Automatic Scouting of Agricultural Fields with High Resolution Satellite Imagery.

High resolution, remotely-sensed data is useful for agricultural producers, giving insights into the current biomass, nutrient status, water stress, and condition of the crops on a field. When agricultural fields are continuously monitored throughout a growing season, comparison of sequential images to each other and to imagery from previous seasons can be used to identify areas of a field experiencing acute, potentially yield-limiting conditions. Continuous monitoring and analysis of fields can enable farmers to prioritize fields in need of scouting, allowing for intervention in areas of a field experiencing treatable stresses such as nutrient deficiencies, weed pressure, or pest infestations.

Images of agricultural fields (approximately 17,500,000 acres) managed by midwestern users of FarmLogs, a farm management software, are regularly collected by the RapidEye satellite constellation. Archival imagery of the same area is available for past growing seasons between 2009 and 2014; the collection of 2015 growing season imagery is ongoing. Field conditions are monitored using the Wide Dynamic Range Vegetative Index (WDRVI), an index utilizing the red and NIR bands. WDRVI pixel values are transformed into percentiles, representing the relative rank of pixels to one another. Large changes in relative pixel rank in a manner that is not consistent with historical trends in the same field indicate a potential stress on the crop. As the magnitude of the change in relative pixel rank increases, the likelihood of a stress on the crop increases. In order to validate field conditions, FarmLogs reached out to farmers with anomalous areas in one or more of their fields, asking them to scout troubled areas. Using this algorithm, we were successfully able to identify pest infestations, areas of flooding, and decreases in soil moisture that lead to lower-than-average yield in the affected areas.