Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model error”

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 685 records · Page 38

Improving Seasonal Forecast Using Probabilistic Deep Learning

The path toward realizing the potential of seasonal forecasting and its socioeconomic benefits relies on improving general circulation model (GCM) based dynamical forecast systems. To improve dynamical seasonal forecasts, it is crucial to set up forecast benchmarks, and clarify forecast limitations posed by model initialization errors, formulation deficiencies, and internal climate variability. With huge costs in generating large forecast ensembles, and limited observations for forecast verification, the seasonal forecast benchmarking and diagnosing task proves challenging. Here, we develop a probabilistic deep learning-based statistical forecast methodology, drawing on a wealth of climate simulations to enhance seasonal forecast capability and forecast diagnosis. By explicitly modeling the internal climate variability and GCM formulation differences, the proposed Conditional Generative Forecasting (CGF) methodology enables bypassing crucial barriers in dynamical forecast, and offers a top-down viewpoint to examine how complicated GCMs encode the seasonal predictability information. We apply the CGF methodology for global seasonal forecast of precipitation and 2 m air temperature, based on a unique data set consisting 52,201 years of climate simulation. Results show that the CGF methodology can faithfully represent the seasonal predictability information encoded in GCMs. We successfully apply this learned relationship in real-world seasonal forecast, achieving competitive performance compared to dynamical forecasts. Using this CGF as benchmark, we reveal the impact of insufficient forecast spread sampling that limits the skill of the considered dynamical forecast system. Finally, we introduce different strategies for composing ensembles using the CGF methodology, highlighting the potential for leveraging the strengths of multiple GCMs to achieve advantgeous seasonal forecast.

54 ENVIRONMENTAL SCIENCES↗

Classical Benchmarks for Variational Quantum Eigensolver Simulations of the Hubbard Model

Simulating the Hubbard model is of great interest to a wide range of applications within condensed matter physics, however its solution on classical computers remains challenging in dimensions larger than one. The relative simplicity of this model, embodied by the sparseness of the Hamiltonian matrix, allows for its efficient implementation on quantum computers, and for its approximate solution using variational algorithms such as the variational quantum eigensolver. While these algorithms have been shown to reproduce the qualitative features of the Hubbard model, their quantitative accuracy in terms of producing true ground state energies and other properties, and the dependence of this accuracy on the system size and interaction strength, the choice of variational ansatz, and the degree of spatial inhomogeneity in the model, remains unknown. Here we present a rigorous classical benchmarking study, demonstrating the potential impact of these factors on the accuracy of the variational solution of the Hubbard model on quantum hardware, for systems with up to 32 qubits. We find that even when using the most accurate wavefunction ansätze for the Hubbard model, the error in its ground state energy and wavefunction plateaus for larger lattices, while stronger electronic correlations magnify this issue. Concurrently, spatially inhomogeneous parameters and the presence of off-site Coulomb interactions only have a small effect on the accuracy of the computed ground state energies. Our study highlights the capabilities and limitations of current approaches for solving the Hubbard model on quantum hardware, and we discuss potential future avenues of research.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Hyper Suprime-Cam Year 3 results: Cosmology from cosmic shear power spectra

Here, we measure weak lensing cosmic shear power spectra from the 3-year galaxy shear catalog of the Hyper Suprime-Cam (HSC) Subaru Strategic Program imaging survey. The shear catalog covers 416 deg 2 of the northern sky, with a mean i-band seeing of 0.59 arcsec and an effective galaxy number density of 15 arcmin —2 within our adopted redshift range. With an i-band magnitude limit of 24.5 mag, and four tomographic redshift bins spanning 0.3 ≤ z ph ≤ 1.5 based on photometric redshifts, we obtain a high-significance measurement of the cosmic shear power spectra, with a signal-to-noise ratio of approximately 26.4 in the multipole range 300 < ℓ < 1800. The accuracy of our power spectrum measurement is tested against realistic mock shear catalogs, and we use these catalogs to get a reliable measurement of the covariance of the power spectrum measurements. We use a robust blinding procedure to avoid confirmation bias, and model various uncertainties and sources of bias in our analysis, including point spread function systematics, redshift distribution uncertainties, the intrinsic alignment of galaxies and the modeling of the matter power spectrum. For a flat Λ CDM model, we find S 8 ≡ σ 8 (Ω m /0.3) 0.5 = ${0.776}_{—0.033}^{+0.032}$, which is in excellent agreement with the constraints from the other HSC Year 3 cosmology analyses, as well as those from a number of other cosmic shear experiments. This result implies a ~2σ-level tension with the Planck 2018 cosmology. We study the effect that various systematic errors and modeling choices could have on this value, and find that they can shift the best-fit value of S 8 by no more than ~ 0.5σ, indicating that our result is robust to such systematics.

