Geology topics

Jared David Smith

Publications and source records attributed to Jared David Smith.

5 recordsLinked to original sources

USGS2026

An approach for identifying drought typologies: Drivers, characteristics, and impacts of streamflow drought across the Delaware River Basin

Connecting drought drivers to drought conditions and drought conditions to consequences are two key components of drought risk assessment. We propose a two-step data-driven approach that contributes to advancing this capability. First, we use machine learning feature selection techniques to transform hydrologic model outputs into calculated indices and determine which features are most predictive of streamflow drought duration and severity. Second, we use the selected features to cluster drought events from 1985–2015 and examine drivers of drought in each cluster, spatial and temporal patterns, frequency over time, event characteristics, and connections to water management metrics. Our approach is implemented in the Delaware River Basin (DRB), a highly managed basin in the northeastern United States with a history of consequential droughts. Feature selection identified 19 features that can predict drought duration and severity better than models fit to all features. K-means clustering identified three distinct drought types: one capturing summer and fall droughts, largely in the lower basin; one capturing winter droughts, concentrated in the upper basin; and one capturing long-duration droughts across regions and seasons. Although annual drought frequency has decreased slightly over the period of record, those declines have not been split equally between clusters. Analysis of water management metrics revealed that soil moisture indices at 3-6 month timescales are the most able to differentiate the more consequential long-duration droughts from other drought types. This work contributes both specific insights for stakeholders in the DRB and a generalizable approach for similar analyses and explorations in other basins.

ESS Open Archive↗

USGS2024

Predictability and behavior of water transfers across basin boundaries

Inter-basin water transfers (IBTs) are important components of water balances of basins, and they can have substantial impact on regional water availability. Flow information is often not available at locations with known IBTs, which is a drawback in several published IBT databases. Few, if any, studies examine whether IBT flow behavior can be generalized, and if these behaviors can be predicted at undocumented locations or known IBT locations with no flow information. In this study, we employ a clustering method based on image matching to identify similar classes of flow behavior of IBTs. Machine learning models are used to assess how well IBT flow characteristics (e.g., average flow) associated with these behaviors can be predicted. These evaluations of IBTs are done for two regions in the United States. Three primary classes of IBTs (seasonal, nonseasonal/not mixed, and seasonal/mixed) are identified across the two regions analyzed. The IBT flow characteristics are accurately predicted in the northeast region. In the Colorado region, however, only the flow characteristics related to timing were accurately predicted. These results indicate that the proposed modeling framework can be used to identify generalizable IBT flow characteristics. This framework is shown to predict flow characteristics with a reasonable amount of accuracy to undocumented locations and improves previously published IBT databases by backfilling flow information to locations with a known IBT presence.

Arizona, Colorado, Delaware, Kansas, Maryland, Mas↗

USGS2024

Predictive understanding of stream salinization in a developed watershed using machine learning

Stream salinization is a global issue, yet few models can provide reliable salinity estimates for unmonitored locations at the time scales required for ecological exposure assessments. Machine learning approaches are presented that use spatially limited high-frequency monitoring and spatially distributed discrete samples to estimate the daily stream-specific conductance across a watershed. We compare the predictive performance of space- and time-unaware Random Forest models and space- and time-aware Recurrent Graph Convolution Neural Network models (KGE: 0.67 and 0.64, respectively) and use explainable artificial intelligence methods to interpret model predictions and understand salinization drivers. These models are applied to the Delaware River Basin, a developed watershed with diverse land uses that experiences anthropogenic salinization from winter deicer applications. These models capture seasonality for the winter first flush of deicers, and the streams with elevated predictions correspond well with indicators of deicer application. This result suggests that these models can be used to identify potential salinity-impaired streams for winter best management practices. Daily salinity predictions are driven primarily by land cover (urbanization) trends that may represent anthropogenic salinization processes and weather at time scales up to three months. Such modeling approaches are likely transferable to other watersheds and can be applied to further understand salinization risks and drivers.

Environmental Science and Technology↗

USGS2024

Deep learning of estuary salinity dynamics is physically accurate at a fraction of hydrodynamic model computational cost

Salinity dynamics in the Delaware Bay estuary are a critical water quality concern as elevated salinity can damage infrastructure and threaten drinking water supplies. Current state-of-the-art modeling approaches use hydrodynamic models, which can produce accurate results but are limited by significant computational costs. We developed a machine learning (ML) model to predict the 250 mg L −1 Cl − isochlor, also known as the “salt front,” using daily river discharge, meteorological drivers, and tidal water level data. We use the ML model to predict the location of the salt front, measured in river miles (RM) along the Delaware River, during the period 2001–2020, and we compare predictions of the ML model to the hydrodynamic Coupled Ocean–Atmosphere-Wave-Sediment Transport (COAWST) model. The ML model predicts the location of the salt front with greater accuracy (root mean squared error [RMSE] = 2.52 RM) than the COAWST model does (RMSE = 5.36); however, the ML model struggles to predict extreme events. Furthermore, we use functional performance and expected gradients, tools from information theory and explainable artificial intelligence, to show that the ML model learns physically realistic relationships between the salt front location and drivers (particularly discharge and tidal water level). These results demonstrate how an ML modeling approach can provide predictive and functional accuracy at a significantly reduced computational cost compared to process-based models. In addition, these results provide support for using ML models in operational forecasting, scenario testing, management decisions, hindcasting, and resulting opportunities to understand past behavior and develop hypotheses.

Limnology and Oceanography↗

USGS2024

New diagnostic assessment of MCMC algorithm effectiveness, efficiency, reliability, and controllability

Markov Chain Monte Carlo (MCMC) is a robust statistical approach for estimating posterior distributions. However, the significant computational cost associated with MCMC presents a considerable challenge, complicating the selection of an appropriate algorithm tailored to the specific problem at hand. This study introduces a novel and comprehensive framework for evaluating the performance of MCMC algorithms, drawing inspiration from diagnostics used for multi-objective evolutionary algorithms. We employ visualizations to evaluate key algorithmic characteristics: Effectiveness (the ability to accurately find representative posterior modes, quantified by the Kullback-Leibler Divergence (KLD) andWasserstein Distance (WD)), Efficiency (the speed of posterior characterization), Reliability (consistency across different random seeds), and Controllability (insensitivity to hyperparameter variation). Evaluating three prominent MCMC algorithms—Metropolis-Hastings (MH), Adaptive Metropolis (AM), and Differential Evolution Adaptive Metropolis (DREAM)—on high-dimensional and bimodal test problems, our analysis uncovers several insights. First, across algorithms, the number of function evaluations most controls performance on the high-dimensional problem, while the number of chains most controls performance on the bimodal problem. While this suggests similar controllability across algorithms, differences emerge on the other algorithmic characteristics. For high numbers of functions evaluations, AM performs best on the high-dimensional problem, while for low (<5) and high (>15) chain counts, MH and AM perform best on the bimodal problem, as measured by KLD. However, outside these specific cases, DREAM consistently demonstrates superior efficiency and reliability, making it a robust choice for both high-dimensional and multimodal problems. These findings can inform MCMC algorithm selection for Bayesian inference applications, as well as hyperparameterization of the chosen algorithm...

IEEE Access↗