Skip to main content


Title: Investigating the feasibility of using neural networks to derive chlorophyll a prediction algorithms for case 2 waters
Author: Isabel Sargent
Document Type: Monograph
The measurement of chlorophyll a in the coastal zone of the United Kingdom is necessary for monitoring the state of the aquatic environment. Multispectral aerial imagery provides a wealth of information about this region but the algorithms that have been developed for estimating chlorophyll a have been found to be useful only for very local regions over short periods of time. This study investigated the ability of neural networks to develop algorithms for chlorophyll a prediction and found: The neural network algorithms were significantly more successful than linear regression algorithms at predicting chlorophyll a in the same region as they were trained; The relationship between water-leaving spectra and chlorophyll a is non-linear; The FLH feature has the most linear relationship to chlorophyll a; The blue wavelengths have a very non-linear relationship to chlorophyll a but hold a great deal of information about the chlorophyll a in water; Non-linear algorithms performed better when applied to a new site.
Publisher: Environment Agency
Publication Date: 1999
Publication Place: [Bristol]
Subject Keywords: Coastal watersAquatic environmentMonitoringChlorophyllsComputer scienceFeasibility studies
Extent: 27; + appendices
Total file downloads: 294

Download PDF    Display PDF in separate tab