*WINNER* Estimating the Condition of Streams & Rivers: An approach using supervised learning methodologies

Abstract

 Conservation efforts are ongoing by various state and federal agencies to improve the health, function, and connectivity of southeastern ecosystems by 10% by 2060. One mechanism for achieving this goal is by looking for where the opportunities to improve the network of lands and waters through land management practices, in the context of wildlife species sustainment. In this study, a habitat system condition index is developed and modeled to be representative of the relative departure of a current wildlife habitat condition from a desired condition to identify where there are conservation opportunities available across the landscape. The resulting index could be used to inform decision makers on where to conduct habitat specific conservation projects. Similar efforts in landscape scale ecology to produce condition indices commonly use an expert system modeling approach or a multi-criterion decision making method. This study explores the application of supervised learning methodologies to produce a more accurate and flexible index that is assessed using elicited field expertise knowledge and continuously validated through its usage by conservation practitioners. Due to the novelty of this approach, this paper discusses the methodologies used to solicit and digitize field knowledge expertise to create the training dataset used.