Why should geologists learn how to code?

2022/03/14

Before the second day of my 4-week “Introduction to Python for Geoscientists” workshop a few weeks ago, I was asked by a student why they needed to learn how to code. After all, they are training to become a geologist, not a geophysicist, and couldn’t see how this would be useful in their future career.

It was a fair question since, by that point, all we had seen were variables and basic data types and operations. Day 2 of the workshop covers loading that with numpy and making plots (following the excellent Software Carpentry “Programming with Python” lesson), which is less abstract and luckily helped show them how this skill is useful, though for now they could still have done the same thing with Excel just as easily. But not being a professional geologist myself, it’s hard to come up with specific examples.

Luckily, I have access to a wide network of professional geologists who code through the Software Underground community! So I asked them “How has coding helped you do your job?”

Here are some of their replies (with slight edits by me marked in []).

Although I am not a geologist specifically, I did have a lot of geology training in my BSc before choosing geophysics. My suggestions would be:

3D model creation using Python interfaces [to graphical software] or GemPy.

Any graph theory application (material fluxes). [For examples, see] Phillips et al. (2015).

Automated fracture or discontinuity mapping. [For examples, see] Vöge et al. (2013) and Prabhakaran et al. (2021).

[Creating] virtual outcrops.

Sped up tectonic forces and erosion.

You could build geological models in Minecraft ([using their] Python interface).

Maybe most importantly: Automate the boring things.

[Here are] a couple of points for motivation that I give my students:

Data analysis: Automation of processing (including graph generation), reproducibility, transparency.

Creativity: Find solutions [to the problems you come across] in a creative way.

Freedom: Don’t be stuck in a program that defines for you what you are able to do.

It is a tool set that makes you more independent in the future job choice.

I also like the general viewpoint of computational thinking as a problem-solving art. Being able to deconstruct complex problems into single tasks can also be transferred to non-computational problems in geology!

But maybe more directly to geology:

More and more data now gathered and automated processing developed, also meaning that a lot of jobs will become obsolete. But if you are able to combine geological knowledge with programming, then a lot of opportunities open up.

As success stories: A lot of my students found jobs because they were able to combine geological knowledge with programming. I received a super nice letter last year from [a former student] where he made it very clear how much that helped and opened his mind.

I also tell the story from our industrial advisory member in the last study program accreditation where the advisor mentioned explicitly that the programming skills we teach are essential for many jobs in industry in the future (was from raw materials, but I am sure you could get similar quotes from geothermal, geotech, etc.).

Of course, the key difficulty I see for our education: we need to teach students both geoscientific skills and knowledge and a set of tools - and there is only limited time. But programming is certainly a part of the standard tool set that is expected (or seen as an advantage) for many industry jobs (or will be in the near future).

Learning to code helped me to get a better grasp of mathematics as well as you can focus on the math part rather then “calculating”. You can build your own tools to solve problems and avoid cumbersome tasks and focus on the fun parts.

In industry (at least in my experience) everything is statistical. Rather then one best estimate we try to cover the range of possible outcomes to base decisions on. Coding allows you to do that relative easy.

But the hardest point to bring across is that [before you start] you don’t know all the things and ideas you will have once you can code. You have a new set of tools at your hand and start to see all kind of problems that could be solved.

What Florian Wellmann mentioned is right. For me learning to code opened a new chapter in my life and how much it changed is hard to put into words here. The easiest thing to describe is that it got me the job I have right now and it was one of the most empowering experiences in my life.

Doing any real (geo)statistics really asks for a good programming knowledge. From data handling to analysis and understanding. There are many things non-coders can not do (they are stuck with commercial software).

When they have to type in their weekly report and all their statistics from their drilling program by hand, then redo when their head of commercial wants it done differently, build a 300 layer 3d project and you have to do 25 of them, organise all your assays from the last 10 years…

When I think of core geology skills I think of making maps. Possibly the last time they’ll make a contour map by hand is in school. After that, they will have to use software to be productive. It is crucial to understand how these algorithms work, when to use them, and their pitfalls. One could learn how krigging works (math), and then code a simple implementation to bring it home (learning math by coding).

Also, coding is great for automation. Even Petrel, [a standard commercial software in the oil and gas industry], has a very important area for developing workflows to automate data preparation, regrid, and all things needed for statical modeling. It is not a real programming language, but it has units of code blocks (for loops, if then else, while) that if you had never been exposed to before, might make hard to grasp their use and connect the dots to solve your problems.

I’m not a graduated geologist yet, but knowing how to code in Python brought me several opportunities in terms of joining research groups and finding interesting internships. I use Python very often for automation on GIS and for data cleaning to input in a few software [tools], like MODFLOW. I’m also developing some geowebapps in order to allow free and open source data processing. So far, I have developed only one, and a few more are on the way! You can check it out here.

