Search NASA⌕ Search

SEARCH · Search NASA

Results for “benchmark testing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,225 records · Page 68

Assessing the accuracy of time-fraction and ductility exhaustion approaches for creep-fatigue damage prediction through feature-test validation

Determining creep-fatigue damage formation is critical for elevated temperature components integrity. This study evaluates creep-fatigue assessments with emphasis on differences between creep damage models. Evaluated are the time-fraction model and the classical and stress-modified ductility exhaustion creep damage models. This work extends the domain of stress-modified ductility exhaustion models by introducing and validating such formalism to Ni-based alloys. The fidelity of the assessments was benchmarked against uniaxial creep-fatigue and multiaxial feature tests of Alloy 617. For uniaxial specimens, best estimate predictions rank ductility exhaustion as the most accurate and precise and time-fraction as markedly conservative. For feature tests, ductility exhaustion predictions are within < 4.0 times difference, whereas time-fraction underpredicts life by factors of 7–14. The observations suggest ductility exhaustion models as alternative to time-fraction models in design codes for situations requiring characterization of the design margin. Further work in employing such models to assess other relevant phenomena (e.g., stress relaxation cracking) is discussed.

creep-fatigue↗

Evaluating Network Performance of Containerized Test Framework for Distributed Space Systems

Distributed space systems are a mission architecture consisting of multiple spacecraft as a cohesive system which provide multipoint sampling, increased mission coverage, or improved sample resolution, while reducing mission risk through redundancy. To fully realize the potential of these systems, eventually scaling to hundreds or thousands of spacecraft, distributed space systems need to be operated as a single entity, which will enable a variety of novel scientific space missions. The Distributed Spacecraft Autonomy (DSA) project is a software project which aims to mature the technology needed for those systems, namely autonomous decision-making and swarm networking. The DSA project leverages a containerized swarm test framework to simulate spacecraft software, which can identify emergent behavior early in development. Container virtualization allows distributed spacecraft systems to be simulated entirely in software on a single computer, avoiding the overhead associated with conventional approaches like hardware facsimiles and virtual machines. For this approach to be effective, the simulated system behavior must not be artificially influenced by the swarm test framework itself. To address this, we present a series of benchmarks to quantify virtual network bandwidth available on a single-host computer and contextualize this against the network and application behavior of the DSA swarm test framework.

Distributed Spacecraft Autonomy↗

iSSH v. Auditd: Intrusion Detection in High Performance Computing

The goal is to provide insight into intrusions in high performance computing, focusing on tracking intruders motions through the system. The current tools, such as pattern matching, do not provide sufficient tracking capabilities. We tested two tools: an instrumented version of SSH (iSSH) and Linux Auditing Framework (Auditd). First discussed is Instrumented Secure Shell (iSSH): a version of SSH developed at Lawrence Berkeley National Laboratory. The goal is to audit user activity within a computer system to increase security. Capabilities are: Keystroke logging, Records user names and authentication information, and Catching suspicious remote and local commands. Strengths for iSSH are: (1) Good for keystroke logging, making it easier to track malicious users by catching suspicious commands; (2) Works with Bro to send alerts; could be configured to send pages to systems administrators; and (3) Creates visibility into SSH sessions. Weaknesses are: (1) Relatively new, so not very well documented; and (2) No capabilities to see if files have been edited, moved, or copied within the system. Second we discuss Auditd, the user component of the Linux Auditing System. It creates logs of user behavior, and monitors systems calls and file accesses. Its goal is to improve system security by keeping track of users actions within the system. Strenghts of Auditd are: (1) Very thorough logs; (2) Wider variety of tracking abilities than iSSH; and (3) Older, so better documented. Weaknesses are: (1) Logs record everything, not just malicious behavior; (2) The size of the logs can lead to overflowing directories; and (3) This level of logging leads to a lot of false alarms. Auditd is better documented than iSSH, which would help administrators during set up and troubleshooting. iSSH has a cleaner notification system, but the logs are not as detailed as Auditd. From our performance testing: (1) File transfer speed using SCP is increased when using iSSH; and (2) Network benchmarks were roughly the same regardless of which tool was running.

