The Data Pipeline: Once the RCMs have finished running, there is still a considerable amount of work that must be done to make the data ready for end users. First, the model output must be post-processed into CF-compliant netCDF. Then it must be transferred from the modeling team's home institution to NCAR. (This may involves shipping terabyte hard-drives overseas.) Once the data arrive, a backup copy is made to the NERSC HPSS for disaster-recovery purposes. Then the data must be checked for correct formatting and data integrity and any errors corrected, which could involve the entire file being regenerated. Finally, an archival copy is made to the NCAR MSS, the data is transferred to its final resting place on disk, and the files are published through the ESG data portal. All of this takes time, so there can be a considerable delay between when the modeling run is finished and when the data is available for download from ESG.

Modeling Run Status

The following table shows the status of the NARCCAP modeling runs. No firm ETAs are currently available for the completion of runs in progress.

Data Availability

The following tables show which datafiles from completed runs have been checked, archived, and published, and are avilable for download through the ESG dataportal. Table 4 files are available here. 3-D Table 5 data has been separated into primary (250, 300, 500, 600, 700, 850, and 950 mb) and other levels, with the primary levels piroritized.

Missing Timesteps: Some files are missing timesteps. This happens for various reasons; it may be impossible for the RCM to run the final day before the GCM data cuts off, for example, or there may have been problems with storage space during the run that caused the data from some timesteps to fail to write to disk. We have done our best to ensure that any known gaps in the data during the official coverage periods have been filled with missing_data. If you find a file with a gap in the time coordinate, let us know and we will fix it. We have tabulated the known timesteps where all values are missing here.

CCSM-driven runs end in 1999: The driving data from the CCSM historical run ends one year short of the NARCCAP coverage period, at the end of 1999 instead of on 2000/12/31. All CCSM-current runs have only 29 years of valid data instead of 30.

ECPC & WRFP are ncep-only: They have been superseded by ECP2 and WRFG, respectively. See the model changes page for more information.

Missing data in HRM3-hadcm3: There are occasional, isolated missing_values (typically no more than 1 per file) scattered throughout these files. They replace 'glitch' values we found during QC that we believe were due to hardware errors associated with data transfer.

HRM3 times: Many, but not all, HRM3 files start at 00:00 UTC, rather than the standard 03:00 UTC. This may cause the number of timesteps per file to be different than expected. Note: it is likely there is a systematic off-by-one error in the time coordinate related to this issue. We are investigating; in the meantime, use caution with analyses that depend on time of day

HRM3 winds: Wind variables (spdmax, uas, vas, ua, va) from HRM3 have been output on a staggered grid, which is offset by half a gridcell relative to the grid used for all the other variables. We plan to interpolate the results to the standard grid for surface variables, but not for the 3-D (Table 5) data.

HRM3-hadcm3 mrro 2056: Data is not properly masked over the oceans in this file; use caution.

WRF-ncep starts late: The NCEP-driven WRF runs (both WRFG and WRFP) begin 1979-09-01, not 1979-01-01. This is an error that results in a shortened spin-up period, but is expected to have minimal effect on variables except those closely tied to soil moisture.

Missing WRFG-ccsm longwave radiation: Some of the longwave radiation variables for the CCSM-driven WRFG runs have extensive gaps or are missing entirely. This is a data problem, not a model problem: the outputs were not saved correctly, but the values were correct within the model while it was running.

GFDL timeslice mrro, mrros: No data is available for these variables because runoff values were not correctly computed in the 50km model.

GFDL timeslice mrfso, rootd: No data is available for these variables, because there are no variables in the model that correspond directly.

GFDL timeslice cli, clw: No data is available for these variables. They were not saved because the storage requirements proved prohibitive.

GFDL timeslice wa: No data is available for this variable because there is no vertical wind component diagnostic variable in this model. Instead, the vertical velocity in pressure coordinates (variable wap, units Pa/sec) is provided.

GFDL timeslice all 3-D variables: 3-D data from the GFDL timeslice is on native pressure levels, not interpolated to 25-mb heights.

GFDL timeslice landtyp, lat, lon: These variables are not yet available due to unresolved questions regarding format and representation.