Evaluation of multi-temporal landsat 8 data for wetland classification in newfoundland, Canada

Wetlands are important natural resources which provide many benefits to the environment. Consequently, mapping and monitoring wetlands has gained a considerable attention in recent years among remote sensing experts. Wetlands undergo a considerable change within a year. Thus, it is important to study how much various wetland types are distinguishable at different dates. This will help in choosing an appropriate image for wetland classification. On the other hands, combining various satellite images acquired on different dates is a promising approach to obtain a more accurate classified map compared to the map obtained by single-date satellite imagery. In this study, wetlands within a pilot sites, located in Newfoundland were first classified using each of the several available Landsat 8 data, captured in the three seasons of Spring, Summer, and Fall. By doing this, the separability of the wetland classes in each season was analyzed. Then, these multi-temporal data were integrated to obtain a more accurate map of wetlands. The overall classification accuracy of the final map was 88%, proving that using multi-temporal remote sensing data was necessary to obtain a more reliable and accurate map of the dynamic wetlands in the province.