Inter-model Comparisons Research Articles

Diastolic Dysfunction Assessed Using Contemporary Guidelines and Prognosis Following Myocardial Infarction

Recent American Society of Echocardiography and European Association of Cardiovascular Imaging guidelines for the assessment of diastolic dysfunction (DD) recommend a simplified approach with four key variables incorporated into a novel diagnostic algorithm. The aim of this study was to assess the prognostic value of significant DD assessed using the algorithm recommended in the 2016 American Society of Echocardiography and European Association of Cardiovascular Imaging guidelines (DD2016) in comparison with the prognostic value of significant DD assessed using the 2009 guidelines (DD2009) as well as the individual parameters incorporated in the 2016 algorithm. Retrospective data on 419 consecutive patients with first ever myocardial infarction were included. Doppler echocardiography was performed within 24hours of admission in all patients. Significant DD was defined as grade 2 or 3 DD. The primary outcome measure was composite major adverse cardiovascular events (MACEs), comprising death, myocardial infarction, and heart failure. At a median follow-up of 24months, there were 61 MACEs. On Kaplan-Meier analysis, DD2016 showed a better association with MACEs than DD2009 (log-rank χ2=21.01 [P<.001] vs 13.13 [P=.001]). On Cox proportional-hazards multivariate analysis incorporating significant clinical predictors and left ventricular ejection fraction, DD2016 (hazard ratio, 2.22; 95% CI, 1.25-3.98; P=.007) was the strongest independent predictor of MACEs, whereas DD2009 (hazard ratio, 1.63; 95% CI, 0.95-2.80; P=.074) was not a significant predictor. Of the four key diastolic parameters, only left atrial volume index was independently associated with MACEs (hazard ratio, 1.79; 95% CI, 1.02-3.14; P=.041) when included in a Cox proportional-hazards multivariate model incorporating significant clinical predictors and left ventricular ejection fraction, although the association was weaker than DD2016. Intermodel comparisons with model χ2 and Harrell's C statistic were satisfactory for DD2016. Significant DD assessed using the 2016 American Society of Echocardiography and European Association of Cardiovascular Imaging guidelines is a robust independent predictor of clinical outcomes following myocardial infarction and compares favorably with DD2009 as well as the individual parameters incorporated in the novel 2016 algorithm.

Numerical modeling of non-breaking, impulsive breaking, and broken wave interaction with elevated coastal structures: Laboratory validation and inter-model comparisons

Quantitative CFD model validation and inter-model comparisons between IHFOAM and ANSYS-FLUENT were performed for pressures and forces on an elevated structure using a 1:10 physical model. Non-breaking, impulsive breaking, and broken wave conditions at the structure's location were simulated in IHFOAM and FLUENT. The calculated time series of water surface elevation and horizontal and vertical pressures and forces were compared with the measured data. We introduced the impulse of residual to quantify the variation of the force and pressure time series. Results indicated that the numerical models performed differently depending on the wave conditions, even for the same initial set up. Non-breaking wave simulations showed the best agreement with experimental data for both models, while broken wave trials showed the largest deviations. Bottom pressures and vertical forces were less sensitive to wave breaking conditions. Results indicate that future benchmarking tests for an elevated structure must consider both horizontal and vertical forces due to various wave breaking conditions. The accuracy of simulated wave shoaling and breaking processes played a key role in precisely calculating the forces and pressures on the structure, and it was difficult for the CFD models to simulate the exact wave breaking conditions as the measurements.

18F]Fludarabine-PET as a promising tool for differentiating CNS lymphoma and glioblastoma: Comparative analysis with [18F]FDG in human xenograft models

