Date of Award

1993

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Earth & Environmental Systems

Abstract

Airborne multispectral video imagery of two soil series (Ragsdale silt loam and Iva silt loam) in western Indiana was acquired on May 17, 1988. Three narrow bands in the yellow-green (543-552 nm), red (644-656 nm), and near infrared (815-827 nm) portion of the spectrum were collected from an altitude of 2400 m using a color multispectral imaging video system developed by the USDA-ARS in Weslaco, Texas. Digital image processing of the multispectral video data was implemented on the microcomputer-based MULTISPEC image analysis system. Several versions of spectral maps were generated from the three original bands, principal component transformed bands, simple ratio bands, and individual band of near infrared, red, and yellow-green. Each spectral map developed was evaluated with regard to its ability to depict discrete soil classes that were separable and provided important soil parameter information, particularly soil organic matter content and soil color. Divergence statistical separability techniques were employed to evaluate and group individual spectral classes into meaningful soil regions. Based on statistical separability and ground measurements/observations, it was found that principal components transformed video data and the original three band video data yielded soil spectral classes that best associated with the soil series patterns in the field. A principal component transformation (PC1) yielded the most detailed soil spectral classes and also was found to contain the most soil information. When soil survey soil series were compared to the spectral soil class map derived from principal component transformation, the latter was more accurate in expressing soil parameter patterns. The spectral maps compiled from PC1 and three band video data delineated six and five classes respectively instead of the two classes identified on the soil series map. Organic matter and soil color were statistically correlated with video data in its various original and transformed formats. Significant correlation coefficients were found at the.01 probability level, with the exception of the YG/R ratio band. These significant correlations indicated that the soil spectral classes delineated were good indicators of the designated soil parameters. The video spectral class data was used to demonstrate its potential value in rural tax base reassessment based on a hypothetical analysis of study site using official tax assessed data. Insights provided by analysis of the video data products indicated that they are likely to have practical applications in the planning of selected farm operations (i.e., herbicide and fertilizer application) and a potential for more accurate placement of soil series boundaries. The results obtained from this research indicate a need to reevaluate the efficacy of the conventional procedures currently employed in soil surveys particularly in small fields where data secured from other information-gathering platforms may not be available.

Download

Share

COinS