Replicability of the EC-Earth3 Earth System Model under a change in computing environment

Massonnet, François; Ménégoz, Martin; Acosta, Mario; Yepes-Arbós, Xavier; Exarchou, Eleftheria; Doblas‐Reyes, F. J.

doi:10.5194/gmd-2019-91

Cited by 9 publications

(11 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, Baker et al [1] have used their approach to identify issues with certain compilers on certain machines. Massonnet et al [9] also applied their index based test as they ported their model to a new machine. We plan to apply the new multivariate tests for such real world scenarios in the near future, which will also help establish an optimum balance between the false positive and false negative rates of the tests.…”

Section: Summary Discussion and Future Workmentioning

confidence: 99%

“…Mahajan et al [7,8] applied a testing approach where they also compare short (1 yr) ensembles (about 30 members) of the control as well as the perturbed simulation to evaluate if the climate statistics of the two ensembles are statistically similar by evaluating the global annual average of each model output variable separately. Massonnet et al [9] compare the distribution of a normalized index of 13 variables to evaluate the reproducibility of a fully coupled model using 5member ensembles of 20-yr simulations.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Multivariate Approach to Ensure Statistical Reproducibility of Climate Model Simulations

Mahajan

Evans

Kennedy

et al. 2019

Proceedings of the Platform for Advanced Scientific Computing Conference

View full text Add to dashboard Cite

Effective utilization of novel hybrid architectures of pre-exascale and exascale machines requires transformations to global climate modeling systems that may not reproduce the original model solution bit-for-bit. Round-off level differences grow rapidly in these non-linear and chaotic systems. This makes it difficult to isolate bugs/errors from innocuous growth expected from round-off level differences. Here, we apply two modern multivariate two sample equality of distribution tests to evaluate statistical reproducibility of global climate model simulations using standard monthly output of short (∼ 1-year) simulation ensembles of a control model and a modified model of US Department of Energy's Energy Exascale Earth System Model (E3SM). Both the tests are able to identify changes induced by modifications to some model tuning parameters. We also conduct formal power analyses of the tests by applying them on designed suites of short simulation ensembles each with an increasingly different climate from the control ensemble. The results are compared against those from another such test. These power analyses provide a framework to quantify the degree of differences that can be detected confidently by the tests for a given ensemble size (sample size). This will allow model developers using the tests to make an informed decision when accepting/rejecting an unintentional non-bit-for-bit change to the model solution. CCS CONCEPTS• Software and its engineering → Software verification and validation; Empirical software validation.

show abstract

Section: Summary Discussion and Future Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Multivariate Approach to Ensure Statistical Reproducibility of Climate Model Simulations

Mahajan

Evans

Kennedy

et al. 2019

Proceedings of the Platform for Advanced Scientific Computing Conference

View full text Add to dashboard Cite

show abstract

“…For scenario MIP, we select three available climate model data with high spatial resolution, namely BCC-CSM2-MR (Wu et al 2019), EC-Earth3 (Massonnet et al 2019) and MPI-ESM1-2-HR (Gutjahr et al 2019), the basic information of which is shown in Table 1. Since the second aspect of this study is mainly to compare the differences of the three calibration strategies in streamflow prediction, the daily scale precipitation, minimum and maximum temperatures of the three climate models for 2021-2050 are analyzed.…”

Section: Climate Model Datamentioning

confidence: 99%

Influence of multisite calibration on streamflow estimation based on the hydrological model with CMADS inputs

Song

Zhang

Lai

2021

Journal of Water and Climate Change

View full text Add to dashboard Cite

Due to the spatial heterogeneity, the hydrological model calibration results only at the total outlet of the basin may not represent the whole basin. To more accurately simulate the historical streamflow process within the Qujiang River Basin, we set up three calibration strategies (single-site, S1; multisite simultaneous, S2; and multisite sequential, S3) for four hydrological stations based on the SWAT (Soil and Water Assessment Tool) model driven by CMADS (China Meteorological Assimilation Driving Datasets for the SWAT model). In addition, the implications of these calibration issues are extended to future streamflow projections using multimodel ensemble data in CMIP6 (Coupled Model Intercomparison Project Phase 6). In the model calibration phase, the SWAT model achieved very satisfactory results in the study area. Compared with S1 and S2, S3 can effectively improve the accuracy of streamflow simulation of stations within the basin and reduce the simulation deviation. Especially at the daily scale, the average NSE values of the four stations with S3 increased by 0.26 and 0.07, and the overall deviation decreased by 0.25 and 6.43%, respectively. Parameter sensitivity analysis also shows that spatial heterogeneity can be more adequately considered when using S3 to calibrate the model. As for the results of future streamflow projection, when using the S3, the average annual streamflow of four stations in the three climate scenarios from 2021 to 2050 is about 44.21, 130.00, 321.55 and 713.24 m3/s, respectively. Correspondingly, the use of S1 and S2 would bring certain risks to future water resource management.

show abstract

“…However, the cray reproducibility tests did not pass and therefore the EC-Earth consortium was advised to use the intel compilers in place of cray compliers on the ECMWF system. The EC-Earth reproducibility protocol was designed by the EC-Earth community (Massonnet et al, 2019) to assess the reproducibility of the EC-Earth model system. This protocol is based on a statistical comparison of standard climate metrics derived from multi-ensembles of multi-decadal control integrations executed in different HPC environments.…”