This paper investigated whether positron emission tomography (PET) imaging with [18F]fludarabine ([18F]FDB) can help to differentiate central nervous system lymphoma (CNSL) from glioblastoma (GBM), which is a crucial issue in the diagnosis and management of patients with these aggressive brain tumors. Multimodal analyses with [18F]fluorodeoxyglucose ([18F]FDG), magnetic resonance imaging (MRI) and histology have also been considered to address the specificity of [18F]FDB for CNSL.Methods: Nude rats were implanted with human MC116 lymphoma-cells (n = 9) or U87 glioma-cells (n = 4). Tumor growth was monitored by MRI, with T2-weighted sequence for anatomical features and T1-weighted with gadolinium (Gd) enhancement for blood brain barrier (BBB) permeability assessment. For PET investigation, [18F]FDB or [18F]FDG (~11 MBq) were injected via tail vein and dynamic PET images were acquired up to 90 min after radiotracer injection. Paired scans of the same rat with the two [18F]-labelled radiotracers were investigated. Initial volumes of interest were manually delineated on T2w images and set on co-registered PET images and tumor-to-background ratio (TBR) was calculated to semi-quantitatively assess the tracer accumulation in the tumor. A tile-based method for image analysis was developed in order to make comparative analysis between radiotracer uptake and values extracted from immunohistochemistry staining.Results: In the lymphoma model, PET time-activity curves (TACs) revealed a differential response of [18F]FDB between tumoral and healthy tissues with average TBR varying from 2.45 to 3.16 between 5 to 90 min post-injection. In contrast, [18F]FDG demonstrated similar uptake profiles for tumoral and normal regions with TBR varying from 0.84 to 1.06 between these two time points. In the glioblastoma (GBM) model, the average TBRs were from 2.14 to 1.01 for [18F]FDB and from 0.95 to 1.65 for [18F]FDG. Therefore, inter-model comparisons showed significantly divergent responses (p < 0.01) of [18F]FDB between lymphoma and GBM, while [18F]FDG demonstrated overlap (p = 0.04) between the groups. Tumor characterization with histology (based mainly on Hoechst and CD79), as well as with MRI was overall in better agreement with [18F]FDB-PET than [18F]FDG with regard to tumor selectivity.Conclusions: [18F]FDB-PET demonstrated considerably greater specificity for CNSL when compared to [18F]FDG. It also permitted a more precise definition of target volume compared to contrast-enhanced MRI. Therefore, the potential of [18F]FDB-PET to distinguish CNSL from GBM is quite evident and will be further investigated in humans.

Holocene temperature evolution in the Northern Hemisphere high latitudes – Model-data comparisons

Heterogeneous Holocene climate evolutions in the Northern Hemisphere high latitudes are primarily determined by orbital-scale insolation variations and melting ice sheets. Previous inter-model comparisons have revealed that multi-simulation consistencies vary spatially. We, therefore, compared multiple model results with proxy-based reconstructions in Fennoscandia, Greenland, north Canada, Alaska and Siberia.Our model-data comparisons reveal that data and models generally agree in Fennoscandia, Greenland and Canada, with the early-Holocene warming and subsequent gradual decrease to 0 ka BP (hereinafter referred as ka). In Fennoscandia, simulations and pollen data suggest a 2 °C warming by 8 ka, but this is less expressed in chironomid data. In Canada, a strong early-Holocene warming is suggested by both the simulations and pollen results. In Greenland, the magnitude of early-Holocene warming ranges from 6 °C in simulations to 8 °C in δ18O-based temperatures.Simulated and reconstructed temperatures are mismatched in Alaska. Pollen data suggest strong early-Holocene warming, while the simulations indicate constant Holocene cooling, and chironomid data show a stable trend. Meanwhile, a high frequency of Alaskan peatland initiation before 9 ka can reflect a either high temperature, high soil moisture or large seasonality. In high-latitude Siberia, although simulations and proxy data depict high Holocene temperatures, these signals are noisy owing to a large spread in the simulations and between pollen and chironomid results. On the whole, the Holocene climate evolutions in most regions (Fennoscandia, Greenland and Canada) are well established and understood, but important questions regarding the Holocene temperature trend and mechanisms remain for Alaska and Siberia.

Intercomparison of global river discharge simulations focusing on dam operation—multiple models analysis in two case-study river basins, Missouri–Mississippi and Green–Colorado

We performed an intercomparison of river discharge regulated by dams under four meteorological forcings among five global hydrological models for a historical period by simulation. This is the first global multimodel intercomparison study on dam-regulated river flow. Although the simulations were conducted globally, the Missouri–Mississippi and Green–Colorado Rivers were chosen as case-study sites in this study. The hydrological models incorporate generic schemes of dam operation, not specific to a certain dam. We examined river discharge on a longitudinal section of river channels to investigate the effects of dams on simulated discharge, especially at the seasonal time scale. We found that the magnitude of dam regulation differed considerably among the hydrological models. The difference was attributable not only to dam operation schemes but also to the magnitude of simulated river discharge flowing into dams. That is, although a similar algorithm of dam operation schemes was incorporated in different hydrological models, the magnitude of dam regulation substantially differed among the models. Intermodel discrepancies tended to decrease toward the lower reaches of these river basins, which means model dependence is less significant toward lower reaches. These case-study results imply that, intermodel comparisons of river discharge should be made at different locations along the river's course to critically examine the performance of hydrological models because the performance can vary with the locations.

