Evaluating the Spatial and Temporal Accuracy of Crop Acreage Estimation from Different Agencies

Crop mapping is one of the most critical tasks in crop identification and yield estimation. In recent decades, the studies of crop field mapping are emerging thanks to the development of Remote Sensing and Geographic Information Science technologies. For example, the United States Department of Agriculture (USDA) publishes Cropland Data Layer (CDL) for agricultural fields mapping across the contiguous United States every year. Soon after CDL was first published on CropScape (A Web service based application for exploring and disseminating CDL data) in the early 2000s, it was widely adopted as one of the key references in many agricultural studies. There are some misclassifications during the production of CDL, which we can observe by exploring the data or studying the accuracy assessment reports. However, the report only provides state-level general accuracy estimation while lacking a comparison between different states. Moreover, the state-level assessment is not able to tell the crop mapping accuracy for other counties. For example, the crop planting mapping accuracy for counties with more agricultural activities might be different than places with less planting activities. This research is aimed to compare the accuracy of CDL in different locations. In addition, the paper also examined the acreage in CDL with other official data sources (Crop Acreage Data from Farm Service Agency and survey data from National Agricultural Statistics Services). The result could support further scientific research by providing a valuable understanding of the spatial variation of CDLs accuracy.