Automatic Features Detection in a Fluvial Environment through Machine Learning Techniques Based on UAVs Multispectral Data

The present work aims to demonstrate how machine learning (ML) techniques can be used for automatic feature detection and extraction in fluvial environments. The use of photogrammetry and machine learning algorithms has improved the understanding of both environmental and anthropic issues. The developed methodology was applied considering the acquisition of multiple photogrammetric images thanks to unmanned aerial vehicles (UAV) carrying multispectral cameras. These surveys were carried out in the Salbertrand area, along the Dora Riparia River, situated in Piedmont (Italy). The authors developed an algorithm able to identify and detect the water table contour concerning the landed areas: the automatic classification in ML found a valid identification of different patterns (water, gravel bars, vegetation, and ground classes) in specific hydraulic and geomatics conditions. Indeed, the RE+NIR data gave us a sharp rise in terms of accuracy by about 11% and 13.5% of F1-score average values in the testing point clouds compared to RGB data. The obtained results about the automatic classification led us to define a new procedure with precise validity conditions.

Implications of taxonomic and numerical resolution on DNA metabarcoding-based inference of benthic macroinvertebrate responses to river restoration

Exploring and clearly defining the level of taxonomic identification and quantification approaches for diversity and biomonitoring studies are essential, given its potential influence on the assessment and interpretation of ecological outcomes. In this study, we evaluated the response of benthic macroinvertebrate communities to the restoration or construction of gravel bars conducted in the dam-impacted Trinity River, with the non-dam influenced tributaries serving as the reference sites. We aim to evaluate the performance of different taxonomic and numerical (i.e., abundance vs. presence/absence data) resolutions of DNA metabarcoding with consequent comparison to morphology-based identification and how it affects assessment outcomes. DNA metabarcoding detected 93% of the morphologically identified individuals and provided finer taxonomic resolution. We also detected significant correlations between morphological sample abundance, biomass, and DNA metabarcoding read abundance. We observed a relatively high and significant congruence in macroinvertebrate community structure and composition between different taxonomic and numerical resolutions of both methods, indicating a satisfactory surrogacy between the two approaches and their varying identification levels and data transformation. Additionally, the community-environmental association were significant for all datasets but showed varying significant associations against the physicochemical parameters. Furthermore, both methods identified Simulium spp. as significant indicators of the dam-impacted gravel bars. Still, only DNA metabarcoding showed significant false discovery rate proving the method’s robustness compared to morphology-based identification. Our observations imply that coarser taxonomic resolution could be highly advantageous to DNA metabarcoding-based applications in situations where the lack of taxonomic information, e.g., poor reference database, might severely affect the quality of biological assessments.

GRAINet: mapping grain size distributions in river beds from UAV images with convolutional neural networks

Grain size analysis is the key to understand the sediment dynamics of river systems. We propose GRAINet, a data-driven approach to analyze grain size distributions of entire gravel bars based on georeferenced UAV images. A convolutional neural network is trained to regress grain size distributions as well as the characteristic mean diameter from raw images. GRAINet allows for the holistic analysis of entire gravel bars, resulting in (i) high-resolution estimates and maps of the spatial grain size distribution at large scale and (ii) robust grading curves for entire gravel bars. To collect an extensive training dataset of 1491 samples, we introduce digital line sampling as a new annotation strategy. Our evaluation on 25 gravel bars along six different rivers in Switzerland yields high accuracy: the resulting maps of mean diameters have a mean absolute error (MAE) of 1.1 cm, with no bias. Robust grading curves for entire gravel bars can be extracted if representative training data are available. At the gravel bar level the MAE of the predicted mean diameter is even reduced to 0.3 cm, for bars with mean diameters ranging from 1.3 to 29.3 cm. Extensive experiments were carried out to study the quality of the digital line samples, the generalization capability of GRAINet to new locations, the model performance with respect to human labeling noise, the limitations of the current model, and the potential of GRAINet to analyze images with low resolutions.

Quantifying how hydrological and geomorphical dynamics in the land-river interface create opportunities and trade-offs for sustainable development

