You are here:
Machine Learning Tools for Predicting Freshwater Fish Populations
Citation:
Patton, D., M. Cyterski, D. Smith, K. Wolfe, B. Rashleigh, JohnM Johnston, AND R. Parmar. Machine Learning Tools for Predicting Freshwater Fish Populations. ICRW7 2020, Athens, GA, November 16 - 19, 2020.
Impact/Purpose:
Advancing Watershed Science using Machine Learning, Diverse Data, and Mechanistic Modeling
Description:
To address the lack of publicly available fish community data for the majority of US, lotic freshwater habitats, we developed scientific software modules and databases for predicting fish populations by NHDPlus (National Hydrography Dataset) segment. We applied machine learning techniques such as generalized boosting and local kernel smoothing regression to survey data from the EPA, USGS, and state agencies. We discuss and assess the interpretability of the models alongside predictive performance. We illustrate applications of the model for supporting management decisions that impact covariates used for fish community prediction, such as the index of watershed integrity (IWI, Thornbrugh et al. 2018). The tools are made available through the browser-based software platform named PiSCES.
URLs/Downloads:
https://www.icrwatersheds.org/seventh-interagency-conference-on-research-in-the-watersheds/agenda-program/