BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

A Professor Ran A Weather Prediction Model On A $50 Computer

Following
This article is more than 4 years old.

This weekend I wrote about a National Oceanic and Atmospheric Administration (NOAA) announcement that the agency is tripling supercomputing capacity for its weather and climate modeling operations. In many circles, this news was well-received. A well-known “weather modeling” arms race (at least in the minds of some people) is underway with the European model (”the Euro”) as the front-runner. I discussed in Forbes why some of the “model A is better than model B” banter is a bit over the top. Since any good forecaster uses all of the models anyhow, the weather enterprise benefits from the “one-upmanship mentality” and recently announced upgrades by U.S., United Kingdom, Canadian, and European modeling centers. Herein, I write about the other end of the spectrum. A meteorology professor ran a numerical weather prediction model on a $50 computer and made a forecast for his state.

Over the weekend I noticed a Tweet from Victor Gensini, a professor in the Department of Geography and Atmospheric Sciences at Northern Illinois University. Gensini is a former doctoral student at the University of Georgia, and I happened to served on his advisory committee. He is one of the brightest young scholars within the weather and climate field and has made a name for himself with his studies on tornado prediction and severe weather - climate change relationships. His Tweet below inspired this article.

Saturday eve. project complete! Raspberry Pi 3 B+ running WRF creating a 48h forecast for IL/IN domain (30-km grid spacing). Took about 20 min wall clock to simulate. Next step is to bring in breadboard and have it draw a meteogram for my house on the LED header.

Professor Victor Gensini, Northern Illinois University

To translate some of that meteorology and computer jargon, he used a Raspberry Pi 3 B+ (more on that in a moment) to run the Weather Research and Forecasting (WRF) model to produce a 2-day forecast for a region encompassing Illinois and Indiana. I am familiar with Raspberry Pi because I bought my 12 year old son one last year. According to the website RaspberryPi.org, “The Raspberry Pi is a low cost, credit-card sized computer that plugs into a computer monitor or TV, and uses a standard keyboard and mouse.” I purchased one for my son so that he could practice coding in Python and other web-based programming. The small computer can do pretty much anything that a larger computer can do and can connect to many external resources. The neat little computer is the brainchild of the The Raspberry Pi Foundation, a registered educational charity in the United Kingdom. The website notes that the Foundation’s goal is “to advance the education of adults and children, particularly in the field of computers, computer science and related subjects.”

I asked Professor Gensini why he decided to do this. He said because it was cool to do, but he also offered some scientific and instructional justification also:

Cool to do, but anyone with a Pi can do it in less than 30 min. It just takes three commands! I think it makes weather modeling accessible to the masses Victo☈ Gensini It’s remarkable that there’s more compute power on that Pi than previous supercomputers

Professor Victor Gensini, Northern Illinois University

Genisini noted that Raspberry Pi is a very powerful educational tool that is accessible to everyone. It can also teach the basics of electricity, computer engineering, or programming. As a young meteorology student at Florida State University, I took a class in FORTRAN programming, a rather ancient computer language these days. Ironically, many weather models still have legacy code in that language. I foresee numerical weather modeling pivoting as artificial intelligence or machine learning concepts emerge. The next generation of meteorologists will need to be increasingly comfortable with the atmospheric sciences, data, and coding so Gensini may be on to something with his Raspberry Pi experiment.

However, its applications are not limited to numerical weather modeling. John Trostel is a scientist and doctoral candidate at the Georgia Tech Research Institute and the Director of its Severe Storms Research Center. When I first bought my son’s Pi, Trostel told me that he had been using it as the Print Server in his house for over two years. Tom Gill is a professor of geology at the University of Texas-El Paso and an atmospheric dust expert Gill told me, “The State of Arizona in conjunction with the NWS is using a network of Raspberry-Pi-based devices to sense particulate matter concentrations in remote locations to give advance warning of dust storms causing hazards on Arizona highways.” Peter Neilly is the IBM Distinguished Engineer and Senior Vice President for Global Forecasting Sciences at The Weather Company. He has Pis connected to various personal weather stations for automation, collection, and visualization of data.

However, there are critics. Some have argued that though Raspberry Pi is cheap and accessible, it still may may be too complex for many students. Calvin Mackie disagrees. Mackie is a personal friend of mine and is an author, inventor, engineer, and former professor. He founded STEM NOLA as a non-profit organization “to expose, inspire and engage communities about the opportunities in Science, Technology, Engineering and Mathematics (STEM)” according to its website. Mackie told me his organization has been using Raspberry Pi in middle schools with great success in New Orleans.

For more information on how to run the WRF model on a Raspberry Pi, the National Center for Atmospheric Research (NCAR) has a website with information on Pi-WRF at this link.

Follow me on TwitterCheck out my website