Blended spectral classification techniques for mapping water surface Transparency and Chlorophyll concentration

An innovative technique for estimating Secchi Disk Transparency and Chlorophyll a concentration is examined using in situ samples and coincidental satellite imagery for West Point Lake, Georgia. The technique is divided into two main components: (1) unsupervised classification to organize and reduce spectral variance, and (2) linear logarithmic modeling to transfer class structure onto primary water quality measurements. In componenf 1, clusters are derived using a non-parametric approach that is computationally unique from the traditional ISODATA algorithm. The method includes focused stratified sampling, non-parametric estimation, and blending of class structure using first-order principal components. In component 2, the class structure is tied to water quality estimation using primary band ratios for visible, near infrared, and middle infrared as independent variables. The results indicate a strong association between the Landsat TM middle infrared band and observed measurements for Secchi Disk nansparency and Chlorophyll a concentration. Logarithmic ratios for the visible green to the visible red are shown to be the second most significant covariates. The resultant models are shown to explain 98 percent of the variance in Secchi Disk Transparency, and 93 percent of the variance in Chlorophyll a concentration using pooled data from 59 sampling stations acquired during two distinct