Geology topics

Scott Douglas Hamshaw

Publications and source records attributed to Scott Douglas Hamshaw.

12 recordsLinked to original sources

USGS2026

Predicting streamflow drought in the conterminous United States using machine learning and a donor-gage approach, 1982–2020

Drought is a highly consequential natural disaster that may likely increase in both severity and extent across the conterminous United States (CONUS). The mechanisms affecting the propagation of drought from the atmosphere to streamflow are complex and interactive, making the prediction of streamflow drought difficult with current modeling approaches. Machine learning is an emerging tool in the field of hydrology that may be well-suited to prediction of streamflow drought across large and topographically diverse areas. Here, we train and analyze 3198 random forest models at U.S. Geological Survey streamgages to understand common meteorological drivers of streamflow drought and to define physiographic characteristics of basins sensitive to these drivers. We also develop a novel dynamic regionalization approach using donor gages to predict daily streamflow drought at pseudo-ungaged locations. Our results show that teleconnections, temperature, evaporative demand, and snow–water equivalent are important drivers of streamflow drought in the West, Southwest, and Northern Rocky Mountains (Northern Rockies) regions of the United States, and precipitation and soil moisture are primary drivers of streamflow drought in the Northeast, Southeast, and the Northwest regions. Prediction using dynamic regionalization shows comparable performance to at-site models. This method may be applicable to other hydrologic prediction problems requiring transferability to ungaged locations.

conterminous United States↗

USGS2026

A roadmap for identifying and interpreting physical processes and national water model prediction bias associated with baseflow index regimes across the contiguous United States

Understanding how groundwater–surface water interactions shape streamflow variability is critical for diagnosing low flow behavior and prediction bias in continental scale hydrologic models. We present a process informed framework that links observed baseflow (BF) dynamics, watershed attributes, and National Water Model (NWM) performance across the contiguous United States. Using daily observed streamflow from 797 reference quality streamgages, we developed monthly baseflow index (BFI) signatures using a streamgage specific, calibrated digital filter. Hierarchical clustering of these signatures identified seven distinct BFI regimes capturing regional and seasonal variability. We evaluated NWM v3.0 retrospective streamflow performance within each regime using multiple hydrograph and flow duration curve-based metrics. Model skill varied systematically across regimes: mixed flow systems were simulated most accurately, while predominantly BF dominated and quickflow dominated regimes exhibited substantially poorer performance. Across nearly all regimes, the NWM underestimated observed BFI magnitude and frequently failed to reproduce seasonal BF patterns, indicating systematic biases in simulated low flow contributions. To relate these regimes to potential process controls, we trained a Random Forest classifier using static watershed attributes and applied Shapley Additive Explanations to identify features most strongly associated with each regime. Results highlight regionally varying influences, including the dominant role of snow fraction and seasonal runoff timing in snow dominated basins and the importance of evapotranspiration and aridity in quickflow dominated systems. Collectively, these findings demonstrate how hydrologic signatures combined with interpretable machine learning can diagnose regime specific model biases and generate process-based hypotheses about limitations in large scale hydrologic prediction systems.

contiguous United States↗

USGS2026

Benchmark dataset of historical annual peak floods classified by causal mechanisms for select US river basins

Considering the causal mechanisms of floods can improve estimates of flood recurrence intervals given that certain flood types can be associated with higher magnitude and more damaging floods. However, few verified datasets of flood types are available to validate the semiautomated and automated classification algorithms needed to apply flood-typing across large hydrologically diverse regions. To address this gap, a benchmark dataset of manually classified flood types was compiled for 1,763 annual maximum flood peaks from 18 stream gauges in six different river basins across the conterminous United States from 1851 to 2022. Within each basin, three representative stream gauges were selected for manual flood typing. A flexible classification framework is introduced that facilitates flood typing across hydrologically diverse regions and accommodates unique combinations of weather and antecedent watershed conditions specific to each region. Floods were manually typed by domain experts using multiple lines of evidence to identify a primary surface water input of each flood (rainfall, snowmelt, or both) and, if relevant, associated storm type and secondary causal mechanisms characterizing antecedent watershed conditions. Across all the study basins, 49% of historical annual maximum flood peaks were attributable to rainfall, 28% to snowmelt, 22% to mixed precipitation, and 1% could not be assigned to a mechanism due to missing or incomplete data. The proposed flood-typing schema supports varying levels of flood typing specificity required for mixed population flood-frequency analysis, flood-type-specific design hydrographs, water quality response studies, and additional applications. This detailed, manually determined benchmark dataset serves as a resource that can be used developing and validating automated or machine learning-based algorithms capable of operationalizing expanded flood peak information.

