SEARCH · Engineering Papers

Results for “statistical modeling”

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.

SourceResearch connectionFrom yearThrough year

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

At least 397 records · Page 22

All sources NASA NTRS DOE OSTI

NASA NTRS1985

Use of observational and model-derived fields and regime model output statistics in mesoscale forecasting

Various empirical and statistical weather-forecasting studies which utilize stratification by weather regime are described. Objective classification was used to determine weather regime in some studies. In other cases the weather pattern was determined on the basis of a parameter representing the physical and dynamical processes relevant to the anticipated mesoscale phenomena, such as low level moisture convergence and convective precipitation, or the Froude number and the occurrence of cold-air damming. For mesoscale phenomena already in existence, new forecasting techniques were developed. The use of cloud models in operational forecasting is discussed. Models to calculate the spatial scales of forcings and resultant response for mesoscale systems are presented. The use of these models to represent the climatologically most prevalent systems, and to perform case-by-case simulations is reviewed. Operational implementation of mesoscale data into weather forecasts, using both actual simulation output and method-output statistics is discussed.

Forbes, G. S.↗

NASA NTRSDate not supplied

Multi-source Estimates of Land / Ocean Moisture Transport Variability over the Satellite Era

It is widely appreciated that atmospheric transport of water from the world’s oceans is a process key to planetary energy balance as well as Earth's habitability. What is not yet clear is the extent of variability in moisture transports, the relative importance of interdecadal variability versus climate change signals, and importantly, our ability to quantify these changes. This work assesses variations in moisture transport variability during the satellite era (~1980 to present) by comparing several different estimates. (i) The most direct estimate is the vertically integrated flux convergence of moisture from reanalyses which use observed wind and moisture information. (ii) One alternative estimate comes from P-ET over land taken from global hydrologic models constrained with precipitation and near-surface meteorology. Here we use an ensemble of six models. An adjunct to this method is to employ satellite derive ET (e.g., GLEAM or DOLCE). (iii) Complementary to this is E-P over the global oceans derived from satellite estimates of P such as TRMM, GPM and GPCP and SeaFlux V3 or J-OFURO3 estimates of E, all relying heavily upon microwave measurements. Transport between land and oceans must essentially balance at monthly scales, i.e., vanish globally. (iv) a fourth perspective comes from estimate of terrestrial RO + storage rate, delta S. G-RUN Ensemble which uses observed streamflow and P measurements to calibrate a statistical model provides the former while GRACE, GRACE-FO provide total water storage anomalies used to calculate storage rate changes. GRACE REC uses GRACE data to train a precipitation-driven statistical model to extend storage estimates before the GRACE era. (All of these alternatives to reanalysis estimates also consider the small atmospheric column water vapor contribution.) We examine the transport changes from these three different methodologies, their relative accuracies and discuss the origin of their differences. Regional trends in moisture flux divergence and their role in multi-decadal trends are considered. Interannual variability arising in connection with ENSO variability is a dominant signal, driven largely by P changes. Trends since 1980 include reductions in moisture delivery to the western U.S., eastern Brazil, and central Africa with recovery of moisture convergence to the Sahel and parts of eastern North America.

Franklin Robertson↗

NASA NTRS2011

Statistical Projection of Solar Cycle 24 for the Exposure Estimates

A solar cycle statistical model has been developed based on the accumulating cycle sunspot data to estimate future levels of the solar cycle activity. Since the current solar cycle 24 has progressed about three years, the cycle activity levels are estimated with an accurately defined solar minimum 24. Then, solar cycle 24 is projected with the cycle activity levels using the statistical model. The projection of solar cycle 24 is then coupled to space related quantities of interest to radiation protection, because the interplanetary plasma and radiation fields are modulated by the degree of disturbance in the solar surface and the radiation doses received by astronauts in interplanetary space are likewise influenced. The resultant projection of solar cycle 24 provides a basis for estimating exposure in future space missions, and projection errors can be corrected as the cycle progresses and observations become available because this model is shown to be self-correcting.

