Predicting river system dynamics using machine learning: a multivariable approach to assess the interactions and impacts of hydrological variables on suspended sediment transport

Salinity and suspended sediment transport in a shallow estuary on the east coast of India

River flows and associated suspended sediment (SS) transport are intermittent processes possessing fluctuations over a large range of time scales and space, making it challenging to develop predictive models that are applicable across timescales and rivers. A concept of “effective timescales of connectivity” has been used to define the timeframe over which sediment (dis)connectivity occurs, whereby parts of the catchment are “switched on and off” as a response of events with varying frequency-magnitude relationships and antecedent soil moisture. These concepts provide excellent frameworks to understand temporal variability and identify relevant timescales for sediment transport, but do not help in the knowledge of mechanisms for temporal variability in SS transport. The complexity and scale dependency of processes driving SS transport stress the need to detect how sediment generation, storage, and transport are linked across different timescales. Furthermore, the mechanisms that produce travel time distributions over many orders of magnitude are not known precisely. To this end, in this study we have considered SS transport as a fractal system. By approaching SS transport dynamics as a fractal system, it is assumed that patterns of variation in SS transport exist over different timescales, while linkages across those temporal scales are expressed as fractal power-laws.This work aims to defines the link between (i) sediment transport and deposition and (ii) fractal geometry and fractal storage time distributions in streams.Here, we present case study where fractals are used to describe and predict patterns over different spatial or temporal scales of dynamics in SSCs. We have considered in these studies the statistics and the dynamics of streamflow, SSCs and associated grain size distribution at event based by considering respectively their probability distribution function and Fourier power spectra.We set up a natural experiment site of a first-order mixed bedrock and alluvial stream channel by using LISST instrument coupling with LIDAR remote sensing measurement. Here we obtain high-resolution observations of streambed topography and continuously long-term measurements of suspended sediment in natural experimental site located in an agricultural watershed of a Chianti area (Florence, Italy).The LISST is a submersible laser diffraction particle size analyzer for measuring suspended particle size (range from 2.50 µm to 500 µm), its volume concentration at different time step and depth. We set up at time interval equal to 5 minutes of sample rate.Preliminary results obtained indicate large fluctuations with heavy tails, and long-range properties, characterized by extreme events much more frequent than what is found for a Gaussian process.Hence, insights into the degree of fractal power of a SS transport system may provide a useful basis to evaluate and develop the most appropriate predictive models and management strategies.

Suspended sediment plays a crucial role in shaping stream morphology and maintaining ecological balance, yet the main controls on its sources and dynamics in forested mountain catchments are still poorly documented. In this study we aimed at assessing suspended sediment spatio-temporal sources and transport dynamics in a Mediterranean mountain catchment. A relevant role might be played by large wood debris in suspended sediment retention and release: large wood structures can significantly influence sediment dynamics by trapping sediments and creating stable habitats for aquatic organisms. This aspect is particularly relevant in forested mountain streams where wood accumulation can alter flow patterns and sediment transport mechanisms.The experimental activities were carried out in the densely forested Re della Pietra catchment located in Tuscany, Central Italy.To assess suspended sediment spatio-temporal dynamics, field measurements were conducted since December 2024, including monitoring of turbidity at the catchment outlet using a high-definition turbidimeter, stream stage measurements, soil moisture measurements at two depths, and the main meteorological variables.Preliminary results show a significant correlation between turbidity, rainfall intensity and stage variation, suggesting that rainfall intensity is crucial in suspended sediment release and transport patterns. Notably, pronounced turbidity peaks were observed during moderate to intense storm events occurred during the wet season but did not correlated to meteorological variables. The analysis of the hysteresis loops between turbidity and stream stage (as a proxy of discharge) reported that the 15% of the loops were clockwise, suggesting that suspended sediment primarily originates from local sources, mostly during the wet season.The study highlights the relationships between suspended sediment transport, large wood debris, and hydrological variables, emphasizing the need for further investigation of the factors affecting suspended sediment transport in forested mountain environments. The determination of flow rating curve is in progress, and future analysis will consider suspended sediment concentration and discharge data. The large wood impact will be studied through the visual analysis of the photographic documentation produced by cameras located at the catchment outlet.