Compensation in Root Water Uptake Models Combined with Three‐Dimensional Root Length Density Distribution

Because root water uptake is a considerable component in the soil water balance, a lot of attention has been paid to defining and applying root water uptake models. These models can be grouped into empirical vs. mechanistic root uptake models. Intermodel comparisons are valuable in understanding the different concepts used. Such a comparison is sometimes difficult because the level of information required by the models is different, for example, information on the root length density distribution. Here a three‐dimensional root length density distribution function is introduced that makes it possible to compare two empirical uptake models with a more mechanistic uptake model. Adding a compensation component to the more empirical model results in prediction of root water uptake distributions in the root zone similar to those predicted by the more complex model.

Assessment of infiltration heat recovery and its impact on energy consumption for residential buildings

Infiltration is a major contributor to the energy consumption of buildings, particularly in homes where it accounts for one-third of the heating and cooling loads. Traditionally, infiltration is calculated independent of the building envelope performance, however, it has been established that a thermal coupling exists between the infiltration and conduction heat transfer of the building envelope. This effect is known as infiltration heat recovery (IHR). Experiments have shown that infiltration heat recovery can typically reduce the infiltration thermal load by 10–20%.Currently, whole-building energy simulation tools do not account for the effect of infiltration heat recovery on heating and cooling loads. In this paper, five steady-state IHR models are described to account for the thermal interaction between infiltration air and building envelope components. In particular, inter-model and experimental comparisons are carried out to assess the prediction accuracy of five IHR models. In addition, the results from a series of sensitivity analyses are presented, including an evaluation of the predictions for heating energy use associated with four audited homes obtained from whole-building energy simulation analysis with implemented infiltration heat recovery models.Experimental comparison of the IHR models reveal that the predictions from all the five models are consistent and are within 2% when 1-D flow and heat transfer conditions are considered. When implementing IHR models to a whole-building simulation environment, a reduction of 5–30% in heating consumption is found for four audited residential homes.

Plasma size and collisionality scaling of ion-temperature-gradient-driven turbulence

Fixed-flux (FF), fixed-gradient (FG) and local fluxtube (FT) gyrokinetic simulations are systematically compared for ion-temperature-gradient (ITG)-driven turbulence. The collisionality (ν*) dependence of ion heat diffusivity is verified through the inter-model comparisons. When the temperature gradient is far from the nonlinear critical value, the FF and FT models give a weak ν*-dependence, while the FG model shows a strong ν*-dependence. The entropy transfer analysis on the zonal-flow saturation mechanisms in the quasi-steady state of the FT simulation provides clear insights on the different ν*-dependence of the turbulent transport and zonal-flow shearing rate in the far-above- and near-critical cases. It has also been revealed that the FG model provides the strong ν*-dependence through the change of ITG-mode stability due to ν*-dependent heating/sink by the adaptive heat source, where the velocity distribution function is deformed. The plasma size (ρ*) scan in the FF simulations show a Bohm-like transport scaling even in a local limit regime, ρ*−1 ⩾ 300, where profile-shear effects are weak. It has been clarified that the transient variations of local power balance are essential mechanisms leading to the Bohm-like heat transport even at similar mean temperature gradients, where the burst amplitude and its frequency increase with the plasma size and the heating power. The mechanism is unique to the FF model. Comparisons of statistical characteristics in the local limit regime show differences in frequency spectra and probability density functions of the heat flux, while zonal-flow structures and avalanche propagations properties are similar among these models.