79 ASTRONOMY AND ASTROPHYSICS↗

ORBITS OF NEARBY PLANETARY NEBULAE AND THEIR INTERACTION WITH THE INTERSTELLAR MEDIUM

We present and analyze the orbits of eight nearby planetary nebulae (PNs) using two different Galactic models. The errors of the derived orbital parameters are determined with a Monte Carlo method. Based on the derived orbital parameters, we find that Sh 2-216, DeHt 5, NGC 7293, A21, and Ton 320 belong to the thin-disk population, and PG 1034 + 001 and A31 belong to the thick-disk population. PuWe 1 probably belongs to the thick-disk population, but its population classification is very uncertain due to the large errors of its derived orbital parameters. The PN-interstellar medium (ISM) interactions are observed for the eight PNs in our sample. The position angles of the proper motions of the PNs are consistent with the directions of the PN-ISM interaction regions. The kinematic ages of PNs are much smaller than the time for them to cross the Galactic plane. Using the models of Borkowski et al. and Soker et al., the PN-ISM interaction can be used to derive the local density of ISM in the vicinity of evolved PNs. According to the three-dimensional hydrodynamic simulations of Wareing et al. (WZO), Sh 2-216, A21, and Ton 320 are in the WZO 3 stage, PG 1034 + 001 and NGC 7293 are in the WZO 1 stage, and PuWe 1 is in the WZO 2 stage.

79 ASTRONOMY AND ASTROPHYSICS↗

Probabilistic error estimation for non-intrusive reduced models learned from data of systems governed by linear parabolic partial differential equations

This work derives a residual-based a posteriori error estimator for reduced models learned with non-intrusive model reduction from data of high-dimensional systems governed by linear parabolic partial differential equations with control inputs. It is shown that quantities that are necessary for the error estimator can be either obtained exactly as the solutions of least-squares problems in a non-intrusive way from data such as initial conditions, control inputs, and high-dimensional solution trajectories or bounded in a probabilistic sense. Here, the computational procedure follows an offline/online decomposition. In the offline (training) phase, the high-dimensional system is judiciously solved in a black-box fashion to generate data and to set up the error estimator. In the online phase, the estimator is used to bound the error of the reduced-model predictions for new initial conditions and new control inputs without recourse to the high-dimensional system. Numerical results demonstrate the workflow of the proposed approach from data to reduced models to certified predictions.

97 MATHEMATICS AND COMPUTING↗

Dark Energy Survey Year 3 results: Constraints on extensions to Λ CDM with weak lensing and galaxy clustering

We constrain six possible extensions to the Λ cold dark matter (CDM) model using measurements from the Dark Energy Survey’s first three years of observations, alone and in combination with external cosmological probes. The DES data are the two-point correlation functions of weak gravitational lensing, galaxy clustering, and their cross-correlation. We use simulated data vectors and blind analyses of real data to validate the robustness of our results to astrophysical and modeling systematic errors. In many cases, constraining power is limited by the absence of theoretical predictions beyond the linear regime that are reliable at our required precision. The Λ CDM extensions are dark energy with a time-dependent equation of state, nonzero spatial curvature, additional relativistic degrees of freedom, sterile neutrinos with eV-scale mass, modifications of gravitational physics, and a binned σ 8 ( z ) model which serves as a phenomenological probe of structure growth. For the time-varying dark energy equation of state evaluated at the pivot redshift we find ( w p , w a ) = ( - 0.9 9 - 0.17 + 0.28 , - 0.9 ± 1.2 ) at 68% confidence with z p = 0.24 from the DES measurements alone, and ( w p , w a ) = ( - 1.0 3 - 0.03 + 0.04 , - 0. 4 - 0.3 + 0.4 ) with z p = 0.21 for the combination of all data considered. Curvature constraints of Ω k = 0.0009 ± 0.0017 and effective relativistic species N eff = 3.1 0 - 0.16 + 0.15 are dominated by external data, though adding DES information to external low-redshift probes tightens the Ω k constraints that can be made without cosmic microwave background observables by 20%. For massive sterile neutrinos, DES combined with external data improves the upper bound on the mass m eff by a factor of 3 compared to previous analyses, giving 95% limits of ( Δ N eff , m eff ) ≤ ( 0.28 , 0.20 eV ) when using priors matching a comparable Planck analysis. For modified gravity, we constrain changes to the lensing and Poisson equations controlled by functions Σ ( k , z ) = Σ 0 Ω Λ ( z ) / Ω Λ , 0 and μ ( k , z ) = μ 0 Ω Λ ( z ) / Ω Λ , 0 , respectively, to Σ 0 = 0. 6 - 0.5 + 0.4 from DES alone and ( Σ 0 , μ 0 ) = ( 0.04 ± 0.05 , 0.0 8 - 0.19 + 0.21 ) for the combination of all data, both at 68% confidence. Overall, we find no significant evidence for physics beyond Λ CDM .