99 GENERAL AND MISCELLANEOUS↗

Monte Carlo Perturbation Theory Estimates of Sensitivities to System Dimensions

Here, Monte Carlo methods are developed using adjoint-based perturbation theory and the differential operator method to compute the sensitivities of the k-eigenvalue, linear functions of the flux (reaction rates), and bilinear functions of the forward and adjoint flux (kinetics parameters) to system dimensions for uniform expansions or contractions. The calculation of sensitivities to system dimensions requires computing scattering and fission sources at material interfaces using collisions occurring at the interface—which is a set of events with infinitesimal probability. Kernel density estimators are used to estimate the source at interfaces using collisions occurring near the interface. The methods for computing sensitivities of linear and bilinear ratios are derived using the differential operator method and adjoint-based perturbation theory and are shown to be equivalent to methods previously developed using a collision history–based approach. The methods for determining sensitivities to system dimensions are tested on a series of fast, intermediate, and thermal critical benchmarks as well as a pressurized water reactor benchmark problem with iterated fission probability used for adjoint-weighting. The estimators are shown to agree within 5% and 3σ of reference solutions obtained using direct perturbations with central differences for the majority of test problems.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

2D radial-azimuthal particle-in-cell benchmark for E × B discharges

In this paper we propose a representative simulation test-case of E × B discharges accounting for plasma wall interactions with the presence of both the electron cyclotron drift instability and the modified-two-stream-instability. Seven independently developed particle-in-cell (PIC) codes have simulated this benchmark case, with the same specified conditions. The characteristics of the different codes and computing times are given. Here, results show that both instabilities were captured in a similar fashion and good agreement between the different PIC codes is reported as main plasma parameters were closely related within a 5% interval. The number of macroparticles per cell was also varied and statistical convergence was reached. Detailed outputs are given in the supplementary data, to be used by other similar groups in the perspective of code verification.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Coupling of Pronghorn and RELAP-7 for a Pebble Bed Reactor

High temperature gas cooled reactors (HTGR) are a candidate for timely Gen-IV reactor technology deployment because of high technology readiness and walk-away safety. Among HTGRs, pebble bed reactors (PBRs) have attractive features such as low excess reactivity and online refueling. Pebble bed reactors pose unique challenges to analysts and reactor designers such as continuous burnup distribution depending on pebble motion and recirculation, radiative heat transfer across a variety of gas-filled gaps, and long design basis transients such as pressurized and depressurized loss of forced circulation. Modeling and simulation is essential for both the PBR’s safety case and design process. In order to verify and validate the new generation codes the Nuclear Energy Agency (NEA) Data bank provide a set of benchmarks data together with solutions calculated by the participants using the state of the art codes of that time. An important milestone to test the new PBR simulation codes is the OECD NEA PBMR-400 benchmark which includes thermal hydraulic and neutron kinetic standalone exercises as well as coupled exercises and transients scenarios. In this work, the reactor multiphysics code MAMMOTH and the thermal hydraulics code Pronghorn, both developed by the Idaho National Laboratory (INL) within the multiphysics object-oriented simulation environment (MOOSE), have been used to solve Phase 1 exercises 1 and 2 of the PBMR-400 benchmark. The steady state results are in agreement with the other participants’ solutions demonstrating the adequacy of MAMMOTH and Pronghorn for simulating PBRs.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Recent MOOSE Update Slides for a DOE CI visit

This is an updated slide deck for which much of the content has been previously cleared. This slide deck incorporates recent NRIC Virtual Test Bed work with several updates on modeling and simulation of benchmark advanced reactor problems. Note that none of the models or results contained within these slides is proprietary. Many of these have been openly published on the Virtual Test Bed.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Visualization of flow interaction in a channel flow produced by a pair of helically coiled rods

