Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “evaluation process”

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 1,585 records · Page 88

Polyethylene/Boron Composites for Radiation Shielding Applications

Multifunctional composites made with boron are absorbers of low energy nuetrons, and could be used for structural shielding materials. Polyethylene/boron carbide composites were fabricated using conventional polymer processing techniques, and were evaluated for mechanical and radiation shielding properties. Addition of neat boron carbide (powder and nanoparticles) to an injection molding grade HPDE showed superior mechanical properties compared to neat HDPE. Radiation shielding measurements of a 2 wt% boron carbide composite were improved over those of the neat polyethylene.

36 MATERIALS SCIENCE↗

Analysis of anion exchange membrane water electrolyzer performance and its evolution over time

Understanding water, evolved gas, and ionic transport in membrane-electrode-assemblies (MEAs) is essential for the development of high performance and durable anion exchange membrane water electrolyzers (AEMWEs). This study evaluates the MEA conditioning process, operating conditions, and short-term stability in a 1 M potassium hydroxide (KOH) electrolyte, focusing on the underlying transport phenomena. We observe a significant initial voltage loss in continuous cell operation, which could be associated with gas bubble accumulation, transport layer or flow field passivation, and changes in the catalyst oxidation state. Further, we investigate the effects of materials and operational configurations, including the membrane type and thickness, and the electrolyte flow rate, including KOH being fed to both electrodes as well as to the anode only. Furthermore, the effect of membrane drying temperature on ex situ as well as in situ electrochemical performance is evaluated. Finally, we discuss 700 h of AEMWE operation at 1 A/cm 2 , highlighting the underlying degradation phenomena.

25 ENERGY STORAGE↗

Assessing the influence of processing parameters and external loading on the nanoporous structure and morphology of nanoporous gold toward catalytic applications

Nanoporous gold structure gains several unique physical and chemical properties from its nanoscale foam structure composed of ligaments and connected pores. The high specific surface area and interconnected pores make its structure ideal for gas catalysis. Catalytic activity arises in part from the high number density of surface steps and kinks at points of high curvature in the nanoporous structure. However, the structure also makes the material macroscopically brittle and unreliable for many applications due to poor mechanical stability. The ligament diameter, which regulates both desirable chemical properties and mechanical instability, can be tuned to over three orders of magnitude, but few morphology features can be adjusted independently of this. In this report we fabricate nanoporous gold with an average ligament diameter of around 10 nm while varying synthesis parameters and quantify the resulting morphology in order to evaluate the effect of processing on the resulting nanoscale foam structure. In situ tensile experiments are performed to observe how both individual ligaments and the foam structure's features change during deformation in real-time. We found two new morphology features that are independently controllable through synthesis parameters and are likely to influence mechanical stability while preserving or enhancing catalytic activity. The curvature of the ligaments can be increased by stirring the electrolyte during dealloying and the width of the ligament diameter size distribution can be adjusted by controlling the coarsening rate. In situ mechanical tests show that grain boundaries can influence crack propagation and that ligaments removed from the crack tip experience both elastic and plastic stress.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

The delta envelope: A technique for dose distribution comparison

The {gamma} index is a tool that compares a dose distribution with a reference distribution by combining dose-difference and distance-to-agreement criteria. It has been widely used for ten years despite its high computational cost. This cost is due to both a search process for each reference point and the necessity to remove overestimations caused by the discrete nature of dose grids. The method proposed in this paper is much faster since it avoids both these problems. It consists in computing the {delta} envelope formed by the {gamma} ellipsoids around the points of the reference distribution. This {delta} envelope provides dose-difference tolerances that are then used to create new indices, called the {delta} indices, that provide useful information to interpret the deviations. Applied to both 1D and 2D test cases and compared to the {gamma} index, the {delta} indices proved to be very accurate and intuitive. Their computational efficiency was evaluated on a 3D case: the {delta} envelope can be computed in 8 s on a 250x250x50 grid. Moreover it can be precomputed if the reference dose is known in advance. Then the {delta} indices are obtained in less than 2 s.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Optical depth measurements by shadow-band radiometers and their uncertainties

Shadow-band radiometers in general, and especially the Multi-Filter Rotating Shadow-band Radiometer (MFRSR), are widely used for atmospheric optical depth measurements. The major programs running MFRSR networks in the U.S. include DOE Atmospheric Radiation Measurement (ARM) Program, USDA UV-B Monitoring and Research Program, NOAA Surface Radiation (SURFRAD) Network, and NASA Solar Irradiance Research Network (SIRN). In this paper we discuss a number of technical issues specific for shadow-band radiometers and their impact on the optical depth measurements. These problems include instrument tilt and misalignment, as well as some data processing artifacts. Techniques for data evaluation and automatic detection of some of these problems are described.

