Remote Multifrequency Reconnaissance of Coastal Environments
E. A. Weisblatt, John L. Culver · Offshore Technology Conference · 1972
Abstract Current research in remotely sensed microwave and infrared radiometry has demonstrated the theoretical capability of discrimination between many kinds of natural surfaces inherent to the coastal zone. In practice, exceedingly high noise levels of instrumentation and atmospheric phenomena tend to preclude detailed separation of environments. As a result, probabilistic, rather than deterministic models of discrimination are attempted. An application of discriminate analysis to multi-frequency microwave and infrared signals is presented in an attempt to identify eight nearshore environments along the southeast Florida coast at Boca Raton. Data were collected during a 1968 NASA mission (66) with objectives unrelated to this study, but which nevertheless provided excellent study data. Graphical summaries and correlations of results are presented with a cursory analysis of the physical causes for the relationships. Conclusions illustrate the potential for environment identification of more than 80 percent of the sample points in the Boca Raton study area when the relative radiometric temperature in four frequencies are used in the analysis. Despite overlapping of environment radiometric signatures, correct identifications are made with a high degree of probability. The study also illustrates the limitations and importance of the environment classification employed in monitoring the coastal zone. Introduction Current emphasis on the coastal zone by both public and private sectors of our society have created a demand for time-sequential, large-scale, broad coverage maps of coastal environments. In part, environmental problems created the need for data, however, the realization during the last twenty-five years that coastal processes are best understood in a synoptic sense, has given the scientific community impetus to develop new data capture techniques. Remote detection methods of data capture have depended primarily on panchromatic aerial photography, however, in recent years multi-frequency sensors have enabled rapid data capture of numerous discrete bands of visible light as well as ultraviolet, infrared, microwave, and radio frequency signals. The most difficult problem in establishing multi-frequency remote sensors as a practical surveillance tool is the development of a method of extracting particular information from a large mass of data. Additionally, an electromagnetic signal, like most other characteristics of a terrain, exhibits a dispersion about a mean value which is a function of numerous surface, atmospheric, and instrument parameters. Because of the large data mass and the statistical nature of multi-frequency signals, a method of discrimination is required which is tailored to recognize the target terrain. Such methods of discriminant analysis are currently operational with varying degrees of sophistication as several investigators have demonstrated (eg. Nalepka, et al., 1970; Weisblatt, 1970). Discussion Primary emphasis in the development of discriminant techniques have focused on agricultural areas where different crop types exhibit discrete terrain covers. Considerably less attention has focused on terrain cover displaying continual gradients. This study attempts to demonstrate the feasibility of automatically processing multispectral data acquired along profiles flown perpendicular to the shoreline at Boca Raton, Florida (Figure 1), with the objective of discriminating major geomorphic zones. Although the sensor configuration utilized was designed for problems unrelated to the task at hand, encouraging results are achieved.