Helically coiled tubes have been considered in the design of steam generators or heat exchangers of next generation nuclear power plants. Its compactness, efficiency, and the capability to absorb thermal expansion are desirable for nuclear reactor designs. However, the complex flow generated by the helically coiled geometry can result in flow induced vibration (FIV). In order to predict the FIV, there have been efforts using Computational Fluid dynamics (CFD)4, 5 to be coupled with multi-mechanics simulations. However, it is required to validate CFD to have confidence in capturing the flow structure characteristics of the transient complex flow behavior and fluctuation. In the present study, a simplified helically coiled steam generator (HCSG) design experiment was built acquiring data for CFD validation purpose. It is not perfect representation of the multiple rods, however, the current test facility produced transient data of the flow interaction between the rods and flow fluctuations near the rods. Particle Image Velocimetry (PIV) was applied to visualize the turbulent flow in a channel with slant rods. The two dimensional turbulent flow fields were analyzed. The velocity components were defined as below. u = U + u' (1) v = V + v' (2) where u is the lateral velocity on x-axis, v is the vertical velocity on y-axis, U and V are the mean velocities, and u' and v' are the fluctuation. Reynolds number, Re, was calculated using the rod diameter, d = 15.9 mm, and the upstream approach velocity, V{sub 0} = 0.5 m/s. Re = V{sub 0} d/η (3) where η is the kinematic viscosity of water. The experiment was conducted at Re ≅ 9,000 considering the currently reasonable computing power. Flow visualization in a flow channel produced by helically coiled rods was performed using PIV. The mean velocity field and streamlines were presented in this paper. The vortical flow structures in the cavity between rods were observed. Because of the asymmetric geometry, the flow profile was skewed. Additional measurements and transient analysis will be conducted to produce benchmark data for FIV study using CFD. Also, a test facility with a 5-rod heat exchanger design is operating to produce more data as a further work.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

A PDF-Based Parameterization of Subgrid-Scale Hydrometeor Transport in Deep Convection

Abstract A parameterization scheme is proposed for the subgrid-scale transport of hydrometeors in an assumed probability density function (PDF) scheme. Joint distributions of vertical velocity and hydrometeor mixing ratios are typically unknown, but marginal (1D) PDFs of these variables are available. The parameterization is developed using high-resolution simulations of continental and tropical deep convection. A 3D cloud-resolving model (CRM) providing benchmark solutions has a horizontal grid spacing of 250 m and employs the Morrison microphysics scheme, which treats prognostically mass and number mixing ratios for four types of precipitating hydrometeors (rain, graupel, snow, and ice) as well as cloud droplet number mixing ratio. The subgrid-scale hydrometeor transport scheme assumes input given in the form of marginal PDFs of vertical velocity and hydrometeor mixing ratios; in this study, these marginal distributions are provided by the cloud-resolving model. Conditional sampling and scaling are then applied to the marginal distributions to account for subplume correlations. The parameterized fluxes tested for four episodes of deep convection show good agreement with benchmark fluxes computed directly from the CRM output. The results demonstrate the potential use of the subgrid-scale hydrometeor transport scheme in an assumed PDF scheme to parameterize the covariances of vertical velocity and hydrometeor mixing ratios.

Wong, May↗

Active Learning in the Era of Big Data

Active learning methods automatically adapt data collection by selecting the most informative samples in order to accelerate machine learning. Because of this, real-world testing and comparing active learning algorithms requires collecting new datasets (adaptively), rather than simply applying algorithms to benchmark datasets, as is the norm in (passive) machine learning research. To facilitate the development, testing and deployment of active learning for real applications, we have built an open-source software system for large-scale active learning research and experimentation. The system, called NEXT, provides a unique platform for realworld, reproducible active learning research. This paper details the challenges of building the system and demonstrates its capabilities with several experiments. The results show how experimentation can help expose strengths and weaknesses of active learning algorithms, in sometimes unexpected and enlightening ways.

Jamieson, Kevin↗

Contra-Rotating Open Rotor Tone Noise Prediction

Reliable prediction of contra-rotating open rotor (CROR) noise is an essential element of any strategy for the development of low-noise open rotor propulsion systems that can meet both the community noise regulations and cabin noise limits. Since CROR noise spectra exhibit a preponderance of tones, significant efforts have been directed towards predicting their tone content. To that end, there has been an ongoing effort at NASA to assess various in-house open rotor tone noise prediction tools using a benchmark CROR blade set for which significant aerodynamic and acoustic data have been acquired in wind tunnel tests. In the work presented here, the focus is on the nearfield noise of the benchmark open rotor blade set at the cruise condition. Using an analytical CROR tone noise model with input from high-fidelity aerodynamic simulations, tone noise spectra have been predicted and compared with the experimental data. Comparisons indicate that the theoretical predictions are in good agreement with the data, especially for the dominant tones and for the overall sound pressure level of tones. The results also indicate that, whereas the individual rotor tones are well predicted by the combination of the thickness and loading sources, for the interaction tones it is essential that the quadrupole source is also included in the analysis.