Water erosion is a current issue, especially in hilly and areas, where driving force such as surface runoff and subsurface flow can mobilize large amounts of sediment to rivers. In fact, how and at which timescale, seasonality precipitation is turned into runoff or streamflow (Q) it is difficult to be predicted without calibrating site-specific models. The potential soil erosion can be assessed through the study of the relationships between sediment sources and sinks in a watershed (i.e., sediment connectivity assessment) and associated suspended sediment (SS) transport in rivers. On the other hand, sediment connectivity, defined as structural (from a geomorphological point of view) and functional connectivity (considering forcing processes), can be evaluated by the using of specific indexes (e.g., Index of connectivity – IC).  SS transport processes are intermittent processes fluctuating over a large range of temporal and spatial scales, making it challenging to develop predictive models applicable across timescales and rivers. While temporal variability in sediment transport is explained by the concept of “effective timescale of connectivity”; the mechanism behind this variability remains unknown. Here we used a data-driven approach considering two years of monitoring Q and SS to develop and demonstrate a proof of concept for automating the classification of event-based sediment dynamics by using a machine learning approach.  For each storm event we i) calculate the sediment connectivity (extreme rainfall events also are considered) and ii) define the link between sediment transport and deposition by considering SS transport as a fractal system (i.e. fractal storage time distributions in streams). Fractals are here used to describe and predict patterns over different temporal scales of dynamics in SS   The statistic and dynamics of Q, SSCs and associated grain size distribution, at event based, were considered by assessing their probability distribution function, Fourier power spectra, and the machine-learning classification of hysteresis index. Indeed, by approaching SS transport dynamics as a fractal system, it is assumed that patterns of variation in SS transport exist over different timescales, while linkages across those temporal scales are expressed as fractal power-laws. The study site, located near Florence in the Chianti area, is a 1 Km2 agricultural watershed with different types of land cover and characterize by a first-order mixed bedrock and alluvial stream channels. The area was mapped at high resolution with a Drone LIDAR scanner and equipped with a submersible laser diffraction particle size analyser (LISST) for long-term measuring suspended particle size and its volume concentration. Preliminary results showed a robust correlation between sediment connectivity, land cover, and sediment connectivity. Q-SS information flows exhibit seasonally varying behaviour consistent with dominant runoff generation mechanisms (catchment connectivity in wet to dry season). However, the timing and the magnitude of runoff also reflect considerable catchment heterogeneity, likely attributable to differences in baseflow contributions from different lithologies, and variation in of preferential flow paths (land use/land cover).  In conclusion, this study allowed to analyse a small catchment area in term of sediment connectivity and related sediment transport to identify potential areas of (dis)connectivity in the basin.

Impact of cold water mass on suspended sediment transport in the South Yellow Sea

Catchment properties controlling suspended sediment transport in wind-water erosion crisscross region

The relative proportion of bedload and suspended sediment transport in rivers features very large variations both in space (within and among rivers) and in time (ordinary vs infrequent floods). However, the current knowledge about the temporal variability in bedload/suspended proportion is limited, as well as of its controlling factors (e.g., water runoff, sediment supply). The present work investigates the partitioning of sediment load into bedload and suspended fractions in the glacier-fed Sulden/Solda River (eastern Italian Alps, drainage area 130 km2), where a monitoring station for water and sediment transport has been operating since 2014.From 2014 to 2020 the station was equipped with one turbidimeter and 8 geophone plates for monitoring suspended sediment and bedload transport, respectively. Calibration curves for deriving suspended sediment concentrations were derived based on 474 water samples, whereas to convert the geophones signal to bedload mass a portable trap mounted on a crane was used (76 samples collected). Two meteorological stations located at about 2800 m and 1900 m a.s.l. recorded precipitation and air temperature within the catchment.A 10-minute interval dataset was established, including suspended load, bedload, water discharge, precipitation, and air temperature, measured from May to October 2014-2020 (2018 data had to be excluded for technical problems of the turbidimeter). Hourly intervals characterised by data gaps regarding geophone plates or turbidimeter were excluded from the analysis, and a total of 18,549 hourly data were analysed. The overall range of water discharge observed in the study period was 0.7 – 80.8 m3/s, but reliable suspended sediment data are available only up to about 40 m3/s (RI about 2 yr). The mean annual discharge of the Sulden River was 6.3 m3/s, while the mean discharge from May to October equalled 10.0 m3/s.  Results show that suspended/bedload partitioning varied with water discharge and time of the year (month) in a complex way. On average, the relative contribution of suspended sediment transport was around 89% for Q<10 m3/s, 96% with 10<Q<20 m3/s, and 97% for Q>20 m3/s. At the lower water discharges (Q<10 m3/s), the suspended sediment fraction diminished over the summer from May/June/July (about 94%) to August/September (88%), and reached its minimum value in October (79%). This trend is possibly due to the intense glacier ablation occurring in August, which increased the coarse sediment supply – transported as bedload in the channels – more markedly than the finer fractions carried in suspension.At higher flow rates (Q>10 m3/s), the percentage of sediments transported in suspension shows an opposite temporal trend, increasing from May (80%) to October (97%). Such a remarkable difference compared to lower discharges may be due to the strong increase in suspended sediment concentration at higher water discharges multiplied by larger water volumes carrying suspended transport through the entire flow depth, differently from bedload. Such complex dynamics is consistent with previous results in the Sulden River, where a low effective discharge for bedload – driven by the supply of coarse sediments during the glacier melt period – was observed for the same period analysed here.