47 OTHER INSTRUMENTATION↗

Summertime Continental Shallow Cumulus Cloud Detection Using GOES‐16 Satellite and Ground‐Based Ceilometer at North Alabama

Abstract Accurate simulations of boundary layer cloud processes remain challenging in Earth system modeling. Observations are essential to evaluate and improve models of such processes. This study introduces a comprehensive validation framework for a satellite‐based detection algorithm of continental shallow cumulus (ShCu) clouds during the daytime, which was initially developed using ground‐based observations of stereo cameras at the Department of Energy Atmospheric Radiation Measurement (ARM) Southern Great Plains site (J. Tian et al., 2021, https://doi.org/10.3390/rs13122309 , 2022, https://doi.org/10.1029/2021gl097070 ). To validate this algorithm, the framework employs ground‐based ceilometer measurements from North Alabama (NA) where ShCu populations are prevalent. This study first generates clear‐sky surface reflectance maps at NA and identifies ShCu pixels with a detection threshold using Geostationary Operational Environmental Satellite (GOES) reflectance data. The obtained cloud fractions (CFs) are then compared against CFs from a ground‐based ceilometer, considering factors such as observed area differences, satellite parallax issue, and systematic biases. We found that with a detection threshold (∆R) of 0.05, the ShCu detection algorithm is effective for NA, enabling the reproduction of hourly ShCu CFs using GOES. Our framework is straightforward and easily repeatable to evaluate the effectiveness of a ∆R threshold for detecting ShCu clouds in various geographic regions where ceilometers are deployed. This satellite detection of ShCu provides a crucial regional context for ground‐based measurements, facilitating the tracking of convection initiation and its coupling with land surface conditions. Integrating localized ground‐based and regional satellite data will enhance our ability to conduct thorough studies of cloud morphology and land‐atmosphere interactions in North Alabama.

54 ENVIRONMENTAL SCIENCES↗

Combining of different data pools for calculating a reliable POD for real defects

Real defects are essential for the evaluation of the reliability of non destructive testing (NDT) methods, especially in relation to the integrity of components. But in most of the cases the amount of available real defects is not sufficient to evaluate the system. Model-assisted and transfer functions are one way to handle that challenge. This study is focused on a combination of different data pools to create a sufficient amount of data for the reliability estimation. A widespread approach for calculating the Probability of Detection (POD) was used on a radiographic testing (RT) method. The highest contrast to noise ratio (CNR) of each indication is usually selected as the signal in the 'â vs. a' (signal-response) approach for RT. By combining real and artificial defects (flat bottom holes, side drill holes, flat bottom squares, notches, etc) in RT the highest signals are close to each other, but the process of creating and evaluating real defects is much more complex. The solution is seen in the combination of real and artificial data using a weighted least square approach. The weights for real or artificial data were based on the importance, the value and the different detection behavior of the different data. For comparison, the alternative combination through the Bayesian Updating was also applied. As verification, a data pool with a large amount of real data was available. In an advanced approach for evaluating the digital RT data, the size of the indication (perpendicular to the X-ray beam) was introduced as additional information. The signal now consists of the CNR and the area of the indication. The detectability is changing depending on the area of the indication, a fact that was ignored in the previous POD calculations for RT. This points out that a weighted least square approach to pool the data might no longer be adequate. The Bayesian Updating of the estimated parameters of the relationship between the signal field (the area of the indication) and the geometry of the defects is seen as the appropriate model to combine the different defect types in a useful and meaningful way. This work was carried out together with the Finnish company for spent nuclear fuel and waste management - Posiva Oy. The digital RT is one of the NDT methods that might be used for the inspection of the weld of the copper canister to be used for the spent nuclear fuel in the Scandinavian concept of final disposal.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Overview of Lattice Design and Evaluation for the APS Upgrade

