Stone SouperComputer Applications

Global Terrestrial Ecosystem Carbon Model

Wilfred M. Post, Anthony W. King, Stan D. Wullschleger, and M. Lynn Tharp

We develop terrestrial biogeochemistry models for analysis of global change impacts on terrestrial systems and feedbacks to the climate system. In the future it will be necessary to study terrestrial carbon processes using models that are coupled to global and regional atmospheric models. The terrestrial process components require the same order of magnitude computational demands as the atmospheric components of such a coupled model. Experience now with large scale parallel computing will be necessary to work with the next generation of integrated models.

Current ESD/ORNL Terrestrial Codes:

LoTEC - kernel for regional and global models. LoTEC is a mechanistic soil-plant-atmosphere model of ecosystem carbon storage and CO₂ and H₂O flux. Canopy photosynthesis is described by a "bigleaf" implementation of either a C3 or C4 biochemical model of photosynthesis combined with a sub-model of stomatal conductance. Maintenance respiration for four plant compartments is a function of tissue nitrogen concentration and temperature, while growth respiration is proportional to the change in compartment size. Canopy photosynthesis and maintenance respiration are calculated hourly; carbon allocation, growth, and growth respiration are calculated daily. Canopy dynamics and phenology are specified by model inputs of change in leaf area index derived from AVHRR/NDVI data. Environmental inputs are solar radiation, air temperature, and precipitation. Precipitation is input to a dynamic model of soil water, and aspects of modeled plant physiology are influenced by soil water potential.

Litter produced in the turnover of vegetation compartments is passed to the decomposable and resistant plant material compartments of the soil module. Decomposition of litter and soil organic carbon and the associated release of CO₂ in heterotrophic respiration are modeled with a daily time step implementation of the Rothamsted soil carbon turnover model. Incoming organic matter passes through the decomposable and resistant plant material compartments once. Both decomposable and resistant material decompose to CO₂ (which is lost to the atmosphere), microbial biomass, and humified organic matter. When microbial biomass and humus decomposes, CO₂ is released and microbial biomass and humus are formed. The microbial biomass and humus are subject to further decomposition. The soil is also assumed to contain a small amount of organic matter that is inert to biological attack. The characteristic decay rates of each soil compartment are modified by environmental factors of temperature and soil moisture deficit. The difference between net primary production (NPP) and the CO₂-carbon released during decomposition is net ecosystem production (NEP) and represents any change in total ecosystem carbon storage and net carbon exchange with the atmosphere.

Figure 1. Historical and future calculations of carbon in vegetation and soil for the Continental U.S. from the ESD/ORNL carbon cycle model. Calculations performed using the ESD StoneSouper Computer and the IBM SP computer.

GTEC - global model contains 21,600 1° terrestrial cells. The carbon dynamics of each vegetated land cell (1.0° latitude × 1.0° longitude resolution) is described by a mechanistic soil-plant-atmosphere model (LoTEC) of ecosystem carbon storage and CO₂ and H₂O flux. Each grid cell is assigned to one of 15 ecosystem types and one of 105 soil types.

vGTEC - continental U.S. model contains 3160 terrestrial 0.5° cells. The carbon dynamics of each vegetated land cell is described by a mechanistic soil-plant-atmosphere model (LoTEC) of ecosystem carbon storage and CO₂ and H₂O flux. Vegetation, soil and climate data are provided by the Vegetation/Ecosystem Modeling and Analysis Project (VEMAP) to facilitate comparison between terrestrial carbon cycle model results. See The VEMAP Project for additional details.

A complete run of these models generally requires 3 phases of simulation:

• Spinup for initial conditions (>2,000 simulation years)
  (The spinup phase does not have to be repeated for every production run except when initial and/or boundary conditions are changed. Even then, it may not take the whole 2,000 years for the long-term carbon pools to equilibrate if we use conditions saved from a previous run that is similar to the new starting conditions.)
• Historical time-dependent (100 simulation years, 1895 to 1993)
• Future scenarios (30-100 simulation years, 1994 onward, provided by atmospheric general circulation model (AGCM) output)

This model is run on both the Stone SouperComputer and the IBM SP. Using 32 nodes of the IBM SP, the worst case spinup period is about 3 days for a GTEC run with actual time-dependent portions completing in 10 hours.

Future Directions:

Land-surface model. We plan to develop GTEC as a land-surface model component for coupling with Atmospheric AGCMs and mesoscale weather models like MM5. In this way, carbon cycles and hydrological cycles are linked and coupled to the atmosphere at the most appropriate scale.

Atmospheric chemistry. Effects of air pollution and influence of biogenic emissions depend on physiological processes of photosynthesis and evapotranspiration. Atmospheric chemistry feedbacks involving tropospheric ozone add another dimension of processes that depend on stomatal conductance for both emissions of VOC's and impacts of ozone.

Return to the Stone SouperComputer Homepage.