Dinos to dev: climate models and ChatGPT
Rhea runs volcanoes and climate models!
Who: Rhea
What: Software developer and climate scientist
Where: Seattle, WA
Industry: Technology
Rhea is a friend of mine (we’ve run around mountains and all the way across Iowa together!), a “real” climate scientist (like, someone with a PhD in atmospheric science!) and an ultramarathon runner. Given her background, it’s interesting to me that she doesn’t work in climate at this time. I wanted to ask her some questions to find out about how her path changed. Here’s what she had to say.
What do you do?
I build business application software leveraging artificial intelligence.
How does it relate to climate?
My current role does not directly relate to climate, but in my previous jobs it was the driving motivation. I've worked on developing software for renewable energy analysis products and rewriting a climate model to run fast on multiple kinds of hardware (e.g., GPUs, CPUs) while being readable and efficient to modify. Both of these efforts relied on atmospheric models, informed by observations, to provide insight and information about climate.
Information from this type of software and models can help a wind farm developer choose the optimal locations for installing wind turbines to maximize energy output, or help regional planners understand and prepare for likely impacts of climate change in their community.
What is your educational background? How did you get interested in studying climate?
I have a PhD in atmospheric science and a bachelor's degree in computer science. My interest in climate started with a seminar I took by Walter Alvarez, whose passion for geology made it my most inspiring and interesting class at Berkeley. (A version of Alvarez’s class, Big History, is available online! - Ed.) The exciting tale of his journey to uncover evidence that an asteroid impact caused the extinction of dinosaurs turned my attention towards earth science.
From there, my enjoyment of math and physics steered me towards atmospheric science with its fun fluid dynamics. At that point, because climate is the global average of processes I was studying, and also such a concerning contemporary topic, I found myself on a path to studying climate.
What led you away from climate work?
Having spent my career so far in climate related work, my recent job transition away from climate was inspired both by a genuine interest in expanding my skill set (that is, learning the engineering problems and solutions of another domain) as well as the reality that funding for my project at the time was ending.
What are some obstacles you encountered along your climate career path?
It can be challenging to get that first job in the field, and also to find ways to maintain dual goals of making a meaningful impact and making a living. Although there are many paths to working on climate related topics, sometimes geographic and compensation flexibility are necessary. The political aspects of the field can cause volatility in the perceived value and availability of career options.
Do you think there is a way for ChatGPT (or other AI tools) to explain climate change to people?
At the risk of this statement already being dated (given how fast the field is moving), ChatGPT could indeed be a tool used in science communication, including climate change, with some precautions in place. Given many of the generative AI models’ known tendencies towards confabulation/hallucination, their answers should be reviewed by a climate expert prior to widespread sharing. Although scientific papers come with abstracts that provide short summaries of key findings, they are often densely worded and full of jargon people outside the field may not be familiar with. Using AI to summarize a scientific paper's results could help scientists share their findings in a way that is more consumable by a wider audience. But, if such a summary gets created and published without confirmation that it’s indeed accurate, such AI tools have the potential to add to the already existing body of misinformation out there.
What are some of your concerns about climate models? Where do you think they fall short?
The large amount of compute time and energy that is needed to run a global climate simulation at resolutions that resolve cloud and other local scale-processes makes them difficult and expensive to run. Supercomputers can and have run at 1-km resolution, but it can take a month of modeling time to simulate 1 year (for context, being able to simulate one year per day of runtime is a typical goal) using ~5000 machines and produce massive amounts of data.
Cheaper, coarser resolution runs do provide a wealth of good information, but fall short when trying to answer climate questions about a particular location. For example, such a model would not capture the airflow patterns around a local mountain range, whose shift over time could impact local precipitation amounts. As a result, regional planning for future climate change impacts and adaptation benefit from higher-resolution results.
There are also several efforts to use AI to replace some of the expensive small-scale physics computations in models, which may make this information easier to come by as well. These definitely need and have guardrails to keep the model from producing unphysical results.
What can people with an existing software engineering background do to transition to work in climate? How much climate expertise do they need, versus coding expertise that can be applied toward climate goals?
The great thing about big problems is that you need a team of people with varying perspectives and experience to make it work. So it’s not usually up to just the software engineer to code up climate-related calculations in a vacuum. There are scientists and analysts and other experts who have a keen interest in your code doing the right thing.
On the job, if it’s related to climate, you would quickly get familiar with the valid data ranges for different types of variables (such as temperature and pressure), which are important for geophysical software to get right. It helps to have an interest in the field and bring an enthusiasm for the topic. Skills around heavy compute processes or large data management can come into play, but a project could also require web user experience expertise, so it highly depends on what is needed! It’s also helpful to meet people in the field who can help you get interviews at companies working in climate. It’s not usually necessary to have an expertise in physics unless you are trying to step into a scientific role.
“The great thing about big problems is that you need a team of people with varying perspectives and experience to make it work.”
- Rhea, computer and climate scientist
What worries you most about climate change?
There are so many worrisome things that will have enormous impacts. The disturbing amount of incorrect and misconstrued information about what the science knows continues to be quite concerning. It blocks and slows mitigation and adaptation solution efforts that are challenging enough.
How do you think about climate equity and climate justice in your work?
I think generally there is not enough focus on justice. There is often keen awareness of the problem, but no clear mechanism for impacting it directly. In general, it’s seemed like a different category of work than I've been involved with. But, improving information about climate and climate impacts should in theory also help prepare to offset climate injustice, if the knowledge is used and motivating.
How do you think about helping others in their climate career paths?
It is a great joy when I can connect people to opportunities, both for the person and also to help elevate the field by introducing more great minds.
What gives you hope when it comes to climate?
It gives me hope that there are so many people who care about climate change and are working on solutions to the many problems we face.
What climate resources do you suggest (books, podcasts, etc)?
RealClimate is a great resource for understanding climate from the perspective of climate scientists. If you want even more detail, the IPCC reports summarize the scientific knowledge we have.
Lessons from Rhea
Fluid dynamics is fun (so are ultramarathons!).
Models are important and require collaboration and time to build.
Computer scientists can play an important role in addressing climate change.
Solving climate problems requires lots of people with unique perspectives and expertise.
You can’t work on climate all the time.
Resources
RealClimate: Climate science from climate scientists
IPCC Reports on climate change