79 ASTRONOMY AND ASTROPHYSICS↗

Life Prediction of Large Lithium-Ion Battery Packs with Active and Passive Balancing

Lithium-ion battery packs take a major part of large-scale stationary energy storage systems. One challenge in reducing battery pack cost is to reduce pack size without compromising pack service performance and lifespan. Prognostic life model can be a powerful tool to handle the state of health (SOH) estimate and enable active life balancing strategy to reduce cell imbalance and extend pack life. This work proposed a life model using both empirical and physical-based approaches. The life model described the compounding effect of different degradations on the entire cell with an empirical model. Then its lower-level submodels considered the complex physical links between testing statistics (state of charge level, C-rate level, duty cycles, etc.) and the degradation reaction rates with respect to specific aging mechanisms. The hybrid approach made the life model generic, robust and stable regardless of battery chemistry and application usage. The model was validated with a custom pack with both passive and active balancing systems implemented, which created four different aging paths in the pack. The life model successfully captured the aging trajectories of all four paths. The life model prediction errors on capacity fade and resistance growth were within +/-3% and +/-5% of the experiment measurements.

25 ENERGY STORAGE↗

On the application of the expected log-likelihood gain to decision making in molecular replacement

Molecular-replacement phasing of macromolecular crystal structures is often fast, but if a molecular-replacement solution is not immediately obtained the crystallographer must judge whether to pursue molecular replacement or to attempt experimental phasing as the quickest path to structure solution. The introduction of the expected log-likelihood gain has given the crystallographer a powerful new tool to aid in making this decision. The eLLG is the log-likelihood gain on intensity expected from a correctly placed model. It is calculated as a sum over the reflections of a function dependent on the fraction of the scattering for which the model accounts, the estimated model coordinate error and the measurement errors in the data. It is shown how the eLLG may be used to answer the question 'can I solve my structure by molecular replacement?'. However, this is only the most obvious of the applications of the eLLG. It is also discussed how the eLLG may be used to determine the search order and minimal data requirements for obtaining a molecular-replacement solution using a given model, and for decision making in fragment-based molecular replacement, single-atom molecular replacement and likelihood-guided model pruning.

59 BASIC BIOLOGICAL SCIENCES↗

Assessment of Crustal Leg Uncertainties for RSTT Path-dependent Uncertainty

The Regional Seismic Travel Time (RSTT) model and software were recently updated to allow for path-dependent uncertainty (PDU) estimates for travel-time (Begnaud et al. 2021; Begnaud et al. 2020). Begnaud et al. (2021) describes the process of improving the current distance-dependent uncertainty parameterization for RSTT using a random effects model (REM) to estimate slowness (inverse velocity) uncertainty as a mean squared error for each model parameter. The REM separates the error between observed slowness and model predicted slowness into bias and random components. The PDU is calculated by integrating these mean squared errors along a seismic-phase ray path.