Kim, Myung-Hee↗

DOE OSTI2023

Neural networks for parameter estimation in intractable models

The goal is to use deep learning models to estimate parameters in statistical models when standard likelihood estimation methods are computationally infeasible. For instance, inference for max-stable processes is exceptionally challenging even with small datasets, but simulation is straightforward. Data from model simulations are used to train deep neural networks and learn statistical parameters from max-stable models. The proposed neural network-based method provides a competitive alternative to current approaches, as demonstrated by considerable accuracy and computational time improvements. Finally, it serves as a proof of concept for deep learning in statistical parameter estimation and can be extended to other estimation problems.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Data assimilation for combustion ignition delay time simulation using schlieren image velocimetry

This study sought to improve the accuracy of simulating spray penetration and combustion ignition delay by means of data assimilation (DA). The simulations were conducted using the Reynolds-averaged Navier–Stokes (RANS) equations and assimilating the schlieren image data. In DA, an ensemble square root filter (EnSRF) was used to build the statistical model, making the simulation results more accurate without any change in the governing equations. Recognizing that the spray-cone injection angle has a large effect on penetration, we created ensemble members with different injection angles. And we applied the two-component velocity distribution calculated via SIV and updated both velocity and temperature by using a DA statistical model derived from RANS ensemble simulations. The ignition delay time is generally known to vary even under the same experimental conditions because it is influenced by many factors. In this study, we attempted the transient DA-assisted RANS simulation to predict the ignition delay time even when the temporal resolution and accuracy of the observation data ware insufficient. Our trials offer an example of how a combination of techniques can be effectively used to assimilate experimental data obtained under restricted conditions.

Combustion simulation↗

DOE OSTI2015

Spatial and Temporal Correlates of Greenhouse Gas Diffusion from a Hydropower Reservoir in the Southern United States

Emissions of CO 2 and CH 4 from freshwater reservoirs constitute a globally significant source of atmospheric greenhouse gases (GHGs), but knowledge gaps remain with regard to spatiotemporal drivers of emissions. We document the spatial and seasonal variation in surface diffusion of CO 2 and CH 4 from Douglas Lake, a hydropower reservoir in Tennessee, USA. Monthly estimates across 13 reservoir sites from January to November 2010 indicated that surface diffusions ranged from 236 to 18,806 mg m -2 day -1 for CO 2 and 0 to 0.95 mg m -2 day -1 for CH 4 . Next, we developed statistical models using spatial and physicochemical variables to predict surface diffusions of CO 2 and CH 4 . Models explained 22.7 and 20.9% of the variation in CO 2 and CH4 diffusions, respectively, and identified pH, temperature, dissolved oxygen, and Julian day as the most informative important predictors. These findings provide baseline estimates of GHG emissions from a reservoir in eastern temperate North America a region for which estimates of reservoir GHGs emissions are limited. Our statistical models effectively characterized non-linear and threshold relationships between physicochemical predictors and GHG emissions. Further refinement of such models will aid in predicting current GHG emissions in unsampled reservoirs and forecasting future GHG emissions.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2021

Validation Metrics for Fixed Effects and Mixed-Effects Calibration

The modern scientific process often involves the development of a predictive computational model. To improve its accuracy, a computational model can be calibrated to a set of experimental data. A variety of validation metrics can be used to quantify this process. Some of these metrics have direct physical interpretations and a history of use, while others, especially those for probabilistic data, are more difficult to interpret. In this work, a variety of validation metrics are used to quantify the accuracy of different calibration methods. Frequentist and Bayesian perspectives are used with both fixed effects and mixed-effects statistical models. Through a quantitative comparison of the resulting distributions, the most accurate calibration method can be selected. Two examples are included which compare the results of various validation metrics for different calibration methods. It is quantitatively shown that, in the presence of significant laboratory biases, a fixed effects calibration is significantly less accurate than a mixed-effects calibration. This is because the mixed-effects statistical model better characterizes the underlying parameter distributions than the fixed effects model. The results suggest that validation metrics can be used to select the most accurate calibration model for a particular empirical model with corresponding experimental data.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2019

Spatial Variability of Winds and HRRR–NCEP Model Error Statistics at Three Doppler-Lidar Sites in the Wind-Energy Generation Region of the Columbia River Basin

