Satellite-based remote sensing of running water habitats at large riverscape scales: Tools to analyze habitat heterogeneity for river ecosystem management

Abstract

We illustrate an approach to quantify patterns in hydraulic habitat composition and local heterogeneity applicable at low cost over very large river extents, with selectable reach window scales. Ongoing developments in remote sensing and geographical information science massively improve efficiencies in analyzing earth surface features. With the development of new satellite sensors and drone platforms and with the lowered cost of high resolution multispectral imagery, fluvial geomorphology is experiencing a revolution in mapping streams at high resolution. Exploiting the power of aerial or satellite imagery is particularly useful in a riverscape research framework (Fausch et al., 2002), where high resolution sampling of fluvial features and very large coverage extents are needed. This study presents a satellite remote sensing method that requires very limited field calibration data to estimate over various scales ranging from 1m to many tens or river kilometers (i) spatial composition metrics for key hydraulic mesohabitat types and (ii) reach-scale wetted habitat heterogeneity indices such as the hydromorphological index of diversity (HMID). When the purpose is hydraulic habitat characterization applied over long river networks, the proposed method (although less accurate) is much less computationally expensive and less data demanding than two dimensional computational fluid dynamics (CFD). Here, we illustrate the tools based on a Worldview 2 satellite image of the Kiamika River, near Mont Laurier, Quebec, Canada, specifically over a 17-km river reach below the Kiamika dam. In the first step, a high resolution water depth (D) map is produced from a spectral band ratio (calculated from the multispectral image), calibrated with limited field measurements. Next, based only on known river discharge and estimated cross section depths at time of image capture, empirical-based pseudo-2D hydraulic rules are used to rapidly generate a two-dimensional map of flow velocity (V) over the 17-km Kiamika reach. The joint distribution of D and V variables over wetted zones then is used to reveal structural patterns in hydraulic habitat availability at patch, reach, and segment scales. Here we analyze 156 bivariate (D, V) density function plots estimated over moving reach windows along the satellite scene extent to extract 14 physical habitat metrics (such as river width, mean and modal depths and velocity, variances and covariance in D and V over 1-m pixels, HMID, entropy). A principal component analysis on the set of metrics is then used to cluster river reaches in regard to similarity in their hydraulic habitat composition and heterogeneity. Applications of this approach can include (i) specific fish habitat detection at riverscape scales (e.g., large areas of riffle spawning beds, deeper pools) for regional management, (ii) studying how river habitat heterogeneity is correlated to fish distribution and (iii) guidance for site location for restoration of key habitats or for post regulation monitoring of representative reaches of various types.