Code Supporting "Could the Last Interglacial Constrain Projections of Future Antarctic Ice Mass Loss and Sea-level Rise?" by Gilford et al. (2020)

This repository is the Python code base for Gilford et al. 2020: "Could the Last Interglacial Constrain Projections of Future Antarctic Ice Mass Loss and Sea-level Rise?" Gilford et al. (2020), published in JGR-Earth Surface.

Project abstract:

Previous studies have interpreted Last Interglacial (LIG; ~129--116 ka) sea-level estimates in multiple different ways to calibrate projections of future Antarctic ice-sheet (AIS) mass loss and associated sea-level rise. This study systematically explores the extent to which LIG constraints could inform future Antarctic contributions to sea-level rise. We develop a Gaussian process emulator of an ice-sheet model to produce continuous probabilistic projections of Antarctic sea-level contributions over the LIG and a future high-emissions scenario. We use a Bayesian approach conditioning emulator projections on a set of LIG constraints to find associated likelihoods of model parameterizations. LIG estimates inform both the probability of past and future ice-sheet instabilities and projections of future sea-level rise through 2150. Although best-available LIG estimates do not meaningfully constrain Antarctic mass loss projections or physical processes until 2060, they become increasingly informative over the next 130 years. Uncertainties of up to 50 cm remain in future projections even if LIG Antarctic mass loss is precisely known (+/-5 cm), indicating there is a limit to how informative the LIG could be for ice-sheet model future projections. The efficacy of LIG constraints on Antarctic mass loss also depends on assumptions about the Greenland ice sheet and LIG sea-level chronology. However, improved field measurements and understanding of LIG sea levels still have potential to improve future sea-level projections, highlighting the importance of continued observational efforts.

If you have any questions, comments, or feedback on this work or code, please contact Daniel or open an Issue in the repository.

Citation

Gilford et al. (2020) has been published and is open access at JGR-Earth Surface. If you use any or all of this code in your work, please include the citation:

Gilford, D. M., Ashe, E. L., DeConto, R. M., Kopp, R. E., Pollard, D., & Rovere, A. (2020). 
Could the Last Interglacial constrain projections of future Antarctic ice mass loss and sea‐level rise?. 
Journal of Geophysical Research: Earth Surface, 125, e2019JF005418. https://doi.org/10.1029/2019JF005418

Data

The ice-sheet data used to construct the emulator used in this work has been archived at Zenodo with the doi:

Other data sources, including from Kopp et al. (2009), are included in this repository and should be appropriately cited if used.

Getting Started

This codebase requires Python version 3.5+ to run. It was written and tested with Python 3.5.2 and Jupyter Notebook 5.4.0. To get it up and running on your system, clone the repository and ensure that you have the required dependencies and Jupyter Notebook.

Dependencies

• Pandas 0.19.2
• GPflow 1.3.0
• Matplotlib 1.5.3
• NumPy 1.16.0
• TensorFlow 1.8.0
• SciPy 1.5.2

File Descriptions

Analysis:

Emulator Construction:

• construct_lig_emulator.ipynb - Notebook creating the Last Interglacial emulator optimized+conditioned on the training data
• construct_rcp_emulator.ipynb - Notebook creating the RCP8.5 emulator optimized+conditioned on the training data
• lig_utilities.py - Utility file containing functions for emulation and analysis
• plot_original_simulations.ipynb - Notebook exploring and visualizing the ice-sheet model ensembles (training data)

Emulator Sampling:

• lhc_design.ipynb - Notebook developing the latin-hypercube design used to sample the emulators
• sample_lig_emulator.ipynb - Notebook specifying the LIG constraint distributions, finding the likelihoods of each CLIFVMAX/CREVLIQ sample, and sampling the LIG emulator weighted by these likelihoods
• sample_rcp_emulator.ipynb - Notebook sampling the RCP8.5 emulator weighted by the likelihoods of each CLIFVMAX/CREVLIQ sample (inform by each specified LIG constraint)
• visualize_lig_error_importance.ipynb - Notebook illustrating how future projection uncertainties vary as a function of LIG constraint uncertainties

Emulator Validation:

• lig_cv_validation.ipynb - Notebook exploring various covariances functions to inform emulator construction
• loo_lig_analyses.ipynb - Notebook analyzing the validity of the LIG emulator with the Bastos and O'Hagan (2009) Leave-one-out method
• loo_rcp_analyses.ipynb - Notebook analyzing the validity of the RCP8.5 emulator with the Bastos and O'Hagan (2009) Leave-one-out method

Data:

• lig_data_dec18.pk1 - Last Interglacial Ensemble of 196 PSU3Dice simulations developed by Rob DeConto in December 2018
• rcp85_data_sept18.pk1 - Representative Concentration Pathway 8.5 Ensemble of 196 PSU3Dice simulations developed by Rob DeConto in September 2018
• K09_data.mat - MATLAB file containing Gaussian Process samples from the Kopp et al. (2009) Last Interglacial probabilistic assessment

Figures:

Main Text (see main manuscript for full captions):