Multiple stressors of ocean ecosystems in the 21st century: projections with CMIP5 models

Abstract. Ocean ecosystems are increasingly stressed by human-induced changes of their physical, chemical and biological environment. Among these changes, warming, acidification, deoxygenation and changes in primary productivity by marine phytoplankton can be considered as four of the major stressors of open ocean ecosystems. Due to rising atmospheric CO2 in the coming decades, these changes will be amplified. Here, we use the most recent simulations performed in the framework of the Coupled Model Intercomparison Project 5 to assess how these stressors may evolve over the course of the 21st century. The 10 Earth system models used here project similar trends in ocean warming, acidification, deoxygenation and reduced primary productivity for each of the IPCC's representative concentration pathways (RCPs) over the 21st century. For the "business-as-usual" scenario RCP8.5, the model-mean changes in the 2090s (compared to the 1990s) for sea surface temperature, sea surface pH, global O2 content and integrated primary productivity amount to &amp;amp;plus;2.73 (±0.72) °C, −0.33 (±0.003) pH unit, −3.45 (±0.44)% and −8.6 (±7.9)%, respectively. For the high mitigation scenario RCP2.6, corresponding changes are +0.71 (±0.45) °C, −0.07 (±0.001) pH unit, −1.81 (±0.31)% and −2.0 (±4.1)%, respectively, illustrating the effectiveness of extreme mitigation strategies. Although these stressors operate globally, they display distinct regional patterns and thus do not change coincidentally. Large decreases in O2 and in pH are simulated in global ocean intermediate and mode waters, whereas large reductions in primary production are simulated in the tropics and in the North Atlantic. Although temperature and pH projections are robust across models, the same does not hold for projections of subsurface O2 concentrations in the tropics and global and regional changes in net primary productivity. These high uncertainties in projections of primary productivity and subsurface oxygen prompt us to continue inter-model comparisons to understand these model differences, while calling for caution when using the CMIP5 models to force regional impact models.

Extratropical storm inundation testbed: Intermodel comparisons in Scituate, Massachusetts

[1] The Integrated Ocean Observing System Super-regional Coastal Modeling Testbed had one objective to evaluate the capabilities of three unstructured-grid fully current-wave coupled ocean models (ADCIRC/SWAN, FVCOM/SWAVE, SELFE/WWM) to simulate extratropical storm-induced inundation in the US northeast coastal region. Scituate Harbor (MA) was chosen as the extratropical storm testbed site, and model simulations were made for the 24–27 May 2005 and 17–20 April 2007 (“Patriot's Day Storm”) nor'easters. For the same unstructured mesh, meteorological forcing, and initial/boundary conditions, intermodel comparisons were made for tidal elevation, surface waves, sea surface elevation, coastal inundation, currents, and volume transport. All three models showed similar accuracy in tidal simulation and consistency in dynamic responses to storm winds in experiments conducted without and with wave-current interaction. The three models also showed that wave-current interaction could (1) change the current direction from the along-shelf direction to the onshore direction over the northern shelf, enlarging the onshore water transport and (2) intensify an anticyclonic eddy in the harbor entrance and a cyclonic eddy in the harbor interior, which could increase the water transport toward the northern peninsula and the southern end and thus enhance flooding in those areas. The testbed intermodel comparisons suggest that major differences in the performance of the three models were caused primarily by (1) the inclusion of wave-current interaction, due to the different discrete algorithms used to solve the three wave models and compute water-current interaction, (2) the criterions used for the wet-dry point treatment of the flooding/drying process simulation, and (3) bottom friction parameterizations.

Comparing Lagrangian and Eulerian models for CO&amp;lt;sub&amp;gt;2&amp;lt;/sub&amp;gt; transport – a step towards Bayesian inverse modeling using WRF/STILT-VPRM