58 GEOSCIENCES↗

Model predictive control and optimization for battery charging and discharging

An apparatus for model predictive control (“MPC”) is disclosed. A method and system also perform the functions of the apparatus. The apparatus includes a measurement module that receives battery status information from one or more sensors receiving information from a battery cell, and a Kalman filter module that uses a Kalman filter and the battery status information to provide a state estimate vector. The apparatus includes a battery model module that inputs the state estimate vector and battery status information into a battery model and calculates a battery model output, the battery model representing the battery cell, and an MPC optimization module that inputs one or more battery model outputs and an error signal in a model predictive control algorithm to calculate an optimal response. The optimal response includes a modification of the error signal.

Trimboli, Michael Scott↗

Lower-tropospheric CO 2 from near-infrared ACOS-GOSAT observations

We present two new products from near-infrared Greenhouse Gases Observing Satellite (GOSAT) observations: lowermost tropospheric (LMT, from 0 to 2.5 km) and upper tropospheric–stratospheric ( U , above 2.5 km) carbon dioxide partial column mixing ratios. We compare these new products to aircraft profiles and remote surface flask measurements and find that the seasonal and year-to-year variations in the new partial column mixing ratios significantly improve upon the Atmospheric CO 2 Observations from Space (ACOS) and GOSAT (ACOS-GOSAT) initial guess and/or a priori, with distinct patterns in the LMT and U seasonal cycles that match validation data. For land monthly averages, we find errors of 1.9, 0.7, and 0.8 ppm for retrieved GOSAT LMT, U , and X CO 2 ; for ocean monthly averages, we find errors of 0.7, 0.5, and 0.5 ppm for retrieved GOSAT LMT, U , and X CO 2 . In the southern hemispheric biomass burning season, the new partial columns show similar patterns to MODIS fire maps and MOPITT multispectral CO for both vertical levels, despite a flat ACOS-GOSAT prior, and a CO–CO 2 emission factor comparable to published values. The difference of LMT and U , useful for evaluation of model transport error, has also been validated with a monthly average error of 0.8 (1.4) ppm for ocean (land). LMT is more locally influenced than U , meaning that local fluxes can now be better separated from CO 2 transported from far away.

54 ENVIRONMENTAL SCIENCES↗

A Grid Intensity Map-Based Source Model for Monte Carlo Modeling of IMRT Field Irradiation - Paper 5

An accurate source model is important for TPS (Treatment Planning System) dose calculation and verification, especially for IMRT (Intensity Modulated Radiotherapy) irregular and inhomogeneous fields. Monte Carlo simulation is the most accurate way to calculate dose distributions in the heterogeneous media like human bodies. A full radiotherapy modeling generally includes Linac modeling and phantom dose calculation. Full Linac modeling is very time-consuming and requires highly detailed knowledge of the accelerator composition and structure, however, which is very difficult to obtain for commercial interests. Often all relevant parameters are not known with sufficient accuracy, and thus a time-consuming tuning of e.g. the electron beam spot size and energy must be performed. A multiple source model (MSM) regards particles coming from different parts of an accelerator as if they are coming from different virtual sources. MSM uses a set of virtual sources representing the treatment beam (e.g. primary photons, head scatter, electron contamination) by a set of analytic functions or histograms with parameters derived from dosimetric measurements or a full MC simulation. Tillikainen et al built a three sub-source model by matching with measurement data using an automatic process to optimize the free model parameters for a Varian 2100 C/D accelerator. Lin et al have built a convenient five-photon-sub-source model for Varian 2300C 6 MV-X beam based on Monte Carlo simulated Linac for ARTS (Accurate Radiotherapy System). The particle generation of MSM is very fast compared to full Linac simulation, and its sampling efficiency is not limited by the size of the phase space file (PSF). However, generally MSM is generated based on regular field PSF and needs an irregular component to form the beam shape meanwhile apt to bringing error Linac modeling and MSM producing from PSF analysis actually both are multi-freedom optimization process by adjusting several parameters (e.g. electron energy and angular divergence). Linac modeling adjusts its parameters from the beginning of electron hitting target through all components to fit the simulation to the measurement. MSM tunes its PSF-derived parameters by optimizing to minimize the deviations between the simulation and the measurement. Since the full Linac modeling and the MSM model both need to fine tune their parameters to match the simulation to the measurement, and sometimes maybe offsets the real accelerator for its multi-parameter optimization process. So, why not fine tune directly all source model parameters just behind all Linac components to fit the simulation to the measurement? A grid intensity-based surface source model was suggested here, which assumed the particle emitting from the bottom surface of Linac last beam trimmer MLC (Multi Leaf Collimator). The particle emitting position is sampled on the grid intensity map optimized by TPS, and the particle weight is adjusted by its original grid intensity. This source model not only takes advantage of the Monte Carlo method in the heterogeneous media, but also avoids effectively the transportation speed shortcoming of Monte Carlo simulation through Linac components. (authors)

