From 2008-2013, the National Agricultural Statistics Service’s (NASS) archive only provided conterminous U.S. crop-type maps. Earth Resources and Observation Science (EROS) scientists Aaron Friesz (Innovate), Bruce Wylie, and Danny Howard wanted to document actual crop distribution going farther back in time to better assess whether crop areas were carbon sinks or carbon sources.
The team formulated a method to extend the NASS conterminous Cropland Data Layer archive back 8 additional years using carbon flux models. They conducted a feasibility assessment on crop typing in the Greater Platte River Valley and Great Plains for mapping crop types with primarily the phenological signal of weekly MODIS Normalized Difference Vegetation Index (NDVI) data. Howard presented their findings at an annual American Geophysical Union conference. The response was promising: “he [Howard] said [that] there were lots of people very excited about it who wanted it for the wider conterminous U.S.,” said Wylie. Hydrologists wanted it to help estimate regional water use among different crops while others saw value in knowing how crop type may affect regional pesticide impacts and determining whether specific crops were long-term regional carbon sources or carbon sinks.
Using more than 11 million records, the scientists trained a classification tree algorithm and developed a crop classification model that extended the NASS CDL archive back 8 years from 2008, producing a consistent time series of major crop-type maps for the CONUS at 250-meter spatial resolution for 2000 to 2013. Their research suggests that it is possible to produce these annual crop-type maps by September 2017. This modeling analysis will provide annual updates of crop types, crop water use, and carbon flux. “Our long-time series also allows the assessment of historic impacts of policy and economics,” Wylie said.