Trappist Exoplanetary System: Creating 3D Maps of the Atmospheres of the Exoplanets
Advisor: Dr Sanhita Dixit
Internship Location: NASA Goddard Space Flight Center, Maryland
Onsite Mentor: Dr. Vincent Kofman
Exoplanet Trappist-1e, according to past research, has a high likelihood of hosting liquid water and possibly having an atmosphere and climate. Studies have investigated the temperature range that the planet’s surface might have and what types of clouds may be present. In particular, the Trappist Habitability Atmospheres Intercomparison (THAI) project investigated how several different global climate models (GCMs) represent Trappist-1e, consisting of three-dimensional atmosphere files that include the abundance of different molecules and the abundance of liquid and ice water clouds at high-spatial resolution across the planets’ atmosphere. In this project, I leverage these GCMs and create spectral simulations with the GlobES module on NASA Planetary Spectrum Generator (PSG) to investigate the effects of the clouds on the visibility of the molecules O2, O3, and H2O in its atmosphere. Detecting these molecules are strong indicators of biosignatures that would bring greater interest to this planet for the larger scientific community. This research will contribute to a greater understanding of the evolution of planetary systems and their potential for supporting life, which can address the question of how life came to be on Earth.
Project Blog
Week 0: Alien, who?
Aliens are cool. You might’ve seen aliens on TV, with UFOs, stealing people and cows from Earth. But for now, IRL alien sightings are only a conspiracy. But can they exist? Can life as we know it exist elsewhere? Stay tuned to my blog to find out.
I wanted to search directly from the source—terrestrial, or rocky, exoplanets. For life as we know it to exist, the exoplanet must be home to liquid water. By observing terrestrial exoplanets that are similar to Earth, my senior research project aims to detect life’s necessary molecules on exoplanets to suggest the potential for extraterrestrial life.
But first—what does an exoplanet look like? This is easily answerable with a quick Google search… But as it turns out, these images are all artist renditions based on available data, much like images of dinosaurs. I’ll be generating software that reads General Circulation Model (GCM) files, which contain data on exoplanet atmospheres, to create 3D models and spectra of exoplanets. This is important work because current analyses of exoplanet-related research are based on 1D models, which do not contain information about the atmosphere, and are therefore less accurate than they could be.
I’m excited for the opportunity to tell you guys about my process, and I’m incredibly grateful to be working with Dr. Vincent Kofman all the way at NASA GSFC and Dr. Dixit, my physics teacher and mentor since the seventh grade.
Don’t let aliens catch you lackin’.
Tanvi Deshmukh
Week 1: Dimensions of Earth
Welcome back! Last week I told you guys about how I’m going to generate images of exoplanets. But there needs to be a starting point… Earth. Using General Circulation Model (GCM) files of Earth, which (as a little recap) include information about the atmospheric composition, and some Python, I created these 2D images, and I’m going to let them speak for themselves a little bit:
Cool, right? Above, we have ourselves a temperature map of the Earth, and below that we have a pressure map. These are from quite a few kilometers above the surface of the planet, and I am able to vary the distance above the planet I image from. For a better understanding, think of the atmosphere like cake, and different information is revealed at different layers.
Now for the most interesting (and most useful) part: relative humidity maps. These are essentially maps of cloud coverage, and being able to include this kind of data for exoplanets is critical for a more accurate analysis. But for now, here’s Earth’s relative humidity map:
With these base generators, I’ll be able to create 2D maps of exoplanet Trappist-1e, which are similar maps to these. However, I expect the resolution of the image will not be as clear since Trappist-1e is 41 lightyears away from Earth… Nonetheless, I’m excited to see what they look like!
“I'm a alien, I feel like changin' the weather” - Yeat, bigger thën everything.
Tanvi Deshmukh
Week 2: Trappist-1e Looks Like a Tennis Ball
BASIS fam and all others, welcome back to your favorite, out-of-this-world 🚀 Senior Project blog by yours truly.
I wanted to start off this week by saying that learning the art of clickbait is incredibly silly. I came up with "Unbelievable! Trappist-1e REVEALED – You Won't Believe What This Exoplanet Resembles! 🌌🎾 #ExoplanetResearch #TennisBallPlanet," but ultimately decided against using it as my title 😐.
