This document describes access to the
data available from recent
direct numerical simulations of turbulent channel flow at Re_tau=180,
550 and 940 in relatively large spatial domains. Fields from these
simulations have been saved for further postprocessing analysis. For
each case, a number of velocity fields (order 100), broadly spaced in
time (75 delta/u_tau for Re_tau=940). Also available are standard
statistical profiles and spectra. Access to the data is here http://turbulence.oden.utexas.edu/data/channels/fields/.
Access to the velocity fields is involved, and is described below.

Each of the turbulence fields is a file consisting of approximately 10Gb of data for the Re_tau=940 case. The entire Re_tau=940 data base totals approximately 0.72 terabytes. For the Re_tau=550 and 180 cases, each field is approximately 2Gb and 130Mb respectively, for a total data base size of 55Gb and 15Gb. We are willing to share the fields with other research groups to facilitate their use in advancing turbulence research. However, the magnitude of the data, and the idiosyncrasies of this particular data set introduce several difficulties to sharing this field data.

The size of the fields produces two problems. First the size of the total data set makes it difficult to transfer the data to the recipient. Second, since an individual field is so large, processing it to extract the desired information requires rather sophisticated data management or very large memory (100's of gigabytes or more depending on the analysis to be performed).

The simulations were performed using the Fourier-spectral numerical method described in Kim, Moin & Moser (1987, J. Fluid Mech. vol 177 pp133-166). The state variables used in the code are the wall-normal vorticity and the Laplacian of the wall-normal velocity, and it is the Fourier/Chebychev transform of these quantities that are stored in the field files. Just to obtain the velocities on a grid, it is necessary to perform significant processing of the data, including solving a Poisson equation for the vertical velocity, determining the horizontal velocities from the vorticity and vertical velocity, and inverse transforming the data to obtain velocities on a grid.

To facilitate the processing of the fields, a detailed description of the structure and semantics of the data is included here http://turbulence.oden.utexas.edu/data/channels/fields which should provide sufficient information to allow the data to be correctly processed. Further, one of our post-processing codes (written in FORTRAN) is provided as an example, it reads the field, determines the velocities, and writes a new file containing the Fourier transform of the velocities. This code is provided as a supplement to the documentation, it may or may not work on your system, and we are unable to provide support for the operation of this code.

Due to the complexity of using this data, there is significant opportunity for error, and there are many practical difficulties in processing the data as well. Further, the costs and effort involved in transferring the data are substantial. It is therefore in the interest of anyone planning to use the data to ensure that they will be able to overcome the processing difficulties before acquiring the entire data set. To this end, we have provided a single field for immediate download here. There is one field from each of the three
cases. For the Re_tau=550
case (http://turbulence.oden.utexas.edu/data/channels/fields/re550),
a number of quantities computed from this
field by are provided. Those
interested in using the fields are asked to first download the sample
field, analyze it to obtain the quantities, and compare
with those provided. In this way, we can all be sure that the
recipient has the computational resources (both hardware and software)
necessary to process the data. The data is in single precision and
the statistical quantities are reported with 8 digits of precision.
The recipient should be able to reproduce these quantities to
round-off error (approximately a part in 10^7).

Once it is clear that the recipient of the data is equipped to process it, arrangements to transfer the data can be made. It is not practical to transfer data over the internet. Instead, the data will be physically transferred. The recipient is asked to purchase/build a disk or RAID disk system sufficiently large to store the desired data (about 0.8 terabytes is required for all three cases, and a RAID array of this size can be put together for a few thousand US dollars). The recipient will format the disk system and ship it to us. Formatting and shipping information is available here: http://turbulence.oden.utexas.edu/content/shipping.html.
We will load
the data set onto your disk system and ship it back. Before purchasing
hardware, or shipping it to us, please contact us as described in
formatting/shipping instructions listed above, so a number of details
can be decided upon.

Finally, we are providing these data as a service to the turbulence research community. We are however unable to provide support for the development of the software required to process the data. As a condition for the use of the data, the recipient agrees to acknowledge the data authors in any publications arising from the analysis of the data. Further, recipients of the data will cite the source of the data. Acknowledgement, and citation details are provided here http://turbulence.oden.utexas.edu/content/acknowledge.html.
Please check for
up-to-date information when finalizing your paper.

Each of the turbulence fields is a file consisting of approximately 10Gb of data for the Re_tau=940 case. The entire Re_tau=940 data base totals approximately 0.72 terabytes. For the Re_tau=550 and 180 cases, each field is approximately 2Gb and 130Mb respectively, for a total data base size of 55Gb and 15Gb. We are willing to share the fields with other research groups to facilitate their use in advancing turbulence research. However, the magnitude of the data, and the idiosyncrasies of this particular data set introduce several difficulties to sharing this field data.

The size of the fields produces two problems. First the size of the total data set makes it difficult to transfer the data to the recipient. Second, since an individual field is so large, processing it to extract the desired information requires rather sophisticated data management or very large memory (100's of gigabytes or more depending on the analysis to be performed).

The simulations were performed using the Fourier-spectral numerical method described in Kim, Moin & Moser (1987, J. Fluid Mech. vol 177 pp133-166). The state variables used in the code are the wall-normal vorticity and the Laplacian of the wall-normal velocity, and it is the Fourier/Chebychev transform of these quantities that are stored in the field files. Just to obtain the velocities on a grid, it is necessary to perform significant processing of the data, including solving a Poisson equation for the vertical velocity, determining the horizontal velocities from the vorticity and vertical velocity, and inverse transforming the data to obtain velocities on a grid.

To facilitate the processing of the fields, a detailed description of the structure and semantics of the data is included here http://turbulence.oden.utexas.edu/data/channels/fields which should provide sufficient information to allow the data to be correctly processed. Further, one of our post-processing codes (written in FORTRAN) is provided as an example, it reads the field, determines the velocities, and writes a new file containing the Fourier transform of the velocities. This code is provided as a supplement to the documentation, it may or may not work on your system, and we are unable to provide support for the operation of this code.

Due to the complexity of using this data, there is significant opportunity for error, and there are many practical difficulties in processing the data as well. Further, the costs and effort involved in transferring the data are substantial. It is therefore in the interest of anyone planning to use the data to ensure that they will be able to overcome the processing difficulties before acquiring the entire data set. To this end, we have provided a single field for immediate download here

Once it is clear that the recipient of the data is equipped to process it, arrangements to transfer the data can be made. It is not practical to transfer data over the internet. Instead, the data will be physically transferred. The recipient is asked to purchase/build a disk or RAID disk system sufficiently large to store the desired data (about 0.8 terabytes is required for all three cases, and a RAID array of this size can be put together for a few thousand US dollars). The recipient will format the disk system and ship it to us. Formatting and shipping information is available here:

Finally, we are providing these data as a service to the turbulence research community. We are however unable to provide support for the development of the software required to process the data. As a condition for the use of the data, the recipient agrees to acknowledge the data authors in any publications arising from the analysis of the data. Further, recipients of the data will cite the source of the data. Acknowledgement, and citation details are provided here