LENS | Known Issues1>
- September 2016: A small systematic difference between the CESM Large Ensemble runs done on the NCAR supercomputer Yellowstone without compression (members 1-30, 32, 34, 35) and CESM Large Ensemble runs done on the University of Toronto supercomputer (members 101-105) has been detected. Specifically, the global mean temperature in the University of Toronto runs (Figure 1, plotted in red) is systematically larger than the global mean temperature in the Yellowstone runs
(Figure 1, plotted in black). The origin of this systematic difference is a research
question. Differences in compiler interpretation or machine architecture may be
responsible. At present, there is no evidence linking this systematic difference to an
error in the implementation of the experimental protocol.
Plotting the first 50 days of the CESM Large Ensemble runs shows the expected
initial tight clustering of ensemble members and the expected increase in ensemble
spread associated with chaotic atmospheric circulation. Over the first 50 days, the
University of Toronto runs (Figure 2, plotted in red) are indistiguishable from
Yellowstone runs (Figure 2, plotted in black). CESM Large Ensemble users should
decide if this known systematic difference impacts their study. We are continuing to
examine the source of this difference as research resources allow. We will provide
any additional updates here.
Figure 1. Globally averaged annual-mean surface temperatures (TS) from the CESM Large Ensemble. Black lines represent members run on the Yellowstone supercomputer (members 1-30, 32, 34 and 35). Red lines represent members run on the University of Toronto supercomputer (members 101-105). Figure courtesy Vineel Yettella, University of Colorado.
Figure 2. Globally averaged daily mean surface temperature (TS). Black lines represent members run on Yellowstone supercomputer (members 1-30, 32, 34 and 35). Red lines represent members run on the University of Toronto supercomputer (members 101-105). Figure courtesy Vineel Yettella, University of Colorado.
- October 2015: There is a discontinuity in prescribed ozone concentration values between December 2080 and January 2081 in ensemble members 2-7 and 9-33. This is due to the fact that the CMIP5 ozone file was used starting in January 2081 instead of the WACCM ozone file. This error is expected to mainly affect the high-latitude Southern Hemisphere stratosphere and troposphere in austral spring and summer. Note: This known issue has no impact on any of the ensemble members before 2081. Note also that ensemble members 1, 8 and 34-40 use WACCM ozone throughout the entire period of simulation (1920-2100).
- October 2013: Ocean biogeochemistry was corrupted in ensemble members #3-8. This issue impacts all ocean biogeochemistry output fields, as well as some of the diagnostic atmospheric CO2 tracers; all other output for ensemble members #3-8 is unaffected. Affected history output has been removed in order to avoid corruption of analysis. While restart files for these ensemble member are available for use, users are advised against using them to initialize sensitivity experiments if their anticipated analysis includes ocean or atmosphere biogeochemistry fields. Note: This known issue has no impact on the physical climate system.