Food Aid Project, Inc. 501(c)(3) is a charitable purpose, nonprofit organization whose mission is to support food and economic security by increasing the efficiency of and access to the bulk food procurement and distribution network. They requested a GISCorps volunteer to assist them in remote sensing and GIS analysis to perform global yield prediction via spatial modeling of dry edible beans.
GISCorps volunteer Jill Stanford assisted the Food Aid Project in 2018-19. Through this project, Stanford designed remote sensing methods, evaluated various satellite imagery sources, digitized bean field locations and conducted spatial analysis on identified field locations for yield prediction.
The project kicked off in November 2018 with steering committee member George Jibilian and Stanford collaborating to identify the project goals. It was determined the project would require satellite or aerial imagery and field boundary polygons where edible beans were previously planted and harvested. This data was required to conduct spatial analysis for yield prediction. Jibilian secured the use of Google Earth Engine for the nonprofit project. Stanford generated a workspace, built Fusion tables, and then generated initial crop models with Landsat and MODIS imagery in Google Earth Engine.
The project required edible bean field boundaries, so for several months on the project Jibilian reached out to several farmers and academics at universities researching agronomy, yield, and disease & pest management of dry edible beans. Unfortunately, only one data set was secured limiting the study fields to a large farmer in the United States. These field boundaries were paper insurance maps, so Stanford digitized the 533 acres of dry beans from the paper maps and used Google Earth Engine to test the crop model. Stanford tested multiple imagery sources including Landsat 7, Landsat 8, and MODIS using imagery that corresponded to the known bean field’s growth stage in 2018. Google Earth Engine proved to be a valuable tool for not only the imagery sources but also the derived NDVI maps used in the analysis. As a result, the dry edible bean field identification and yield estimation was successful at a small regional scale, but without additional test data sets it was difficult to expand the analysis to a larger regional or global scale. Therefore, the project was concluded December 2019, but Jibilian will continue the data collection phase in order to identify additional edible bean fields.