Scenario: Data Generated by Computer Models

Tools at a Glance

This case study describes research investigating the sensitivity of global climate models to increases in model resolution. The comparison involves output from several models, stored in formats including GRIB and netCDF (with and without the Climate and Forecast (CF) metadata conventions). Output files are converted between formats using NCEP's Unified Post Processor or the NCAR Command Language.

Project Background

Researchers at North Carolina State University (NCSU) led by PI Dr. Gary Lackmann investigated the possible changes in the North Atlantic Storm Track associated with climate change, specifically with taking into account enhanced condensational latent heating which accompanies warming. The team combined century-scale global climate modeling results with high-resolution synoptic dynamics to study possible changes to the expected storm track.

The research was supported by a National Science Foundation grant awarded in 2010, before Foundation-wide data management policies were implemented. Interested readers can peruse “The Importance of Resolving Mesoscale Latent Heating in the North Atlantic Storm Track”, which captures some of the results of the project.

Data Collected

Output from the Global Forecast System (GFS) model final analysis (FNL), retrieved from the NCAR Research Data Archive (RDA) in GRIB format, was used to establish initial and boundary conditions for a Weather Research and Forecasting (WRF) model simulation of a tropical cyclone that began December 30th, 2001. Similarly, a range of projections from the Coupled Model Intercomparison Project phase 3 (CMIP3) and phase 5 (CMIP5), in CF-compliant netCDF files, were used to initialize the WRF model for simulations of ten winter seasons in the North Atlantic.

In all cases, output from the WRF model took the form of staggered-grid netCDF-format files.

Tools Used for Data Processing and Analysis

Staggered-grid output files from the WRF model runs were processed using the NCEP Unified Post Processor software, which:

Interpolates the forecasts from the model's native vertical coordinate to NWS standard output levels;
Destaggers the forecasts from the staggered native grid to a regular non-staggered grid;
Computes diagnostic output quantities;
Outputs the results in NWS and WMO standard GRIB format.

The resulting GRIB and GEMPAK format files were then analyzed using Matlab and other software tools. In some cases, the NCAR Command Language was used to extract subsets of the data for detailed analysis.

Data Storage Strategy and Tools Used

The WRF model output was initially stored directly on large-capacity (3-5 terabyte) external hard drives connected to the NCSU network.

Data Access Strategy and Tools Used

Initially, data sharing was accomplished manually, with the research team using NCAR Command Language (NCL) programs to select subsets of the WRF model output, which were then made accessible to selected individuals via FTP.

Rather than sharing the entire raw WRF model output, the team provided the initial and boundary condition files, the WRF model namelist, along with the WRF model instructions, which allowed others to replicate the WRF model run.

Working with Unidata, Dr. Lackmann is also making the initial boundary layer conditions, the model, and model instructions available via a RAMADDA server running at NCSU. Portions of the dataset and results are also on the NCSU RAMADDA server.

Researchers who have data needs beyond what is publicly available on the RAMADDA server should contact Dr. Lackmann.

Data Archiving Strategy and Tools Used

Dr. Lackmann's current plans call for keeping the project files at NCSU, available to the scientific community via the NSCU RAMADDA server.