61 RADIATION PROTECTION AND DOSIMETRY↗

Quantifying structural errors in cloud condensation nuclei activity from reduced representation of aerosol size distributions

Aerosol effects on clouds and radiation are the dominant contribution to uncertainty in radiative forcing relative to the pre-industrial atmosphere. While previous studies have assessed the impact of parametric uncertainty on modeled forcing, structural errors from the numerical representation of particle distributions have not been well quantified. Here we present a framework for quantifying error in aerosol size distributions and cloud condensation nuclei activity, which we apply to the widely used 4-mode version of the Modal Aerosol Module (MAM4). Box model predictions from the MAM4 are evaluated against the Particle Monte Carlo Model for Simulating Aerosol Interactions and Chemistry (PartMC-MOSAIC), a benchmark model that tracks the evolution of individual particles. We show that size distributions simulated by MAM4 diverge from those simulated by PartMC-MOSAIC after only a few hours of aging by condensation and coagulation in polluted conditions, which leads to large errors in modeled cloud condensation nuclei concentrations. We find that differences between MAM4 and PartMC-MOSAIC are largest under polluted conditions, where the size distribution evolves rapidly though aging by condensation of semi-volatile substances and coagulation among particles. These findings suggest that structural error in modeled aerosol properties contributes to the large inter-model variability in aerosol radiative forcing.

Fierce, Laura M.↗

A Logarithmic Bayesian Approach to Quantum Error Detection

We consider the problem of continuous quantum error correction from a Bayesian perspective, proposing a pair of digital filters using logarithmic probabilities that are able to achieve near-optimal performance on a three-qubit bit-flip code, while still being reasonable to implement on low-latency hardware. These practical filters are approximations of an optimal filter that we derive explicitly for finite time steps, in contrast with previous work that has relied on stochastic differential equations such as the Wonham filter. By utilizing logarithmic probabilities, we are able to eliminate the need for explicit normalization and can reduce the Gaussian noise distribution to a simple quadratic expression. The state transitions induced by the bit-flip errors are modeled using a Markov chain, which for log-probabilities must be evaluated using a LogSumExp function. We develop the two versions of our filter by constraining this LogSumExp to have either one or two inputs, which favors either simplicity or accuracy, respectively. Using simulated data, we demonstrate that the single-term and two-term filters are able to significantly outperform both a double threshold scheme and a linearized version of the Wonham filter in tests of error detection under a wide variety of error rates and time steps.

97 MATHEMATICS AND COMPUTING↗

Applications of the Dulmage–Mendelsohn decomposition for debugging nonlinear optimization problems

Nonlinear modeling and optimization is a valuable tool for aiding decisions by engineering practitioners, but programming an optimization problem based on a complex electrical, mechanical, or chemical process is a time-consuming and error-prone activity. Therefore, there is a need for model analysis and debugging tools that can detect and diagnose modeling errors. One such tool is the Dulmage–Mendelsohn decomposition, which identifies structurally under- and over-determined subsets in systems of equations and variables by partitioning the bipartite graph of the system. This work provides the necessary background to understand the Dulmage–Mendelsohn decomposition and its application to the analysis of nonlinear optimization problems, demonstrates its use in diagnosing a variety of modeling errors, and introduces software implementations for analyzing nonlinear optimization problems in the Pyomo and JuMP algebraic modeling languages.

42 ENGINEERING↗

Symmetry tuning via controlled crossed-beam energy transfer on the National Ignition Facility

The hohlraum energetics experimental campaign started in the summer of 2009 on the national Ignition Facility (NIF) (cf. N. Meezan et al., Phys. Plasmas, this issue). These experiments have demonstrated controlled crossed-beam transfer as an efficient and robust tool to tune the implosion symmetry of ignition capsules. A new linear model calculating crossed-beam energy transfer between cones of beams on the NIF has been developed. The model has been applied to the emulator targets shot for the National Ignition Campaign in 2009. These targets have no lip liner at the laser entrance holes, which reduces the amount of energy transfer between laser beams and requires a larger wavelength separation between the cones of laser beams in order to achieve the best symmetry (with no net transfer between the inner and outer cones). A good agreement is found between the calculations and measurements; remaining discrepancies can be explained by a 10% error in modeling the flow at the laser entrance hole of ignitions hohlraums.