Annually and seasonally averaged wind profiles from three Doppler lidars were gathered from sites in the Columbia River basin of east-central Oregon and Washington, a major region of wind-energy production, for the Second Wind Forecast Improvement Project (WFIP2) experiment. The profile data are used to quantify the spatial variability of wind flows in this area of complex terrain, to assess the HRRR-NCEP model's ability to capture spatial and temporal variability of wind profiles, and to evaluate model errors. Annually averaged measured wind speed differences over the 70-km extent of the lidar measurements reached 1 m s -1 within the wind-turbine rotor layer, and 2 m s -1 for 200-500 m AGL. Stronger wind speeds in the lowest 500 m occurred at sites higher in elevation, farther from the river, and farther west - closer to the Cascade Mountain barrier. Validating against the lidar data, the HRRR model underestimated strong wind speeds (>12 m s -1 ) and, consequently, their frequency of occurrence, especially at the two lowest-elevation sites, producing annual low biases in rotor-layer wind speed of 0.5 m s -1 . The RMSE between measured and modeled winds at all sites was about 3 m s -1 and did not degrade significantly with forecast lead time. The nature of the model errors was different for different seasons. Furthermore, although the three sites were located in the same basin terrain, the nature of the model errors was different at each site. Thus, if only one of the sites had been instrumented, different conclusions would have been drawn as to the major sources of model error, depending on where the measurements were made.

17 WIND ENERGY↗

NASA NTRS2021

Ecosystem age-class dynamics and distribution in the LPJ-wsl v2.0 global ecosystem model

Forest ecosystem processes follow classic responses with age, peaking production around canopy closure and declining thereafter. Although age dynamics might be more dominant in certain regions over others, demographic effects on net primary production (NPP) and heterotrophic respiration (Rh) are bound to exist. Yet, explicit representation of ecosystem demography is notably absent in many global ecosystem models. This is concerning because the global community relies on these models to regularly update our collective understanding of the global carbon cycle. This paper aims to present the technical developments of a computationally efficient approach for representing age-class dynamics within a global ecosystem model, the Lund–Potsdam–Jena – Wald, Schnee, Landschaft version 2.0 (LPJ-wsl v2.0) dynamic global vegetation model and to determine if explicit representation of demography influenced ecosystem stocks and fluxes at global scales or at the level of a grid cell. The modeled age classes are initially created by simulated fire and prescribed wood harvesting or abandonment of managed land, otherwise aging naturally until an additional disturbance is simulated or prescribed. In this paper, we show that the age module can capture classic demographic patterns in stem density and tree height compared to inventory data, and that simulated patterns of ecosystem function follow classic responses with age. We also present two scientific applications of the model to assess the modeled age-class distribution over time and to determine the demographic effect on ecosystem fluxes relative to climate. Simulations show that, between 1860 and 2016, zonal age distribution on Earth was driven predominately by fire, causing a 45- to 60-year difference in ages between older boreal (50–90° N) and younger tropical (23° S–23° N) ecosystems. Between simulation years 1860 and 2016, land-use change and land management were responsible for a decrease in zonal age by −6 years in boreal and by −21 years in both temperate (23–50° N) and tropical latitudes, with the anthropogenic effect on zonal age distribution increasing over time. A statistical model helped to reduce LPJ-wsl v2.0 complexity by predicting per-grid-cell annual NPP and Rh fluxes by three terms: precipitation, temperature, and age class; at global scales, R2 was between 0.95 and 0.98. As determined by the statistical model, the demographic effect on ecosystem function was often less than 0.10 kg C/sq. myr but as high as 0.60 kg C/sq. myr where the effect was greatest. In the eastern forests of North America, the simulated demographic effect was of similar magnitude, or greater than, the effects of climate; simulated demographic effects were similarly important in large regions of every vegetated continent. Simulated spatial datasets are provided for global ecosystem ages and the estimated coefficients for effects of precipitation, temperature and demography on ecosystem function. The discussion focuses on our finding of an increasing role of demography in the global carbon cycle, the effect of demography on relaxation times (resilience) following a disturbance event and its implications at global scales, and a finding of a 40 Pg C increase in biomass turnover when including age dynamics at global scales. Whereas time is the only mechanism that increases ecosystem age, any additional disturbance not explicitly modeled will decrease age. The LPJ-wsl v2.0 age module represents another step forward towards understanding the role of demography in global ecosystems.

