Modeling the Future

Supercomputers produce breakthroughs in climate research.

Heat waves and heavy rainfall will be more frequent. Sea ice near Earth's poles will shrink. Tropical hurricanes will become more intense, with higher wind speeds and heavier precipitation.

Earth's surface will continue to warm, increasing on average 1- 3°C globally, 7-10° in the polar regions and virtually 0°C near the equator. The oceans will continue to rise slowly over the next millennium as a result of past and projected carbon dioxide emissions from human activities, such as destruction of forests and fossil fuel combustion in power plants and transportation vehicles. Each CO₂ molecule generated will survive for nearly 100 years in the atmosphere before being permanently removed by terrestrial photosynthesis or diffusion in the ocean.

In 2004-05 the Department of Energy's Center for Computational Sciences at Oak Ridge National Laboratory provided an important contribution to a suite of climate modeling simulations from 15 nations, helping develop predictions in response to questions posed in an international investigation of Earth's climate. In February 2007, the first report of the fourth assessment of the United Nations' Intergovernmental Panel on Climate Change was released in Paris, France. The report, which received wide coverage by the world's news media, concluded that evidence of a global warming trend is "unequivocal" and that increased human emissions of greenhouse gases to the atmosphere has "very likely" been the driving force in that change over the past 50 years.

"What must be done to respond to a dangerous buildup of CO₂ in the atmosphere can be viewed as drastic," remarks John Drake, who leads climate modeling at ORNL, site of the world's most powerful open-science supercomputer. "Humankind must start reducing CO₂ production, returning to almost preindustrial emission levels of 1870."

In both scientific and societal terms, the challenge is extraordinary. The atmospheric concentration of CO₂ has risen exponentially from 280 parts per million in 1870 to 385 ppm today. Some models predict atmospheric CO₂ doubling by 2100. Even if humankind emitted no more CO₂ after 2100, Earth's temperature would still rise by 2°C, making a warmer ocean so acidic from dissolved CO₂ that coral reefs, which shelter fish and protect shores against beach erosion, could vanish.

Cheetah's calculations

Many of the Intergovernmental Panel's projections concerning present and future climate change and the roles of human activity resulted from runs in 2004-2005 of climate models on ORNL's "Cheetah" supercomputer. The simulations supplied calculations for more than 200 papers, including several from ORNL, for the first report of the fourth IPCC assessment. Cheetah is an IBM Power4 supercomputer with a peak capacity of 6.4 tera-flops, or 6.4 trillion calculations per second.

According to the 2007 report, projections of computer models closely match observed temperature increases of "about 0.2°C (0.36°F) per decade, strengthening confidence in near-term projections." Observed trends indicating climatic warming include more moisture in the air, less ice and snow on land, less Arctic sea ice and rising sea level.

Two U.S. organizations supporting the Intergovernmental Panel on Climate Change developed models for running energy scenarios that assumed both high and low greenhouse gas emissions. The goal was to predict changes in temperature and precipitation up to the year 2100. One group was a National Oceanic and Atmospheric Administration center in Princeton, N.J. The other organization, which interacted with university, ORNL and other national lab researchers, was the National Center for Atmospheric Research in Boulder, Colo.

With funding from the Department of Energy, ORNL researchers led by John Drake and David Erickson dedicated fully one-third of Cheetah's processing time to climate modeling in 2004-2005. Over the same period, NCAR's runs, funded by the National Science Foundation and DOE, consumed an additional one-third of the supercomputer's processing time.

"Every nation ran the same emissions scenarios as the U.S. using their own models," says Drake. "In that way we addressed a range of uncertainties." ORNL's primary responsibilities included developing computer codes and fine-tuning the U.S. models to make sure they could run the agreed-upon scenarios on Cheetah. The project produced 100 terabytes of data that were posted and stored on DOE's Earth System Grid. Some data are archived at ORNL's Leadership Computing Facility, a node on this grid.

Code contributions

From 2002 to 2004 Drake and his ORNL team contributed to the development of the Community Climate System Model used by researchers worldwide in support of the Intergovernmental Panel on Climate Change. University researchers can obtain the model and participate in its development and use for simulations.

