Date

2022-2-03

Author

Özer, Erdem

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

2399
views

167
downloads

Cite This

Wetlands constitute one of the wealthiest and most productive ecosystems on earth. These areas are sophisticated aquatic habitats serving not only the locals but also the whole Earth system on a broad range. Following tropical forests, they have the highest biological diversity. These ecosystems are viable nourishment, reproduction, and sheltering environments for a whole range of living beings and are therefore accepted as natural wealth museums of the world. Monitoring such valuable areas and obtaining crucial information from them, in this regard, has been the primary motivation of the studies performed during the preparation of this thesis. When the sizes, geographic distribution, and total coverage of wetlands across the earth are taken into account, remote sensing shines out as the most economically and technically feasible method to realize the goals related to the mentioned motivation. Concerning the utilization of medium resolution satellite images as the input, the pixel-level approach falls short of understanding the wetland dynamics since vast amounts of pixels in such areas have mixed content. In this study, the soft classification of wetlands is aimed in order to determine all ground characteristics and their exact proportions. The path to achieving this goal passes through conducting an investigation within correct boundaries. Hence, detection of the wetland extent prior to sub-pixel analysis is addressed as a critical pre-processing step for realizing the subject motivation. The extent determination part includes calculating Tasseled Cap Water Index (TCWI) values on time series and modeling variations throughout the year by fitting a double-sided sigmoid function. This information is coupled with Digital Terrain Model (DTM) thresholding to extract the final extent. The sub-pixel analysis covers adopting a systematic approach using a three-element (soil, vegetation, water) scheme for establishing wetland ontology and implementing supervised spectral unmixing enhanced by the band and endmember optimizations. Balıkdamı, one of the most impressive wetlands of Turkey, is chosen as the test area. Open access optical satellite data, acquired by Sentinel-2 Multispectral Instrument (MSI), are utilized as the primary input. Since the abundance values of land cover classes in each Sentinel-2 pixel are estimated, reference abundance data with a 10 m grid interval are generated using four-band aerial images having a 30 cm ground sampling distance (GSD) for the verification stage. A new metric entitled "Abundance Confusion Matrix (ACOMA)" is introduced for the comparison and detailed assessment of reference and estimated fractional land cover. Experimental results demonstrate that the extent determination is addressed with a sensitivity of 93.55% and a precision of 99.21%. Moreover, abundance values of land cover classes are determined with overall accuracies of 66.17% and 66.27% for the monotemporal and multitemporal cases, respectively. In addition to a 2% overall accuracy increase compared to the hard classification, the detectability of sparse land cover classes is demonstrated that are vanished while using pixel-based approaches. Furthermore, gradients are able to be observed, particularly at watersides. As a result, the proposed method proves to be a valuable tool for the detailed monitoring of wetlands.

Subject Keywords

Wetlands, Spectral Unmixing, Sentinel-2, Abundance Confusion Matrix, Fractional Land Cover

URI

https://hdl.handle.net/11511/98554

Collections

Graduate School of Natural and Applied Sciences, Thesis