Shared publicly  - 
2
Josh Pan's profile photoJaymin Patel's profile photoDaniel Rothenberg's profile photoPeter Wilczynski's profile photo
10 comments
 
Nick Barnes is actually my mentor in the GSoC this summer. The biggest problem we have in the scientific community isn't the scientists and researchers. I gave a talk at the National Climatic Data Center last week about using Python in their operational work, and everybody was thrilled at the possibilities scientific programming in Python brings to the table. But the truth is, few if any of those scientists will adopt Python because their IT group is not capable of providing them with the necessary tools; apparently, it took the IT guys half a year to get SciPy installed, and at the end of it, they had installed an outdated and not-as-useful version of the package.

I spend a lot of time evangelizing Python and better software engineering to the scientific community, but the truth is we're just not training our science students well enough in this discipline. They simply don't know the difference between good coding and bad coding, and since in meteorology we train them in Fortran, they'll never develop good programming and software engineering habits and skills.

If you're interested, Max, I'd be happy to detail how huge this problem is in Atmospheric Science, and discuss with you some of the things that I and others are doing to alleviate the problem.
 
There are people in my office who deal with what happens when these research groups take their zany homebrew code and then try to get it running on Kraken. Fun times.
 
literally just talking about this at lunch...interesting article.
 
Good timing on this post! I've spent the last week and a half working with Matlab on a code I'm gonna use for some biomed research that my PI and I intend to publish. Given that it's just Matlab, I probably haven't screwed up anything major, but I'm paranoid of the casual +/- sign switches possibly lurking in one of my algorithms.
 
totally agree. working on climate models this summer and the documentation for the code is severely lacking.
 
Yeah - at SilviaTerra we've spent a good chunk of time transitioning our codebase from my co-founder's research project to production-ready code. It is a huge problem in the scientific community for sure.

Daniel - would be great to catch up! Shoot me an email with some times you're free for a call. Also, a quick intro - Josh has been my college roommate/suitemate for the past 3 years. He's a math major with interests in biology and has recently gotten into programming and environmental science. He's working this summer on atmospheric modeling (but he can tell you more about that). Josh - Daniel is a good friend of mine from high school who went to Cornell and does a lot of computing/environment work.
 
Right now I'm working on a simple 2-box climate model to simulate temperature as a function of radiative forcing for CO2 and O3. It's been used in papers to calculate the global temperature potentials for different economic sectors, and different transport sectors. I'm adapting it to calculate the direct and indirect effects of tropospheric ozone. There was a paper in Nature (Sitch 2008?) which showed that while ozone itself contributes to forcing, ozone damage on plants leads to less CO2 used in photosynthesis, which leads to higher CO2 levels, which actually may have a higher RF impact than the ozone itself.
 
+Max Uhlenhuth, that would be great... by any chance are you in Louisville at all this summer? I'm here right now, but headed up to Cambridge at the end of August for grad school. I was interested in what Josh is doing because my research focus is actually atmospheric modeling - I've worked with Dave Randall at CSU/CMMAP in the past, and actually I'm headed to MIT for grad school to do a PhD with a focus on atmospheric modeling (but mostly microphysical/aerosol stuff as opposed to GFD). +Josh Pan, if you ever need a consult on atmospheric science/modeling, I'd be happy to lend a hand; sounds like you're working on a cool project. I mostly work with the big dynamical cores and coupled models, but I've done stuff with theoretical models like you're working with in the past.
Add a comment...