<p>Hydrological and geomorphological processes within the land-river interface (LRI) can be directly linked to several Sustainable Development Goals (SDG). The transfer of water and material along the LRI provides a range ecosystem services that support environmental, economic and social needs. However, the LRI is also very dynamic from a hydrologic and geomorphic perspective. Benefits can turn into hazards and vice versa, depending on natural and human-induced variations in flow and associated geomorphic activity. This study aimed to identify these critical areas by (i) quantifying the natural and human controlled variation in hydrology and geomorphology, and (ii) mapping associated SDG-related opportunities and trade-offs. The upper reaches of the Himalayan Beas River (India) were used as a case study, where the LRI is characterised by three main sections: (i) a free-flowing confined upper valley, (ii) a heavily regulated confined middle valley, and (iii) and a valley with wide floodplains flowing into the Pong Reservoir. Remote sensing imagery from Sentinel-2 (ESA) (2016-2019) were used to quantify the monthly spatial recurrence of river channels and gravel bars. In addition, data was collected on human and natural infrastructure within the catchment (including road network, urban areas, cropland, national parks, etc.). Combination of both datasets indicated that hydrological and river geomorphological processes in the upper part are the most spatially and temporally variable, leading to fertile soils (SDG 1,2), but also the highest risk of flooding in urban areas and cropland (SDG 11, 13) . The middle part is characterised by stable river channels (i.e. no lateral movement) due to the presence of two dams and confines valleys, leading to limited interaction with the surrounding land, except for the provision of water (SDG 6) and a higher risk of landslides (SDG 1,11). Finally, the lower part is again more dynamic in terms of geomorphological processes, with wide gravel bars and side channels. These dynamics allow larger urban areas and cropland to develop (SDG 1, 11), but also exposes cropland to flooding and erosion (SDG 2, 6). By quantifying the spatiotemporal dimension of hydrological and geomorphological processes and how these relate to LRI characteristics, this study provides a dynamic baseline to identify opportunities and trade-offs in optimising the role of the LRI in driving sustainable development.</p>

Theoretical Investigation of Equilibrium Dynamics in Braided Gravel Beds for the Preservation of a Sustainable Fluvial Environment

Gravel bars have an important role in the exchange between surface and subsurface waters, in preventing and mitigating riverbank erosion, in allowing the recreational use of rivers, and in preserving fluvial or riparian habitats for species of fishes, invertebrates, plants, and birds. In many cases, gravel bars constitute an important substrate for the establishment and development of ground flora and woody vegetation and guarantee higher plant diversity. A sustainable management of braided rivers should, therefore, ensure their ecological potential and biodiversity by preserving a suitable braiding structure over time. In the present study, we propose an analytical–numerical model for predicting the evolution of gravel bars in conditions of dynamical equilibrium. The model is based on the combination of sediment balance equation and a regression formula relating dimensionless unit bedload rate and stream power. The results highlight the dependence of the evolving sediment particles’ pattern on the ratio of initial macro-bedforms longitudinal dimension to river width, which determines the gradual transition from advective and highly braiding to diffusive transport regime. Specifically, the tendency to maintain braiding and flow bifurcation is associated with equilibrium average bed profiles and, therefore, equilibrium average stream power characterized by the maximum period that does not exceed transverse channel dimension.

Bathymetric Detection of Fluvial Environments through UASs and Machine Learning Systems

In recent decades, photogrammetric and machine learning technologies have become essential for a better understanding of environmental and anthropic issues. The present work aims to respond one of the most topical problems in environmental photogrammetry, i.e., the automatic classification of dense point clouds using the machine learning (ML) technology for the refraction correction on the fluvial water table. The applied methodology for the acquisition of multiple photogrammetric flights was made through UAV drones, also in RTK configuration, for various locations along the Orco River, sited in Piedmont (Italy) and georeferenced with GNSS—RTK topographic method. The authors considered five topographic fluvial cross-sections to set the correction methodology. The automatic classification in ML has found a valid identification of different patterns (Water, Gravel bars, Vegetation, and Ground classes), in specific hydraulic and geomatic conditions. The obtained results about the automatic classification and refraction reduction led us the definition of a new procedure, with precise conditions of validity.

A model study of the combined effect of above and below ground plant traits on the ecomorphodynamics of gravel bars

Both above- and below-ground plant traits are known to modulate feedbacks between vegetation and river morphodynamic processes. However, how they collectively influence vegetation establishment on gravel bars remains less clear. Here we develop a numerical model that couples above- and below-ground vegetation dynamics with hydromorphological processes. The model dynamically links plant growth rate to water table fluctuations and includes plant mortality by uprooting and burial. We considered a realistic hydrological regime and used the model to simulate the coevolution of alternate gravel bars and vegetation that displays trade-offs in investment of above- and below-ground biomass. We found that a balanced plant growth above- and below-ground facilitates vegetation to establish on steady, stable bars, because it allows plants to develop traits that maximise growth performance during low flow periods and thus survival during floods. Regardless of the growth strategy, vegetation could not establish on migrating bars because of large plant loss by uprooting during floods. These findings add on previous studies suggesting that morphodynamic processes play a key role on determining plant trait distributions and highlight the importance of including the dynamics of both above- and below-ground plant traits for predicting shifts between bare and vegetated states in river bars.