Abstract. We present simulations of atmospheric CO2 concentrations provided by two modeling systems, run at high spatial resolution: the Eulerian-based Weather Research Forecasting (WRF) model and the Lagrangian-based Stochastic Time-Inverted Lagrangian Transport (STILT) model, both of which are coupled to a diagnostic biospheric model, the Vegetation Photosynthesis and Respiration Model (VPRM). The consistency of the simulations is assessed with special attention paid to the details of horizontal as well as vertical transport and mixing of CO2 concentrations in the atmosphere. The dependence of model mismatch (Eulerian vs. Lagrangian) on models' spatial resolution is further investigated. A case study using airborne measurements during which two models showed large deviations from each other is analyzed in detail as an extreme case. Using aircraft observations and pulse release simulations, we identified differences in the representation of details in the interaction between turbulent mixing and advection through wind shear as the main cause of discrepancies between WRF and STILT transport at a spatial resolution such as 2 and 6 km. Based on observations and inter-model comparisons of atmospheric CO2 concentrations, we show that a refinement of the parameterization of turbulent velocity variance and Lagrangian time-scale in STILT is needed to achieve a better match between the Eulerian and the Lagrangian transport at such a high spatial resolution (e.g. 2 and 6 km). Nevertheless, the inter-model differences in simulated CO2 time series for a tall tower observatory at Ochsenkopf in Germany are about a factor of two smaller than the model-data mismatch and about a factor of three smaller than the mismatch between the current global model simulations and the data.

Crop residue removal effects on soil carbon: Measured and inter-model comparisons

Crop residues can be a viable source for biofuel production and other industrial products; however, their removal from agricultural land may negatively impact productivity and environmental quality. In this study three process-based models (CENTURY, DAYCENT, and DNDC) and the CAMPBELL empirical model were used to simulate soil organic carbon (SOC) change and were compared to observations from 14 residue removal experiments within the temperate climate areas of Canada and the Midwestern USA. The experimental results indicated that residue removal effects on SOC were more likely to be observed (i) with greater rates of residue removal, (ii) after longer periods, and (iii) with greater rates of (N) fertilizer. All four models simulated the hypothesized decline in SOC when residues were removed. The average measured SOC change for residue removal across all experimental sites and durations was −235gm−2 (i.e., 3.3% of SOC in top 20cm) whereas the SOC change as estimated by the models were −423, −417, −201, and −218gm−2 for the CENTURY, DAYCENT, DNDC, and CAMPBELL models, respectively. All model predictions were close or within the range of uncertainty of estimates derived from measurements.

Evaluation of cloud fraction and its radiative effect simulated by IPCC AR4 global models against ARM surface observations

Abstract. Cloud Fraction (CF) is the dominant modulator of radiative fluxes. In this study, we evaluate CF simulated in the IPCC AR4 GCMs against ARM long-term ground-based measurements, with a focus on the vertical structure, total amount of cloud and its effect on cloud shortwave transmissivity. Comparisons are performed for three climate regimes as represented by the Department of Energy Atmospheric Radiation Measurement (ARM) sites: Southern Great Plains (SGP), Manus, Papua New Guinea and North Slope of Alaska (NSA). Our intercomparisons of three independent measurements of CF or sky-cover reveal that the relative differences are usually less than 10% (5%) for multi-year monthly (annual) mean values, while daily differences are quite significant. The total sky imager (TSI) produces smaller total cloud fraction (TCF) compared to a radar/lidar dataset for highly cloudy days (CF &gt; 0.8), but produces a larger TCF value than the radar/lidar for less cloudy conditions (CF &lt; 0.3). The compensating errors in lower and higher CF days result in small biases of TCF between the vertically pointing radar/lidar dataset and the hemispheric TSI measurements as multi-year data is averaged. The unique radar/lidar CF measurements enable us to evaluate seasonal variation of cloud vertical structures in the GCMs. Both inter-model deviation and model bias against observation are investigated in this study. Another unique aspect of this study is that we use simultaneous measurements of CF and surface radiative fluxes to diagnose potential discrepancies among the GCMs in representing other cloud optical properties than TCF. The results show that the model-observation and inter-model deviations have similar magnitudes for the TCF and the normalized cloud effect, and these deviations are larger than those in surface downward solar radiation and cloud transmissivity. This implies that other dimensions of cloud in addition to cloud amount, such as cloud optical thickness and/or cloud height, have a similar magnitude of disparity as TCF within the GCMs, and suggests that the better agreement among GCMs in solar radiative fluxes could be a result of compensating effects from errors in cloud vertical structure, overlap assumption, cloud optical depth and/or cloud fraction. The internal variability of CF simulated in ensemble runs with the same model is minimal. Similar deviation patterns between inter-model and model-measurement comparisons suggest that the climate models tend to generate larger biases against observations for those variables with larger inter-model deviation. The GCM performance in simulating the probability distribution, transmissivity and vertical profiles of cloud are comprehensively evaluated over the three ARM sites. The GCMs perform better at SGP than at the other two sites in simulating the seasonal variation and probability distribution of TCF. However, the models remarkably underpredict the TCF at SGP and cloud transmissivity is less susceptible to the change of TCF than observed. In the tropics, most of the GCMs tend to underpredict CF and fail to capture the seasonal variation of CF at middle and low levels. The high-level CF is much larger in the GCMs than the observations and the inter-model variability of CF also reaches a maximum at high levels in the tropics, indicating discrepancies in the representation of ice cloud associated with convection in the models. While the GCMs generally capture the maximum CF in the boundary layer and vertical variability, the inter-model deviation is largest near the surface over the Arctic.