LPJ-wsl↗

DOE OSTI2016

A NEW STATISTICAL PERSPECTIVE TO THE COSMIC VOID DISTRIBUTION

In this study, we obtain the size distribution of voids as a three-parameter redshift-independent log-normal void probability function (VPF) directly from the Cosmic Void Catalog (CVC). Although many statistical models of void distributions are based on the counts in randomly placed cells, the log-normal VPF that we obtain here is independent of the shape of the voids due to the parameter-free void finder of the CVC. We use three void populations drawn from the CVC generated by the Halo Occupation Distribution (HOD) Mocks, which are tuned to three mock SDSS samples to investigate the void distribution statistically and to investigate the effects of the environments on the size distribution. As a result, it is shown that void size distributions obtained from the HOD Mock samples are satisfied by the three-parameter log-normal distribution. In addition, we find that there may be a relation between the hierarchical formation, skewness, and kurtosis of the log-normal distribution for each catalog. We also show that the shape of the three-parameter distribution from the samples is strikingly similar to the galaxy log-normal mass distribution obtained from numerical studies. This similarity between void size and galaxy mass distributions may possibly indicate evidence of nonlinear mechanisms affecting both voids and galaxies, such as large-scale accretion and tidal effects. Considering the fact that in this study, all voids are generated by galaxy mocks and show hierarchical structures in different levels, it may be possible that the same nonlinear mechanisms of mass distribution affect the void size distribution.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2016

Proton radius from electron scattering data

Background: The proton charge radius extracted from recent muonic hydrogen Lamb shift measurements is significantly smaller than that extracted from atomic hydrogen and electron scattering measurements. The discrepancy has become known as the proton radius puzzle. Purpose: In an attempt to understand the discrepancy, we review high-precision electron scattering results from Mainz, Jefferson Lab, Saskatoon and Stanford. Methods: We make use of stepwise regression techniques using the F-test as well as the Akaike information criterion to systematically determine the predictive variables to use for a given set and range of electron scattering data as well as to provide multivariate error estimates. Results: Starting with the precision, low four-momentum transfer (Q 2 ) data from Mainz (1980) and Saskatoon (1974), we find that a stepwise regression of the Maclaurin series using the F-test as well as the Akaike information criterion justify using a linear extrapolation which yields a value for the proton radius that is consistent with the result obtained from muonic hydrogen measurements. Applying the same Maclaurin series and statistical criteria to the 2014 Rosenbluth results on GE from Mainz, we again find that the stepwise regression tends to favor a radius consistent with the muonic hydrogen radius but produces results that are extremely sensitive to the range of data included in the fit. Making use of the high-Q 2 data on G E to select functions which extrapolate to high Q 2 , we find that a Pad´e (N = M = 1) statistical model works remarkably well, as does a dipole function with a 0.84 fm radius, G E (Q 2 ) = (1 + Q 2 /0.66 GeV 2 ) -2 . Conclusions: Rigorous applications of stepwise regression techniques and multivariate error estimates result in the extraction of a proton charge radius that is consistent with the muonic hydrogen result of 0.84 fm; either from linear extrapolation of the extreme low-Q 2 data or by use of the Pad´e approximant for extrapolation using a larger range of data. Thus, based on a purely statistical analysis of electron scattering data, we conclude that the electron scattering result and the muonic hydrogen result are consistent. Lastly, it is the atomic hydrogen results that are the outliers.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2021

International collaboration framework for the calculation of performance loss rates: Data quality, benchmarks, and trends (towards a uniform methodology)

Abstract The IEA PVPS Task 13 group, experts who focus on photovoltaic performance, operation, and reliability from several leading R&D centers, universities, and industrial companies, is developing a framework for the calculation of performance loss rates of a large number of commercial and research photovoltaic (PV) power plants and their related weather data coming across various climatic zones. The general steps to calculate the performance loss rate are (i) input data cleaning and grading; (ii) data filtering; (iii) performance metric selection, corrections, and aggregation; and finally, (iv) application of a statistical modeling method to determine the performance loss rate value. In this study, several high‐quality power and irradiance datasets have been shared, and the participants of the study were asked to calculate the performance loss rate of each individual system using their preferred methodologies. The data are used for benchmarking activities and to define capabilities and uncertainties of all the various methods. The combination of data filtering, metrics (performance ratio or power based), and statistical modeling methods are benchmarked in terms of (i) their deviation from the average value and (ii) their uncertainty, standard error, and confidence intervals. It was observed that careful data filtering is an essential foundation for reliable performance loss rate calculations. Furthermore, the selection of the calculation steps filter/metric/statistical method is highly dependent on one another, and the steps should not be assessed individually.

