ARS | Publication request: Evaluation of satellite data and spatial transferability for improved crop residue models

Submitted to: Institute of Electrical and Electronics Engineers
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: May 5, 2012
Publication Date: N/A

Interpretive Summary: Much has been learned from the more than decade long Earth Observing One (EO-1) Satellite Mission, including increased appreciation for imaging spectrometer data in science studies of ecosystems, agriculture, volcanoes, and coasts. EO-1 has also led to considerable technological developments in hardware, software, communications, calibration methods, and other areas. An active area of satellite-based research over the past decade has involved methods to estimate crop residue cover using remotely sensed data acquired by airborne and space-based sensors. Crop residue is an important factor in the recycling of plant nutrients, the primary source of organic material added to the soil, and an important constituent for the stability of agricultural ecosystems. This study is conducted in two Midwestern agricultural watersheds in Indiana, USA: the Mace subwatershed located within the Walnut Fork Creek watershed north of Indianapolis, IN, and the McFarland/Otterbein subwatershed located within the Little Pine Creek watershed west of Purdue University in West Lafayette, IN. Both watersheds are primarily agricultural, with corn and soybean production dominating landuse. The results of this study provide an alternative method for improved spatially explicit remotely sensed estimates of crop residue cover which should allow users/stakeholders to identify hot spots, evaluate the benefits and the risks of biofuel production, and evaluate progress in terms of crop residue management, thereby providing improvements to current decision making processes and allowing new guidelines for best management practices to be developed for sustainable residue removal.

Technical Abstract: Various studies have demonstrated that spectral indices derived from Landsat TM data can be used effectively to quantify crop residue cover if adequately calibrated using in situ data. However, recent developments in remote sensing technologies may possibly accelerate research related to residue coverage estimation and the spatial distribution of the residue to obtain more accurate quantifications. The capability of two multispectral sensors, Landsat Thematic Mapper (TM) and Advanced Land Imager (ALI) and two hyperspectral sensors, Hyperion and airborne SpecTIR were evaluated for estimation of crop residue cover based on the Normalized Difference Tillage Index (NDTI) and Cellulose Absorption Index (CAI). Hyperspectral indices computed from Hyperion and airborne data were also evaluated in a multi-season, multi-watershed study in North central Indiana. The Cumulative Distribution Function (CDF) matching method was used as an upscaling method and possible means for obtaining spatially transferrable observation operators by integrating multispectral (NDTI) and hyperspectral (CAI) data. Results show that the ALI index (NDTI) consistently yields crop residue models with lower Root Mean Square Error (RMSE) values than the Landsat multispectral data. The results strongly indicate that the ALI sensor data improve the ability to detect land cover characteristics relative to Landsat TM data. The observation operators obtained in this study from the linear piece-wise CDF matching method were found to be spatially robust, providing the capability to utilize hyperspectral data to “calibrate” multispectral indices for assessing crop residue cover estimates.