conterminous United States↗

USGS2026

Machine learning generated streamflow drought forecasts for the conterminous United States (CONUS): developing and evaluating an operational tool to enhance sub-seasonal to seasonal streamflow drought early warning for gaged locations

Forecasts of streamflow drought, when streamflow declines below typical levels, are notably less available than for floods or meteorological drought, despite widespread impacts. We apply machine learning (ML) models to forecast streamflow drought 1–13 weeks ahead at 3,219 streamgages across the conterminous United States. We applied two ML methods (Long short-term memory neural networks; Light Gradient-Boosting Machine) and two benchmark models (persistence; Autoregressive Integrated Moving Average) to predict weekly streamflow percentiles with independent models for each forecast horizon. ML models outperformed benchmarks in predicting continuous streamflow percentiles below 30%. ML models generally performed worse than persistence models for discrete classification (moderate, severe, extreme) but exceeded the benchmark models for drought onset/termination. Performance was better for less intense droughts and shorter horizons, with predictive power for 1–4 weeks for severe droughts (10% threshold). This work highlights challenges and opportunities to advance hydrological drought forecasting and supports a new experimental forecasting tool.

conterminous United States↗

USGS2025

Disentangling geomorphic equifinality in sediment and hydrologic connectivity through the analyses of landscape drivers of hysteresis

Sources, transport mechanisms and pathways of fine sediment in river systems are dependent on a multitude of climatic, geomorphic and anthropogenic factors, resulting in geomorphic equifinality, in which it is difficult to parse how different landscape processes affect sediment transport across different spatiotemporal scales. The objectives of this study are to 1) provide a conceptual model to consider how differing spatial distributions and hydrologic timing of sediment sources, both upland and in-channel, can result in equifinal sediment transport outcomes, and 2) utilize analytical methods with widely available environmental datasets to infer sediment processes from stream gaging data. Hysteretic patterns of observed storm events were classified based on their direction and timing of peak sediment concentration, relative to streamflow, using records from 35 U.S. Geological Survey stream gages in the period between 2007 and 2023 within two different physiographic regions: the Mid-Atlantic Delaware River Basin (DRB) and the Midwestern Illinois River Basin (IRB). The DRB contains mixed forest, urban, suburban and agricultural watersheds over diverse topography, and the IRB is primarily an intensively managed agricultural watershed on flat terrain. We use principal component analysis and linear discriminant analysis to infer regional hydrologic relations with turbidity dynamics, and to identify the primary hydrologic and land surface characteristics most effective at distinguishing between hysteretic classes in each region. These analyses reveal underlying regional relations in storm event hydrodynamics and landscape characteristics that contribute to varying patterns in sediment dynamics. Incorporating these sediment dynamic relations with spatial distributions and hydrologic timing of sediment sources could help to improve process understanding and predictive capability of fine sediment transport in watersheds.

Delaware, Illinois, Indiana, New Jersey, New York,↗

USGS2025

A spatiotemporal interrogation of hydrologic drought model performance for machine learning model interpretability

The predictive accuracy of regional hydrologic models often varies across both time and space. Interpreting relationships between watershed characteristics, hydrologic regimes, and model performance can reveal potential areas for model improvement. In this study, we use machine learning to assess model performance of a regional hydrologic model to forecast the occurrence of streamflow drought. We demonstrate our methodology using a regional long short-term memory (LSTM) deep learning model developed by the U.S. Geological Survey (USGS) and data from 384 streamgages across the Colorado River Basin region. Performance was assessed by clustering catchments using: (a) physical and climatological catchment attributes, and (b) streamflow drought signatures time series. We examined the association of USGS LSTM model error measures with clusters generated by both approaches to interpret meaningful spatial and temporal information about LSTM model performance. Clustering static catchment attributes identified elevation, degree of streamflow regulation, baseflow contribution, catchment aridity, and drainage area as the most influential attributes to model performance. Clustering gages by their drought signatures revealed that catchments with significant seasonal peak runoff between January and June generally exhibited better model performance. Additionally, a Random Forest classifier was trained to successfully predict LSTM model performance (F1 score of 0.72) based on physical and climatological catchment attributes. Low degree of flow regulation was identified as a key indicator of better LSTM model performance. These findings point to the opportunities for improving the USGS LSTM model performance in future hydrologic drought prediction efforts across regional and CONUS scales.

