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 #360278

Research Project: Improving Agroecosystem Services by Measuring, Modeling, and Assessing Conservation Practices

Location: Hydrology and Remote Sensing Laboratory

Title: Comparison of methods for modeling fractional cover using simulated satellite hyperspectral imager spectra

item DENNISON, P.E. - University Of Utah
item KOKALY, R.F. - Us Geological Survey (USGS)
item THOMPSON, D.R. - Jet Propulsion Laboratory
item QI, Y. - Nebraska Department Of Natural Resources
item Daughtry, Craig
item MEERDINK, S.K. - University Of California
item QUEMADA, M. - Universidad Politécnica De Madrid
item GADER, P.D. - University Of Florida
item ROBERTSON, D.A. - University Of California
item WETHERLEY, E.B. - University Of California

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/2/2019
Publication Date: 9/4/2019
Citation: Dennison, P., Kokaly, R., Thompson, D., Qi, Y., Daughtry, C.S., Meerdink, S., Quemada, M., Gader, P., Robertson, D., Wetherley, E. 2019. Comparison of methods for modeling fractional cover using simulated satellite hyperspectral imager spectra. Remote Sensing of Environment.

Interpretive Summary: Terrestrial vegetation is dynamic both spatially and temporally. Natural and agricultural ecosystems have three fractional components that sum to 100% cover: photosynthesizing or “green” vegetation (GV), non-photosynthesizing or “brown” vegetation (NPV), and bare soil. Brown vegetation includes dead and senescent leaves and needles, plant litter, and non-photosynthesizing branch and stem tissues. The transition from green vegetation to brown vegetation or soil cover is a hallmark of both seasonal and/or drought changes. In croplands, green vegetation cover typically increases through the growing season, followed by increases in brown vegetation and soil fractions after harvest. Crop residue cover is an important form brown vegetation cover that often indicates soil tillage intensity and conservation practices. Hyperspectral sensors have narrow spectral bands capable of fully resolving different “brown” absorption features and can accurately map fractional cover. As a result, a new generation of satellite hyperspectral sensors have been proposed that should improve mapping the fractions of the green vegetation, brown vegetation, and soil.

Technical Abstract: Remotely sensed data can be used to model the fractional cover of green vegetation (GV), non-photosynthetic vegetation (NPV), and soil in natural and agricultural ecosystems. NPV and soil cover have traditionally been difficult to estimate accurately, since the primary distinguishing feature of NPV, absorption by lignin, cellulose, and other organic molecules in the shortwave infrared (SWIR), cannot be resolved by broadband multispectral data. A new generation of satellite hyperspectral imagers will provide contiguous narrowband coverage across the visible-through-SWIR spectrum, enabling new, more accurate, and potentially global fractional cover products. We used four field spectroscopy datasets with corresponding assessments of fractional cover to simulate satellite hyperspectral data, including sensor noise and atmospheric correction artifacts. The combined dataset was used to compare index-based and spectroscopic methods for estimating GV, NPV, and soil fractional cover. Index-based metrics included vegetation indices like normalized difference vegetation index (NDVI) for GV cover and cellulose absorption index (CAI) for NPV cover. Spectroscopic methods included partial least squares (PLS) regression, spectral feature analysis (SFA), and multiple endmember spectral mixture analysis (MESMA). Vegetation indices like NDVI performed well for estimating GV cover, demonstrating that within the constraints of our dataset, multispectral data are sufficient for accurate mapping of GV cover. PLS and SFA were consistently more accurate than index-based metrics for estimating NPV and soil cover, demonstrating the need for narrow, contiguous bands to accurately resolve these two cover types. Using an independent validation dataset, the lowest root mean squared error (RMSE) values were 0.078 for GV using NDVI, 0.154 for NPV using PLS, and 0.136 for soil using PLS. SFA had the lowest RMSE averaged across all three cover types (0.129). This work highlights the need for more extensive and diverse fine spatial scale assessment of fractional cover, to improve methodologies for estimating cover in preparation for future hyperspectral global monitoring missions.