The Advanced Photon Source (APS) is a 7-GeV synchrotron light source that has been in operation since 1996. Since that time, the effective emittance has been decreased from 8 nm to 3.1 nm, which is very competitive for a 3rd-generation light source. However, newer facilities such as PETRA-III, NSLS-II, and MAX-IV are pushing the emittance to significantly smaller values. MAX-IV in particular has set the current benchmark with an emittance of about 300 pm at 3 GeV. This was accomplished by use of a multi-bend achromat lattice, which takes advantage of the 1/M3 scaling of the emittance with respect to the number of dipoles M. In order to ensure that our facility remains competitive, APS is pursuing a major upgrade, which involves replacement of the existing double-bend lattice with a seven-bend achromat lattice, promising a 40-fold reduction in emittance. This paper describes the process of developing and evaluating candidate lattice designs. Two candidate 6-GeV lattices are described: one providing a natural emittance of 67 pm and the other providing 41 pm. Our analysis includes single-particle dynamics as well as single- and multi-bunch collective effects.

07 ISOTOPE AND RADIATION SOURCES↗

Options To Cleanup Site-wide Vadose Zone Contamination At The Hanford Site, WA, State

The U.S. Department of Energy (DOE) Hanford Site in south central Washington State lies along the Columbia River and is one of DOE's largest legacy waste management sites. Enormous radionuclide and chemical inventories exist below-ground. These include Resource Conservation and Recovery Act (RCRA) storage facilities where hazardous and radioactive contaminants were discharged and leaked to the soil surface and to the deep vadose zone and groundwater. The vadose zone is also contaminated from facilities regulated by the RCRA and Comprehensive Environmental Response Compensation and Liability Act (CERCLA) Act. Hanford now contains as much as 28,300 cubic meters of soil contaminated with radionuclides from liquid wastes released near processing facilities. The Hanford Federal Facility Agreement and Consent Order, Tri-Party Agreement (TPA) has set the completion of the cleanup of these sites by 2024. There are numerous technical and regulatory challenges to cleanup of the vadose zone at the Hanford site. This paper attempts to identify the categories of deep vadose zone problem and identifies a few possible regulatory options to clean up the site under the mix of state and federal regulatory authorities. There are four major categories of vadose contamination areas at the Hanford Site. The first is laterally extensive with intermediate depth (ground surface to about 45 meters depth) mostly related to high volume effluent discharge into cribs, ponds and ditches of designated CERCLA facilities. The second is dominated by laterally less extensive mostly related to leaks from RCRA tank farms. The later contamination is often commingled at depth with wastes from adjacent CERCLA facilities. The third category is from the high volume CERCLA facilities extending from the surface to more than 60 meters below ground. Contamination from the later category crosses the entire thickness of the vadose zone and reached groundwater. The fourth category is the lower volume waste sites. There are multiple management options to clean up the above four categories of vadose zones sites. The following are some of the options considered for detailed evaluation: - Maintain separate decision processes for each RCRA and CERCLA units/waste sites with a more accommodating schedule. - Create new vadose zone operable units with limited geographical boundaries regardless of site category/origin and make an integrated decision. - Expand the existing CERCLA groundwater operable units to include the deep vadose zone - Use a combination of the above. Each option has pros and cons and regulatory limitations. Detailed evaluation of these options is required to support a cost effective expedited cleanup. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Small molecule crystals with 1D water wires modulate electronic properties of surface water networks

This work evaluates the effect of water vapor pressure on the surface conductivity of a hydrated small molecule crystal under dry and wet conditions. The crystal contains a discrete metal complex with waters of hydration in structural positions and in 1D water wires. Under dry conditions, the surface of the hydrated crystal behaves as a near ideal capacitor. Under wet conditions, the waters of hydration serve as a template to assemble surface water networks by organization of water from the environment. The surface water provides pathways for charge transfer attributed to proton mobility that increases conductivity by > 3 orders in magnitude. Modulation between wet and dry conditions tunes the conductivity of the crystal surface via a two-step process, which has been evaluated kinetically as a function of the water vapor pressure.