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

USGS2025

Assessing and enhancing national water model streamflow predictions for montane catchments in the northeastern United States

This study evaluates National Water Model (NWM) performance in low-order montane catchments across the northeastern United States by comparing retrospective simulations to measured observations. To address deficiencies, we develop a machine learning (ML) correction model for selected sites using LightGBM, a different approach from conventional bias correction methods. Montane, low-order streams play a crucial role in water quality and flood generation but pose challenges for streamflow prediction and are under-represented in the national streamgaging network. NWM provides streamflow forecasts across the United States; yet a focused assessment of its performance in these settings has not been comprehensively undertaken. Results indicate NWM performance varied seasonally, with the best performance during the fall and particularly poor performance during snowmelt, spring runoff, and high flow events, with a tendency towards flow underestimation. The ML correction model markedly improved hourly streamflow prediction accuracy based on continuous time series and runoff event-based metrics. Including antecedent water level measurements as input, even from distant sites, greatly improved model performance, demonstrating the potential to improve predictions by deploying supplemental low-cost water level sensors. We demonstrate that NWM performance can be improved in these complex watersheds using ML tools. This approach could be implemented elsewhere to improve NWM streamflow predictions.

Maine, Massachusetts, New Hampshire, New York, Pen↗

USGS2025

Leveraging high-frequency sensor data and U.S. National Water Model output to forecast turbidity in a drinking water supply basin

As high-frequency sensor networks increasingly enhance data-driven models of water quality, process-based models like the U.S. National Water Model (NWM) are generating accessible forecasts of streamflow at increasingly dense scales. There is now an opportunity to combine these products to construct actionable water quality forecasts. To that end, we couple streamflow forecasts from the NWM to a gradient-boosted decision tree algorithm (LightGBM) trained on 5+ years of high-frequency monitoring data to forecast in-stream turbidity levels in the Catskill Mountains, NY, USA. Results indicate LightGBM models are capable of relatively skillful predictions, which enable robust forecasts for 1–3 days lead times. LightGBM models offer improvements over a simplified linear model across the entire forecast horizon, and more spatially complex models are more resilient to error at shorter lead times (1–3 days). Moreover, interpretation of model features emphasizes high flows as a driver of turbidity in the region. Results suggest that interpretable, flexible, and efficient machine learning algorithms can produce capable water quality forecasts from streamflow forecasts and expand understanding of process dynamics. The use case illustrated here—to our knowledge the first NWM-based water quality forecast—underscores the potential to employ the NWM to expand national water quality forecasting capacity and can overall serve as a guide for similar efforts in basins across the country.

New York↗

USGS2025

Machine learning generated streamflow drought forecasts for the Conterminous United States (CONUS): Developing and evaluating an operational tool to enhance sub-seasonal to seasonal streamflow drought early warning for gaged locations

Forecasts of streamflow drought, when streamflow declines below typical levels, are notably less available than for floods or meteorological drought, despite widespread impacts. To address this gap, we apply machine learning (ML) models to forecast streamflow drought 1-13 weeks into the future at > 3,000 streamgage locations across the conterminous United States (CONUS). We applied two ML methods (Long short-term memory (LSTM) neural networks; Light Gradient-Boosting Machine - LightGBM) and two benchmark model approaches (persistence; Autoregressive Integrated Moving Average - ARIMA) to predict weekly streamflow percentiles with independent models for each forecast horizon. To explore whether a training focus on dry weeks improved performance, both ML models were trained using all percentiles (LSTM-all, LightGBM-all) and only percentiles below 30% (LSTM<30, LightGBM<30). We evaluated model performance regionally and nationally for drought occurrence (the classification performance for a future date) and for drought onset/termination (performance identifying drought starts and ends). ML models generally performed worse than the persistence model for discrete classification (moderate, severe, extreme drought) of drought occurrence but exceeded the benchmark models for onset/termination. ML models outperformed benchmarks in predicting continuous streamflow percentiles below 30%. Occurrence performance was better for less intense droughts and shorter forecast horizons, with the ML models having predictive power at 1-4 week horizons for severe droughts (10th percentile threshold). All models struggled to forecast onset, though the best ML model was the LSTM<30 (sensitivity of 22%). Termination performance was greater, with the drought termination performance greatest for the LightGBM-all model. When estimating model uncertainty, the LSTM<30 model had the narrowest 90% percentile interval with closest to optimal capture. This work highlights the challenges and opportunities to further advance hydrological drought forecasting and supports an experimental operational streamflow drought assessment and forecast tool.