Propeller Noise↗

Thermally Conductive Backsheets (TCB) for Reduced Operating Temperatures

The performance of photovoltaic (PV) modules is primarily influenced by the plane of array (POA) irradiance and their operating temperature. The lower the module operating temperature, the higher is its performance. Increasing a module’s operating temperature reduces its power output due to an increase in the semiconductor’s intrinsic carrier density, leading to a reduction in the open-circuit voltage and, subsequently, reducing the maximum power point voltage. Reducing the operating temperature lowers the levelized cost of energy (LCOE) via increased energy production (i.e., more kilowatt hour over the life of the asset) and increased service lifetime (years in the field). The module operating temperature is dictated by the weather parameters and the thermal properties of module construction materials. The primary weather parameters influencing the PV module temperature are: the ambient temperature (directly related), irradiance (directly related), and wind speed (inversely related). The primary thermal properties of module construction materials influencing the PV module temperature are: the optical transmissivity and thermal conductivity of superstrate, the thermal conductivity of substrate, the thermal conductivity of encapsulant and the efficiency of the encapsulated solar cells. The primary goal of this study is to reduce the operating temperature of PV modules using thermally conductive backsheet (TCB) substrates developed by the industry. The temperature of PV modules installed in a large multi-MW plant can also be influenced by their physical locations in the plants due to the influence of wind direction. A thermal mapping of PV plants has also been investigated and presented in this study. The most commonly used PV backsheet material has been polyvinyl fluoride (DuPont trade name “Tedlar”)-based backsheet, which typically has three layers: Tedlar–polyethylene terephthalate (PET)–Tedlar (TPT). The primary goal of this project is to reduce the module operating temperature below the benchmark TPT backsheet temperature using newly developed TCBs by the industry. The overall approach of this project involves six focus areas: i) select and procure one TPT and thirteen promising TCBs available in the industry; ii) construct a large number of 1-cell and 9-cell modules using benchmark TPT and selected TCBs [all other construction materials are fixed and they are: glass superstrate, EVA encapsulant, mono-Si cell and junction box]; iii) to identify best performing TCBs, monitor the long-term field operating temperatures of the constructed modules which are installed in three climatic conditions (hot-low wind [Mesa, AZ], hot-high wind [Maricopa County, AZ] and temperate [Cumberland County, NC]; iv) to identify the most durable TCB, perform extended accelerated reliability testing of modules constructed with best three TCBs and benchmark TPT; v) to predict module temperature, develop physical and empirical thermal models with and without thermal properties of all construction materials in the modules, respectively; and vi) to determine the influence of wind speed and direction on the module operating temperature in a plant, and perform thermal mapping of PV modules in two large PV plants [Maricopa County, AZ and Cumberland County, NC]. The key achievements of this project are listed below: Backsheet selection: 14 backsheets were procured after an extensive literature and industry search. Finally, based on the thermal and other selection criteria, 1 TPT and 4 thermally conductive backsheets were selected for this project. Fabrication of mini modules: 50 one-cell modules were fabricated in year 1 for thermal performance evaluation. The module design was changed to nine-cell module at start of year 2 to mimic commercial design. 20 nine-cell modules were fabricated in year 2 for thermal performance evaluation. 28 four-cell modules were fabricated in year 3 for accelerated stress-testing. Thermal conductivity of backsheets: Year 1 TCBs showed at least 8% higher thermal conductivities than TPT. Year 2 and 3 TCBs showed at least 20% higher thermal conductivities than TPT. Installation of mini PV modules at three sites: 6 one-cell modules and 8 nine-cell modules were installed at AZ-1 site [Mesa, AZ]. 12 one-cell modules and 6 nine-cell modules were installed at each of two powerplants (AZ-2 [Maricopa County, AZ] and NC sites). Nominal operating cell temperatures (NOCT) of mini PV modules: NOCT of TCBs were observed to be lower than TPT with differences as high as 2°C. Comparison of thermal performance: The median operating temperature of one-cell modules was found to be reduced by 2°C by using TCB_A modules compared to TPT modules on several days in a month for few months in a year. The median temperature improvement of 4oC was achieved at several instances on a few days, but not all days, in a year using multiple TCBs compared to TPT. The final project milestone of achieving median temperature improvement of 5oC was achieved at several instances on a few days, but not all days, in a year using TCB_A compared to TPT. Empirical thermal model (without the need of construction material properties): Empirical thermal model for one-cell module was developed based on the input data of irradiance, wind speed and ambient temperatures. The primary input data differentiator for temperature differences between TPT and TCBs appears to be the irradiance level (solar gain). Accelerated durability testing: 20 four-cell modules using 4 backsheets (1 TPT and 3 TCBs) were exposed to Damp Heat 1000 hours (8 modules), UV/Thermal cycling (50 cycles)/Humidity freeze (10 cycles) test sequence (8 modules) and UV Weathering for qualification (8 modules). All the modules with four selected backsheets passed the stress tests per the retest guidelines of IEC 61215 standard for the backsheet change. Simplified comparative LCOE analysis: Using backsheet-specific power degradation levels in the accelerated tests (for TCB_A and TPT only), efficiency, service lifetime and assuming same cost for all backsheets, the LCOE of modules using TCB_A is determined to be lower than TPT. Due to lower operating temperature of TCB_A-based modules compared to TPT-based modules, the operating efficiency and service life of TCB_A-based modules are calculated to be higher than TPT-based modules. Temperature mapping of PV power plants: A distinct temperature difference (as high as 4°C) has been observed between the modules located in various parts of a single PV plant.

