From site to system: Approaches for producing system‐wide estimates of fish habitat in large rivers

Abstract

Worldwide, many productive rivers are dam-regulated and rely on flow management strategies that must balance support of ecological processes with human water use.One component of evaluating this balance is to understand ecological consequences of alternative flow management strategies, which has often been accomplished by coupling population dynamics models with models that relate streamflow to habitat availability and quality. Numerous methods assign habitat availability to locations within a river basin: These include fine-scale field-measured values that are extrapolated to other locations within the basin having similar physical characteristics or equationdriven values created by functions of model-predicted values of physical characteristics.The array of options for creating habitat models is evolving rapidly as high-resolution remote-sensing data becomes more accessible and computational capacity improves.Our objective was to identify trade-offs among approaches that assign habitat relationships to large rivers and to create a decision support tool to supplement choices of extent and granularity. Using a selection of case studies that represent a breadth of scales and diverse trade-offs, we demonstrate the need for a transparent process of data evaluation and assessment to determine the appropriate fit for model scope or context that best supports management needs and recognize sources of uncertainty.The structured approach proposed here aims at improving future model development and refine population dynamics models that inform the management of rivers.

Random forest models to estimate bankfull and low flow channel widths and depths across the conterminous United States

“…Depending on available resources and data, researchers and managers may have access to a variety of methodologies for estimating channel dimensions and habitat capacity, ranging from submeter resolution ground‐based GPS surveys (e.g., Tonina et al, 2019) to broad regional regression models (e.g., Bieger et al, 2015; Faustini et al, 2009). Thus, our models may be of particular value to regional and national habitat applications where headwater stream data are unavailable through remote sensing (e.g., LIDAR) or when managers are constrained by current computational limitations or costs (Robinson et al, 2022).…”