Detecting the drivers of suspended sediment transport in an intermittent river: An event-based analysis

— Muti-temporal images of total suspended matter (TSM) derived from MODIS (Moderate Resolution Imaging Spectrometer) Terra 250 m red-band reflectance data were used to set the initial conditions and validate output of a hydrodynamic and sediment transport model ( ECOMSED) developed for Lake Pontchartrain, LA USA. During this initial study, the moderately high spatial resolution of the 250 m images provided a significant improvement in model definition when compared to model runs calibrated using 1 km data available from most ocean viewing instruments (e.g., AVHRR and SeaWiFS). The combination of daily MODIS imagery and model simulations offer a more robust monitoring and prediction system of suspended sediments than available from either system alone. The additional use of Aqua MODIS imagery should afford greater refinement in model paramerization and accuracy. Keywords-suspended sediments; sediment transport; MODIS; ECOMSED; water quality; monitoring. I. I NTRODUCTION Coastal waters, especially bays and estuaries, are often characterized by high concentrations of suspended materials derived from freshwater inflow and/or by the resuspension of bottom sediments. High concentrations of suspended sediments can govern many important water column parameters and processes including total water quality, benthic and phytoplankton productivity, and the redistribution and transport of water-borne pathogens, pollutants, and other materials. For example, several investigators [1, 2] have reported that the distribution and concentration of fecal coliform are directly associated with sediment resuspension and suspended sediment transport. These findings establish a direct link between sediment dynamics and public health. Hence, there is considerable interest in monitoring the transport and fate of suspended sediments from a broad range of investigators, environmental managers, and policy makers. Unfortunately, in many coastal aquatic systems, the use of traditional field measurements is inadequate to effectively monitor suspended sediments at the desired spatial and temporal scales due to the presence of dynamic physical processes (e.g., wind driven resuspension, river flow) or the large size of the system. In contrast, observations from remote sensing instruments and numerical models can provide unique information regarding the distribution and transport of suspended sediments that cannot be obtained using traditional field sampling techniques. Remote sensing from airborne and space-based instruments provide a large-scale synoptic view; numerical models provide a better understanding of the underlying system processes as 0-7803-9119-5/05/$20.00 (C) 2005 IEEE

In this study, the effects of cascade sediment-storage dams (SSDs) on the suspended sediment (SS) transport in the Sera Lake Watershed (northeast Turkey) were investigated. Ten SS monitoring stations from upstream (S01) to downstream (S10) were selected for the study considering the locations of the SSDs and Sera Lake. In a gauging station, where S05 was selected, the water stage monitoring studies were also conducted, and daily discharges and SS loads were computed. The Spearman correlation coefficient between SS and discharges data was computed as 0.592. The deterioration of the naturalness of the watershed with external interventions was effective on this value. While the Sera Lake trapped 33% of the SS transport, this value varied between –25.8 and 44% for SSDs. Although the SSDs have been proportionately succeed in trapping the SS, they have been insufficient in preventing sedimentation and problems caused by sediment transport in the watershed. In addition, the annual average soil erosion amount was determined by the revised universal soil loss equation (RUSLE) method in the watershed. According to the RUSLE method, the annual average soil loss of the watershed is 1.65 tons/ha. The SDR value of the watershed was calculated as 0.104. Integrating different environmental management tools to achieve a balanced and sustainable decision, a strategic environmental management plan should be constituted.

