Details, scripts, and examples for using the Horel-Group object archive on CHPC's Pando system.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

MesoWest Logo, why doesn't this work??

Using the Horel S3 Archive Buckets

Author: Brian Blaylock
Date: February 22, 2017
updated September 22, 2017
updated February 6, 2018 after New Pando rebuild


In January 2017, CHPC allocated the Horel Group 60 TB on the Pando S3 (Simple Storage Service) archive space. This space is used for the Horel archive. Presently, it houses the HRRR archive (> 45 TB), and GOES-16 archive. Pando failed in January 2018, and the entire archive was lost. The archive has been rebuilt and we continued to archive HRRR and GOES data. We are currently working with The Climate Corporation who have generously offered to help us restore the lost HRRR sfc files. We hope to restore about 75% of what was lost.

Pando, latin for I spread, is named after the vast network of aspen trees in Utah, thought to be the largest and oldest living organism on earth.

You can view and access data objects on Pando via:

Generic web access to all buckets and objects on Pando:

Access data via rclone

rclone allows you to sync files and directories between your linux computer and the S3 buckets (and other cloud services). Before getting started, first review the CHPC rclone tutorial.


  1. The easiest way is to load rclone with modules. Load rclone (I do this is in my .custom.csh file). Some versions of rclone are different depending on the host's RedHat version.

    module load rclone

  2. Set up the config file: Note: These are the settings used for the mesohorse user

    Type rclone config. You will be asked a series of questions. Use these options:

    1. Select n for new remote
    2. Enter a name. You will reference the S3 archive with this name. I used horelS3
    3. Type: Select option 2 for S3
    4. "Get AWS credentials from runtime": Set to False
    5. Enter the access key: Ask Brian or John for this, unless you know where to find it
    6. Enter the secret key: You'll have to ask for this, too
    7. Region: leave blank (press enter to skip)
    8. Endpoint: Enter
    9. Location: Select option 1 for none

Completing this setup makes a .rclone.conf file in your home directory

Basic Command Examples

The full usage documentation for rclone is found at The following examples are some of the more useful. These examples can be used if you named the archive source horelS3 as described in the configuration step above. If you named your source differently when you configured rclone, simply replace the name before the colon.

What do you want to do? Command Notes
make a new bucket rclone mkdir horelS3:hrrr
make a new bucket/path rclone mkdir horelS3:hrrr/oper/sfc/ copy will make the directory if it doesn't exist, so it isn't necessary to mkdir before copying
list top-level buckets rclone lsd horelS3: lsd Only lists the directories in the path
list buckets in bucket rclone lsd horelS3:hrrr
list buckets in path rclone lsd horelS3:hrrr/oper
list bucket contents rclone ls horelS3:hrrr ls will list everything in the bucket including all directory's contents, so this particular example isn't very useful
list bucket/path contents rclone ls horelS3:hrrr/oper/sfc/20171201 currently, no way to sort alpha-numerically, unless you pipe the output to sort. Add the following: `
list bucket contents rclone lsl horelS3:hrrr/oper/sfc/20161213 lsl will list more details than ls
copy file from your computer to S3 rclone copy ./file/name/on/linux/system horelS3:path/you/want/to/copy/to/ You have to use the newest version of rclone to rename files when you copy. With version 1.39 (not installed on the meso boxes), use copyto or moveto in order to rename files when transferring to Pando
copy file from S3 to your current directory rclone copy horelS3:HRRR/oper/sfc/20161201/hrrr.t12z.wrfsfcf16.grib2 .
delete a file or directory on S3 I'm not going to tell you how to do this because there is no undo button!!!

With rclone version 1.39, you can do a little more like rename a file on S3. This version is not an installed module on the meso boxes because they don't have the updated RedHat software. But you can use rclone-v1.39 that is located here: /uufs/

What do you want to do? Command Notes
move file from computer to S3 and rename ./rclone-v1.39/rclone moveto /this/path/and/file horelS3:HRRR/path/and/new-name will overwrite existing file?
copy file from computer to S3 and rename ./rclone-v1.39/rclone copyto /this/path/and/file horelS3:HRRR/path/and/new-name will not overwrite if file exists??

Access via URL and curl commands

You can view some of the file contents here: The trouble is that it shows everything in the HRRR bucket without letting you view the files for each specific directory. Also, not every file is listed because the list is limited to 1000 files.

Download a file:

Download a file from a browser URL

Download with wget


Download with cURL

curl -O

Download with cURL and rename

curl -o hrrr20170101_00zf00.grib2

Download a single variable with cURL

GRIB2 files have a useful ability. If you know the byte range of the variable you are interested, you can get just that variable rather than the full file by using cURL.

Byte ranges for each variable are located on Pando. Just add a .idx to the end of the file name you are interested:

For example, to get TMP:2 m temperature from a file:

curl -o 20180101_00zf00_2mTemp.grib2 --range 34884036-36136433

NOTE: If you can't view the .idx files from Pando in your browser, and instead prompts a download, then you many need to remove the .idx file from your list of default apps. I had to remove .idx from my Windows registry.

Pando Archive Contents and URL Structure

Important Dates What happened? Notes
2015-Apr-18 Began downloading HRRR sfc and prs analyses HRRRv1 Some days/hours may be missing
2015-May-30 Began downloading HRRR Bufr soundings for KSLC, KODG, and KPVU
2016-Jul-27 Began downloading HRRR sfc 15 hr forecasts
2016-Sep-01 Taylor began downloading HRRR-Alaska prs analyses and sfc 36 hr forecasts Runs occur every three hours, but becuase it's an experimental model, runs are not always availalbe.
2016-Aug-23 HRRRv2 implemented at NCEP starting with 12z run
2016-Aug-24 Began downloading HRRR sfc 18 hr forecasts HRRRv2 increased forecasts from 15 to 18 hours.
2016-Dec-01 Began downloading experimental HRRR sfc analyses HRRRv3: Runs aren't always available becuase this is an experimental model.
2017-Oct-01 Stopped downloading sub-hourly files will start again when fire season begins (May 2018)
2018-Jan Pando Failed and Rebuilt Start the archive again beginning January 1, 2018. Hope to recover past years with data from The Climate Company.


