Date of this Version
Agriculture, Ecosystems & Environment 161:17–26; doi: 10.1016/j.agee.2012.07.014
With the increasing biofuel demands in recent years, the cultivated lands for biofuel crops, i.e., corn and soybeans, the major sources of ethanol and biodiesel, have been greatly expanding in the northern Great Plains states of the United States. Simulating the spatio-temporal dynamics of biofuel croplands can provide critical information required for assessing the impacts of land-use change on wildlife conservation and water quality. But, yearly agricultural practices such as crop rotations often complicate the spatially explicit modeling of specific crops’ expansion. Our research focused on developing a geospatial modeling framework that is able to distinguish long-term, regional changes in croplands from short-term, local fluctuations (such as rotations), using geographic information systems (GIS) and the land transformation model (LTM). The USDA Cropland Data Layers (CDLs) of North Dakota for 1999, 2000, 2004, 2005, 2010, and 2011 were spatially and temporally aggregated to generate a series of biofuel cropland maps. The historical cropland data for 1999/2000 and 2004/2005, together with a collection of environmental factors (i.e., topography, soil fertility, and climate), were used to calibrate the neural network embedded in the LTM. Validation analysis was then conducted by simulating the biofuel cropland change during the period of 2004/2005–2010/2011 using the calibrated LTM and comparing the simulation result with the observed change for the same time period, resulting in 6.3% allocation disagreement (0% quantity disagreement) and 27.4% figure of merit. Future forecast for 2020 showed that biofuel croplands would be expanding northwestward from southeastern North Dakota.