Mapping rice in the USA with Earth Observations in real time

Wednesday, 17 December 2014
Nathan Torbick1, William Salas1, Rick Mueller2, Matthew Hanson1, Megan Corbiere1 and Andy McKenzie3, (1)Applied Geosolutions, LLC, Durham, NH, United States, (2)USDA Washington DC, Washington, DC, United States, (3)University of Arkansas, Agricultural Economics, Fayetteville, AR, United States
The USA is a major rice growing nation and one of the top rice exporters. Weather variability, water resources, and price volatility are current risks to rice production. To support risk management the USDA National Agricultural Statistics Service and Economic Research Service are tasked with providing area statistics and production estimates. A Decision Support Tool (DST) is being developed to provide real-time estimates of rice extent and indicators of condition. The DST is largely driven by multi-scale Earth Observations including Landsat and MODIS that provide daily and 8-day indices that are sensitive to rice growth status and management practices. A multitemporal Classification And Regression Tree approach ingests multiscale imagery in real time to provide rice crop metrics. We hindcast the archives of Landsat (1984-2014) and MODIS (2002-2014) for California and achieve >80% accuracy by June and >95% accuracy by end of July as compared to the Crop Data Layer and county statistics. Outcomes are similar for the Midsouth rice region. The DST was utilized to assess the impact of current drought in California on rice. We predicted a 20% reduction in rice area based on our near time rice extent projections and assuming yields similar to 2013 and recent USDA average farm price estimates, 2014 production losses associated with the drought will amount to approximately $175 million. Additional results using Radarsat-2 in the Midsouth in preparation of ALOS-2 and Sentintel will be shared.