GOES-16 Level 2 data (multiband format) from the Amazon AWS NOAA archive.

  • ABI-L2-MCMIPC/ Advanced Baseline Imager, Level 2, multiband format Cloud Moisture products

      Example File:
      File description on Amazon.

horelS3:hrrr/ Operational HRRR

  • sfc/ Surface fields

      • Analysis and forecast hours (f00-f18) for all hours (00-23).
      • File example: hrrr.t00.wrfsfcf00.grib2
  • prs/ Pressure fields

      • Analysis hour (f00) only for all hours (00-23).
      • File example: hrrr.t00.wrfprsf00.grib2

hrrrX/ Experimental HRRR

  • sfc/ Surface fields
      • Analysis hour (f00) for all hours, if available.
      • File example: hrrrX.t00.wrfsfcf00.grib2

hrrrak/ HRRR Alaska (Operattional after May ??, 2018)

  • sfc/ Surface fields
      • Analysis and 36 hour forecasts (f00-f36), if available. Runs initialize every three hours at 0z, 3z, 6z, 9z, 12z, 15z, 18z, 21z.
      • File example: hrrrAK.t00.wrfsfcf00.grib2
  • prs/ Pressure fields
      • Analysis hours (f00) for run hours, if available
      • File example: hrrrAK.t00.wrfsfcf00.grib2

A visulatizaion of HRRR file available on the S3 archive can be explored on the HRRR download page.

Contents of this repository


Download scripts for the GOES-16 data from the noaa-goes16 bucket on Amazon S3. Run by cron every 15 minutes.


Download scripts for the HRRR data. These are run by cron every four hours (00:20, 06:20, 12:20, 18:20)

  • main download script
  • controls HRRR downloads from NOMADS
  • controls HRRR downloads from ESRL
  • sends me an email to confirm the files are now on the Pando archive.


Contains the version of rclone you should use so we don't get stuck when CHPC updates rclone versions.


Contains s3cmd which is used to change permissions of files on S3 from private to public, and vice versa. (see below Q&A for usage)


Old scripts used before the Pando Failure in January 2018


Scripts shared to me by users of the archive

Other scripts:


Other Questions and Answers:

How do I rename a file when I copy it to S3?

Use the copyto and moveto commands only available in the newest rclone version. The rclone version best to use is the one installed here: /uufs/

How do I list files on Pando in alpha-numeric order with rclone?

rclone will not sort file names for you, but you can pipe the output to the sort command. For example:

rclone ls horelS3:HRRR/oper/sfc/20170109/ | sort -k 2`

Where the "k" specifies which field to sort by. The first field is file size and the second field (2) is the file name.

You can sort direcotry contents like this:

rclone lsd horelS3:HRRR/oper/sfc | sort -k 4

How do you get the total size of a bucket or directory?

With some creative linux commands...

How big is a bucket, in Terabytes?

rclone ls horelS3:HRRR | cut -c 1-10 | awk '{total += $0} END{print "sum(TB)="total/1000000000000}'

How big is a directory, in Gigabytes?

rclone ls horelS3:HRRR/oper/sfc/20161213 | cut -c 1-10 | awk '{total += $0} END{print "sum(GB)="total/1000000000}'

How do you make a directory or files public/private?

You have to use s3cmd to change the files from public to private. You would want to do this for each file added to the S3 archive that you want to be downloadable from the download URL.

s3cmd is installed here: /uufs/
NOTE: In order to set bucket names that are all lower case to public, I had to modify the configuration file. In my .s3cfg file on the host_bucket line, remove the “s” after $(bucket). Once I did this I can could and make public whatever bucket name I want.

First navigate to /uufs/ directory.

Make public

A new bucket: ./s3cmd setacl s3://GOES16 --acl-public
A single file: ./s3cmd setacl s3://hrrr/sfc/20180101/filename.grib2 --acl-public
A directory: ./s3cmd setacl s3://hrrr/sfc/20180101/ --acl-public --recursive

Make private

A new bucket: ./s3cmd setacl s3://GOES16 --acl-private
A single file: ./s3cmd setacl s3://hrrr/sfc/20180101/filename.grib2 --acl-private
A directory: ./s3cmd setacl s3://hrrr/sfc/20180101/ --acl-private --recursive

How is rclone and s3cmd configured?

Configuration files for the mesohorse user:

How much space is left and when will the S3 archive fill up?

The Pando Usage Web Display (PUWD) shows the the Pando allocation and usage for each bucket. The script that creates this display is run once a day by Brian on meso4 and is located on GitHub.

Where can I find examples on how to download HRRR data with a script?

Check out the scripting tips here: Scripting Tips

How do I configure rclone to access the NOAA's GOES-16 archive on Amazon AWS?

Since the NOAA GOES-16 archive is a public and free bucket, it is really easy to access the data via rclone.

Configure rclone

rclone config

Name the remote something like goes16AWS. Select 2 for Amazon S3 access and press enter to select empty or default values. When asked if it is right, type y for yes.

You are now on your way to accessing the Amazon GOES16 archive. To list the buckets in the noaa-goes16 archive, type:

rclone lsd goes16AWS:noaa-goes16 

Contact: Brian Blaylock (