Submitted to: Remote Sensing of Environment
Publication Type: Peer reviewed journal
Publication Acceptance Date: 9/21/2006
Publication Date: 3/27/2007
Citation: Wylie, B.K., Fosnight, E.A., Gilmanov, T.G., Frank, A.B., Morgan, J.A., Haferkamp, M.R., Myers, T.P. 2007. Adaptive data-driven models for estimating carbon fluxes in the northern great plains. Remote Sensing of Environment 106:399-413. Interpretive Summary: With greenhouse gases increasing in Earth’s atmosphere and the potential for global climate change, there is an urgent need to understand the sources and sinks of these greenhouse gases if we are intelligently manage our activities to curtail their emissions into the atmosphere. Carbon dioxide (CO2) is one of the most important of those greenhouse gases, and agriculture plays an important role in the land-atmosphere exchange of this gas since CO2 is assimilated in plants through photosynthesis, and as is emitted back into the atmosphere through soil microbial, plant and animal respiration. Although the exchange rates of CO2 by arid and semi-arid rangelands in the western United States are relatively low compared to more productive ecosystems and croplands, the extensive areas which rangelands occupy suggest they have the potential to be an important sink and/or source of CO2. This experiment takes land measured CO2 fluxes of different rangelands in the Northern Great Plains, and through modeling and mapping, scales those fluxes across a vast region of the Northern Great Plains. The results suggest that the Northern Great Plains of the United States is near equilibrium for CO2 fluxes, which means that as a region, neither a net loss of gain of carbon is currently occurring. Thus, while this region does not appear to be sequestering C, neither does it represent a problem in terms of being a source of the greenhouse gas CO2.
Technical Abstract: Rangeland carbon fluxes are highly variable in both space and time. Given the expansive areas of rangelands and their soil organic matter stocks, how rangelands respond to climatic variation, management and soil potential is important to understanding carbon dynamics. Rangeland carbon fluxes associated with Net Ecosystem Exchange (NEE) were measured from multiple year data sets at five flux tower locations in the Northern Great Plains. These flux tower measurements were combined with 1-km2 spatial data sets of Photosynthetically Active Radiation (PAR), Normalized Difference Vegetation Index (NDVI), temperature, precipitation, seasonal metrics, and soil characteristics. Flux tower measurements are used to train and select variables for a rule-based piece-wise regression model. The accuracy and stability of the model was assessed through random cross-validation and cross-validation by site and year. Estimates of NEE were produced for each ten-day period during each growing season from 1998 to 2001. Growing season carbon flux estimates were combined with winter fluxes estimates to derive and map annual carbon budgets. The rule-based piece-wise regression model is a dynamic adaptive model that captures the relationships of the spatial data to NEE as conditions evolve throughout the growing season. The carbon dynamics in the Northern Great Plains proved to be in near equilibrium, serving as a small carbon source in three years and as a small carbon sink in 1999. Patterns of carbon sinks and sources are very complex with the carbon dynamics tilting toward sources in the drier west and toward sinks in the east and near the mountains in the extreme west. Significant local variability exists which initial investigations suggest are likely related to soil properties and management.