Advances in Watershed Management and Fluvial Hazard Mitigation Using Artificial Neural Networks and Remote Sensing

by Lance Besaw,
Keith Pelletier,
Donna M. Rizzo,
Leslie Morrissey,
Michael Kline,

Document Type: Proceeding Paper

Part of: World Environmental and Water Resources Congress 2008: Ahupua'A


Watershed and channel management across spatial and temporal scales requires an informed, interdisciplinary approach by experts and stakeholders, who often have disparate goals and objectives. The Vermont Agency of Natural Resources' (VTANR) is tasked with solving multi-objective problems associated with Vermont's dynamic waterways, such as mitigating loss of property from stream bank erosion and flooding, minimizing aquatic habitat threats due to geomorphic instability and reducing fluvial erosion hazards as well as sediment and nutrient loading, among others. In an attempt to solve such complex problems, the VTANR River Management Program has been developing and testing field-based and remote sensing data collection protocols and a Geographical Information Systems (GIS)-based data management system. In collaboration with the VTANR, this research focuses on 1) developing methods to predict channel geomorphic condition and inherent sensitivity and 2) exploring the feasibility of incorporating high resolution remotely sensed information (including LiDAR) into the protocols. We have developed a hierarchical system of data-driven artificial neural networks (ANNs) that can incorporate large amounts of disparate data for use in the operational management of channels and watersheds. These ANNs can be used to further the development of region-specific fluvial geomorphic classification systems and/or predict stream geomorphic condition and inherent sensitivity. The VTANR defines stream sensitivity (or the susceptibility of a reach to lateral and vertical adjustment) based on 1) susceptibility associated with stream inherent vulnerability (i.e. hydraulic geometry, sediment regime and stream type) and 2) increased susceptibility associated with current stream adjustment processes and diminished geomorphic condition. Incorporating ANNs and high resolution remotely sensed data has the potential to better quantify stream adjustment properties while providing greater insight to a stream's state of dynamic equilibrium with higher accuracy than traditional methods.

Subject Headings: Rivers and streams | Sensors and sensing | Watersheds | Information management | Systems management | Neural networks | Hydrologic data | Vermont | United States

Services: Buy this book/Buy this article


Return to search