Skip to main content
ARS Home » Research » Publications at this Location » Publication #312466
Title: Downscaling Landsat 7 canopy reflectance employing a multi soil sensor platform

Author
item SCUDIERO, ELIA - University Of California
item Corwin, Dennis
item Wienhold, Brian
item BOSLEY, BRUCE - Colorado State University
item SHANAHAN, JOHN - Dupont Pioneer Hi-Bred
item JOHNSON, CINTHIA - Plainview Farms, Inc

Submitted to: Precision Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/30/2015
Publication Date: 6/30/2015
Citation: Scudiero, E., Corwin, D.L., Wienhold, B.J., Bosley, B., Shanahan, J.F., Johnson, C.K. 2015. Downscaling Landsat 7 canopy reflectance employing a multi soil sensor platform. Precision Agriculture. doi: 10.1007/s11119-015-9406-9.

Interpretive Summary: The delineation of input prescription maps in precision agriculture (e.g., map of site-specific management units for fertilizer, irrigation water, soil amendments etc.) relies on a careful study of the spatial variability of soil-plant relationships. Monitoring crop health at the farm scale (e.g., hundreds of hectares) at high spatial resolution is still a challenge. Unfortunately, good quality yield maps are not extensively used due to high costs. Additionally, yield maps do not provide information on plant development through the growing season. Remote sensing can be used effectively to monitor crop health and predict yield. However, in many cases freely available data is characterized by coarse spatial resolution which does not fully represent patterns in crop spatial variability. On the other hand, the price of very-high resolution satellite data is prohibitive. The scope of this paper is to provide an inexpensive option for the use of high resolution canopy reflectance. In particular, a downscaling procedure that allows sharpening the L7 data from a spatial resolution of 30×30 m to that typical of yield maps (e.g., 5×5 m, or higher), using spatial information provided by multiple soil sensors, is presented. The results provided reliable high spatial resolution, reflectance that may be used for precision agriculture practices at the farm scale. In fact, the relationships between Landsat 7 reflectance and crop-yield maps (used as ground truth for assessing the quality of the downscaling procedure) were similar (or improved) using high resolution simulated reflectance, as opposed to the original 900m**2 resolution Landsat 7 data. Land resource managers, producers, agriculture consultants, extension specialists, and Natural Resource Conservation Service field staff can benefit from this fairly inexpensive downscaled L7 reflectance in a number of precision agriculture practices, including: variable inputs maps, plant-stress monitoring, and yield prediction.

Technical Abstract: Crop growth and yield can be efficiently monitored using canopy reflectance. The spatial resolution of freely available remote sensing data is however too coarse to fully understand spatial dynamics of crop status. In this manuscript Landsat 7 (L7) reflectance is downscaled from the native resolution of 30×30 m to that typical of yield maps (ca. 5×5 m) over two fields in northeastern Colorado, USA. The fields were cultivated with winter wheat (Triticum aestivum L.) in the 2002-2003 growing season. Geospatial yield measurements were available (1 per ca. 20 m**2). Geophysical (apparent electrical conductivity and bare-soil imagery) and terrain (micro-elevation) data was acquired (resolution <5×5 m) to characterize soil spatial variability. In a first step, geographically weighted regressions (GWRs) were used to study the relationships between L7 reflectance and the geophysical and terrain data at the 30×30 m scale. Geophysical and terrain sensors could describe a large portion of the L7 reflectance spatial variability (0.83 < R2 <0.94). Maps for regression parameters and intercept were obtained at 30×30 m and used in the second step of the procedure to estimate the L7 reflectance at 5×5 m resolution. To independently assess the quality of the downscaling procedure, yield maps were used. In both fields, the 5×5 m estimated reflectance showed stronger correlations (average increase in explained variance = 3.2%) with yield than in at the 30×30 m resolution. Land resource managers, producers, agriculture consultants, extension specialists, and Natural Resource Conservation Service field staff are the beneficiaries of fairly inexpensive downscaled L7 reflectance data.