14 SOLAR ENERGY↗

Advanced trust region optimization strategies for glass box / black box models

We present an improved trust region filter (TRF) method for optimization of combined glass box/black box systems. Glass box systems refer to models that are easily expressed in an algebraic modeling language, providing cheap and accurate derivative information. By contrast, black box systems may be computationally expensive and derivatives are unavailable. The TRF method, as first introduced in our previous work (Eason and Biegler, AIChE J . 2016; 62:3124–3136), is able to handle hybrid systems containing both glass and black box components, which can frequently arise in chemical engineering, for example, when a multiphase reactor model is included in a flow sheet optimization problem. We discuss several recent modifications in the algorithm such as the sampling region, which maintains the algorithm's global convergence properties without requiring the trust region to shrink to zero in the limit. To benchmark the development of this optimization method, a test set of problems is generated based on modified problems from the CUTEr and COPS sets. The modified algorithm demonstrates improved performance using the test problem set. Finally, the algorithm is implemented within the Pyomo environment and demonstrated on a rigorous process optimization case study for carbon capture. © 2018 American Institute of Chemical Engineers AIChE J , 64: 3934–3943, 2018

Eason, John P.↗

The GPU-enabled divide-expand-consolidate RI-MP2 method (DEC-RI-MP2)

We report porting of the Divide-Expand-Consolidate Resolution of the Identity second-order Møller–Plesset perturbation (DEC-RI-MP2) method to the graphic processing units (GPUs) using OpenACC compiler directives. It is shown that the OpenACC compiler directives implementation efficiently accelerates the rate-determining step of the DEC-RI-MP2 method with minor implementation effort. Moreover, the GPU acceleration results in a better load balance and thus in an overall scaling improvement of the DEC algorithm. The resulting cross-platform hybrid MPI/OpenMP/OpenACC implementation has scalable and portable performance on heterogeneous HPC architectures. The GPU-enabled code was benchmarked using a reduced version of the S12L test set of Stefan Grimme (Grimme, Chem. Eur. J. 2012, 18, 9955) consisting of supramolecular complexes up to 158 atoms and 4292 contracted basis functions (cc-pVTZ). The test set results demonstrate the general applicability of the DEC-RI-MP2 method showing results consistent with the DEC-RI-MP2 introductory paper (Baudin et al., J. Chem. Phys. 2016, 144, 054102) on molecules of complicated electronic structures. © 2016 Wiley Periodicals, Inc.

Chemistry↗

Community detection and unveiling of hierarchy in networks: a density-based clustering approach

Abstract The unveiling of communities within a network or graph, and the hierarchization of its members that results is of utmost importance in areas ranging from social to biochemical networks, from electronic circuits to cybersecurity. We present a statistical mechanics approach that uses a normalized Gaussian function which captures the impact of a node within its neighborhood and leads to a density-ranking of nodes by considering the distance between nodes as punishment. A hill-climbing procedure is applied to determine the density attractors and identify the unique parent (leader) of each member as well as the group leader. This organization of the nodes results in a tree-like network with multiple clusters, the community tree. The method is tested using synthetic networks generated by the LFR benchmarking algorithm for network sizes between 500 and 30,000 nodes and mixing parameter between 0.1 and 0.9 . Our results show a reasonable agreement with the LFR results for low to medium values of the mixing parameter and indicate a very mild dependence on the size of the network.