Tidal asymmetry in sediment resuspension on a macrotidal beach in northwestern Australia

The upper Driva drainage basin system in central Norway (Oppdal) is situated in a cold climate and mountainous environment and ranges with a total drainage basin area of 1630 km2 from 220 to 2286 m a.s.l. The mean annual air temperature at Oppdal (545 m a.s.l.) is 4.3°C, and mean annual precipitation at Oppdal amounts to 532 mm. The lithology in the drainage basin is complex and varied, and is clearly dominated by metamorphic rocks (mostly gneisses and schists). Vegetation cover varies between tundra vegetation in the high and rather flat areas of the uppermost drainage basin area, situated at elevations around 900-1200 m a.s.l., tree vegetation (mostly birch and pine) in the lower parts of the incised tributary valleys of the Driva main river and grasslands in the agriculturally used areas close to the lower sections of the main river Driva. Relevant geomorphological processes include chemical and mechanical weathering, rockfalls, snow avalanches, debris flows, slides, wash processes, fluvial erosion, fluvial streambank erosion and down-cutting, and fluvial solute, suspended sediment and bedload transport.This ongoing GFL research on environmental drivers, quantitative rates and future trends of chemical and mechanical denudation includes detailed field and remotely sensed geomorphological mapping, permafrost mapping, and computing of morphometric catchment parameters. This work is combined with the detailed statistical analysis of high-resolution meteorological and ground temperature data, and the continuous observation and year-round monitoring of sediment transfers, runoff and fluvial solute and sediment transport using a range of different techniques. Specific focus is on six selected tributary systems (Svone, Kaldvella, Stølåa, Tronda, Vinstra, Ålma) of the upper Driva drainage basin system. Stationary hydrological stations are monitoring continuously and year-round runoff, fluvial solute and suspended sediment transport. The analysis of fluvial bedload transport includes the application of different tracer techniques together with underwater video filming and Helley Smith and impact sensor measurements.Discharge in the upper Driva drainage basin occurs year-round with a nival runoff regime and a mean annual runoff of 576 mm. The temporal variability of sediment transfers, runoff and fluvial transport are largely controlled by thermally and/or pluvially determined events. The selected tributary systems display varying solute and sediment yields which are explained by different lithologies, valley morphometries and sediment availabilities. The activation of sediment sources and mechanical denudation are strongly determined by thermally and/or pluvially induced events. The highest share of annual sediment transport occurs during the snowmelt period in spring. Altogether, drainage-basin wide chemical denudation dominates over drainage-basin wide mechanical fluvial denudation. It is expected that global warming and the connected shifts in the ratio of snow and rain, the increased frequency of heavy rainfall events, and the continued thawing of permafrost will have complex effects on denudation, with an increasing importance of pluvially induced denudational events, a decreasing importance of snowmelt induced denudation processes, and an increasing dominance of chemical denudation over mechanical denudation.

<p>The upper Driva drainage basin in central Norway (Oppdal) is situated in a cold climate and mountainous environment and ranges with a total drainage basin area of 1630 km<sup>2</sup> from 220 to 2286 m a.s.l. The mean annual air temperature at Oppdal (545 m a.s.l.) is 4.3°C, and mean annual precipitation at Oppdal amounts to 532 mm. The lithology in the drainage basin is complex and varied, and is clearly dominated by metamorphic rocks (mostly gneisses and shists). Vegetation cover varies between tundra vegetation in the high and rather flat areas of the uppermost drainage basin area, situated at elevations around 900-1200 m a.s.l., tree vegetation (mostly birch and pine) in the lower parts of the incised tributary valleys of the Driva main river and grasslands in the agriculturally used areas close to the lower sections of the main river Driva. Relevant geomorphological processes include chemical and mechanical weathering, rockfalls, snow avalanches, debris flows, slides, wash processes, fluvial erosion, fluvial streambank erosion and down-cutting,  and fluvial solute, suspended sediment and bedload transport.</p><p>This ongoing GFL research on controls and spatiotemporal variability of contemporary chemical and mechanical denudation includes detailed field and remotely sensed geomorphological mapping, permafrost mapping, and computing of morphometric catchment parameters combined with the detailed statistical analysis of high-resolution meteorological and ground temperature data, and the continuous observation and year-round monitoring of sediment transfers, runoff and fluvial solute and sediment transport using a range of different techniques. Specific focus is on six selected tributary systems (Svone, Kaldvella, Stølåa, Tronda, Vinstra, Ålma) of the upper Driva drainage basin system. Stationary hydrological stations are monitoring continuously and year-round runoff, fluvial solute and suspended sediment transport. The analysis of fluvial bedload transport includes the application of different tracer techniques together with underwater video filming and Helley Smith and impact sensor measurements. Discharge in the upper Driva drainage basin occurs year-round with a nival runoff regime and a mean annual runoff of 576 mm. The temporal variability of sediment transfers, runoff and fluvial transport are largely controlled by thermally and/or pluvially determined events. The selected tributary systems display varying solute and sediment yields which is explained by different lithologies, valley morphometries and sediment availabilities. The activation of sediment sources and mechanical denudation are strongly determined by thermally and/or pluvially induced events. The highest share of annual sediment transport occurs during the snowmelt period in spring. Altogether, drainage-basin wide chemical denudation dominates over drainage-basin wide mechanical fluvial denudation.</p>