14 SOLAR ENERGY↗

NASA NTRS2001

The AutoBayes Program Synthesis System: System Description

AUTOBAYES is a fully automatic program synthesis system for the statistical data analysis domain. Its input is a concise description of a data analysis problem in the form of a statistical model; its output is optimized and fully documented C/C++ code which can be linked dynamically into the Matlab and Octave environments. AUTOBAYES synthesizes code by a schema-guided deductive process. Schemas (i.e., code templates with associated semantic constraints) are applied to the original problem and recursively to emerging subproblems. AUTOBAYES complements this approach by symbolic computation to derive closed-form solutions whenever possible. In this paper, we concentrate on the interaction between the symbolic computations and the deductive synthesis process. A statistical model specifies for each problem variable (i.e., data or parameter) its properties and dependencies in the form of a probability distribution, A typical data analysis task is to estimate the best possible parameter values from the given observations or measurements. The following example models normal-distributed data but takes prior information (e.g., from previous experiments) on the data's mean value and variance into account.

Fischer, Bernd↗

DOE OSTI2015

SPINE: SParse eIgengene NEtwork linking gene expression clusters in Dehalococcoides mccartyi to perturbations in experimental conditions

We present a statistical model designed to identify the effect of experimental perturbations on the aggregate behavior of the transcriptome expressed by the bacterium Dehalococcoides mccartyi strain 195. Strains of Dehalococcoides are used in sub-surface bioremediation applications because they organohalorespire tetrachloroethene and trichloroethene (common chlorinated solvents that contaminate the environment) to non-toxic ethene. However, the biochemical mechanism of this process remains incompletely described. Additionally, the response of Dehalococcoides to stress-inducing conditions that may be encountered at field-sites is not well understood. The constructed statistical model captured the aggregate behavior of gene expression phenotypes by modeling the distinct eigengenes of 100 transcript clusters, determining stable relationships among these clusters of gene transcripts with a sparse network-inference algorithm, and directly modeling the effect of changes in experimental conditions by constructing networks conditioned on the experimental state. Based on the model predictions, we discovered new response mechanisms for DMC, notably when the bacterium is exposed to solvent toxicity. The network identified a cluster containing thirteen gene transcripts directly connected to the solvent toxicity condition. Transcripts in this cluster include an iron-dependent regulator (DET0096-97) and a methylglyoxal synthase (DET0137). To validate these predictions, additional experiments were performed. Continuously fed cultures were exposed to saturating levels of tetrachloethene, thereby causing solvent toxicity, and transcripts that were predicted to be linked to solvent toxicity were monitored by quantitative reverse-transcription polymerase chain reaction. Twelve hours after being shocked with saturating levels of tetrachloroethene, the control transcripts (encoding for a key hydrogenase and the 16S rRNA) did not significantly change. By contrast, transcripts for DET0137 and DET0097 displayed a 46.8±11.5 and 14.6±9.3 fold up-regulation, respectively, supporting the model. This is the first study to identify transcripts in Dehalococcoides that potentially respond to tetrachloroethene solvent-toxicity conditions that may be encountered near contamination source zones in sub-surface environments.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2014

Pathways to Identity. Using Visualization to Aid Law Enforcement in Identification Tasks

The nature of identity has changed dramatically in recent years and has grown in complexity. Identities are defined in multiple domains: biological and psychological elements strongly contribute, but biographical and cyber elements also are necessary to complete the picture. Law enforcement is beginning to adjust to these changes, recognizing identity’s importance in criminal justice. The SuperIdentity project seeks to aid law enforcement officials in their identification tasks through research of techniques for discovering identity traits, generation of statistical models of identity and analysis of identity traits through visualization. We present use cases compiled through user interviews in multiple fields, including law enforcement, and describe the modeling and visualization tools design to aid in those use cases.