The amount I’ve been coding the past week, you would never believe me if I told you I dislike coding. But it’s a part of the process, and the results are quite visual (as you’ll see), so all the coding is definitely worth the angst.
As a recap from last week, I coded software to generate a 2D relative humidity map (fancy-pants science terminology for cloud coverage map) of the Earth. Earlier this week, I applied this software to Trappist-1e. Here’s what I got:
As expected, the resolution of the image is not as high as it was for Earth. But it’s still pretty good, and we’ll make do, right guys? 🙂
Here, it’s worth mentioning that there are two types of clouds that are accounted for by this simulation: water clouds and water-ice clouds. Also, it was challenging to figure out the new latitude/longitude parameters, since there are a different number of spectral points than the data we have for Earth. Once I figured that out, this step was smooth sailing ⛵.
One of my final objectives in terms of generating figures is to create a composite image combining simulations without clouds/atmosphere and with clouds/atmosphere for Trappist-1e. For Earth, this looks like the (super cool) following image, which is incredibly realistic and entirely based off atmospheric data for Earth:
For Trappist-1e, the relative humidity 2D map is essential to establish cloud-saturated (middle) simulations. However, the first step to create an accurate cloudy simulation (right) is generating a visual for the planet without an atmosphere and without clouds (left). I have done just that:
Since Trappist-1e is tidally locked, the backside of the planet is not illuminated by its star, so it is dark. The blue is ocean and the white is snow, so we can see that Trappist-1e is a water-world with polar ice caps.
We must also take into account that Trappist-1e orbits a red M-dwarf star, which which has a temperature about half that of our Sun. Incorporating that into our simulation, we see:
…It high-key looks like a tennis ball. The upcoming figures will all be simulated to have blue oceans instead of red oceans, since blue is nicer to look at, and blue oceans are easier to comprehend than red oceans. Let me know if you disagree.
Next week, I’ll be generating graphs of spectral radiance versus wavelength for this model of Trappist-1e without clouds and without an atmosphere. I’ll also start working to include the atmosphere in the simulations, so you can see the difference between simulations with and without an atmosphere. See you then!
Underwater aliens maybe 👽?
Tanvi Deshmukh
Week 3: Standing on Rayleigh Scattering
Hey guys, welcome back to my blog! Last week, I showed you guys what the exoplanet Trappist-1e looks like without an atmosphere. For a visible image, there must be a simulated light source (the star Trappist-1). Now, let’s snap our fingers and add the atmosphere back.
But first, what is an atmosphere anyway? It’s an envelope of gases that surrounds a celestial body. Most celestial bodies—planets, moons, stars, even comets—have atmospheres. In the case of most planetary/exoplanetary atmospheres, the gravity, molar mass, pressures, and temperatures define the vertical structure. When said structure is predominantly gravity-dependent, the atmosphere is considered to be in hydrostatic equilibrium—hence the term ‘hydrostatic equilibrated atmosphere.’ This is the kind of atmosphere Trappist-1e has.
So, what happens when photons strike atmospheric particulate matter? Let’s visualize the scenario: a ray of light (from a star, say Trappist-1) travels through empty space and suddenly collides with a particle (which is, say, in the atmosphere of Trappist-1e). The outcome of this collision varies depending on the size of the particle and the wavelength of the light. If we consider small particles like atoms and molecules, the Rayleigh Scattering model best predicts the light’s subsequent behavior.
As of right now, I’m struggling with some errors in the simulation that includes the atmosphere, as you can see here:
This indicates that the software is not corresponding the correct information with each pixel. Keep your fingers crossed that next week I’ll be able to show you the Rayleigh scattering in more detail for this exoplanet.
Lastly, for a more structured progression, I’d like to generate graphs of spectral radiance versus wavelength for the cases of Trappist-1e with and without an atmosphere. After that, I’ll attempt to add clouds to the fray.
I wonder if aliens would enjoy eating pie as much as I do. Happy Pi Day, and see you next week!
