PIRCS Beta-test Dataset

PIRCS Boundary Conditions Beta-Test Data Set

To access the betatest dataset:

ftp ftp.pircs.iastate.edu

passwd: your e-mail address

cd /pub/PIRCS/BETA.RELEASE

"Get" the files

	README (contains information similar to what is in this document)
	tarCODE (tar file contain code to read the data files,
	also some interpolation routines)


	and one of the following:
	tarGLOBAL (contains data on a global 2.5 degree grid (i))
	tarREGDOM (contains data on PIRCS 0.5 degree grid (ii))
	tarPSTDOM (contains data on PIRCS polar stereographic grid (iii))
	(depending on which set of LBC's are appropriate for your model)

Note:

To extract files from a tarfile (tarCODE for example) type the command tar xvf tarCODE
Sometimes the network can be slow, so if problems are encountered transferring files, try again later, if the problem persists, please contact us.

2 - DATA PACKING

Users of the PIRCS Beta-Test dataset should be aware that the provided variables are not stagggered, i.e., they are all at the same location. Because participant models may use staggered grids horizontally the width of the forcing zone will differ between models. For example, for the Arakawa B grid, the width for T and q will be 8 grid spacings, but 9 for U and V.

It is up to the participant to determine which PIRCS dataset suits their needs best. It is recommended that participants use a bilinear interpolation method to interpolate the PIRCS grid-point data to their own model's grid. A bilinear interpolation subroutine based on bint.f (described in Manning and Haagenson, 1992) is part of program indat_gen.f included in the distributed dataset. Various programs that read the files containing the data are available. Hopefully these programs are designed so that they can be easily adapted by the participants in their efforts to convert the LBC's to their specific model configuration. These codes are in the file tarCODE in the directory /pub/PIRCS/BETA.RELEASE at the FTP site. A listing and brief description of each program is included in the README file of the directory READ.CODE (the directrory which the files are in, when tarCODE is untarred).

The data are packed in integer format, and are in slices ('horizontal') of constant pressure. (25 mb being first, 1050 being last.) The data are ordered as follows: surface pressure (Variable code 001; units Pa), temperature (011; K), specific humidity (051; kg/kg), u component of the wind (033; m/s), and the v component of the wind (034; m/s). This code "unpacks" the data, as they are stored in a positive integer format; i.e., for each field (horizontal slice) a maximum value is determined, and this (and every other data point) is multiplied by some scale factor to ensure that the desired number of significant digits are to the left of the decimal point. The original data are extracted by subtracting the stored data from the base value and then dividing by the scale factor. The space requirements are minimized through the storage of data as positive integers. To further minimize the space requirements, data were compacted using the standard UNIX "compress" (a modified Lempel-Ziv algorithm popularized in Welch, 1984) command. While this choice of packing procedure is not a meteorological "standard" like GRIB or BUFR, we think that fewer problems will be experienced in the short term.

For each 'horizontal' slice the data are ordered as follows, as you look at it on the screen (Note, this applies for the ordering of the forcing data as well.)

             
  SW Corner  (1,1) .........................(XMAX,1) SE Corner

            	(1,2) .........................(XMAX,2)

             	...........(I,J)...............

            	...............................

  NW corner  (1,YMAX).......................(XMAX,YMAX) NE Corner

Note, in the read program, the I index is E-W, and the J index is N-S.

3 - ISU BETATEST - SOME RESULTS

We ran the beta test at ISU using REGCM2 to produce the output appearing here.

( Figure 1.1(0.00833 MB) shows the time series of the domain averaged temperature at a near-surface level (sigma=0.9) for the BETAtest period as simulated by REGCM2, and ( Figure 1.2(0.00833 MB) shows the time series of the domain averaged wind speed at the same level. Figure 1.3 (below) shows the simulated precipitation for the period.


Fig. 1.3 - Total rainfall (mm) for the 5 day period
from 00 UTC July 1 to
00 UTC July 6, 1988 as simulated by REGCM2 using the
PIRCS BETAtest
dataset.

It has been our experience that a 5 day integration of REGCM2 (see Giorgi, et al., 1994) configured for the PIRCS domain takes about 17 hours of cpu time on a DEC AXP 3000 workstation with 32 MB of assigned memory, and 32 hours with 10 MB assigned.

The variables provided are surface pressure, T, q, u, & v. They are provided every 25 mb from 25 mb to 1050 mb. Also provided are fields of the NCEP analysis topography for all domains and topographic datasets for the limited domains (ii) and (iii). The latter were derived from a 30 minute dataset, and they have more detail than the analysis topography. Land-use datasets are also provided for each of the limited domains, as well as global SST's, soil moisture for the limited domains, and the Great Lakes surface temperatures.

REFERENCES

Manning, K. W., and P. L. Haagenson, 1992: Data ingest and objective analysis for the PSU/NCAR modeling system: Programs DATAGRID and RAWINS. NCAR Technical note, NCAR/TN- 376+IA, 209 pp.