Sensitivity of interglacial Greenland temperature and δ&amp;lt;sup&amp;gt;18&amp;lt;/sup&amp;gt;O: ice core data, orbital and increased CO&amp;lt;sub&amp;gt;2&amp;lt;/sub&amp;gt; climate simulations

Abstract. The sensitivity of interglacial Greenland temperature to orbital and CO2 forcing is investigated using the NorthGRIP ice core data and coupled ocean-atmosphere IPSL-CM4 model simulations. These simulations were conducted in response to different interglacial orbital configurations, and to increased CO2 concentrations. These different forcings cause very distinct simulated seasonal and latitudinal temperature and water cycle changes, limiting the analogies between the last interglacial and future climate. However, the IPSL-CM4 model shows similar magnitudes of Arctic summer warming and climate feedbacks in response to 2 × CO2 and orbital forcing of the last interglacial period (126 000 years ago). The IPSL-CM4 model produces a remarkably linear relationship between TOA incoming summer solar radiation and simulated changes in summer and annual mean central Greenland temperature. This contrasts with the stable isotope record from the Greenland ice cores, showing a multi-millennial lagged response to summer insolation. During the early part of interglacials, the observed lags may be explained by ice sheet-ocean feedbacks linked with changes in ice sheet elevation and the impact of meltwater on ocean circulation, as investigated with sensitivity studies. A quantitative comparison between ice core data and climate simulations requires stability of the stable isotope – temperature relationship to be explored. Atmospheric simulations including water stable isotopes have been conducted with the LMDZiso model under different boundary conditions. This set of simulations allows calculation of a temporal Greenland isotope-temperature slope (0.3–0.4‰ per °C) during warmer-than-present Arctic climates, in response to increased CO2, increased ocean temperature and orbital forcing. This temporal slope appears half as large as the modern spatial gradient and is consistent with other ice core estimates. It may, however, be model-dependent, as indicated by preliminary comparison with other models. This suggests that further simulations and detailed inter-model comparisons are also likely to be of benefit. Comparisons with Greenland ice core stable isotope data reveals that IPSL-CM4/LMDZiso simulations strongly underestimate the amplitude of the ice core signal during the last interglacial, which could reach +8–10 °C at fixed-elevation. While the model-data mismatch may result from missing positive feedbacks (e.g. vegetation), it could also be explained by a reduced elevation of the central Greenland ice sheet surface by 300–400 m.

Process-oriented tests for validation of baroclinic shallow water models: The lock-exchange problem

A first step often taken to validate prognostic baroclinic codes is a series of process-oriented tests, as those suggested by Haidvogel and Beckmann [Haidvogel, D., Beckmann, A., 1999. Numerical Ocean Circulation Modeling. Imperial College Press, London], among others. One of these tests is the so-called “lock-exchange” test or “dam break” problem, wherein water of different densities is separated by a vertical barrier, which is removed at time zero. Validation against these tests has primarily consisted of comparing the propagation speed of the wave front, as predicted by various theoretical and experimental results, to model output. In addition, inter-model comparisons of the lock-exchange test have been used to validate codes. Herein, we present a high resolution data set, taken from a laboratory-scale model, for direct and quantitative comparison of experimental and numerical results throughout the domain, not just the wave front. Data is captured every 0.2 s using high resolution digital photography, with salt concentration extracted by comparing pixel intensity of the dyed fluid against calibration standards. Two scenarios are discussed in this paper, symmetric and asymmetric mixing, depending on the proportion of dense/light water (17.5 ppt/0.0 ppt) in the experiment; the Boussinesq approximation applies to both. Front speeds, cast in terms of the dimensionless Froude number, show excellent agreement with literature-reported values. Data are also used to quantify the degree of mixing, as measured by the front thickness, which also provides an error band on the front speed. Finally, experimental results are used to validate baroclinic enhancements to the barotropic shallow water ADvanced CIRCulation (ADCIRC) model, including the effect of the vertical mixing scheme on simulation results. Based on salinity data, the model provides an average root-mean-square (rms) error of 3.43 ppt for the symmetric case and 3.74 ppt for the asymmetric case, most of which can be attributed to the shear instabilities along the density front, which are not resolved by this hydrostatic model. Front tracking and mixed layer thickness rms errors average 34.9 mm (15.7% relative to total fluid depth) and 10.2 mm (4.9%) for the symmetric test and 37.2 mm (16.8%) and 14.0 mm (6.2%) for the asymmetric test, respectively. All non-constant vertical closure schemes are able to capture the front speed, but the Mellor-Yamada 2.5 scheme most accurately captures the synoptic behavior of the mixing layer thickness.

