducted about 2500 years of simulation at LANL and NERSC over the last 18 months.

Data Availability to the Community

We have made the data available to the community. For example, the simulations of climate change will be a part of the ongoing IPCC assessment as well as the U. S. National Assessment. We have also provided data to the Scripps Oceanographic Institution, University of Washington, GFDL, etc. Arrangements have been made to make all of the atmospheric, ocean and sea ice data available to any scientist through the DOE’s PCMDI capability. The data will be on a large file server system at PCMDI (Lawrence Livermore National Laboratory) after the data is transferred from various computational sites.


Meehl, G.A., P. Gent, J.M. Arblaster, B. Otto-Bliesner, E. Brady, and A. Craig, 2000: Factors that affect amplitude of El Niņo in global coupled climate models. In preparation for Climate Dynamics.

Semtner A.J., 2000: Ocean and Climate Modeling on Advanced Parallel Computers: Progress and Prospects. Communications of the ACM (in press).

Washington, W. M. et al., 2000: Parallel Climate Model (PCM): Control and Transient Simulations. Accepted to Climate Dynamics.

Weatherly, J.W. and Y. Zhang, 2000: The response of the polar climate to increasing CO2 in a global climate model with elastic-viscous-plastic sea ice. Accepted to Journal of Climate.