Widely adopted models for estimating channel geometry attributes rely on simplistic power-law (hydraulic geometry) equations. This study presents a new generation of channel geometry models based on a hybrid approach combining traditional statistical methods (Multi-Linear Regression (MLR)) and advanced tree-based Machine Learning (ML) algorithms (Random Forest Regression (RFR) and eXtreme Gradient Boosting Regression (XGBR)), utilizing novel datasets. To achieve this, a new preprocessing method was applied to refine an extensive observational dataset, namely the HYDRoacoustic dataset supporting Surface Water Oceanographic Topography (HYDRoSWOT). This process improved data quality and identified observations representing bankfull and mean-flow conditions. A compiled dataset, combining the preprocessed dataset with datasets containing additional catchment attributes like the National Hydrography Dataset Plus (NHDplusv2.1), was then used to train a suite of models to predict channel width and depth under bankfull and mean-flow conditions. The analysis shows that tree-based ML algorithms outperform traditional statistical methods in accuracy and handling the data but face limitations in prediction capabilities for streams with characteristics outside the training range. Consequently, a hybrid method was selected, combining XGBR for streams within the dataset range and MLR for those outside it. Two tiers of models were developed for each attribute using discharges derived from distinct sources (HYDRoSWOT and NHDPlusV2.1, respectively), where the second tier of models offers applicability across approximately 2.6 million streams within NHDplusv2.1. Comprehensive independent evaluations are conducted to assess the capability of the developed models in providing stream/reach-averaged (rather than at-a-station) predictions for locations outside the training and testing datasets.

Bedload flux is notoriously challenging to measure and model. The dynamics of bedload therefore remains largely unknown in most fluvial systems worldwide. We present a global scale bedload flux model as part of the WBMsed modeling framework. Our results show that the model can very well predict the distribution of water discharge and suspended sediment and well predict bedload. We analyze the model’s bedload predictions sensitivity to river slope, particle size, discharge, river width and suspended sediment. We found that the model is most responsive to spatial dynamics in river discharge and slope. We analyze the relationship between bedload and total sediment flux globally and in representative longitudinal river profiles (Amazon, Mississippi, and Lena Rivers). We show that while, as expected, the proportion of bedload is decreasing from headwater to the coasts, there is considerable variability between basins and along river corridors. The latter is largely responsive to changes in suspended sediment and river slope due to dams and reservoirs. We provide a new estimate of water and sediment fluxes to global oceans from 2,067 largest river outlets (draining 67% of the global continental mass). Estimated water discharge (30,579 km3/y) corresponds well to past estimates however sediment flux is considerably higher. Of the total 22 Gt/y estimated average sediment flux to global oceans, 19 Gt/y is transported as washload, 1 Gt/y as bedload, and 2 Gt/y as suspended bed material. The largest 25 rivers are predicted to transport over 55% of total sediment flux to global oceans.

Sediment trapping behind dams is currently a major source of bias in large-scale hydro-geomorphic models, hindering robust analyses of anthropogenic influences on sediment fluxes in freshwater and coastal systems. This study focuses on developing a new reservoir trapping efficiency (Te) parameter to account for the impacts of dams in hydrological models. This goal was achieved by harnessing a novel remote sensing data product which offers high-resolution and spatially continuous maps of suspended sediment concentration across the Contiguous United States (CONUS). Validation of remote sensing-derived surface sediment fluxes against USGS depth-averaged sediment fluxes showed that this remote sensing dataset can be used to calculate Te with high accuracy (R2 = 0.98). Te calculated for 116 dams across the CONUS, using upstream and downstream sediment fluxes from their reservoirs, range from 0.3% to 98% with a mean of 43%. Contrary to the previous understanding that large reservoirs have larger Te and vice versa, these data reveal that large reservoirs can have a wide range of Te values. A suite of 21 explanatory variables were used to develop an empirical Te model using multiple regression. The strongest model predicts Te using five variables: dam height, incoming sediment flux, outgoing water discharge, reservoir length, and Aridity Index. A global model was also developed using explanatory variables obtained from a global dam database to conduct a global-scale analysis of Te. These CONUS- and global-scale Te models can be integrated into hydro-geomorphic models to more accurately predict river sediment transport by representing sediment trapping in reservoirs.