Section: Implementation Of Ec-earth On Ichec and Ecmwf Supercomputing Systemsmentioning

confidence: 99%

EC-Earth Global Climate Simulations: Ireland’s Contributions to CMIP6

Nolan¹,

McKinstry²

2020

Preprint

View full text Add to dashboard Cite

The global climate simulations described in this report constitute Ireland’s contribution to the Coupled Model Intercomparison Project (CMIP) (phase 6) (CMIP6) and will be included for assessment in the United Nations Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report (AR6). Since 1995, CMIP has co-ordinated climate model experiments involving multiple international modelling teams. The CMIP project has led to a better understanding of past, present and future climate, and CMIP model experiments have routinely been the basis for future climate change assessments carried out by the IPCC. The CMIP phase 5 (CMIP5) simulations have demonstrated the added value of improved models and enhanced resolution when compared with outputs from the CMIP phase 3 (CMIP3) project. This improvement in skill is expected to continue with the CMIP6 simulations. The EC-Earth consortium participated in CMIP5 and is currently participating in CMIP6 using a model that includes biogeochemical cycles and atmospheric chemistry. The current version of EC-Earth is based on the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast System (IFS) atmospheric model, the Nucleus for European Modelling of the Ocean (NEMO) model, the Louvainla-Neuve sea ice model (LIM3), the atmospheric Tracer Model version 5 (TM5), the Lund-Potsdam-Jena General Ecosystem Simulator (LPJ-GUESS) vegetation model and the Pelagic Interactions Scheme for Carbon and Ecosystem Studies (PISCES) ocean biogeochemistry model. Coupling is provided by OASIS3-MCT (the Ocean Atmosphere Sea Ice Soil – OASIS – coupler interfaced with the Model Coupling Toolkit – MCT).As part of the current project, the EC-Earth Atmosphere–Ocean General Circulation Model (AOGCM) configuration was employed. The atmosphere was simulated with ~79-km horizontal grid spacings (T255) and 91 vertical levels. The ocean was simulated with 1-degree horizontal resolution and 75 vertical levels. In total, five historical (1850–2014) and 20 Scenario Model Intercomparison Project (ScenarioMIP) simulations (2015–2100) were run. The future climate was simulated under the full range of ScenarioMIP “tier 1” shared socioeconomic pathways (SSPs); SSP1–2.6, SSP2–4.5, SSP3–7.0 and SSP5–8.5. For one ensemble member, all model levels were archived, allowing for regional downscaling using regional climate models and participation in the CMIP6 Coordinated Regional Downscaling Experiment (CORDEX) Model Intercomparison Project (MIP). All CMIP6 data were published on the ICHEC Earth System Grid Federation (ESGF) node. Chapter 2 provides an overview of validations of 2-m temperature, precipitation, 10-m wind speed, mean sea level pressure (MSLP), total cloud cover, snowfall, sea surface temperature and sea ice fraction. The EC-Earth historical data were compared with Climatic Research Unit observational datasets and ERA5 global reanalysis data (ERA5 is the fifth generation of the ECMWF global climate reanalysis dataset). Results confirm the ability of the EC-Earth model to simulate the global climate with a high level of accuracy. Chapter 3 provides an overview of EC-Earth global climate projections. The future global climate was simulated to the year 2100 under each of the four SSPs (SSP1–2.6, SSP2–4.5, SSP3–7.0 and SSP5–8.5). This results in 20 future global climate experiments (five ensembles multiplied by four SSPs). Projections of climate change were assessed by comparing the two 30-year future periods 2041–2070 and 2071–2100 with the 30-year historical period 1981–2010. Climate projections are presented for the Northern Hemisphere winter (December, January and February), Northern Hemisphere summer (June, July and August) and over the full year. Results show large projected increases in temperature; the largest are noted over the land masses, in particular the northern-most regions and the Arctic. Projected temperature increases range from ~0.5°C over the Southern Hemisphere oceans for SSP1–2.6 (2041–2070) to ~18°C over the Arctic for SSP5–8.5 (2071–2100). By the year 2100, the global mean temperature is projected to increase by approximately 1.5°C, 2.8°C, 4.2°C and 5.5°C for SSP1–2.6, SSP2–4.5, SSP3–7.0 and SSP5–8.5, respectively. For precipitation, all ensemble members show a steady increase in mean global precipitation from around 2000, with a noticeable divergence between the SSPs around 2060. By the year 2100, global mean precipitation is projected to increase by approximately 4%, 6%, 8% and 10% for SSP1–2.6, SSP2–4.5, SSP3–7.0 and SSP5–8.5, respectively. Northern Hemisphere sea ice is projected to disappear in the September months by 2071–2100 under SSP2–4.5, SSP3–7.0 and SSP5–8.5. Projections of 10-m wind speed, MSLP, total cloud cover, snowfall and sea surface temperature are also presented in Chapter 3

show abstract

Replicability of the EC-Earth3 Earth System Model under a change in computing environment

Cited by 9 publications

References 22 publications

A Multivariate Approach to Ensure Statistical Reproducibility of Climate Model Simulations

A Multivariate Approach to Ensure Statistical Reproducibility of Climate Model Simulations

Influence of multisite calibration on streamflow estimation based on the hydrological model with CMADS inputs

EC-Earth Global Climate Simulations: Ireland’s Contributions to CMIP6

Contact Info

Product

Resources

About