Petascale atmospheric models for the Community Climate System Model: new developments and evaluation of scalable dynamical cores

We present results from the integration and evaluation of the spectral finite-element method into the atmospheric component of the Community Climate System Model (CCSM). This method overcomes the atmospheric scalability bottleneck by allowing the use of a true two-dimensional domain decomposition for the first time in the CCSM. Scalability is demonstrated out to 86,200 processors with an average grid spacing of 0.25° (25 km). We present initial evaluations results using a standardized test problem with the full suite of CCSM atmospheric model forcings and subgrid parametrizations but without the CCSM land, ice, or ocean models. For this realistic setting, the true solution is unknown. Even convergence under mesh refinement is not expected, so we cannot rely on high-resolution reference solutions. Instead we focus on intermodel comparisons and use the Williamson equivalent resolution methodology to evaluate the results.

Evaluation of the Alberta air infiltration model using measurements and inter-model comparisons

The Alberta air infiltration model (AIM-2) is a simplified single-zone model for predicting building air infiltration rates with a number of salient features. This paper presents an empirical study of this model using measured data from 16 detached houses in Ottawa, Canada. A single-fan depressurization test was first conducted for each house to determine its leakage characteristics. Then, the tracer gas concentration decay technique was employed to measure air infiltration rates under a wide range of weather conditions. The AIM-2 model was used to predict air infiltration rates for each of the 16 houses for the measured weather conditions. These model predictions were then compared with the air infiltration rates determined with the tracer gas tests. Additionally, the predictions of the AIM-2 model were compared with those of another model, the Lawrence Berkeley Laboratory (LBL) model. The AIM-2 model tended to underestimate air infiltration rates but performed better than the LBL model. On average, the AIM-2 model has an error of 19% while the LBL model has an error of 25%. The AIM-2 model requires an estimation of the house's leakage distribution, which may have contributed to some of this disagreement. An attempt was made to use genetic algorithms to reduce the uncertainty caused by estimating these model inputs.

Mesh-morphing algorithms for specimen-specific finite element modeling

Despite recent advances in software for meshing specimen-specific geometries, considerable effort is still often required to produce and analyze specimen-specific models suitable for biomechanical analysis through finite element modeling. We hypothesize that it is possible to obtain accurate models by adapting a pre-existing geometry to represent a target specimen using morphing techniques. Here we present two algorithms for morphing, automated wrapping (AW) and manual landmarks (ML), and demonstrate their use to prepare specimen-specific models of caudal rat vertebrae. We evaluate the algorithms by measuring the distance between target and morphed geometries and by comparing response to axial loading simulated with finite element (FE) methods. First a traditional reconstruction process based on μCT was used to obtain two natural specimen-specific FE models. Next, the two morphing algorithms were used to compute mappings from the surface of one model, the source, to the other, the target, and to use this mapping to morph the source mesh to produce a target mesh. The μCT images were then used to assign element-specific material properties. In AW the mappings were obtained by wrapping the source and target surfaces with an auxiliary triangulated surface. In ML, landmarks were manually placed on corresponding locations on the surfaces of both source and target. Both morphing algorithms were successful in reproducing the shape of the target vertebra with a median distance between natural and morphed models of 18.8 and 32.2 μm, respectively, for AW and ML. Whereas AW–morphing produced a surface more closely resembling that of the target, ML guaranteed correspondence of the landmark locations between source and target. Morphing preserved the quality of the mesh producing models suitable for FE simulation. Moreover, there were only minor differences between natural and morphed models in predictions of deformation, strain and stress. We therefore conclude that it is possible to use mesh-morphing techniques to produce accurate specimen-specific FE models of caudal rat vertebrae. Mesh morphing techniques provide advantages over conventional specimen-specific finite element modeling by reducing the effort required to generate multiple target specimen models, facilitating intermodel comparisons through correspondence of nodes and maintenance of connectivity, and lends itself to parametric evaluation of “artificial” geometries with a focus on optimizing reconstruction.