Climate models represent the complex systems and interactions of the atmosphere, oceans, land and ice fed by solar energy. The physics parts of the models predict Earth's surface temperature based on the degrees to which solar radiation warms Earth's surface, provides heat absorbed by the oceans, reflects off land surfaces, ice sheets and atmospheric particulates, and warms the atmosphere. This warming occurs as carbon dioxide and other greenhouse gases in the atmosphere trap solar infrared radiation that Earth returns to the air. The accuracy of the models is tested by seeding them with actual greenhouse gas data from the past 30 years and determining how closely the predicted results match recorded temperatures.

Climate models also simulate dynamics, particularly atmospheric and ocean circulation that drives changes in temperature and precipitation. ORNL has focused on atmospheric circulation and land interactions. DOE's Los Alamos National Laboratory has modeled ocean-air interactions. General circulation models calculate the three-dimensional distribution of winds and currents that move CO₂ and other chemicals through the climate system.

For the Community Climate System Model, Oak Ridge researchers wrote algorithms, or mathematical methods, for solving flow equations, an area of ORNL expertise. ORNL researchers developed flow equations for simulating winds blowing in the atmosphere and the River Transport Code, a river routing scheme that calculates the freshwater balance.

"Ocean circulation depends on salinity as much as on temperature and wind," Drake says. "Salt influences water density, which affects water flow rate. In a coupled land-ocean model, this code collects rain falling in the land model and routes it through the simulated river system to the right places in the ocean. By determining freshwater input into the ocean, the model can accurately calculate the relative proportion of salt in the ocean. Developing a sound hydrology prediction in the model was a major contribution."

Heating and cooling

ORNL's David Erickson asked an important question: "Could climate change actually accelerate if people use more energy to stay cool as temperatures rise?"

Erickson and ORNL engineer Stan Hadley were the first to integrate energy-economic and climate models to evaluate the influence of climatic warming on energy use. They plugged a standard climate model's temperature predictions for 2000 to 2025 into the National Energy Modeling System developed by DOE. The researchers published their results in the August 2006 issue of Geophysical Research Letters.

The modeling system calculates energy requirements for heating and cooling all buildings in every county of the United States. The program takes into account each locale's climate, building styles and fuel sources for heat and electricity. The results range from predictable to surprising.

Predictably, the study projected that CO₂ emissions would rise as more coal is burned to produce additional electricity required for air conditioning during ever-warmer summers in the South and West. However, despite climatic warming in the Northeast, the winters would not be warm enough to reduce significantly residents' need for heat from natural gas and heating oil. The net result: over the next 25 years, increases in carbon emissions from higher air conditioning demands in the South and West will more than offset decreases in carbon emissions from reduced heating needs in the Northeast. Thus, CO₂ emissions will edge up even more in three regions of the United States.

Erickson is interested in improving detailed climate models by including another aspect of heating and cooling. He has written algorithms for simulating the heating and cooling effects of aerosols, particles formed from collections of atoms and molecules that compose clouds and float in the air.

His model will simulate interactions between different wave-lengths of solar radiation—from infrared to visible to ultraviolet light—and a variety of atmospheric aerosols. Some sulfate aerosols from coal-fueled power plants reflect solar radiation in most regions of the spectrum, cooling the atmosphere locally. Other aerosols absorb certain radiation wavelengths, exhibiting a warming effect. Erickson takes into account a variety of aerosols, ranging from sea salts and African desert dust blowing across the ocean to air pollution sulfates and haze-producing organic isoprene shed by tree leaves that gives the Great Smoky Mountains its name.

Earth systems models

Earth System Models

ORNL climate researchers are developing "earth systems models" that simulate a fully interactive carbon cycle. By merging biogeochemical models with a climate model, researchers are developing a model with several layers of complexity. The Earth Systems Model will represent not only the physics of solar energy and the atmosphere and the dynamics of air and ocean circulation but also biogeochemical cycles and socioeconomic data, such as population density and land-use statistics.

