How to Tackle the Astronomy Problem Using a High-Tech Approach

What is your background? Can you please tell us a little bit about your career path.

I am a junior at UC Berkeley, majoring in physics and astrophysics. I’m doing research with NASA’s Outer Planet Atmospheres Legacy to do an annual survey of Jupiter, Saturn, Uranus, and Neptune. I’m also very involved with the Society of Physics Students as their webmaster. This is a club that shows interesting demos to kids twice a month, holds student-faculty lunches and has weekly meetings. I spent almost the entirety of 2018 coming up with my own first author publication to investigate and theorize what Neptune’s dark spot lifetime and occurrence rate is. I used the Hubble Telescope’s Neptune blue wavelength data from 1994 to 2018 to help me create a Monte-Carlo simulation using IDL. The simulation randomly chooses a time in 1994 to 2018 to have a dark spot born and then chooses a random lifetime to see how many of the actual observations made in 1994 to 2018 would have detected it.

Now all of 2018’s work has paid off and I made a theory that Neptune’s dark spots occur every four years and have a lifetime of two years. My publication about it has been accepted into the Astronomical Journal and it has a DOI of 10.3847/1538–3881/ab0747.

Searching “Andrew Hsu Neptune” on Google will yield 50+ results of articles talking about this project as well as what the Outer Planet Atmospheres Legacy does overall.

This January I applied to about 50 software engineering and data science internships. However, I learned the harsh lesson that although I did a grand project that involves programming in IDL throughout 2018 that led to a publication, all the employers except on refused to give me even an interview opportunity. I learned that apparently the industry doesn’t really trust that a physics/astrophysics major who knows how to program to do their research can help their company. It is tough to communicate how my project and background translates into coding ability.

As a result of these experiences, I decided that the only viable option for me after UC Berkeley is to go to an astronomy graduate school to pursue a PhD in astronomy. My dream job is to be a research scientist at NASA and that requires a PhD. I love doing research on Neptune and the other outer planets and I understand that trying to get a job after getting only a bachelor’s in physics is going to put me at a big disadvantage as employers apparently tossed my resume in the trash can because I’m not a CS major.

What is your current specialization?

Right now, I am primarily focused on doing well in my astronomy and physics classes. One of the classes is called Radio Astronomy Lab. In Radio Astronomy Lab there are four groups of four people where we take data collaboratively using the Python Jupyter notebook and do the analysis individually. We use Jupyter to take data from the radio telescope in our Campbell Hall (the astronomy building) and did a Fast Fourier Transform and some Doppler shift calculations to plot our analysis (all on Python Jupyter notebook/iPython).

As part of the Outer Planet Atmospheres Legacy, I used IDL for data processing and attached latitude, longitude, and emission and incident angles to the new images of the outer planets that Hubble takes. I also wrote an IDL program to limb-darken the images so that the center is not too bright and the edges are not too dark. I have done this every month since January 2017. Recently, the November 2018 images of Neptune showed a new great dark spot. We are only able to see the dark spot because I data processed those images using IDL.

As webmaster for the Society of Physics Students, I write codes of HTML as well as some CSS here and there on GitHub to edit the website. I incorporate and keep the Google calendar attached on our website up to date. I add pictures and descriptions of the outreach that we do.

These experiences matter for the industry because these tasks all involve programming in either Python, IDL, or HTML. I believe that I’m qualified to help a company maintain its website using HTML, such as keeping a calendar up to date, changing the buttons, or by adding new pictures and texts. I can perform monte-carlo simulations in either Python or IDL on any data that the company has about its customers to make a prediction about their behavior, such as their visiting rate and duration, just like I had done with Neptune’s dark spot occurrence rate and lifetime. Using Python and plotting and analyzing tons of radio telescope data helps the industry as a whole, since I have the Python and IDL programming skills to do the same thing with customer data from any company.