36 MATERIALS SCIENCE↗

A new stationary gridline artifact suppression method based on the 2D discrete wavelet transform

Purpose: In digital x-ray radiography, an antiscatter grid is inserted between the patient and the image receptor to reduce scattered radiation. If the antiscatter grid is used in a stationary way, gridline artifacts will appear in the final image. In most of the gridline removal image processing methods, the useful information with spatial frequencies close to that of the gridline is usually lost or degraded. In this study, a new stationary gridline suppression method is designed to preserve more of the useful information. Methods: The method is as follows. The input image is first recursively decomposed into several smaller subimages using a multiscale 2D discrete wavelet transform. The decomposition process stops when the gridline signal is found to be greater than a threshold in one or several of these subimages using a gridline detection module. An automatic Gaussian band-stop filter is then applied to the detected subimages to remove the gridline signal. Finally, the restored image is achieved using the corresponding 2D inverse discrete wavelet transform. Results: The processed images show that the proposed method can remove the gridline signal efficiently while maintaining the image details. The spectra of a 1D Fourier transform of the processed images demonstrate that, compared with some existing gridline removal methods, the proposed method has better information preservation after the removal of the gridline artifacts. Additionally, the performance speed is relatively high. Conclusions: The experimental results demonstrate the efficiency of the proposed method. Compared with some existing gridline removal methods, the proposed method can preserve more information within an acceptable execution time.

60 APPLIED LIFE SCIENCES↗

Feasibility of processing the experimental breeder reactor-II driver fuel from the Idaho National Laboratory through Savannah River Site's H-Canyon facility

Savannah River National Laboratory (SRNL) was requested to evaluate the potential to receive and process the Idaho National Laboratory (INL) uranium (U) recovered from the Experimental Breeder Reactor II (EBR-II) driver fuel through the Savannah River Site’s (SRS) H-Canyon as a way to disposition the material. INL recovers the uranium from the sodium bonded metallic fuel irradiated in the EBR-II reactor using an electrorefining process. There were two compositions of EBR-II driver fuel. The early generation fuel was U-5Fs, which consisted of 95% U metal alloyed with 5% noble metal elements “fissium” (2.5% molybdenum, 2.0% ruthenium, 0.3% rhodium, 0.1% palladium, and 0.1% zirconium), while the later generation was U-10Zr which was 90% U metal alloyed with 10% zirconium. A potential concern during the H-Canyon nitric acid dissolution process of the U metal containing zirconium (Zr) is the explosive behavior that has been reported for alloys of these materials. For this reason, this evaluation was focused on the ability to process the lower Zr content materials, the U-5Fs material.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Effects of stratification and charge cooling on combustion in a gasoline direct-injection compression ignition (GDCI) engine

With the development of low temperature engine combustion strategies, performance of gasoline-type fuels under compression ignition conditions has attracted extensive research interest. Meanwhile, for the sake of co-optimization of engines and fuels for future ground transportation, identification and evaluation of general fuel properties should be a core research priority instead of endless testing of specific fuels. In this study, the roles of fuel octane sensitivity in characterizing the ignition performance of gasoline surrogates have been systematically investigated under typical gasoline direct ignition compression ignition (GDCI) engine conditions using 3D combustion CFD simulation, especially considering the subsequent in-cylinder charge stratification and charge cooling. Two different operating conditions, high boost pressure low boost temperature (beyond-RON) case and low boost pressure high boost temperature (beyond-MON) case, were considered. By comparing with our previous zero-dimensional chemical kinetic study of gasoline surrogates in advanced compression ignition (ACI) engines, the effects of stratification and charge cooling on the combustion processes are investigated. It is found that different fuel octane sensitivities lead to slight difference in equivalence ratio stratification and charge cooling due to differences in volatility. However, fuel reactivity is still the more dominant factor than the stratification and charge cooling effects in determining combustion phasing. The present results help to justify the P-T domain framework for engine autoignition analysis of overlapping pressure-temperature trajectory with ignition delay iso-contour. Here, the results also provide useful guidance to the understanding of GCI combustion process, and to the evaluation of controlling fuel properties and the selection of alternative fuels in GCI engines.