42 ENGINEERING↗

An open-access CMIP5 pattern library for temperature and precipitation: Description and methodology

Pattern scaling is used to efficiently emulate general circulation models and explore uncertainty in climate projections under multiple forcing scenarios. Pattern scaling methods assume that local climate changes scale with a global mean temperature increase, allowing for spatial patterns to be generated for multiple models for any future emission scenario. For uncertainty quantification and probabilistic statistical analysis, a library of patterns with descriptive statistics for each file would be beneficial, but such a library does not presently exist. Of the possible techniques used to generate patterns, the two most prominent are the delta and least squared regression methods. We explore the differences and statistical significance between patterns generated by each method and assess performance of the generated patterns across methods and scenarios. Differences in patterns across seasons between methods and epochs were largest in high latitudes (60-90°N/S). Bias and mean errors between modeled and pattern predicted output from the linear regression method were smaller than patterns generated by the delta method. Across scenarios, differences in the linear regression method patterns were more statistically significant, especially at high latitudes. We found that pattern generation methodologies were able to approximate the forced signal of change to within ≤ 0.5°C, but choice of pattern generation methodology for pattern scaling purposes should be informed by user goals and criteria. As a result, this paper describes our library of least squared regression patterns from all CMIP5 models for temperature and precipitation on an annual and sub-annual basis, along with the code used to generate these patterns.

54 ENVIRONMENTAL SCIENCES↗

Scalable deep learning for watershed model calibration

Watershed models such as the Soil and Water Assessment Tool (SWAT) consist of high-dimensional physical and empirical parameters. These parameters often need to be estimated/calibrated through inverse modeling to produce reliable predictions on hydrological fluxes and states. Existing parameter estimation methods can be time consuming, inefficient, and computationally expensive for high-dimensional problems. In this paper, we present an accurate and robust method to calibrate the SWAT model (i.e., 20 parameters) using scalable deep learning (DL). We developed inverse models based on convolutional neural networks (CNN) to assimilate observed streamflow data and estimate the SWAT model parameters. Scalable hyperparameter tuning is performed using high-performance computing resources to identify the top 50 optimal neural network architectures. We used ensemble SWAT simulations to train, validate, and test the CNN models. We estimated the parameters of the SWAT model using observed streamflow data and assessed the impact of measurement errors on SWAT model calibration. We tested and validated the proposed scalable DL methodology on the American River Watershed, located in the Pacific Northwest-based Yakima River basin. Our results show that the CNN-based calibration is better than two popular parameter estimation methods (i.e., the generalized likelihood uncertainty estimation [GLUE] and the dynamically dimensioned search [DDS], which is a global optimization algorithm). For the set of parameters that are sensitive to the observations, our proposed method yields narrower ranges than the GLUE method but broader ranges than values produced using the DDS method within the sampling range even under high relative observational errors. The SWAT model calibration performance using the CNNs, GLUE, and DDS methods are compared using R 2 and a set of efficiency metrics, including Nash-Sutcliffe, logarithmic Nash-Sutcliffe, Kling-Gupta, modified Kling-Gupta, and non-parametric Kling-Gupta scores, computed on the observed and simulated watershed responses. The best CNN-based calibrated set has scores of 0.71, 0.75, 0.85, 0.85, 0.86, and 0.91. The best DDS-based calibrated set has scores of 0.62, 0.69, 0.8, 0.77, 0.79, and 0.82. The best GLUE-based calibrated set has scores of 0.56, 0.58, 0.71, 0.7, 0.71, and 0.8. The scores above show that the CNN-based calibration leads to more accurate low and high streamflow predictions than the GLUE and DDS sets. Our research demonstrates that the proposed method has high potential to improve our current practice in calibrating large-scale integrated hydrologic models.

54 ENVIRONMENTAL SCIENCES↗