My work so far has changed the world because I contributed to discovering a new bright cloud on Uranus after using an IDL program to data process and adjust the brightness of the image of Uranus and Neptune, I found two dark spots on Neptune, I came up with a theory about Neptune’s dark spot occurrence rate and lifetime, and I inspired a lot of kids into strongly considering physics as a career path. The world knows that Earth’s land is very important, as it stops the hurricanes and typhoons from lasting for two years as they do in Neptune.

What are the most important problems your customers are met with?

The customers I can think of are all undergraduate physics majors. I think of Society of Physics Students as being run by undergraduate physics majors and this organization primarily helps freshmen physics major “customers” learn how to get research opportunities and know what projects the professors are working on. The most important problem our “customers” face is how to get research opportunities and knowing what lies ahead of UC Berkeley in terms of both industry and academia. We address them by having about five panels a year where experienced people that were a part of Berkeley SPS answer questions and give their life backgrounds. The website is helpful in getting the customers to know when these panels are and summarizing what happened during these panels for those who can’t make it.

What is the most remarkable experience you’ve had during your professional career in this field?

The best experience I had during my two years working in the Outer Planet Atmospheres Legacy research group was throughout 2018 when I did my own mostly independent projects to investigate Neptune’s dark spot occurrence rate and lifetime. Nobody in the astronomy world has made any bold theory/prediction about this yet. Any previous attempts did not take into account observations from 1994 to 2018. They did not simulate dark spots that last more than one month. The problem is that we don’t quite understand how often dark spots form and the consequence of not having annual Neptune observations in 2012 to 2014. My project concludes that it is pretty likely that there was a dark spot during this time period of no observations based on my monte-carlo simulation. Another assumption is that real dark spots live for at least one year or longer, perhaps even six years. I have this assumption because there is a recent dark spot born in 2015 that didn’t disappear until 2017 so we did our simulation with the lifetime spanning from one to six years. Me and my project mentor knew that the dark spot disappeared in 2017 because we used IDL to data process the new 2017 Hubble images of Neptune and used IDL to make graphs of the brightness in the old latitude that the dark spot resided in.

The hardest part of the problem is to communicate what I did in the paper. It was also hard to communicate what programming skills are used. The main approach to the problem is using IDL to do a monte-carlo simulation where each simulated dark spot has a lifetime of 1 to 6 years, a random latitude based on surface area, a random drift rate, a random birthday within 1994–2018, and a random longitude. There are a total of 8,000 simulated dark spots for each integer lifetime. So a total of 8,000*6 = 48,000 simulated dark spots. For each simulated dark spot, my program calculates how many actual Hubble observations during 1994–2018 would have detected the simulated spot.

Conclusion: Dark spots on Neptune have a lifetime of one to two years, more likely two years. New dark spots occur every four to six years on any part of Neptune.

This finding provides constraints that may lead to an improved understanding of Neptune’s wind field, stratification, and humidity.

What are the sources of knowledge you use to improve your skills?

I honestly find most of the physics and astrophysics courses at Berkeley to be not very useful for either the industry or academia. I don’t need to know all the subfields of physics or all the areas in astronomy to do research on Neptune’s dark spots. The most useful courses are the few courses that taught programming in Python. Namely, Physics 77, Physics 151, and Radio Astronomy Lab. These courses teach me the skills applicable to the industry and any kind of data science in general. Radio Astronomy Lab is useful since it teaches you how to collaborate to do research as well as give you experience obtaining data using Fourier Transform, adjusting for Doppler velocity, and plot graphs using iPython and the Jupyter notebook.

In Physics 77 I used Python to write a smart computer that plays tic-tac-toe based on a monte-carlo simulation of what’s the move that is most likely to win or least likely to lose. In Physics 151 I used Python to do data analysis, such as a monte-carlo Markov chain, training and testing data set, Bayesian inference, Kullback-Leibler divergence, Fisher information matrix, Fourier methods, and extrapolation.

The most useful books are inspirational books that have stories of someone in a disadvantageous position and eventually becoming prosperous and successful.

My plan is to complete a PHD program in astronomy.

Questions asked by Alex Savinkin

Former number cruncher in investment funds & strategy consulting. One of Geekforge Founding Fathers. Blockchain and technical singularity true believer.