33 ADVANCED PROPULSION SYSTEMS↗

Citric Acid Functionalized Natural Fibers to Enhance Thermal Stability and Moisture Resistance in Polylactic Acid Composites

Cellulosic fibers can impart many unique benefits into composite applications, such as reduced weight or structural reinforcement; however, these materials also increase hygroscopicity and decrease thermal stability, restricting broader applications. The present work adapted an experimental process for functionalizing the cellulose surface using citric acid (CA) for three fibers: a 100% cellulose bleached soft Kraft pulp (e.g., creafill) and two natural fibers with similar composition but different fiber morphology, flax fiber and banana fiber. The process uses CA with a sodium hypophosphite (SHP) catalyst to chemically functionalize fiber surfaces, and the reaction mechanism was investigated through Fourier Transform Infrared Spectroscopy (FTIR), which suggested a grafting mechanism rather than a surface-based crosslinking between neighboring sites. Functionalized fibers were compounded into polylactic acid (PLA) at 20 wt.% to better understand how this functionalization might impact critical performance properties like thermal stability, crystallization, thermal mechanical properties, and water uptake of these composites. The study demonstrated varying levels of efficacy for the functionalization of cellulosic fibers with CA/SHP and the fiber with the most open microstructure, e.g., banana fiber, exhibited the largest change in its properties with a 38% reduction in water uptake compared to untreated banana fiber composites. Parallel evaluation of the functionalization process for different fibers demonstrates the importance of fiber morphology on surface modification and can enable their use in composites by demonstrating the efficacy of this potentially low-cost, low-toxicity method for reducing hygroscopicity and improving thermal stability.

citric acid↗

Simulation of Space Experiments for Nuclear Planetology: Measurement of Relative Intensities of Lines of Gamma Ray Emitted upon Thermal-Neutron Capture by Nuclei

The results obtained by experimentally studying gamma rays emitted by samples prepared as analogs of planetary matter and irradiated with thermal neutrons are presented. The intensities of spectral lines of gamma rays emitted by such samples differing in chemical composition are compared. These results will be used in processing data on gamma-ray spectra of the Moon and Mercury from measurements performed onboard spacecrafts with the aim of studying the composition of the surface of these celestial bodies.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Modular Energy Storage System for Alternative Energy Vehicles

An electrical vehicle environment was established to promote research and technology development in the area of high power energy management. The project incorporates a topology that permits parallel development of an alternative energy delivery system and an energy storage system. The objective of the project is to develop technologies, specifically power electronics, energy storage electronics and controls that provide efficient and effective energy management between electrically powered devices in alternative energy vehicles plugin electric vehicles, hybrid vehicles, range extended vehicles, and hydrogen-based fuel cell vehicles. In order to meet the project objectives, the Vehicle Energy Management System (VEMS) was defined and subsystem requirements were obtained. Afterwards, power electronics, energy storage electronics and controls were designed. Finally, these subsystems were built, tested individually, and integrated into an electric vehicle system to evaluate and optimize the subsystems performance. Phase 1 of the program established the fundamental test bed to support development of an electrical environment ideal for fuel cell application and the mitigation of many shortcomings of current fuel cell technology. Phase 2, continued development from Phase 1, focusing on implementing subsystem requirements, design and construction of the energy management subsystem, and the integration of this subsystem into the surrogate electric vehicle. Phase 2 also required the development of an Alternative Energy System (AES) capable of emulating electrical characteristics of fuel cells, battery, gen set, etc. Under the scope of the project, a boost converter that couples the alternate energy delivery system to the energy storage system was developed, constructed and tested. Modeling tools were utilized during the design process to optimize both component and system design. This model driven design process enabled an iterative process to track and evaluate the impact of design alternatives and the impact of changes. Refinement of models was accomplished through correlation studies to measured data obtained from functioning hardware. Specifically, correlation and characterization of the boost converter resulted in a model that was effectively used to determine overall VEMS performance. The successful development of the boost converter can be attributed to utilization of previously proven technologies and adapting to meet the VEMS requirements. This program provided significant improvement in development time of various generations of boost converters. The software strategies and testing results support the development of current energy management systems and directly contribute to the future of similar, commercial products at Magna E-Car Systems. Because of this development project, Magna E-Car Systems is able to offer automotive customers a boost converter system with reduced time to market and decreased product cost, thus transferring the cost and timing benefits to the end use consumer.

