| Summary: | The increasing availability and diversity of remotely sensed imagery, particularly high-resolution imagery has seen a rapid evolution in the use of this technology for generating detailed land cover mapping for protected areas. With a growing selection of different image products available, the choice of imagery will impact the level of detail, accuracy and utility of landcover mapping. Taking into consideration spatial, spectral and temporal resolution, geographical scale, target features, and cost, the choice is often a compromise between these characteristics. This study compares the performance and effectiveness of 10-metre spatial resolution Sentinel-2 and 2-metre spatial resolution WorldView-2 imagery for the identification and mapping of landcover within the 100 Wild Islands Conservation Area, Canada. A reproducible two-phase hybrid image classification workflow incorporating the strengths of object-based image analysis and a random forest machine learning algorithm was used to generate an eleven-class landcover map. Statistical assessment of the results shows that overall WorldView-2 was 16.2% higher in terms of classification accuracy than Sentinel-2. A visual assessment shows that WorldView-2 more accurately and effectively identified the boundaries of landcover patches, particularly small, narrow patches important for understanding habitat distribution and connectivity, as well as patch level metrics. In the case of rockweed (Fucus vesiculosus), a commercially harvested species important in the life cycle of common eider, WorldView-2 more effectively discriminated the size and distribution of patches, providing information critical for informed and effective management of this habitat. While the results of this study show that WorldView-2 imagery is more suitable for landcover mapping within the 100 Wild Islands Conservation Area, for larger protected areas, high-resolution imagery may be impractical when considering cost, availability and processing time. However, a strategic combination of imagery ...
|
|---|