Ecosystem Carbon and Greenhouse Gas Modeling
The General Ensemble Biogeochemical Modeling System (GEMS) is one of the core modeling tools in the LandCarbon project, which is used to assess the effects of natural process and human management on ecosystem C/GHG fluxes at multiple spatial scales.
GEMS accounts for the impacts of atmospheric and climatic changes (e.g., changes in precipitation and temperature, CO2 enrichment, nitrogen deposition), land use change (e.g., urbanization, forest logging, fire fuel treatment, biofuel crops, tillage, reforestation), natural disturbances and extreme events (land fire, insect outbreak, drought, flooding, hurricane) on ecosystem C balances and GHG emissions. GEMS maximally uses highest quality data available from diverse sources, including climate (e.g., NCAR CCSM, CRUTS, CGCM), soil (e.g., STATSGO, SSURGO), vegetation (e.g., GLCC, NLCD, LANDFIRE vegetation maps), land cover/use change and disturbances (e.g., data from FORE-SCE and BLM, MTBS, LANDFIRE), remotely sensed and in-situ observations (e.g., MODIS, Landsat, FIA, CRP, NASWQ, NASS, Fluxnet).
GEMS is an evolving modeling framework that is designed to:
- Encapsulate well-established ecosystem models (e.g. CENTURY, EDCM, BIOME-BGC, IBIS, USPED), and setup the scopes and options of simulations
- Adopt a flexible approach dealing with spatial model simulation units, using wall-to-wall, pixel sampling, or a land patching approach (also called Joint Frequency Distribution (JFD) approach)
- Consider input data uncertainty using Monte Carlo approach, and reconcile multi-scale observations for ecosystem modeling (e.g. STATSGO soil data, FIA data, county- or state-level agricultural census data)
- Perform Monte Carlo ensemble simulations to incorporate the variability/uncertainty of driving variables and model parameters
- Perform data assimilation using ensemble Kalman Filter, MCMC, and PEST to optimize model parameters and/or estimates of ecosystem state variables
- Simulate the lateral movements of materials (e.g., water, erosion and sedimentation of soils and carbon)
- Aggregate and summarize model outputs with uncertainty following IPCC GHG Inventory Guidance
- Facilitate data sharing and visualization using the network common data form (NetCDF) interface.
One key property of GEMS is the ability to generate uncertainty estimates for all the output variables. Three sources of uncertainty can be quantified and transferred to the output via modeling:
- Input data uncertainty. In a nonlinear model, the average of the responses under various conditions is not the same as the response for the average conditions. Any difference between the model scale and the spatial resolution of the data may introduce biases caused by model nonlinearity. An ensemble approach has been implemented in GEMS to perform multiple Monte Carlo simulations to sample fine-scale heterogeneities in various databases in order to reduce potential biases.
- Model parameter uncertainty. Model parameter uncertainty information are derived from data assimilation using sequential and/or nonsequential approaches in GEMS. For example, we can derive model parameter uncertainty using MCMC or PEST from flux tower observations or long-term inventory plots, and then apply this information to regional GEMS simulations.
- Model structure uncertainty. GEMS has an interface (Any Model Interface or AMI) to facilitate setting up an automatic linkage between GEMS and any site-scale models. Once linked, GEMS takes control of these site-scale models, automatically parameterizes them according to biophysical conditions of any land parcel, and deploys them across space. Because GEMS can encapsulate multiple models, and parameterize and drive these models with the same data, it provides an ideal environment of platform to identify and address issues and uncertainty related to model structure and mathematical representations of biophysical processes.
Liu, S. Quantifying the spatial details of carbon sequestration potential and performance. Geographical Monograph Series, American Geophysical Union Book Chapter. (GEMS structure).
Tan, T., S. Liu, C.A. Johnston, T.R. Loveland, L.L. Tieszen, J. Liu, R. Kurtz. 2005. Soil organic carbon dynamics as related to land use history in the northwestern Great Plains. Global Biogeochemical Cycles (19) GB3011, doi:10.1029/2005GB002536. (GEMS applications in an agricultural area).
Zhao, S., S. Liu, Z. Li, T.L. Sohl. 2009. Ignoring detailed fast-changing dynamics of land use overestimates regional terrestrial carbon sequestration. Biogeosciences (6) 1647-1654. (GEMS applications in a forested area)