Predicting carbon sinks, regionally
A team scientists have developed a new method for estimating soil organic carbon sinks across large regions of varying landscapes. The technique builds on efforts to better understand the global carbon cycle and identify potential carbon sink sources. Mitigating the risks of climate change may depend on removing atmospheric carbon dioxide through photosynthesis in plants, and then storing it in soils. Areas of soil capable of absorbing carbon are often referred to as carbon sinks or pools.
The prediction accuracy of this new approach was compared with the two other commonly used methods. The predictions of the new method, called geographically weighted regression, had higher contrast and wider variability of soil organic carbon pools with lower global prediction errors in comparison to other approaches.
Land use has a significant effect on whether soils are a potential source or sink of carbon. The study showed the largest pools of carbon are:
- Forests (46%)
- Cropland (37%)
- Developed wetlands (11%)
- Barren land (0.3%)
Croplands show the highest variability in soil organic carbon pools, likely due to the variable land management practices adopted by farmers in the region, according to the authors.
The research was conducted by scientists from the Ohio State University and Ghent University in Belgium, led by Dr. Umakant Mishra, now at University of California-Berkeley. The new approach used spatially adjusted geographically weighted regression to predict the soil organic pool in seven states in the Midwestern United States: Ohio,Kentucky, Michigan, Indiana, Pennsylvania, West Virginia, and Maryland. Results from this study were reported in the May-June 2010 issue of the Soil Science Society of America Journal, published by the Soil Science Society of America.
The authors used the range of spatial autocorrelation in soil organic carbon observations to define a search radius in geographically weighted regression. This method takes into account the geographic distance of each data point from the regression point. By incorporating environmental variables such as terrain attributes, climate data, land use data, bedrock geology, and normalized difference vegetation index, the method provided a predicted resolution of 30 meters. Greater localized resolution will likely make this method useful to land managers.
The common view of scientists in the discipline is that a single model applicable to different soil landscapes in regional-scale studies was unlikely to be developed. This research shows a relatively simple methodology to predict the soil organic carbon pool at regional scales that land managers can readily use. This new method can play a vital role in improving the prediction ability of soil organic carbon pools on a regional scale.