GSA Annual Meeting in Phoenix, Arizona, USA - 2019

Paper No. 111-1
Presentation Time: 9:00 AM-6:30 PM


SENDROWSKI, Alicia, Geosciences, Colorado State University, 400 University Ave, Fort Collins, CO 80523 and WOHL, Ellen, Department of Geosciences, Colorado State University, 1482 Campus Delivery, Fort Collins, CO 80523-1482

The hydrology of Arctic rivers is influenced by the timing and magnitude of processes including river discharge and ice breakup and jam formation. Arctic rivers are unique in that many maintain natural wood dynamics, while several other rivers have experienced selective wood removal. Large wood transport and deposition exert an influence on system dynamics by affecting channel morphology and stream biochemistry, and are also controlled by river and ice drivers. An understanding of the controls on wood deposition in Arctic rivers would provide further insight into the hydrological influence on material transport and mapping of wood volumes could provide a first order estimate of carbon stored as wood in these systems. Analyses of wood dynamics in rivers tend to focus on smaller scale reaches with field access or use aerial imagery to identify and assess wood deposits. However, the availability of high resolution (50cm) satellite imagery and advanced remote sensing techniques now allows for broader scale patterns of wood deposition to be measured across larger remote areas. Yet, systematic approaches are still needed to capture wood from remotely sensed images given the complexity of wood deposit morphologies in the landscape.

In this work, remote sensing of high resolution GeoEye and WorldView imagery is used to measure wood deposits in the Tatshenshini and Alsek rivers in Alaska, USA and the Yukon, Canada, and the Mackenzie River and Delta in Northwest Territories, Canada. These rivers cover different drainage areas, discharge regime, and channel morphologies. Images collected over several months in 2018 are analyzed that encompass ice jam breakup and river floods. To measure wood deposits, maps of the normalized difference vegetation and water index (NDVI and NDWI, respectively) are combined with principal component analysis (PCA), supervised classification, and machine learning approaches to classify wood. Deposits across these river systems are mapped and a distribution of wood over different spatial extents are collected. The development of a pipeline for wood deposit extraction is a new contribution to wood research. These approaches could also be applied for remote sensing of other hydrological and geomorphological features in the landscape.