Mapping of salmon habitat parameters using digital airborne imagery

• Main Author: Puestow, Thomas, 1967-
• Other Authors: Memorial University of Newfoundland. Dept. of Geography
• Format: Text
• Language: English
• Published: 1998
• Subjects: Atlantic salmon > Habitat > Newfoundland and Labrador > Remote sensing; Newfoundland; Atlantic salmon; Newfoundland studies; University of Newfoundland
• Online Access: http://collections.mun.ca/cdm/ref/collection/theses4/id/152633

Thesis (M. Sc.), Memorial University of Newfoundland, 1999. Geography Bibliography: p. 105-113 This study focuses on the application of airborne remote sensing and image classification to the mapping of bottom substrate, channel pattern and land cover as important freshwater habitat parameters for Atlantic salmon. A Compact Airborne Spectrometric Imager (CASI) was used to collect multispectral image data with approximately 20 nm wide bands centred at wavelengths of 510, 590, 660 and 730 nm. Image preprocessing included a first order atmospheric correction for path radiance and geometric registration to the UTM reference system. Numerical transforms on the imagery included principal component transformations on original and logarithmized spectral bands, as well as the derivation of a normalized difference vegetation index (NDVI). Ancillary information consisted of valley gradient and stream width. Valley gradient was derived from elevation data contained in a 1:50.000 digital map sheet. Stream width was extracted from the image data. The river course was divided in sections of approximately equal length (30 m), and the average width of each segment was calculated from its length and area. The importance of individual predictor variables for the extraction of the habitat parameters was established using the mean response for each predictor variable, standardized distance matrices and plots of group variability. Separate image classifications were carried out for substrate type, channel pattern and land cover using a hierarchical decision tree algorithm. The end nodes of the final classification trees were implemented as classification rules in a FORTRAN program. Classification accuracy was assessed using an independently collected test sample. The observed overall classification accuracies were 66.87 %, 38.11 % and 84.91 % for substrate type, channel pattern and land cover, respectively. Overall accuracy was significantly improved for the habitat parameters substrate type and channel pattern by combining categories of these variables according to their significance in designating suitable spawning habitat. The revised overall accuracy values for these habitat parameters were 73.76 % and 64.47 %, respectively. Finally, substrate type and channel pattern were combined to create composite maps of spawning habitat suitability. The resulting stratification of salmon spawning habitats corresponds well with the findings of earlier investigations. Therefore, the value of the methodology developed in this study for the management and protection of freshwater salmon habitat was successfully demonstrated.