Grassland and Cropland Net Ecosystem Production of the U.S. Great Plains

Abstract:

At observation sites throughout the world, carbon dioxide (CO₂) levels and other ecosystem resources are measured by instruments known as flux towers. Although flux towers only measure the surrounding vicinity or spatial footprint of their placement ecosystem, the data recorded at these towers can be up-scaled to much greater levels through the use of comprehensive remote sensing data and advanced computer modeling. The purpose of this study was to develop ecological net ecosystem production (NEP) models capable of producing weekly cropland and grassland NEP maps of the U.S. Great Plains at 250 meter resolution for 2000 - 2008. Separate NEP regression tree models were developed for each land cover type (cropland and grassland) with 15 flux towers supporting the grassland model and 13 towers supporting the cropland model. The NEP regression tree models were established through training based on data from the supporting flux towers, remote sensing data, and other biogeophysical inputs. Map results of this study indicate, as anticipated, grassland ecosystems generally perform as net carbon (C) sinks, absorbing and storing C from the atmosphere, and conversely, croplands generally as net C sources (crop yields were not taken into account), releasing C, in the form of CO₂, into the atmosphere. The models were evaluated by implementing a leave-one-out cross validation method, which withholds data form one particular year or site for testing a model developed with the remaining data. The cropland model validation analysis received an average Pearson’s correlation coefficient (r) of 0.85 for the yearly validation and an average r = 0.73 for the site withholding. The grassland model validation analysis received an average r = 0.86 for the yearly validation and an average r = 0.83 for the site withholding.