How many measurements are required to construct an accurate sand budget in a large river? Insights from analyses of signal and noise

Paul E. Grams; Daniel D. Buscombe; David J. Topping; Matthew Kaplinski; Joseph Hazel

doi:10.1002/esp.4489

Home/Archive/USGS

USGS · 70213230

How many measurements are required to construct an accurate sand budget in a large river? Insights from analyses of signal and noise

Paul E. Grams·Daniel D. Buscombe·David J. Topping·Matthew Kaplinski·Joseph Hazel

View original source ↗Export metadata ↓

Abstract

Morphological change in river channels is frequently evaluated in the context of mass balance sediment budgets. In a closed sediment budget, measurements of sediment influx and efflux are coupled with measured changes in channel topography to provide both spatial and temporal resolution, and independent estimates of the mass balance. For sediment budgets constructed over long river segments (~10 channel widths or greater) and long periods (~2 years or longer), spatial and temporal accumulation of measurement uncertainty, compounded by inadequate sampling frequency or spatial coverage, may produce indeterminate results. The degree of indeterminacy may be evaluated in the context of a signal-to-noise ratio (SNR), which is a function of the magnitude of the mass balance and the magnitudes of potential systematic uncertainties associated with measurements and incomplete sampling. We report on a closed sand budget consisting of measurements of flux and two morphological surveys for a 50-km segment of a large river over a 3-year period. Accurate reporting of the magnitude and sign of the change in sand storage was only possible by using state-of-the-art techniques with high temporal frequency and large spatial extent. Together, a sand flux and morphological mass balance revealed that sand evacuation was temporally concentrated (~100% of mass change occurred during 19% of the study period) and highly localized (70% of mass change occurred in 12% of the study segment). A SNR analysis revealed that uncertainty resulting from undersampling may approach or exceed that caused by measurement uncertainty and that daily sampling of suspended-sand concentration or repeat mapping of at least 50% of the river segment was required to determine the sand budget with SNR > 1. The approach used here to analyze sand budget uncertainty is especially applicable to other river systems with large temporal variability in sediment transport and large spatial variability in erosion and deposition.

Explore related subjects

Earth Surface Processes and Landforms Article

Source-reported bounding extent: 36.18887535558557° to 36.801587788085314° latitude; -111.90536499023438° to -111.61422729492188° longitude. This indicates report coverage, not an exact sampling location. View area on OpenStreetMap.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Paul E. Grams, Daniel D. Buscombe, David J. Topping, Matthew Kaplinski, Joseph Hazel. 2018-09-24. How many measurements are required to construct an accurate sand budget in a large river? Insights from analyses of signal and noise. https://doi.org/10.1002/esp.4489

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Remote detection of streambank erosion and channel incision using high-resolution, repeat lidar-derived metrics in Raleigh, North Carolina

This study evaluated the use of lidar data to detect streambank erosion and channel incision throughout Raleigh, North Carolina. The study included five components. (1) Generation of lidar-derived rasters, including a digital elevation model of difference that represented channel erosion and positive openness rasters that represented channel incision. (2) Completion of site-based field assessments throughout the study area. (3) Spatial analysis of raster datasets at three spatial scales: erosion patches, channel segments, and subbasins. (4) Comparison of field assessment results with lidar-derived metrics. (5) Pairing of landscape attributes with lidar-derived metrics to evaluate potential causative factors associated with variations in channel incision. Field assessment results indicated 74% of surveyed sites had Bank Erosion Hazard Index ratings ranging from ‘high’ to ‘extreme’ and 77% had ‘high’ or ‘severe’ degrees of channel incision. Select field measures showed agreement with lidar-derived metrics. Lidar-derived erosion patches, ranging in size from 20 to 948 m 2 , were detected in 24.3% of channel segments, and patch density increased with Strahler stream order. Channel segments, ranging from 50 to 150 m, had inconsistent spatial patterns of erosion and incision. Erosion in channel segments increased with Strahler stream order and 36% of channel segments became more incised during the study period. Lidar-derived metrics across subbasins revealed variability of erosion and incision throughout the study area. The assessment of landscape drivers could not explain variability at the channel-segment scale, but subbasins with fewer forests tended to have greater erosion and incision. However, unexplained variability remained, demonstrating the complexity of stream stability in urban watersheds. As lidar datasets become more widespread, pairing remote data with field assessments may be a viable approach to characterize and understand stream geomorphology in urban settings.

North Carolina↗

USGS2026

Deep learning models for mapping surficial geology in selected physiographic regions of New York

Surficial geologic mapping is required for decisions including infrastructure and water-resource development and management. Deep learning, a type of machine learning that uses training data to self-learn and perform tasks, is explored as a tool to help increase efficiency of the labour- and time-intensive mapping process. Deep learning models were trained, and their potential to aid in mapping surficial geology was explored for two physiographic regions in New York: the high-relief Allegheny Plateau and the low-relief Erie-Ontario Lowlands. Key to the development of deep learning models is the availability of highly detailed surficial geologic maps in each of these regions from which the models can learn. Through exploring different groupings of surficial deposits and their corresponding spatial data, deep learning models were iteratively developed to reproduce published mapping for greater than 79% of the training areas, and with similar accuracy in test areas within the same physiographic region. These models were trained using only two inputs: high-resolution lidar data and previously published surficial geologic maps. This straightforward approach is intended to make the methods and models created easily reproducible using widely accessible datasets. This research describes the strengths and limitations of lidar-derived surficial geologic models and how broadly grouping surficial types, characteristics of different physiographic regions, and testing models in physiographic regions outside of their training areas affect model creation and performance. In addition, the models created could be used as a tool for mapping surficial geology in similar physiographic regions, in addition to establishing a framework for creating similar models elsewhere.

New York↗

USGS2026

Regional models for postfire debris-flow likelihood and rainfall thresholds across the western United States

The U.S. Geological Survey (USGS) uses an empirical model developed with logistic regression (the ‘M1’ model) to rapidly assess debris-flow likelihood and to identify quantitative rainfall thresholds for debris flows after wildfire in the western United States. The M1 model was calibrated to a debris-flow inventory from southern California (United States) and has been applied throughout the western United States. Limited spatial coverage in the calibration dataset has motivated evaluation of M1 model accuracy outside the calibration region (e.g., the Sierra Nevada or the eastern Cascade Range, United States). Previous test cases showed that M1 overpredicts debris-flow likelihood and underpredicts rainfall thresholds for some locations (e.g., Arizona, northern California, Colorado, New Mexico, United States). We sought to improve the regional applicability of a debris-flow likelihood model by expanding the debris-flow inventory used for calibration, testing multiple potential models and generating an updated model framework. The updated inventory includes 3788 observations from 67 burned areas paired with short duration rainfall ratios. The updated model framework consists of a modified model structure and sets of coefficients calibrated separately to the entire updated inventory and to subsets of the inventory that intersect three Environmental Protection Agency (EPA) Level 2 ecoregions (Mediterranean California, Upper Gila Mountains and Western Cordillera). Comparisons of predictions from the updated models with observed rainfall and debris-flow activity show that the updated models outperform the M1 model by ~15%–60% and improve the uniformity of predictive performance across the western United States. The updated models also reduce false positive rates relative to M1 and generate rainfall thresholds that are better aligned with relative differences in regional climatology and debris-flow activity.

Arizona, California, Colorado, Idaho, Montana, Nev↗