Skip to main content
ARS Home » Research » Publications at this Location » Publication #212489

Title: Comparison of Airborne Multispectral and Hyperspectral Imagery for Yield Estimation

Author
item Yang, Chenghai
item Everitt, James
item Bradford, Joe

Submitted to: Proceedings of the American Society of Agricultural and Biological Engineers International (ASABE)
Publication Type: Proceedings
Publication Acceptance Date: 6/30/2007
Publication Date: 7/30/2007
Citation: Yang, C., Everitt, J.H., Bradford, J.M. 2007. Comparison of airborne multispectral and hyperspectral imagery for yield estimation. Proceedings of the American Society of Agricultural and Biological Engineers International (ASABE). ASABE Paper No. 071058.

Interpretive Summary: Multispectral imagery and hyperspectral imagery are two types of remote sensing imagery that are being used to monitor crop conditions and map yield variability, but limited research has been conducted to compare their differences for assessing crop growth and yield. This study employed different imaging processing and analysis techniques to compare airborne multispectral imagery with airborne hyperspectral imagery for mapping yield variability in grain sorghum fields. Statistical analysis showed that the hyperspectral imagery accounted for more variability in yield than multispectral imagery. These results indicate that hyperspectral imagery has the potential for improving yield estimation accuracy.

Technical Abstract: Multispectral and hyperspectral imagery is being used to monitor crop conditions and map yield variability. However, limited research has been conducted to compare the differences between these two types of imagery for assessing crop growth and yield. The objective of this study was to compare airborne multispectral imagery with airborne hyperspectral imagery for mapping yield variability in grain sorghum fields. Airborne color-infrared (CIR) and hyperspectral imagery and yield monitor data collected from four fields were used in this study. Three-band imagery with wavebands corresponding to the collected CIR imagery and four-band imagery with wavebands similar to QuickBird imagery were generated from the hyperspectral imagery. All four types of imagery (two original and two simulated) were aggregated to increase pixel size to match the yield data resolution. Principle components and normalized difference vegetation indices (NDVIs) were derived from each type of imagery and related to yield. Statistical analysis showed that the hyperspectral imagery accounted for more variability in yield than the other three types of multispectral imagery and that best narrow-band NDVIs explained more variability than best broad-band NDVIs derived from the multispectral imagery. These results indicate that hyperspectral imagery has the potential for improving yield estimation accuracy. However, further research is needed to determine the advantages and disadvantages of using hyperspectral imagery for yield estimation as compared with multispectral imagery.