A Prototypical Remote-sensing-based Flood Crop Loss Assessment Service System (RF-CLASS) for Crop Risk Management

Flooding often causes severe crop loss over large agricultural areas. In order to make proper polices and provide fair protection to farmers, two agencies in the United States Department of Agriculture (USDA), the National Agricultural Statistics Service (NASS) and the Risk Management Agency (RMA), require the following information timely and accurately in their policy and decision making: (1) time series of flood occurrence and frequency in an area of interest; (2) detection of flooded crop acreage; and (3) actual loss of crop yield due to flood. In order to demonstrate the feasibility of deriving such information timely and accurately from remote sensing data and delivering it to the decision makers effectively through Web service technology, the prototypical RF-CLASS has been designed, developed, implemented, and deployed at http://dss.csiss.gmu.edu/rfclass. The system has the following preliminary functions: (1) production of crop condition indices and flood maps from NASA MODIS data using a workflow of geospatial Web Processing Service (WPS); (2) on-demand geospatial queries, data retrieval and rendering through the Web-based online interactive portal; and (3) online time series analysis of crop condition and progress required by decision-making activities. The system supports online on-demand production and visualization of annual crop condition profile, multiple year crop condition profiles, median crop condition profile of multiple years, cross year crop condition comparative profile, and comparison chart of crop condition profiles vs. historical median. Two representative flooding cases, one in early growing season and another in late growing season, have been selected to demonstrate the feasibility. In the two use cases, the system has successfully (1) identified flood extent and duration, (2) generated crop condition profiles (representing the complete crop growing season and covering pre-flood, during-flood, and post-flood), and (3) estimated crop loss by using regression analysis and crop growth simulation model to correlate the changes of accumulated crop condition index to crop yield. The two use cases show that: (1) flood events can be detected from satellite images; (2) flood damage can be estimated using the crop condition profiles; (3) high spatial resolution image (e.g. Landsat ETM+) provides accurate delineation of flood extent; and (4) flooding frequency can be computed from multi-year flood maps to support decision on improving rate cropland areas in RMA. The use cases demonstrate that it is feasible to provide timely and accurately flood information to support policy and decision making in USDA agencies by using satellite remote sensing data and Web service technology. The next step will be to develop and deploy an operational RF-CLASS to support the missions of those agencies.