High‐frequency water discharge and suspended sediment concentration (SSC) databases were collected for 3 years on four contrasted watersheds: the Asse and the Bléone (two Mediterranean rainfall regime watersheds) and the Romanche and the Ferrand (two rainfall–snowmelt regime watersheds). SSCs were calculated from turbidity recordings (1‐h time step), converted into SSC values. The rating curve was calculated by means of simultaneous SSC measurement taken by water sampling and turbidity recording. Violent storms during springtime and autumn were responsible for suspended sediment transport on the Asse and the Bléone rivers. On the Ferrand and the Romanche, a large share of suspended sediment transport was also caused by local storms, but 30% of annual fluxes results from snowmelt or icemelt which occurred from April to October. On each watershed, SSC up to 50 g l−1 were observed. Annual specific fluxes ranged from 450 to 800 t km−2 year−1 and 40–80% of annual suspended sediment fluxes occurred within 2% of the time. These general indicators clearly demonstrate the intensity of suspended sediment transport on these types of watersheds. Suspended sediment fluxes proved to be highly variable at the annual scale (inter‐annual variability of specific fluxes) as well as at the event scale (through a hysteresis loop in the SSC/Q relationship) on these watersheds. In both cases, water discharge and precipitations were the main processes involved in suspended sediment production and transport. The temporal and spatial variability of hydro‐meteorological processes on the watershed provides a better understanding of suspended sediment dynamics. Copyright © 2009 John Wiley & Sons, Ltd.

1. The planned restoration of the lowermost 18 km of the Skjern river system (catchment area 2490 km2) through re-meandering the river to its former course and the creation of a shallow lake and ponds is the largest river restoration project in Europe. An important aspect of the project planning and design has been to measure suspended sediment (SS) and total phosphorus (TP) transport in the project area, and to assess the inter-annual variation. 2. SS and TP concentrations were measured continuously (every fourth hour) from 1993 to 1995 in the River Skjern and its main tributary, the River Omme, using automatic sampling equipment (ISCO). In addition, discrete samples were collected monthly in the remaining five smaller tributaries. Estimated SS transport in the Skjern river system in 1994 and 1995 determined on the basis of continuous sampling was approximately 60% greater than that determined on the basis of discrete sampling. Empirical models for SS and TP transport were developed based on the data collected in this study and applied to a 31-year time series of daily discharge values. Mean annual transport amounted to 12 220 t SS and 100 t TP corresponding to 5 t SS km−2 yr−1 and 41 kg TP km−2 yr−1, respectively. 3. Assessment of the effects of the planned restoration project, based on measured transport and estimated SS and TP retention rates for different areas of the lower river system, revealed that SS and TP transport in the river will be reduced by 37% and 20%, respectively. Restoration will therefore considerably enhance the natural self-purification capacity of the river system. In addition, restoration will reduce nitrogen and ochre loading of Ringkjøbing Fjord, thereby improving environmental conditions, and re-meandering will improve habitat quality and diversity in the river system. The study stresses the importance of considering streams and riparian areas as an entity when evaluating the effects of restoration activities. ©1997 John Wiley & Sons, Ltd.