Scaling up Satellite-Based Crop Yield Mapping at the Field Scale: Recent Progress and Testing

Abstract:

Accurate, low-cost crop yield estimates at the scale of individual fields or finer would be helpful for a range of applications. To date, progress has been limited by various factors including (i) a lack of available imagery with relatively fine (<100m) spatial resolution, (ii) time and expense needed to process imagery, (iii) methods that rely too much on local calibration to be useful in other settings; and (iv) lack of extensive independent datasets to test remote sensing estimates. Here we report on a new approach that relies on an easily scalable crop-model based method for yield estimation and leverages Google Earth Engine to access atmospherically corrected Landsat data for multiple sites and years. Estimates for yields in the U.S. Corn Belt are obtained and compared against >10,000 field-specific records of crop yields based on insurance records from the USDA. Possibilities for extending this approach to other sensors and regions will be discussed.