The climate model will start in 1870 and include CO₂ emissions and land-use changes up until the present. The sophisticated model will compute, from preindustrial to the present, CO₂ emissions from forest fires and decomposing fallen trees, as well as the amount of CO₂ absorbed for photosynthesis every 20 minutes as different plants throughout the world respond to changing emission, temperature and moisture levels. The goal of the project is to determine whether the model will compute Earth's current atmospheric CO₂ level of 385 ppm based on the incorporated carbon cycle data and algorithms. To accurately predict the current atmospheric CO₂ level will be a challenge that will require collecting all the needed data and measuring the carbon pools with precision.

ORNL researchers are collecting data from all over the world about the types of plants that have grown in various regions between 1870 and the present. They also are seeding the model with flux tower data on air-surface exchanges of CO₂, water vapor and energy (see Tower of Knowledge article). They will calibrate earth systems models for 15 different ecosystems, types and rates of inputs and environmental conditions and then make global-scale calculations on the CO₂ exchanges between the terrestrial ecosystems and the atmosphere.

Coupled climate and carbon cycle models run on more powerful supercomputers will likely be used to predict global temperature changes between 1870 and 2100. The predictions will be based on greenhouse gas emissions and absorption through photosynthesis, respiration and historical and projected releases of CO₂ from fossil fuel combustion and deforestation.

Merged models using data from ORNL sources will be used to predict how much a warmer climate and an atmosphere with elevated levels of CO₂ will raise rates of photosynthesis and soil decomposition, which in turn will boost atmospheric CO₂ levels even more.

Enhanced photosynthesis by increased atmospheric CO₂ concentrations is limited by the amount of nitrogen available in the soil to form the enzyme that drives photosynthesis. ORNL's climate team seeks to model a tree's response that may be hampered by reduced nitrogen availability except for regions receiving anthropogenic deposits from transportation vehicle exhaust and fossil fuel power plant emissions.

Their effort includes computer models that globally simulate leaf fall and the death of forest trees and grasses, their decomposition by bacteria and fungi in the soil, physical protection of the soil carbon and the eventual release of the soil carbon back to the atmosphere. The polar regions have three times as much soil carbon as the Southeast. With warming, models will predict a shift in the carbon balance as polar soils decompose faster, release more carbon and host many new plants that take in atmospheric CO₂ for photosynthesis.

For the last runs for the Intergovernmental Panel on Climate Change, the modelers had adequate data to predict heat waves for various regions. What the massive computers lack is sufficient resolution to predict accurately the frequency of hurricanes over many years. Hurricanes contribute a significant amount of rainfall to the Southeast. Projecting their frequency correctly over a century would help modelers predict precipitation with greater accuracy in the Southeast so that simulated green plants would grow in the right places.

In one modeling run, clearing of rainforests along the Amazon River reduced transpiration, resulting in fewer clouds and less rainfall. The data generated fear that Amazonian rainforests could become so warm and dry by mid-century they will essentially collapse. As a result, through respiration the dying forest plants would accelerate releases of CO₂ to the atmosphere.

New questions

As more data become available, the questions in the climate discussion are shifting. Initially, scientists sought to understand what might happen to the climate in the next 100 to 500 years. With increasing frequency, the same scientists are attempting to answer how, over the next 20 to 50 years, society will be forced to adapt to the environmental impact of a continued buildup of CO₂ in the atmosphere.

ORNL's John Drake says, "People are now seeing the effects on local scales. We may not see the effects in the Southeast in our lifetimes. But in higher latitudes such as Alaska, Canada and Greenland, people are seeing effects now, such as more rapid melting of ice sheets."

Drake anticipates that ORNL's climate modelers will continue to be asked for additional projections on the effects on climate of different levels of greenhouse gas emissions from various energy scenarios. Policymakers seeking evaluations of different mitigation options may request scenarios that include replacing fossil fuel power plants with nuclear power plants, using carbon capture technologies on coal-fueled power plants and growing vast tracts of switchgrass for producing ethanol to replace gasoline. As the debate continues, ORNL researchers will be providing scientifically grounded predictions about one of the most important issues of our time.—Carolyn Krause

Contact: John B. Drake

Web site provided by Oak Ridge National Laboratory's Communications and External Relations
ORNL is a multi-program research and development facility managed by UT-Battelle for the US Department of Energy
[ORNL Home] [SNS Home] [CAER Home] [Privacy and Security Disclaimer]