In the spring of 2019, Northern Illinois University (NIU) applied for and received a
Unidata Community Equipment
Award
grant for a project titled “Bringing back weather.niu.edu: A multifaced
server at Northern Illinois University.”
The NIU Meteorology department (now Geographic and Atmospheric Sciences) has been involved
in the dissemination of meteorological data since the late 1990s, when
Russell L. DeSouza Award
winner
Mr. Gilbert Sebenste set up the “NIU Weather” server at weather.niu.edu.
The server
relayed data to dozens of Universities via the LDM
and had a popular “storm machine” website that provided some of the earliest
model forecast soundings. After Mr. Sebenste's departure from NIU in 2017, the server was
taken off-line.
NIU's project aimed to bring back this community resource, recreating the server as an LDM relay, THREDDS Data Server (TDS), and public-facing web site. Community members accessing the web site will have access to a variety of Unidata-derived products including GOES 16/17 imagery, model data and images, analysis data, and textual products. The TDS will initially host an archive of North American Regional Reanalysis (NARR) data. And in addition to these public-facing features, the equipment will run a JupyterLab server supporting student access to tools and data at NIU.
(click to enlarge)
In its initial incarnation, the new weather.niu.edu server has been configured as an LDM relay, feeding data to the nearby College of DuPage, and as a JupyterLab server supporting GEOG-493 (Programming for the Geographic and Atmospheric Sciences) during the spring 2020 semester. The server is accessible to students via an internet browser, allowing them to run iPython notebooks remotely and view real-time weather and climate data. It has also been used by students in the MET-421 (Synoptic Meteorology), MET-300 (Meteorology), MET-475 (Practicum in Weather Analysis and Forecasting), and MET-360 (Radar Meteorology) courses.
Existing resource scripts are being transferred to weather.niu.edu, and the public-facing web site is currently under development. Plans call for the public face of the server to provide weather and climate data that are difficult to locate elsewhere or recreate. In providing these data publicly, we hope to build a repository of real-time and archived weather maps and data that the community can use in their teaching, research, and outreach endeavors. We hope to make the TDS serving the NARR dataset available in the fall of 2020; in the meantime, we are happy to provide LDM feeds to other colleges and universities upon request.
The new server was first installed in September 2019. It is a capable server, with 2 Intel Xeon 3.1GHz 18-core CPUs, 192 GB of RAM, a 10 GB ethernet network card, 2 256GB SSD drives for running the OS in RAID 10 configuration, and 8 12-TB HDD drives in RAID 5 configuration for data storage and curation. The server is currently running CENTOS 7, and will use the Wordpress software to serve public-facing web pages.
In addition to serving the community and aiding in teaching, weather.niu.edu will also help serve the needs of research in the department. In spring 2020, the department agreed to purchase a GeForce RTX 2080 Ti 11 GB graphics card to upgrade the GPU in weather.niu.edu. This GPU upgrade will significantly expand the research capability of the server, especially as it pertains to machine learning. Initial testing using Python and tensorflow/keras for the development of a Convolutional Neural Network for image classification showed a 1000x speedup over using CPUs for processing. Given the growing importance of using such research methods in the geosciences, this additional hardware represents a significant cost-sharing investment from the Department of Geographic and Atmospheric Sciences at NIU.
Unidata Community Equipment Award grants make funds available to colleges and universities to purchase equipment or cloud-based computing services that will enhance their participation in the Unidata program. For additional information on program, visit the Equipment Awards page.