Tanvi Deshmukh
Week 4: Laugh Now Cry Later (feat. Lil Durk)
This week I laughed. Next week I will cry. Simulations take so long.
The past couple weeks, I worked on generating simulations of exoplanet Trappist-1e's appearance with and without clouds. Since I succeeded simulating Trappist-1e's atmosphere without clouds, I took a break from simulations to write code to graph Spectral Radiance vs. Wavelength for Trappist-1e's simulation generated data.
Spectral radiance is the amount of light (electromagnetic radiation) emitted or reflected by a surface per unit area, per unit wavelength, per unit solid angle. The units of spectral radiance is [W/m^2/nm/sr]. For the simulation, spectral radiance depends on the size of the FOV (field of view) and the projected size of the planet in it.
There were many challenges on this journey. The wavelength range needed to generate meaningful simulations was: 300-20000 nm. With such a large range the simulations took a lot more time than before. I can only imagine how long each simulation will take when I add in clouds to modelling parameters.
Second, the units. Bytes of data must be properly matched to a color to fill the correct pixel. For this, the simulated units must be correct as well. Initially, I made the error of using 0.3-20 um instead of 300-20000 nm, which is the same range, but the different unit conversion threw off the program. This left me with an oddly colored and mismatched image shown below. As you can see, I used a lower resolution to simulate faster.
Once that was fixed, the main course dish on my erroneous platter was dealing with the number of wavelength points. This is an important parameter to ensure the simulations map the correct information to the correct pixels. This particular problem was exasperating because it was a relatively quick fix that took me so long to understand. Because of this problem, my graphs and images were inaccurate. Unfortunately, I thought it was a problem with the data itself, so I kept regenerating byte files that held correct information, which was a lengthy process. Thankfully, I solved that problem. From next week onwards, I'll be making this software more usable for applications besides Trappist-1e. The first step for that is to make a function that tells me the number of wavelength points (instead of a user having to deep dive into the data to find it themselves). Below is my final spectra:
Now take a nice, long look at the graph image, and spot the mistake. First to comment the answer down below gets $10.
Tanvi Deshmukh
Week 5: My Laptop Gets Sleep Deprivation and I Don't (anymore)
But it's all for a good cause! I promise! That way I can code by day and simulate by night. Poor little laptop.
Welcome back to my blog! Hope you’re enjoying the journey as much as I am (and my laptop isn't). This week, my laptop crashed a grand total of 3 times mid-simulation. Obvious by inspection ☝🤓, it's not very happy with me. By the way, each simulation takes 6 hours, so interruptions are no fun. But I, Batman, have one power, and that is never giving up.
After hours of tedious labor, I succeeded in creating the images I started on last week. As you can see, there is quite a bit of Rayleigh scattering!
With that out of the way, I got to work on creating spectra for the cloudy condition. However, I ran into quite a few errors with the simulation (that took 6 hours to generate, as mentioned above).
As you can tell, the wavelength range is not correct. With some alteration, and more laptop sleep deprivation, I was able to fix both the spectra and the image, resulting in the following correct model. Woop woop!
Goodbye my fellow alien enthusiasts! Next week, I'll discuss more about the significance about these findings, and hopefully not have to deprive my laptop of so much sleep. Until then!
Week 6: Too Many Nights (feat. Don Toliver & with Future)
Hope everyone had a wonderful and restful Spring Break week!
Welcome back to my blog where I track my progress on simulating exoplanets. As a recap, last-last week I demonstrated Rayleigh scattering and generated spectra for the fully saturated cloudy condition of Trappist-1e. These past couple weeks, I worked to adjust the water content of the planet to a more reasonable quantity and FINALLY created accurate cloud-coverage simulations! Which is a lot more exciting than it sounds. Let’s get into it.
Point blank, the parameter I used, relative humidity, is the wrong parameter for this general circulation model (GCM) file. There are three types of humidity: absolute humidity, relative humidity, and special humidity. Rather than relative humidity, special humidity would have made calculations accurate within the Planetary Spectrum Generator (PSG) system.
