Remotely sensed data for bathymetric mapping and ecohydraulics modelling in rivers

Abstract

River management decisions must be based on the best knowledge available. But acquiring information on river conditions and translating it into sufficient mitigation measures can be both time- and resource demanding. Remote sensing data can be an important source of information on the current state of rivers, and may also be used in scenario assessments of future river conditions. This PhD-study has focused on the use of remote sensing data as a source of information on bathymetry, seasonal mesohabitats for fish, scenarios of mitigation measures, and strategies on environmental eDNA sampling. Each of these four information elements are often central parts in river assessments. The results from the PhD-study are presented in four scientific papers and include remote sensing data from green and red LIDAR, multispectral satellite imagery and aerial photos collected and applied in four Norwegian rivers. For the assessment of seasonal mesohabitats for fish, two remote sensing technologies were used: LIDAR and aerial photos. The LIDAR data enabled a 10x10 cm resolution bathymetry used as input to a 2D hydraulic model for the simulation of hydraulic heterogeneity. The aerial photos were used as both a source of calibration data and for assessing spatial heterogeneity along the river banks. The results show that both spatial heterogeneity and hydraulic heterogeneity are significant mesoscale habitat characteristics for European Grayling during its spawning period. No significant relationships were found on spatial and hydraulic heterogeneity for brown trout, emphasising the need for species-specific assessments in terms of mesoscale habitat characteristics. While LIDAR data may be a source of high-resolution bathymetry, collecting and processing LIDAR data can be costly and time consuming. An alternative remote sensing data source for river assessments is multispectral imagery or (the previously mentioned) aerial photos. The application of multispectral imagery from two satellite platforms and aerial photos for mapping of bathymetry by linear models was tested in four rivers within the same geographical region. LIDAR and SONAR data was used in the study for establishing linear image-to-depth models and for verifying the modelled bathymetry. Platform-specific regional models were then established and tested by combining intercept and slope coefficients from the linear models in the four rivers. The results showed that while the final quality of platform-specific regional model bathymetry did not fully match the quality of LIDAR-based bathymetry, overall model performance was adequate for depth calculations. For Worldview-2 images and aerial photos, coefficients of determination (R2) were in the 0.52-0.82 and 0.73-0.91 range, respectively. By adjusting the regional models with estimated local depth and a brightness factor, results on depth calculations improved slightly when compared to the LIDAR-based bathymetry, with coefficients of determination in the 0.47-0.84 and 0.71- 0.91 range, respectively. Point cloud format LIDAR data can easily be modified e.g., for use in assessment of different mitigation measure scenarios. Modified LIDAR data was tested in a public preference study on scenarios for mitigation measures related to weir adjustments and changes in flow-dependent water covered areas in weir basins. The study was conducted in a bypass section (i.e., a river section with water withdrawal) with a 65-70% reduction in yearly discharge due to hydropower production. Findings from the study show that LIDAR data are highly useful for scenario-based adjustments and modelling of weirs. As the potential dry/wet interface at or around the weirs can be hydraulically complex, these locations may require high-density mapping using a combination of red and green LIDAR. Results from the study also include a structured and standardized framework for public preference assessments in rivers which includes the potential use of remote sensing data as input to mitigation measure scenarios. A successful green LIDAR scan depends on a range of factors e.g., the technology applied in the scan process, signal pathway length and disturbances, river conditions and post-processing capabilities. Failure to address each of these factors when applying LIDAR may result in inadequate point cloud classification or coverage. In a study of strategic environmental eDNA sampling for biomonitoring in a river dominated by weirs and flow alteration, a regional adjusted image-to-depth model was applied on remote sensing imagery for river sections where two consecutive LIDAR scans returned inadequate coverage of bathymetry. Results on simulated depths and velocities adequately matched corresponding in-situ measurements of the same variables. Results from generalized mixed models testing the effect of spatial and abiotic variables on sample eDNA concentrations showed that sample location habitat type and season significantly affected eDNA concentrations of brown trout and minnow. Furthermore, weirs were found to be barriers for the dispersion of eDNA during autumn, while no significant effects of weirs were found during spring.