Inter-model comparison of CFD and neural network analysis of natural convection heat transfer in a partitioned enclosure

CFD analysis of heat and mass flow due to natural convection in partitioned enclosures has recently been the focus of many CFD researchers. In some cases, it was reported in the literature that different CFD solutions (due to different numerical stability characteristics) were obtained for different mesh quality, time step, and discretization order. The objective of this paper is to investigate the feasibility of using neural networks (NNs) as a means lending support to the authenticity of steady-state CFD solutions for such ill-posed problems through inter-model comparisons. Attention is focused on using NNs trained on a database generated by numerically-stable CFD analysis to predict flow variables for the aforementioned ill-posed cases, thereby giving confidence in steady-state CFD results for these cases. Three types of NNs were evaluated and parametric studies were performed to optimize network designs for best predictions of the flow variables. A validated CFD code was used to generate a database that covered the range of Rayleigh number Ra from 10 4 to 4.7 × 10 6 . The CFD analysis was used to calculate the normalized temperature θ and stream function ψ throughout a partitioned enclosure. The results of the CFD were used to train and test the neural networks. The robustness of the trained NNs was tested by applying them to a number of “production” data sets that the networks have never “seen” before. The trained NNs were used to verify CFD solutions in cases of higher Ra ranges for which CFD simulations produce different solutions for different discretization schemes. The results showed that it is possible to design and train certain architectures of NNs to accurately predict θ and ψ distribution within the enclosure, and hence impart confidence in the legitimacy of CFD solutions for new cases.

The continuing challenges of testing species distribution models

Summary Species distribution models could bring manifold benefits across ecology, but require careful testing to prove their reliability and guide users. Shortcomings in testing are often evident, failing to reflect recent methodological developments and changes in the way models are applied. We considered some of the fundamental issues. Generalizability is a basic requirement for predictive models, describing their capacity to produce accurate predictions with new data, i.e. in real applications beyond model training. Tests of generalizability should be as rigorous as possible: ideally using a large number of independent test sites (≥ 200–300) that represent anticipated applications. Bootstrapping identifies the role of overfitting of the training data in limiting a model's generalizability. Predictions from most distribution models are continuous variables. Their accuracy may be described by discrimination and calibration components. Discriminatory ability describes how well a model separates occupied from unoccupied sites. It is independent of species prevalence and is readily comparable between models. Rank correlation coefficients, such as the concordance index, are effective measures. Calibration describes the numerical accuracy of predictions (e.g. whether 40% of sites with predicted probabilities of 0·40 are occupied) but is frequently overlooked in model testing. Poor calibration could mislead any conservation efforts utilizing models to estimate the ‘value’ of different sites for a given species. Effective assessments can be made using smoothed calibration plots. The effects of species prevalence on nominal presence–absence predictions are well known. The currently preferred accuracy measure, Cohen's κ, has weaknesses. We argue that mutual information measures, based in information theory, may be more appropriate. Synthesis and applications. Model evaluation must be informative and should ideally: (i) define generalizability in detail; (ii) separate the discrimination and calibration components of accuracy and test both; (iii) adopt assessment techniques that permit more valid intermodel comparisons; (iv) avoid nominal presence–absence evaluation where possible and consider information‐theoretic measures; and (v) utilize the full range of techniques to help diagnose the causes of prediction problems. Few modellers in applied ecology and conservation biology satisfy these needs, making it difficult for others to evaluate models and identify potential misuses. The problems are real, and if uncorrected will damage conservation efforts through the inaccurate assessment of distribution and habitat preferences of important organisms.