Relative humidity is the ratio of the water vapor content in the air to the water vapor content the air could potentially contain at a given temperature, while special humidity is the ratio of the mass of water vapor in a container/section of the atmosphere to the mass of dry air in that same section. These parameters are very different, resulting in faulty images for conditions that include the atmosphere. Thankfully it was only a problem with the GCM translation, and all the image simulation code still worked perfectly. So after a day of exasperation generating the GCM, I calmly returned to depriving my laptop of sleep and running new simulations.
Worry not, there is an image-reveal at the end of my spiel this week. SO KEEP READING. OR ELSE. If you need a breather, pretend this is an intermission at the theater, refill your popcorn, and return.
So now that I have accurate full-saturation cloudy conditions, I’m able to create accurate cloud maps by combining the “yes atmosphere, no clouds” simulation with the “yes atmosphere, yes clouds” simulation. Now say that three times fast 😉.
Without further ado, I present the final cloudy combination simulation. From left to right: Yes atmosphere, no clouds; Yes atmosphere, yes clouds; Special Humidity map; and last but not least, the WOMBO-COMBO!!!!
Since this blog is already technical as it is, I’ll explain the math behind the combination simulation next week, so stay tuned.
Yearn yap yawn,
Tanvi 😐
Week 7: Ad Astra
It’s a bird! It’s a plane! Wrong… It’s a flying Tanvi??? This week I learned how to do a backflip. It was pretty cool. I am also on the east coast, and I’ve been on a tight schedule. I don’t know how I used to go to school from 8:25 AM to 4 PM every day…
Folks, can you believe we’re already on Week 7 👽 ? Welcome back to my blog! Aside from doing backflips, I worked on creating a poster and presentation, which are still in the works. And as promised from last week, I will tell you guys about the calculations to overlay clouds and get an accurate depiction of my second-favorite planet, Trappist-1e.
I have been informed that my blogs have been too technical, so I hope this is easier to understand. The program overlays the cloud-saturated image onto the cloud-free image. However, a simple overlay won't suffice. We need to adjust the opacity (color, in this case) of the cloud layer based on the corresponding humidity levels from the map. Regions/pixels/bins with higher humidity levels receive a higher opacity setting for the cloud layer, indicating denser cloud cover. Conversely, regions with lower humidity have lower opacity, suggesting clearer skies.
A couple of the inaccuracies of the program are related to Trappist-1e’s ocean. It doesn’t resolve for volumetric reflection, which is essentially lower layers of the ocean reflecting and scattering light from the star. It also doesn’t incorporate wave effects and glints that result from wind patterns and the motion of the ocean. These factor into an ocean that is less bright than it should be. But that’s okay! If it ain’t broke, don’t fix it.
Goodbye, it was nice knowing you. See you back in the Bay next week!
Tanvi
Week 8: (;´༎ຶٹ༎ຶ`) Pale Brown Dot ?
Welcome back folks. Aside from going bankrupt this week, I mostly completed my poster for the Senior Project Symposium, leaving space for upcoming developments. May will be a busy month!
This week was a simple, albeit exasperating week. I still haven’t committed to a college yet and that is weighing heavily on my mind 🤠. This week, I remodeled the planet for the cloudy condition but simulated it with its own star, no longer our sun (womp womp 🌞). Typically, a cloudy simulation with 18 phases would take about 6 hours; however, this time, the simulation with only 6 phases took about 9 hours… (;´༎ຶٹ༎ຶ`) which was not very nice. Results are worth it though, displayed below.
Now, this project can take a couple different routes.
Create more General Circulation Models (GCMs) from different data sources that come from different instruments. The Trappist Habitability Atmospheric Intercomparison (THAI) project, where I was able to find the data, compares four models: ExoCAM, LMD-Generic, ROCKE-3D and the UM. Currently, I use the ROCKE-3D model, which is specially adapted for non-modern-Earth simulations. Next, I’d opt to use ExoCam, which supports a greater range of parameters than ROCKE-3D.
Model a different case! Trappist-1e is a transiting planet**, which means that we observe from Earth the planet passes in front of its host star, blocking some light. I could potentially find a different planet that is non-transiting, which seems to be all the rage these days.
