Progress Towards a Pan-CMIP OSSE Capability for CLARREO 1 Daniel Feldman 1 Bill Collins 1,2, Chaincy Kuo 1, Bruce Wielicki 3, Yolanda Roberts 3, Constantine Lukashin 3 December 2, 2015 CLARREO Fall Science Team Meeting 1 Lawrence Berkeley National Laboratory, Earth Sciences Division 2 University of California-Berkeley, Department of Earth & Planetary Science 3 NASA Langley Research Center

Strengthening of International Collaborations. Recent research highlight. NRC Decadal Survey White Paper Summary. Overview and goals of a pan CMIP OSSE capability. Pan CMIP OSSE development update. –First phase models. The aerosol challenge. Summary. Discussion. 2 Outline

International Collaborations Building off PNAS (LBNL), GRL (Umich), and JC (Umich) papers on science value of surface emissivity measurements, collaboration with UK partners strengthened through NERC proposal. Investigation will analyze CIRCCREX data over Greenland and the Denmark Strait and retrieve far-IR surface emissivity over ice-sheet and ocean. Compare with calculations in emissivity database. Investigation will quantify ice-emissivity feedback within models. Preliminary diagnosis using kernel techniques for CESM1.2 is W/m 2 /K (~1/3 ice albedo feedback), but is state-dependent. Investigation will improve understanding of the controls on the polar radiative energy budget. 3

Highlight: Pan-Spectral OSSEs Feldman, D.R., W.D. Collins, J.L. Paige (2014) Pan-Spectral Observing System Simulation Experiments of Shortwave Reflectance and Longwave Radiance for Climate Model Evaluation, Geoscientific Model Development, 8, , doi: /gmd /gmd Summary of simulations of both the shortwave reflectance and longwave radiance clear-sky and all-sky CLARREO (OSSE), indicating the complementarity of the signals. Globally-averaged shortwave and longwave clear- and all-sky spectra from the CLARREO OSSE. Globally-averaged shortwave and longwave clear- and all-sky spectral trends from the CLARREO OSSE. 4

Highlight continued … Feldman, D.R., W.D. Collins, J.L. Paige (2014) Pan-Spectral Observing System Simulation Experiments of Shortwave Reflectance and Longwave Radiance for Climate Model Evaluation, Geoscientific Model Development, 8, , doi: /gmd /gmd We compare averages of random sampling of OSSE grid-cells vs calculations based on all grid cells. Sampling results indicate that acceptable radiometric error is incurred with 1000x fewer calculations are needed for global 10-year averages, 60x fewer calculations for regional averages. Globally-averaged shortwave and longwave clear- and all-sky spectra from the CLARREO OSSE. Globally-averaged shortwave and longwave clear- and all-sky spectral trends from the CLARREO OSSE. 5

NRC White Paper Summary Born out of ongoing conversation with Bruce, Yolanda, Costy and others, and the CFMIP meeting in June, White Paper Submitted November 2, Co-authors William Collins, Bruce Wielicki, Jonathan Jiang (JPL), Isaac Moradi (UMD), Nikki Prive (GSFC). Summarized the informal state of OSSEs for mission design and implementation. Recommendations: Formal, independent, and intramural OSSE capability for NASA. Encouragement of participation from modelers: Scientifically: Use OSSEs to explore hypothesis testing. Financially: Provide resources for modelers to participate in OSSEs. 6

Pan CMIP5 OSSE Capability At Spring CLARREO meeting, Berkeley group received guidance to focus on development of a pan CMIP5 OSSE capability in support of CLARREO. Berkeley group will build off current OSSE, develop model-agnostic OSSE capability for SW reflectance and LW radiance for CMIP5 and CMIP6 models. Goals: Determine relationship between model ECS and pan-spectral trends. Establish pan-spectral variability across a broad range of climate models, with an eye towards observational constraint. Support Decadal Survey discussion. 7

Progress Report First Phase Models Overview. Data Acquisition. Boundary Conditions. Vertical Interpolation. Benchmark Testing Execution on NAS facilities. 8

First Phase Models Model vertical-coordinate idiosyncrasies and data availability from Earth System Grid and mirroring servers have limited the feasible range of models for the first phase of the pan CMIP OSSE. First phase contains nearly complete range of diagnosed ECS across full ensemble. Model NameDiagnosed ECS (°K/2xCO2) CESM1-CAM54.10 CanESM23.69 MIROC-ESM4.67 MIROC52.72 MRI-CGM32.60 MRI-ESM12.11 inmcm

Data Acquisition Thermodynamic and condensate profiles downloaded from Earth System Grid and its mirrors. MODIS 16-day averages of 7 bands of BRDF and albedo, and 6 bands of emissivity gathered for Model-agnostic NCL and Matlab routines developed to concatenate fields into appropriate input files for OSSE. If you expect to work with multiple CMIP5 or CMIP6 fields, these routines will save you time. 10

Boundary Condition Interpolation Key component of Berkeley’s pan-spectral OSSE is surface boundary conditions built off of MODIS land-surface products. Gridding required for MODIS Climate Modeling Grid (CMG) data for each model’s horizontal resolution. Highly computationally-intensive, but parallelizable. 11 VisibleNear-IR

Interpolation OSSE is now model-agnostic for horizontal resolution, but it is less flexible for vertical resolution and inflexible on using layers, instead of levels. Vertical interpolation routines are required for thermodynamic and condensate profiles. Routines need to be mathematically stable, no edge cases. Construction of a positive-definite, monotonic function is necessary for gaseous and condensate species. For cloud amount, we use saturated specific humidity to create saturated precipitable water path, then multiplying by cloud amount to create function onto which interpolation should occur. 12

Benchmark Testing Multiple scattering line-by-line radiative transfer calculations (LBLRTM+CHARTS) coupled to a database of ice cloud optical properties enable versatile spectral flux calculations. NERSC and NAS resources take advantage of embarrassingly parallel nature of these calculations. Currently calculations are disk limited … upcoming AGU poster. 13

Execution on NAS MPI library updates on NASA Advanced Supercomputing (NAS) facilities enable better scaling of OSSE to larger numbers of processors (i.e., 2048, 4096). OSSE has been configured for new MPI library and now Berkeley group received new 1-Million CPU NASA High-End Computing Award received for for execution on NAS. The vertical interpolation routines still need to be tested. 14

The Aerosol Challenge CMIP5 models generally do not archive sufficient information on aerosols to include them in the pan CMIP OSSE. Clean-sky simulations are tractable, otherwise an aerosol placeholder is necessary. Aerosol forcing, and clear-sky reflectance simulations will be inconsistent with CMIP5 model results. CMIP6 will probably be better. 15

International collaboration with Imperial College and UMich partners yielding exciting science in the far-IR. Decadal Survey white paper recommends independent OSSE capability that entrains modelers. Pieces for a pan CMIP OSSE are now in place. –Final testing and benchmarking needed. –With EVM2 done, we expect preliminary OSSE first phase model results of RCP8.5 in early –Time-series analysis of pan-spectral signals to follow. 16 Acknowledgements: This research was supported by NASA grants NNH11AQ75I and NNH14ZDA001N-ACSCS. NERSC and NASA High-End Computing grants provided computational resources. Summary

Discussion Feedback from CLARREO Science Team on what CMIP5 OSSE runs will lead to synergistic research activities. Ongoing collaborations with NASA LaRC on PCRTM and fingerprinting should also be considered. Thoughts on how the pan CMIP OSSE should address the under-specification of aerosols in CMIP5 are appreciated. Feedback can be directly conveyed to Earth System Grid Federation at next week’s meeting. 17