• Figure 1a - Timeseries of LIG ice-sheet model simulations (0-5ka), color-coded by CLIFVMAX
• Figure 1b - Timeseries of RCP8.5 ice-sheet model simulations (2000-2150), color-coded by CLIFVMAX
• Figure 2a - LIG ice-sheet model simulations and emulator output across the ice-sheet model parameter space
• Figure 2b - RCP8.5 ice-sheet model simulations and emulator output (in 2100) across the ice-sheet model parameter space
• Figure 3 - Timeseries of the RCP8.5 emulated prior probability distribution of future AIS mass loss
• Figure 4a - Unconstrained and specified LIG constraint distributions of AIS mass loss in sea-level equivalent
• Figure 4b - Posterior RCP8.5 probability densities in 2100 as a function of (conditional on) the LIG
• Figure 5a - Posterior marginal probability distributions of CLIFVMAX as a function of LIG AIS mass loss
• Figure 5b - Posterior marginal probability distributions of CREVLIQ as a function of LIG AIS mass loss
• Figure 6a - Unconstrained and posterior projected probability distributions of AIS mass loss in 2100
• Figure 6b - Unconstrained and posterior projected probability distributions of AIS mass loss in 2150
• Figure 7a - Timeseries of the unconstrained and posterior probability distribution of AIS mass loss assuming the D20-U is narrowed by 50%
• Figure 7b - Timeseries of the posterior probability distributions of AIS mass loss assuming LIG contributions were High (>6m) or Low (<3.5m)
• Figure 8A - Range of RCP8.5 95% credible interval in 2100 as a function of the LIG uncertainty range, only Field constraints (8+/-0.2m)
• Figure 8B - Range of RCP8.5 95% credible interval in 2100 as a function of the LIG uncertainty range, corrected for GIA/fingerprints (7.6+/-1.7m)
• Figure 8C - Range of RCP8.5 95% credible interval in 2100 as a function of the LIG uncertainty range, corrected for GIA/fingerprints+DT (8.4+/-2.5m)
• Figure 8D - Posterior projected probability distributions of AIS mass loss in 2100 associated with different assumed GrIS sea-level contributions

Supplemental (see supplement for full captions):

• Figure S1a - Timeseries of LIG ice-sheet model simulations (0-5ka), color-coded by CREVLIQ
• Figure S1b - Timeseries of RCP8.5 ice-sheet model simulations (2000-2150), color-coded by CREVLIQ
• Figure S2 - Contours of LIG and RCP8.5 (2100) simulations across the ice-sheet model parameter space
• Figure S3 - Prediction errors from leave-one-out analysis for LIG emulator validation
• Figure S4 - Prediction errors from leave-one-out analysis for RCP8.5 emulator validation (at 2000, 2050, 2100, 2150)
• Figure S5a - Likelihood functions of CLIFVMAX/CREVLIQ set samples from the D20-U constraint
• Figure S5b - Likelihood functions of CLIFVMAX/CREVLIQ set samples from the D20-N constraint
• Figure S5c - Likelihood functions of CLIFVMAX/CREVLIQ set samples from the E19-U constraint
• Figure S5d - Likelihood functions of CLIFVMAX/CREVLIQ set samples from the K09-125ka constraint
• Figure S5e - Likelihood functions of CLIFVMAX/CREVLIQ set samples from the K09-Max-3kyrSmooth constraint
• Figure S6 - Posterior RCP8.5 probability densities in 2150 as a function of (conditional on) the LIG
• Figure S7 - LIG emulator variance across the ice-sheet model parameter space
• Figure S8 - LIG probability distributions comparing the final model with an alternative model form. While not explicitly created in the accompanying scripts, this can be reproduced by replacing the emulator covariance functions in each notebooj with an alternative covariance structure (see paper supplement for details)
• Figure S9 - Posterior RCP8.5 probability densities in 2100 as a function of (conditional on) the LIG with an alternative model form. While not explicitly created in the accompanying scripts, this can be reproduced by replacing the emulator covariance functions in each notebooj with an alternative covariance structure (see paper supplement for details)
• Figure S10 - Posterior projected probability distributions of AIS mass loss in 2100 associated with narrowing the D20-U LIG constraint

Miscellaneous:

• /colormaps/ - Colormap files provided by F. Crameri (2018)
• requirements.txt - List of packages required to compile and run code within this repository

Authors

Contributors

• Daniel M. Gilford, PhD - Primary Author, Development & Maintenance - GitHub
• Erica Ashe, PhD - Co-author, Bayesian Statistics - GitHub

Co-authors

• Prof. Bob Kopp
• Prof. Rob DeConto
• Prof. David Pollard
• Prof. Alessio Rovere

License

This project is licensed under the MIT License - see the LICENSE file for details

Acknowledgments

We thank Andrea Dutton, Anna Ruth Halberstadt, Jacky Austermann, three anonymous reviewers, and the editors for helpful comments which improved this manuscript. DG, RK, RD, and DP were supported by NSF Grant ICER-1663807 and NASA Grant 80NSSC17K0698. EA and RK were supported by NSF Grant OCE-1702587. AR is supported by the European Research Council, under the European Union's Horizon 2020 research and innovation programme (grant agreement n. 802414). This project grew, in part, out of discussions at the 2018 PALeo constraints on SEA level rise (PALSEA) annual meeting. PALSEA is a working group of the International Union for Quaternary Sciences (INQUA) and Past Global Changes (PAGES), which in turn received support from the Swiss Academy of Sciences and the Chinese Academy of Sciences. GP regression was performed with GPflow (Matthews et al., 2017); some color maps provided by Crameri (2018).