Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #381791

Research Project: Integrating Remote Sensing, Measurements and Modeling for Multi-Scale Assessment of Water Availability, Use, and Quality in Agroecosystems

Location: Hydrology and Remote Sensing Laboratory

Title: Evaluating NISAR’s cropland area algorithm over the conterminous United States using Sentinel-1 data

item ROSE, S. - University Of Massachusetts, Amherst
item KRAATZ, S. - University Of Massachusetts, Amherst
item KELLNDORFER, J. - Collaborator
item Cosh, Michael
item TORBICK, N. - Applied Geosolutions, Llc
item HUANG, X. - Applied Geosolutions, Llc
item SIQUEIRA, P. - University Of Massachusetts, Amherst

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/27/2021
Publication Date: 7/1/2021
Citation: Rose, S., Kraatz, S., Kellndorfer, J., Cosh, M.H., Torbick, N., Huang, X., Siqueira, P. 2021. Evaluating NISAR’s cropland area algorithm over the conterminous United States using Sentinel-1 data. Remote Sensing of Environment. 260:112472.

Interpretive Summary: Mapping of crop area is important for monitoring and forecasting food security status. Active microwave remote sensing is a valuable for this endeavor, because it is not subject to cloud interference, like visible and infrared remote sensing. An analysis of a crop area mapping algorithm on the European Sentinel 1-C radar data was evaluated for accuracy. It was determined that common crops like corn, soybean, and cotton estimates had good accuracy, but small grain crops and pasture/rangelands were more difficult to achieve good results. It was also determined that there was a strong geographic dependence on the accuracy of the algorithm. This will be useful for the refinement of the algorithm as satellite missions move forward.

Technical Abstract: Accurate knowledge of the distribution, breadth and change in agricultural activity is important to food security and the related trade and policy mechanisms. Routine observations afforded by spaceborne Synthetic Aperture Radar (SAR) allows for high-fidelity monitoring of agricultural parameters at the field scale. Here we evaluate the approach to be used for generating NASA’s upcoming NASA ISRO SAR (NISAR) mission’s L-band cropland area product using Sentinel-1 C-band data. This study uses all ascending Sentinel-1A/B data collected over the conterminous United States (CONUS) in 2017 to compute the coefficient of variation (CV) at 150 m x 150 m resolution and evaluates the accuracy of CV-based crop/non-crop classifications at 100 one-by-one degree tiles. We compare two approaches of determining the crop/non-crop threshold CV values (CVthr) consisting of a constant threshold set at a literature-recommended value of CVthr=0.5 and a more computationally expensive receiver operating curve approach using Youden’s J statistic (YJS). Comparisons of the thresholds and their respective accuracies inform on how closely the accuracies of the non-optimized but less computationally expensive approach resemble the optimized values. The constant threshold and ROC approach respectively achieved 81.5% and 86.8% accuracy. A breakdown by census geographic region, showed that classifications exceeded 80% (90%) in the South and Midwest, but were only 76.1% (73.5%) in the West, when using CVthr=0.5 (YJS). This improvement for YJS mainly stems from tiles that have 40% or more crop coverage, where accuracies for the YJS-optimized CVthr values were usually about 10% greater. While most non-crop classes were accurately detected (>80%) as non-crop, we found that the approach had difficulty in accurately classifying grasslands/pasture, especially in the West. Corn, soybeans and cotton were accurately detected as crop in all regions, but accuracies for alfalfa, fallow/idle cropland, spring and winter wheat greatly varied by region (ranging from about 20% to 80%). CV values of the crop classes were distinct from the non-crop classes, showing that the approach is suitable for crop versus non-crop classifications. We also found a strong geographic dependence for the optimized CVthr values: values range from about 0.2 at the coasts and gradually increase to about 0.6 in the Central United States, most often falling close to 0.3 and 0.5.