Abstract

Remote sensing techniques could enable remarkable advances in characterizing rivers hydromorphology by providing spatially and temporally explicit information. Remote mapping of hydromorphology can play a decisive role in a wide range of river science and management applications including habitat modeling and river restoration. High resolution satellite imagery (HRSI) has recently emerged as potentially powerful means of mapping riverine environments. This research aims to develop advanced methodologies for processing HRSI to map and quantify a set of key hydromorphological attributes including: (1) river boundaries, (2) bathymetry and (3) riverbed types and compositions. Boundary pixels of rivers are subject to spectral mixture that limits the accuracy of river areas extraction using conventional hard classifiers. To address this problem, unmixing and super resolution mapping (SRM) are focused as two steps, respectively, for estimation and then spatial allocation of water fractions within the mixed pixels. Optimal band analysis for NDWI (OBA-NDWI) is proposed to identify the pair of bands for which the NDWI values yield the highest correlation with water fractions. The OBA-NDWI then incorporates the optimal NDWI as a predictor of water fractions through a regression model. Water fractions obtained from the OBA-NDWI method are benchmarked against the results of simplex projection unmixing (SPU) algorithm. The pixel swapping (PS) and interpolation-based algorithms are applied on water fractions for SRM. In addition, a simple modified binary PS (MBPS) algorithm is proposed to reduce the computational time of the original PS method. Water fractions obtained from the proposed OBA-NDWI method are demonstrated to be in good agreement with those of SPU algorithm (R2=90%, RMSE=7% for WorldView-2 (WV-2) image and R2=87%, RMSE=9% for Geoeye image). The spectral bands of WV-2 provide a wealth of choices through the proposed OBA-NDWI to estimate water fractions. The interpolation-based and MBPS methods lead to sub-pixel maps comparable with those obtained using the PS algorithm, while they are computationally more effective. SRM algorithms improve user/producer accuracies of river areas about 10% with respect to conventional hard classification. This research introduces multiple optimal depth predictors analysis (MODPA) that combines previously developed depth predictors along with other measures such as the intensity components of HSI color space. To avoid over-fitting of the linear model, statistically optimal predictors are selected based on one of partial least square (PLS), stepwise and principal component (PC) regressions. The primary focus of this study is on shallow and clearly flowing streams where substrate variability could have pronounced effect on depth retrievals. Spectroscopic experiments are performed in controlled condition of a hydraulic laboratory to examine the robustness of bathymetry models with respect to changes in bottom types. Further, simulations from radiative transfer modeling are used to extend the analysis by isolating the effect of inherent optical properties (IOPs) and also by investigating the performance of bathymetry models in optically complex and also deeper streams. Bathymetry of Sarca, a shallow river in Italian Alps, is also mapped using a WorldView-2 (WV-2) image where the atmospheric compensation (AComp) product is evaluated for the first time. Results indicate the robustness of multiple-predictor models particularly MODPA rather than single-predictor models such as optimal band ratio analysis (OBRA) with respect to heterogeneity of bottom types, IOPs and atmospheric effects. This study suggests extra predictors when the multiple regression is assisted with an optimal predictors selection process (e.g. MODPA). The extra predictors enhance the accuracy of depth retrievals particularly in optically complex waters and also for low spectral resolution imagery (e.g. GeoEye). Further, enhanced spectral resolution of WV-2 compared to GeoEye improves the bathymetry retrievals. MODPA based on PLS regression provided improvements on the order of 0.05 R2 and 0.7 cm RMSE compared to multiple Lyzenga and 0.18 R2 and 2 cm RMSE compared to OBRA using AComp reflectances of WV-2 for Sarca River with a maximum 0.8 m depth. In addition, a theoretical approach namely hydraulically assisted bathymetry (HAB) is assessed and further modified for calibration of bathymetry models that provided comparable results with the empirical calibration approach. Substrate mapping in fluvial systems has not received as much attention as that in nearshore optically shallow waters of inland and coastal areas. The research to date has been primarily based on surface spectral reflectance data without accounting for water column attenuations. This study aims at retrieving the bottom reflectances in shallow rivers and then examining the effectiveness of inferred bottom spectra in mapping of substrate types. Bathymetry and diffuse attenuation coefficient (kd) are derived from above-water reflectances for which some in-situ/known depths are required. Following the retrievals of depth and kd, bottom reflectances are estimated based on a water column correction method. Moreover, the efficacy of vegetation indices (VIs) is examined for making distinction among the densities of submerged aquatic vegetation (SAV) using either above-water or retrievals of bottom reflectances. This research benefits, for the first time, from three different approaches including controlled spectroscopic measurements in a hydraulic lab, simulations from radiative transfer modeling and an 8-band WordView-3 (WV-3) image. The results indicate the significant enhancements of streambed mapping using inferred bottom reflectances than using above-water spectra. This is evident, for instance, on clustering of three bottom types using simulated spectra with 20% enhancement of overall accuracy. Deep-water correction demonstrated to have most of an impact on retrievals of bottom reflectances only in NIR bands when the water column is relatively thick (> 0.5 m) and/or when the water is turbid. The red-edge (RE) band of WV-3/WV-2 improves remarkably the detection of SAV densities based on the VIs either using above-water or retrieved bottom spectra. Further, the simulated spectra suggest that enhanced spectral resolution of 8-band WV-3 leads to improvements in streambed mapping compared to traditional 4-band imagery. This study demonstrated the feasibility of retrieving bottom reflectances and mapping SAV densities from space in a shallow river using the WV-3 image (user and producer accuracies of 67% and 60% in average for three levels of SAV densities). Moreover, the feasibility of mapping grain size classes is assessed using spectral information based on laboratory experiments coupled with simulations. The changes in grain sizes affect the magnitude of reflectances while the shape of spectra remains almost identical. This characteristic feature demonstrated high potentials for mapping grain size classes by retrieving the bottom reflectances. In summary, HRSI provided promising results and effective means of mapping the selected hydromorphological attributes of shallow rivers in spatially continuous and in large extents.