Felfli, Zineb (ORCID:0000000275294635)↗

Dynamically robust coordinated set point tracking of distributed DERs at point of common coupling

Low-inertia operation of small-scale power systems, such as a microgrid or a portion of a long feeder, requires careful coordination of the controller performance of the constituting devices. This challenge is exacerbated in microgrids serving the functionalities of a conventional synchronous-based generation unit while comprised of smaller DERs operating mainly interfaced through power electronics converters. This paper builds on the idea of set point modulation and proposes a two-level control strategy that aims to achieve superior performance at the point of common coupling (PCC) of microgrids by combining a local control level with a distributed and coordinated level. Several case studies on both AC and DC systems, the CIGRE low-voltage benchmark system as the AC system and a test DC microgrid, validate the performance of the proposed approach. The real-world applicability of the approach is established via a high-fidelity power hardware-in-the-loop (PHIL) experimental setup and an application case study on grid frequency regulation. The proposed approach enables a microgrid to participate in ancillary service provisions where speed and quality of regulation are critical.

24 POWER TRANSMISSION AND DISTRIBUTION↗

The Random Ray Method for neutral particle transport

A new approach to solving partial differential equations (PDEs) based on the method of characteristics (MOC) is presented. The Random Ray Method (TRRM) uses a stochastic rather than deterministic discretization of characteristic tracks to integrate the phase space of a problem. TRRM is potentially applicable in a number of transport simulation fields where long characteristic methods are used, such as neutron transport and gamma ray transport in reactor physics as well as radiative transfer in astrophysics. In this study, TRRM is developed and then tested on a series of exemplar reactor physics benchmark problems. The results show extreme improvements in memory efficiency compared to deterministic MOC methods, while also reducing algorithmic complexity, allowing for a sparser computational grid to be used while maintaining accuracy.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Knowledge-guided graph machine learning for spatially distributed prediction of daily discharge and nitrogen export dynamics

Spatially distributed prediction of streamflow and nitrogen export dynamics is essential for precision management of agricultural watersheds. While temporal deep learning models such as Long Short-Term Memory (LSTM) have shown strong performance at basin scales, their ability to generalize spatially is limited by insufficient representation of spatial dependencies and flow paths, particularly under data-scarce conditions. To address this gap, we propose HydroGraphNet, a knowledge-guided graph machine learning framework that integrates process-based knowledge and explicit spatial learning into temporal modeling. This framework incorporates directed graph topology to encode watershed connectivity and upstream inflows, with mass balance constraints to improve physical consistency. To enhance generalization in sparsely monitored regions, HydroGraphNet is pretrained on synthetic data generated by the SWAT+ (Soil and Water Assessment Tool Plus) model. We evaluated HydroGraphNet in the Upper Sangamon River Basin (44 HUC-12 subwatersheds, 2001–2020) against two LSTM baselines: a lumped basin-level model and a distributed variant. When benchmarked on SWAT+ simulations in pretraining, HydroGraphNet improved test NSEs by 8.9% (discharge) and 13.7% (NO₃–N load) in temporal extrapolation, and by 27.1% and 34.7% in spatial extrapolation, relative to the Lumped LSTM baseline. After fine-tuning with USGS monitoring data, the model achieved mean test NSE (KGE) scores of 0.768 (0.861) for discharge and 0.626 (0.664) for NO₃–N load, substantially outperforming baselines. Attribution analysis further highlighted the importance of upstream inflow representation and graph-based spatial learning in capturing cross-subwatershed dependencies. The model also reproduced seasonal hydrological and biogeochemical patterns consistent with known processes, demonstrating its robustness and process fidelity for spatially distributed prediction. Altogether, HydroGraphNet advances the integration of physical knowledge and spatially explicit learning in hydrological modeling, offering a generalizable framework for distributed modeling to support spatially targeted water quality management in data-scarce watersheds.

54 ENVIRONMENTAL SCIENCES↗