25 ENERGY STORAGE↗

Laser-based Adhesion Strength Measurements for Advanced Manufactured Sensor Adhesion Characterization

As part of the advanced manufacturing (AM) focus of the Nuclear Energy Enabling Technologies (NEET) Advanced Sensors and Instrumentation (ASI) Program, new in-house capabilities to fabricate and test novel active and passive in-pile sensors are being developed at the Idaho National Laboratory (INL). These advanced-manufactured sensors, such as peak temperature monitors and strain gauges, must be miniaturized and be able to provide real-time feedback, all while withstanding the coupled extremes of temperature and radiation typically observed in the environment of a nuclear test reactor. Fabricating durable and robust in-pile sensors with high fidelity requires a fundamental understanding of the optimal process parameters required in the AM technique, as well as the effect of deviations from optimal process parameters on the sensor-substrate adhesion and the ultimate robustness of the sensor. This report presents the use of a laser-based, non-contact approach for characterizing the adhesion of sensors manufactured using aerosol jet printing (AJP) that comprise of silver nanoparticle inks deposited and sintered on austenitic stainless steel (SS) 316L substrates. The effect of substrate surface roughness, surface energy, ink sintering time and sintering temperature on the printed sensor topography and the sensor-substrate adhesion has been systematically investigated. The sensor-substrate interfacial adhesion strength was determined from the critical laser pulse energy required to ablate/ detach the printed sensor from the substrate. Measurements of the sensor topography using optical profilometry revealed that post-deposition sintering at 400°C for 60 minutes resulted in a reduced sensor thickness, while plasma treating the substrate prior to printing yielded sensors with uniform thickness profiles transverse to the print direction. Sintering temperature and time were found to be the dominant parameters that affected sensor-substrate adhesion, followed by substrate surface roughness. The laser-based and locally-destructive approach for adhesion strength measurements shows promise in overcoming the challenges and limitations of current standardized peel tests and can be developed into a post-fabrication process control protocol for evaluating the durability of AM printed sensors in a high throughput fashion. Future studies will investigate the use of laser-induced spallation techniques as well as non-destructive laser ultrasonic methods for sensor-substrate adhesion characterization on sensors manufactured using AJP and other AM techniques such as plasma jet printing (PJP).

42 ENGINEERING↗

Predicting lake surface water phosphorus dynamics using process-guided machine learning

Phosphorus (P) loading to lakes is degrading the quality and usability of water globally. Accurate predictions of lake P dynamics are needed to understand whole-ecosystem P budgets, as well as the consequences of changing lake P concentrations for water quality. However, complex biophysical processes within lakes, along with limited observational data, challenge our capacity to reproduce short-term lake dynamics needed for water quality predictions, as well as long-term dynamics needed to understand broad scale controls over lake P. Here we use an emerging paradigm in modeling, process-guided machine learning (PGML), to produce a phosphorus budget for Lake Mendota (Wisconsin, USA) and to accurately predict epilimnetic phosphorus over a time range of days to decades. In our implementation of PGML, which we term a Process-Guided Recurrent Neural Network (PGRNN), we combine a process-based model for lake P with a recurrent neural network, and then constrain the predictions with ecological principles. We test independently the process-based model, the recurrent neural network, and the PGRNN to evaluate the overall approach. The process-based model accounted for most of the observed pattern in lake P; however it missed the long-term trend in lake P and had the worst performance in predicting winter and summer P in surface waters. The root mean square error (RMSE) for the process-based model, the recurrent neural network, and the PGRNN was 33.0 μg P L -1 , 22.7 μg P L -1 , and 20.7 μg P L -1 , respectively. All models performed better during summer, with RMSE values for the three models (same order) equal to 14.3 μg P L -1 , 10.9 μg P L -1 , and 10.7 μg P L -1 . Although the PGRNN had only marginally better RMSE during summer, it had lower bias and reproduced long-term decreases in lake P missed by the other two models. For all seasons and all years, the recurrent neural network had better predictions than process alone, with root mean square error (RMSE) of 23.8 μg P L -1 and 28.0 μg P L -1 , respectively. The output of PGRNN indicated that new processes related to water temperature, thermal stratification, and long term changes in external loads are needed to improve the process model. By using ecological knowledge, as well as the information content of complex data, PGML shows promise as a technique for accurate prediction in messy, real-world ecological dynamics, while providing valuable information that can improve our understanding of process.

54 ENVIRONMENTAL SCIENCES↗