99 GENERAL AND MISCELLANEOUS↗

DOE OSTI2014

Data-Driven Benchmarking of Building Energy Efficiency Utilizing Statistical Frontier Models

Frontier methods quantify the energy efficiency of buildings by forming an efficient frontier (best-practice technology) and by comparing all buildings against that frontier. Because energy consumption fluctuates over time, the efficiency scores are stochastic random variables. Existing applications of frontier methods in energy efficiency either treat efficiency scores as deterministic values or estimate their uncertainty by resampling from one set of measurements. Availability of smart meter data (repeated measurements of energy consumption of buildings) enables using actual data to estimate the uncertainty in efficiency scores. Additionally, existing applications assume a linear form for an efficient frontier; i.e.,they assume that the best-practice technology scales up and down proportionally with building characteristics. However, previous research shows that buildings are nonlinear systems. This paper proposes a statistical method called stochastic energy efficiency frontier (SEEF) to estimate a bias-corrected efficiency score and its confidence intervals from measured data. The paper proposes an algorithm to specify the functional form of the frontier, identify the probability distribution of the efficiency score of each building using measured data, and rank buildings based on their energy efficiency. To illustrate the power of SEEF, this paper presents the results from applying SEEF on a smart meter data set of 307 residential buildings in the United States. SEEF efficiency scores are used to rank individual buildings based on energy efficiency, to compare subpopulations of buildings, and to identify irregular behavior of buildings across different time-of-use periods. SEEF is an improvement to the energy-intensity method (comparing kWh/sq.ft.): whereas SEEF identifies efficient buildings across the entire spectrum of building sizes, the energy-intensity method showed bias toward smaller buildings. The results of this research are expected to assist researchers and practitioners compare and rank (i.e.,benchmark) buildings more robustly and over a wider range of building types and sizes. Eventually, doing so is expected to result in improved resource allocation in energy-efficiency programs.

Kavousian, A↗

DOE OSTI2020

Enantioselective Allenoate-Claisen Rearrangement Using Chiral Phosphate Catalysts

Herein we report the first highly enantioselective allenoate-Claisen rearrangement using doubly axially chiral phosphate sodium salts as catalysts. This synthetic method provides access to β-amino acid derivatives with vicinal stereocenters in up to 95% ee. We also investigated the mechanism of enantioinduction by transition state (TS) computations with DFT as well as statistical modeling of the relationship between selectivity and the molecular features of both the catalyst and substrate. The mutual interactions of charge-separated regions in both the zwitterionic intermediate generated by reaction of an amine to the allenoate and the Na + -salt of the chiral phosphate leads to an orientation of the TS in the catalytic pocket that maximizes favorable noncovalent interactions. Crucial arene-arene interactions at the periphery of the catalyst lead to a differentiation of the TS diastereomers. Finally, these interactions were interrogated using DFT calculations and validated through statistical modeling of parameters describing noncovalent interactions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2024

Predicting September Arctic Sea Ice: A Multimodel Seasonal Skill Comparison

This study quantifies the state of the art in the rapidly growing field of seasonal Arctic sea ice prediction. A novel multimodel dataset of retrospective seasonal predictions of September Arctic sea ice is created and analyzed, consisting of community contributions from 17 statistical models and 17 dynamical models. Prediction skill is compared over the period 2001–20 for predictions of pan-Arctic sea ice extent (SIE), regional SIE, and local sea ice concentration (SIC) initialized on 1 June, 1 July, 1 August, and 1 September. This diverse set of statistical and dynamical models can individually predict linearly detrended pan-Arctic SIE anomalies with skill, and a multimodel median prediction has correlation coefficients of 0.79, 0.86, 0.92, and 0.99 at these respective initialization times. Regional SIE predictions have similar skill to pan-Arctic predictions in the Alaskan and Siberian regions, whereas regional skill is lower in the Canadian, Atlantic, and central Arctic sectors. The skill of dynamical and statistical models is generally comparable for pan-Arctic SIE, whereas dynamical models outperform their statistical counterparts for regional and local predictions. The prediction systems are found to provide the most value added relative to basic reference forecasts in the extreme SIE years of 1996, 2007, and 2012. SIE prediction errors do not show clear trends over time, suggesting that there has been minimal change in inherent sea ice predictability over the satellite era. Overall, this study demonstrates that there are bright prospects for skillful operational predictions of September sea ice at least 3 months in advance.

54 ENVIRONMENTAL SCIENCES↗