Continental United States↗

USGS2024

Solute export patterns across the contiguous USA

Understanding controls on solute export to streams is challenging because heterogeneous catchments can respond uniquely to drivers of environmental change. To understand general solute export patterns, we used a large-scale inductive approach to evaluate concentration–discharge (C–Q) metrics across catchments spanning a broad range of catchment attributes and hydroclimatic drivers. We leveraged paired C–Q data for 11 solutes from CAMELS-Chem, a database built upon an existing dataset of catchment and hydroclimatic attributes from relatively undisturbed catchments across the contiguous USA. Because C–Q relationships with Q thresholds reflect a shift in solute export dynamics and are poorly characterized across solutes and diverse catchments, we analysed C–Q relationships using Bayesian segmented regression to quantify Q thresholds in the C–Q relationship. Threshold responses were rare, representing only 12% of C–Q relationships, 56% of which occurred for solutes predominantly sourced from bedrock. Further, solutes were dominated by one or two C–Q patterns that reflected vertical solute–source distributions. Specifically, solutes predominantly sourced from bedrock had diluting C–Q responses in 43%–70% of catchments, and solutes predominantly sourced from soils had more enrichment responses in 35%–51% of catchments. We also linked C–Q relationships to catchment and hydroclimatic attributes to understand controls on export patterns. The relationships were generally weak despite the diversity of solutes and attribute types considered. However, catchment and hydroclimatic attributes in the central USA typically drove the most divergent export behaviour for solutes. Further, we illustrate how our inductive approach generated new hypotheses that can be tested at discrete, representative catchments using deductive approaches to better understand the processes underlying solute export patterns. Finally, given these long-term C–Q relationships are from minimally disturbed catchments, our findings can be used as benchmarks for change in more disturbed catchments.

contiguous United States↗

USGS2023

Evaluating deep learning architecture and data assimilation for improving water temperature forecasts at unmonitored locations

Deep learning (DL) models are increasingly used to forecast water quality variables for use in decision making. Ingesting recent observations of the forecasted variable has been shown to greatly increase model performance at monitored locations; however, observations are not collected at all locations, and methods are not yet well developed for DL models for optimally ingesting recent observations from other sites to inform focal sites. In this paper, we evaluate two different DL model structures, a long short-term memory neural network (LSTM) and a recurrent graph convolutional neural network (RGCN), both with and without data assimilation for forecasting daily maximum stream temperature 7 days into the future at monitored and unmonitored locations in a 70-segment stream network. All our DL models performed well when forecasting stream temperature as the root mean squared error (RMSE) across all models ranged from 2.03 to 2.11°C for 1-day lead times in the validation period, with substantially better performance at gaged locations (RMSE = 1.45–1.52°C) compared to ungaged locations (RMSE = 3.18–3.27°C). Forecast uncertainty characterization was near-perfect for gaged locations but all DL models were overconfident (i.e., uncertainty bounds too narrow) for ungaged locations. Our results show that the RGCN with data assimilation performed best for ungaged locations and especially at higher temperatures (>18°C) which is important for management decisions in our study location. This indicates that the networked model structure and data assimilation techniques may help borrow information from nearby monitored sites to improve forecasts at unmonitored locations. Results from this study can help guide DL modeling decisions when forecasting other important environmental variables.

Frontiers in Water↗

USGS2023

Regional streamflow drought forecasting in the Colorado River Basin using Deep Neural Network models

Process-based, large-scale (e.g., conterminous United States [CONUS]) hydrologic models have struggled to achieve reliable streamflow drought performance in arid regions and for low-flow periods. Deep learning has recently seen broad implementation in streamflow prediction and forecasting research projects throughout the world with performance often equaling or exceeding that of process-based models. Deep learning models are a possible approach to increase the accuracy of streamflow drought predictions and to expand the spatial coverage of river locations with available streamflow drought forecasts. As part of a multi-component Data-Driven Drought Prediction project, the U.S. Geological Survey is developing and testing deep learning models for streamflow drought forecasting. In this work, we present preliminary results of a deep learning model capable of predicting streamflow drought occurrence at ungaged locations for the Colorado River Basin (CRB). A long short-term memory (LSTM) neural network model was trained using 40 years (1980-2020) of daily streamflow data from 425 streamgages within and surrounding the CRB using static watershed attributes as well as meteorological and remotely sensed dynamic forcing inputs. Model tests were performed to evaluate model accuracy for now-casting streamflow drought conditions at ungaged locations and for forecasting drought conditions at lead times ranging from 0 to 14 days. Nearly all model configurations showed behavioral performance for predicting daily streamflow percentiles. Comparisons of LSTM model performance for predicting drought using fixed drought thresholds (calculated over all days and years) and variable drought thresholds (unique threshold calculated for each day of the year) identify differences in model skill between locations with implications for model design.

Colorado River basin↗