Lastly, I’ve been thinking about different ways to display and model the planet. I could try to 3D model and print it, but I’m worried that I have too little experience with 3D printing to be able to model it in time. I could also try papier-mache, but I am unsure how accurate the model would turn out. Let me know if you have any suggestions! Greatly appreciated:)
**Just wanted to point out “Longitude: 2” and “Longitude: 182” on the images. These phases bypass the transit cases 0 degrees (behind the star) and 180 degrees (in front of the star). Also, since the planet is tidally locked, the permanent night side is illuminated for your viewing purposes.
If you’re struggling to fall asleep, count aliens 🧛♂️.
Tanvi
Week 9: Welcome to [the Beginning of the End]❌ [My Head]✔
I invite you to live in my head for the duration of this blog. With that, welcome to this week’s edition of Tanvi’s 15 hours of tedious labor!
I’m pleased to say that the week started with writing a simple text message (imagine the notification audio in your head please… ding!). It was from me (you?!) to a trustworthy junior. A capable junior. An incredibly busy junior who is interested in the biology of living systems, not the atmospheric composition of exoplanets. But that’s okay! I have organized a writeup to discuss with her to aid her in presenting my poster 🗣🔥. And no, there are no sneak-peeks for you lodged in my brain as they have been redacted. I encourage you to visit my stand in-person at the symposium to view my poster 🤓. My representative is a lovely individual. I assure you she doesn’t bite.
Then, as per my onsite advisor’s suggestion, I began the image concoction process for the exoplanet Proxima Centauri b. Why ProxCenb? Many reasons. For one, it is a super-Earth, indicating possible habitability. Secondly, it orbits the star Proxima Centauri, which happens to be the closest star to the sun at about four lightyears away. Hopefully in a few years, as scientists and engineers develop methods to directly image exoplanets with telescopes like HabEx and Gaia, we will be able to compare my upcoming ProxCenb simulation with the directly imaged model! Third and most importantly, ProxCenb is a great choice because it is a non-transiting exoplanet. This means that from the observation point of Earth, the planet does not pass in front of its star.
But how do we observe and image these non-transiting planets? This is one of the big questions the Sellers Exoplanet Environments Collaboration, which I am proud to be a part of, seeks to address in upcoming research. Big thank you to my onsite advisors for the plug!
So to get started with ProxCenb, I found a really interesting paper that I have linked below that references another paper (phew!) that links the data for an LMD GCM, a different style of GCM than ROCKE-3D GCMs. This makes it non-ideal to work with, as I’d have to grasp a new format of information within this GCM to process for the simulation steps. I’m looking forward to it:)
Knock knock! Who’s ther-INTERRUPTING PSA alarms reminding me to begin writing my research paper! Okay. After refining my wording, the abstract and introduction are complete. Next week, I’ll have to elaborate on the methodology and results sections. Maybe I will get published. Who knows.
We senior project folk are approaching the date of the symposium at an alarming rate (it’s May already??). In the words of BIF’s lovely senior project coordinator, Ms. Nagpal, there are many questions to address in our grand presentations. A deceptively simple 10-minute discussion of our challenge-riddled progression that led to the final result of each of our projects. So… How do I condense a three-minute-long about-me into a 30-second one? How do I balance formal and technical vocabulary with apt explanations to help my audience understand the significance of my work? And the most challenging question of all… What have I learned about myself over the course of this project?
I’m sure these answers will come to me at some point. Hopefully before Monday, my first presentation practice session.
Now if you would kindly exit my head, or pay rent until next week. The final blog. I am going broke. Anythin helpz 😐.
Tanvi 👽
Week 10: I can't make this sentimental enough.
Hear ye, hear ye. This blog is now in session. Welcome one, welcome all.
Let’s get crackalackin’. This week, I worked on improving my presentation by giving it more structure, generating essential images, and practicing.
I also worked on completing my report/paper. I completed the Intro and Methodology sections, leaving the Results section for this evening. As of this month, I will not be able to publish this report in public journals/articles, since quite a few references are from my advisor’s paper which is still under review.
I am grateful to my internal advisor, Dr. Sanhita, for always looking out for me and making my blogs better!
‘Twas a wonderful senior project run innit. I look forward to seeing you at the symposium.
Many a gracious & grateful remark,
Tanvi