Quantitative analysis and evaluation of AVHRR and terrian data for small scale soil pattern recognition

Endre Zsolt Dobos, Purdue University

Abstract

Previous research has demonstrated that data from air- and spaceborne sensors can be used effectively in identifying and delineating soil patterns and certain soil properties. However, little is known about the potential use of small scale satellite data, such as AVHRR (Advanced Very High Resolution Radiometer) in extracting soil information. The general objective of this study is to evaluate the use of coarse spatial resolution satellite imagery and digital terrain data as potential data sources for delineating meaningful soil information in support of small scale soil database development. An effort was made to quantify and characterize the AVHRR-soil relationship. A multitemporal and multispectral database of AVHRR was used for a statistical analysis. The results show that the amount of extractable information depends on the spectral characteristics of the band, the acquisition date of the image and the environmental conditions of the observed area at the time of data acquisition. The thermal bands and the vegetation index were found to be the best for delineating soil patterns. The results suggest that the soil class identification in small scale endeavors is often more likely to be based on the environmental characteristics, the delineation of the so called soil-forming environment patterns, than on the characteristics of the soil itself. Based on the results reported here, it was concluded, that remotely sensed data are greatly influenced by terrain variability, however, such data do not represent all the soil variability that occurs in the landscape. The integration of digital terrain data of appropriate spatial resolution has greatly improved the classification performance. A new terrain descriptor function, namely the potential drainage density function was developed and integrated to the model for the characterization of a plain landscape with the use of coarse spatial resolution digital terrain data. With the use of integrated AVHRR and digital terrain data, numerous data processing and classification algorithms were tested and evaluated. The best result was achieved with the use of the discriminant analysis feature extraction procedure and a spatial-spectral classifier, namely the ECHO.

Degree

Ph.D.

Advisors

Baumgardner, Purdue University.

Subject Area

Agronomy|Environmental science|Remote sensing

COinS