Machine Learning for a Heterogeneous Water Modeling Framework

JAWRA Journal of the American Water Resources Association, Journal Year: 2025, Volume and Issue: 61(1)

Published: Feb. 1, 2025

ABSTRACT This technical note describes recent efforts to integrate machine learning (ML) models, specifically long short‐term memory (LSTM) networks and differentiable parameter conceptual hydrological models (δ models), into the next‐generation water resources modeling framework (Nextgen) enhance future versions of U.S. National Water Model (NWM). We address three specific methodology gaps this new framework: (1) assess model performance across many ungauged catchments, (2) diagnostic‐based selection, (3) regionalization based on catchment attributes. demonstrate that an LSTM trained CAMELS catchments can make large‐scale predictions with Nextgen New England region match average flow duration curve observed by stream gauges for streamflow low exceedance probability (high flows), but diverges from mean in high (low flows). improvements peak when using δ model, results also suggest increases may come at a cost accurately representing hydrologic states within model. propose novel approach ML predict most performant mosaic improved distributions efficiency scores throughout large sample basins. Our findings advocate development capabilities advancing operational modeling.

Language: Английский

---

Enhancing Streamflow Reanalysis Across the Conterminous US Leveraging Multiple Gridded Precipitation Data Sets

Water Resources Research, Journal Year: 2025, Volume and Issue: 61(3)

Published: March 1, 2025

Abstract Streamflow observations, essential for various water resource applications, are often unavailable at critical locations in need. Although different models have been proposed to enhance streamflow predictability ungauged locations, the challenge extends beyond model fidelity. Differences meteorologic forcing data sets, precipitation particular, can significantly affect accuracy of hydrologic predictions. This intensifies across regions characterized by diverse hydro‐climatological and geographical conditions, such as conterminous US (CONUS) where a single product struggles consistently replicate observed hydrographs, particularly peak flow dynamics. To predictions, we utilize VIC‐RAPID modeling framework driven multiple commonly used meteorological Daymet, PRISM, ST4, AORC, their hybrids create sets 40‐year (1980–2019) hourly, daily, monthly reanalysis, Dayflow Version 2, 2.7 million river reaches CONUS. Most forcings lead skillful performance, except ST4 mountainous west, severe radar blockage adversely affects accuracy. The evaluation using over 6,000 hourly stream gauges shows that AORC improved annual performance Daymet—driven (Dayflow V1), smaller basins, highlighting value high temporal resolution Compared with other benchmark like National Water Model V3.0, AORC‐driven exhibits regional comparable representation. We envision multi‐forcing reanalysis inform need enhancement, diagnose benefit applications.

Language: Английский

---

Spatial Hydrographs of River Flow and Their Analysis for Peak Event Detection in the Context of Satellite Sampling

Water Resources Research, Journal Year: 2025, Volume and Issue: 61(4)

Published: April 1, 2025

Abstract The study of river dynamics has long relied on the analysis traditional in situ hydrographs. This graphical representation temporal variability at a given location is so ubiquitous that mere term “hydrograph” widely recognized as time series. While such “temporal hydrograph” well suited for data analysis, it fails to represent hydrologic across space time; perspective characterizes satellite‐based observations. Here we argue concept “spatial should be focus its own dedicated scrutiny. We build “space series” discharge and present their context peak flow event detection. propose use spatial coverage, referred “length”, an analog duration. Our performed Mississippi basin using dense network. reveal events range length from around 75 1,800 km with median (mean) value 330 (520) along basin's largest rivers. also suggests sampling needs factor 4 (2) finer resolution than lengths detect 81% ± 13% (70% 20%) estimate within 84% 3% (67% 12%) accuracy. evaluate connection between scales flows show longer durations affect larger extents. finally discuss implications design satellite missions concerned capturing floods space.

Language: Английский

---

CONCN: a high-resolution, integrated surface water–groundwater ParFlow modeling platform of continental China

Hydrology and earth system sciences, Journal Year: 2025, Volume and Issue: 29(9), P. 2201 - 2218

Published: May 12, 2025

Abstract. Large-scale hydrologic modeling at the national scale is an increasingly important effort worldwide to tackle ecohydrologic issues induced by global water scarcity. In this study, a surface water–groundwater integrated platform was built using ParFlow, covering entirety of continental China with resolution 30 arcsec. This model, CONCN 1.0, offers full treatment 3D variably saturated groundwater solving Richards' equation, along shallow-water equation ground surface. The performance 1.0 rigorously evaluated both data products and observations. RSR values (the ratio root mean squared error standard deviation observations) show satisfying regard streamflow, yet streamflow lower in endorheic, Hai, Liao rivers due uncertainties potential recharge. also indicate terms table depth model. intermediate compared two models, highlighting that persist current large-scale modeling. Our work comprehensive evaluation workflow for continental-scale ParFlow could be good starting point other regions worldwide, even when different systems. More specifically, vast arid semi-arid substantial sinks (i.e., endpoints endorheic rivers) large recharge pose challenges numerical solution model performance, respectively. Incompatibilities between such as mismatch spatial resolutions models shorter, less frequent observation records, necessitate further refinement enable fast not only establishes first efficient resources management but will benefit improvement next-generation worldwide.

Language: Английский

---

Global Cloud Biases in Optical Satellite Remote Sensing of Rivers

Geophysical Research Letters, Journal Year: 2024, Volume and Issue: 51(16)

Published: Aug. 15, 2024

Abstract Satellite imagery provides a global perspective for studying river hydrology and water quality, but clouds remain fundamental limitation of optical sensors. Explicit studies this problem were limited to specific locations or regions. In study, we characterize the severity by analyzing 22 years daily satellite cloud cover data modeled discharge sample 21,642 reaches diverse sizes climates. Our results show that bias in observed is highly organized space, particularly affecting Tropical Arctic rivers. Given nature limitation, satellites will always provide biased representation conditions. We discuss several strategies mitigate bias, including modeling, fusion, temporal averaging, yet these methods introduce their own challenges uncertainties.

Language: Английский

---

Spatial Hydrographs of River Flow and their Analysis for Peak Event Detection in the Context of Satellite Sampling

Authorea (Authorea), Journal Year: 2024, Volume and Issue: unknown

Published: July 29, 2024

The study of river dynamics has long relied on the analysis traditional in situ hydrographs. This graphical representation temporal variability at a given location is so ubiquitous that mere term "hydrograph" widely recognized as time series. While such "temporal hydrograph" well suited for data analysis, it fails to represent hydrologic across space time; perspective characterizes satellite-based observations. Here we argue concept "spatial should be focus its own dedicated scrutiny. We build "space series" discharge and present their context peak flow event detection. propose use spatial coverage, referred "length", an analog duration. Our performed Mississippi basin using dense network. reveal events range length from around 75 1,800 km with median (mean) value 330 (520) along basin's largest rivers. also suggests sampling needs factor 4 (2) finer resolution than lengths detect 81±13% (70±20%) estimate within 84±3% (67±12%) accuracy. evaluate connection between scales flows show longer durations affect larger extents. finally discuss implications design satellite missions concerned capturing floods space.

Language: Английский

---