PS 46-105 - Multi-scale Field and Remotely Sensed Characterization and Habitat Classification of the Extensive Barguzin Valley Wetlands Protecting Lake Baikal, Russia

Wednesday, August 14, 2019
Exhibit Hall, Kentucky International Convention Center
Charles R. Lane1, Tedros M. Berhane2, Oleg A. Anenkhonov3, Victor V. Chepinoga4,5 and Bradley C. Autrey1, (1)Office of Research and Development, U.S. Environmental Protection Agency, Cincinnati, OH, (2)Pegasus Technical Services, Inc., Cincinnati, OH, (3)Laboratory of Floristics and Geobotany, Institute of General and Experimental Biology, Russian Academy of Sciences - Siberian Branch, Ulan-Ude, Russian Federation, (4)Laboratory of Physical Geography and Biogeography, V.B. Sochava Institute of Geography, Russian Academy of Sciences - Siberian Branch, Irkutsk, Russian Federation, (5)Department of Botany, Irkutsk State University, Irkutsk, Russian Federation
Background/Question/Methods

Nearly 9% of hydrologic inputs into Lake Baikal, Russia, flow through the nearly 500 km2 of Barguzin Valley wetlands. These wetlands are important habitats that provide sedimentation, flow regulation, and biogeochemical processing that protect Lake Baikal – the oldest, deepest, and most voluminous freshwater lake in the world. Like the Selenga River to the south, the Barguzin River’s 21,000 km2 watershed is increasingly modified by forestry activity and hydrologic abstraction, as well as global climate change altering precipitation patterning and flows. These modifications not only affect Lake Baikal, but also affect the characteristics, structure, and functioning of the Barguzin Valley wetlands. However, only coarse-grained typological classifications and understanding of the Barguzin Valley wetland habitats exist, which hampers monitoring and management efforts. To alleviate this, we acquired four-band Quickbird multispectral satellite imagery and conducted an initial unsupervised classification. We also collected field data from 142 sampling plots (100 m2) identifying species abundance with >5% cover. Subsequently, we performed a pixel-based random forest supervised classification using a parsimonious three-layer stack (Quickbird band 3, water ratio index, mean texture). Four increasingly fine-grained ecological classifications were created, ranging from two to 18 identified habitats per level. We used Indicator Species Analysis (ISA) to associate plant communities and habitats for each class by hierarchical level, and mapped class distribution across the wetland area.

Results/Conclusions

Our classification overall accuracy (OA) ranged from 87.9% to 99.8%, depending on hierarchical level (i.e., two habitat classes OA = 99.8%; five habitat classes OA = 94.7%; 13 habitat classes OA = 95.9%; 18 habitat classes OA = 87.9%). Habitat indicators were identified through ISA with sufficient fidelity and specificity for the majority – but not all – of the classes across hierarchical levels ranging from one to 15 indicator taxa per habitat (e.g., most detailed hierarchical level Class 1 (of 18) Utricularia spp. [Class 1, p = 0.0422]; Class 2 (of 18) Batrachium sp. p = 0.0392 and Myriophyllum sp. p = 0.0001], etc.). Mapped habitat extent ranged from 0.9% - 15.3% of the wetland area. Outcomes of this research provide both spectral, satellite-based geospatially explicit classes and ecological, field-based indicators of wetland habitats in the Barguzin Valley that are easily scaled to the four different hierarchical management levels. Armed with this information, baseline assessments of structure and/or function across the wetland can be conducted remotely or in-the-field at any of four different hierarchical levels.