Validation and refinement of cropland data layer using a spatial-temporal decision tree algorithm

Li Lin,Anna Yang,Yahui Di,Liying Guo,Liping Di,Hui Li,Chen Zhang

doi:10.1038/s41597-022-01169-w

Abstract

Space-based crop identification and acreage estimation have played a significant role in agricultural studies in recent years, due to the development of Remote Sensing technology. The Cropland Data Layer (CDL), which was developed by the U.S. Department of Agriculture (USDA), has been widely used in agricultural studies and achieved massive success in recent years. Although the CDL’s accuracy assessments report high overall accuracy on various crops classifications, misclassification is still common and easy to discern from visual inspection. This study is aimed to identify and resolve inaccurate crop classification in CDL. A decision tree method was employed to find questionable pixels and refine them with spatial and temporal crop information. The refined data was then evaluated with high-resolution satellite images and official acreage estimates from USDA. Two validation experiments were also developed to examine the data at both the pixel and county level. Data generated from this research was published online in two repositories, while both applications allow users to download the entire dataset at no cost.

Highlights

Background & SummaryMonitoring agricultural activities can provide valuable information to farmers, investors, and decision makers[1,2], and crop identification and mapping is the fundamental basis to support agricultural studies[3–6]
The confidence layers cannot be used directly to represent the accuracy of the Cropland Data Layer (CDL), the accuracy assessments are crucial in evaluating the quality of classification results for the CDL at the pixel level[16]
We found that refined CDL (R-CDL) has very similar R2 metrics with CDL compared to National Agricultural Statistics Service (NASS) acreage estimates, which indicates the R-CDL is reducing the classification noises

Summary

Introduction

Background & SummaryMonitoring agricultural activities can provide valuable information to farmers, investors, and decision makers[1,2], and crop identification and mapping is the fundamental basis to support agricultural studies[3–6]. The U.S Department of Agriculture (USDA) has published the annual Cropland Data Layer (CDL), which covers the contiguous United States, since 20082,10–12. Since the production of CDL mainly relies on remote sensing datasets, criticisms have arisen in past years regarding the quality of CDL products[16,17]. To address this concern, the USDA National Agricultural Statistics Service (NASS) has published an accuracy assessment for the CDL since 200812. The confidence layers cannot be used directly to represent the accuracy of the CDL, the accuracy assessments are crucial in evaluating the quality of classification results for the CDL at the pixel level[16]. We acknowledge the usage and importance of both measurements; such an assessment cannot be utilized in this paper because the accuracy assessment and confidence layer are byproducts of the CDL production process, and this research is designed to conduct an independent review on CDL17,19

Objectives

Methods

Results