I think I’ve been in their shoes. As a second year geology undergrad, I decided to take Computer Science 101 because someone told me it would be useful. I more or less decided it was a waste of my time, mostly because I couldn’t connect the dots to how it would apply to my studies or career. Currently, I think the points that Florian Wellmann brought up around data analysis (automation of processing, reproducibility, transparency) are some of the most important reasons to code. But to someone whose definition of “big data” might be data that takes more than a few minutes to transcribe from their field notebook to Excel, this might be a bit abstract.

What did start to make it click for me some years later were finding examples of geologic workflows done in Python that I could build on top of. Great examples are:

[Agile Scientific’s] X Lines of Python series

Jesse Pisel’s 5 minutes of Python

Michael Pyrcz’s Python Numerical Demos (check out the readme of that repo for 9 reasons scientists should learn to code)

Brendon Hall’s famous Facies classification using machine learning tutorial

But to answer your actual question: the two biggest ways coding has helped me do my job as a geologist in industry (apart from data analysis) are task automation and extension of proprietary software workflows.

A lot of tasks assigned to entry level, and non-entry level for that matter, geologists are repetitive. For example, you might make some geologic interpretations and file them in a standard report or form that you will then pass on to a colleague to make some engineering designs. You may have to do this 10s or more times in a week. A lot of this type of work can be automated with code.

Second, much of the work done by geologists, in my experience (O&G) , is done inside of some large proprietary software package like Petrel, DSG, ArcMap etc. Why should we learn to code if we have these made for us? Rafael Pinto made some good points about this already. I will add that these software packages have great workflows, and you should use them if you are able to. But many times they can’t do exactly what you need. In these cases it’s a superpower to be able to read the data and your interpretations from these packages into Python, and do a few steps of analysis to get you exactly what you need.

It’s already been said by others (and probably much better) but for me it’s automation and manipulating very large datasets quickly and repeatably. I’d also say that for me, coding is a bit like cooking. There are two approaches:

The recipe approach: you follow instructions to do something new, or that you would normally achieve in expensive software. The cooking analogy here being going out for dinner vs cooking a fancy meal at home.

The skills based approach: coding/programming teaches quite a lot of problem solving and lateral thinking. My approach to data curation and storage is a lot better from coding. The cooking analogy here being making a nice meal with a bunch of random leftovers or ingredients with some improvisation.

I wholeheartedly agree with the others (geology mayor here)! To add in on the automation aspect: The Data Deluge is real in geoscience - not only in geophysics/hydrogeology/map making - but also in some of the more “qualitative” or “old school” geology disciplines. Instrumentation improves, but so does the amount of data! Both width and length increases e.g:

Micro XRF makes it possible to measure almost the whole periodic table every mm along a core.

Mass spec and laser ablation now make it possible to get composition as a line across a mineral in thin section.

In situ loggers makes it possible to measure soil water content, ph, temp, gas fluxes, on a minute scale.

All this is super awesome, and (most) equipment output to a CSV. But the times where a [\geoscientist] could manage all their data in Excel are long gone!

Only through coding can I find and eventually cope with broken data. It might not be easy to appreciate as a student, but “real data” is all too regularly broken in weird and wonderful ways and it never stops. So being able to identify, triage and then treat brokenness can only be done at scale with some degree of computational literacy. Data is the foundation of any technical work, but fixing brokenness is undervalued, unsexy work. So it’s an opportunity, and I am not convinced folks outside the subsurface domain are suited to do the fixing. At least not alone.

I think the best answer to this is that by learning to code you can do more work, easier and better. Not very concrete but let me give you some examples.

By creating a code based pipeline for a certain type of work I do, I can generate $10,000 in consulting revenue in 5-10 mins. This is money that goes right in my pocket. The same work used to take me days doing it manually. I can literally pay myself a full years salary in 60 mins of work a year. Talk about freedom.

A friend of mine had a geology job where he had to pull down hourly data, summarize and email it out to everyone. He automated it to the point where everything was time triggered so he spent literally all his time hanging out and learning from different people in the company rather than doing his monotonous job. Amazing.

Coding means you can do more. It really gives you freedom to drive your career rather than having it dictated to you. I don’t think every geoscientist has to or even needs to code, but it makes things so much better.

Please check out www.discovervolve.com. I’m sure you’ll find a lot of ideas to inspire your students:

Here are [two examples] that discuss about use of coding in handling well logs:

Understanding the WITSML format in the Volve Oilfield dataset

Navigating Well log formats in the Volve Oilfield data with the Welly library

Here’s a machine learning concept: Development of a Machine Learning model based on feature selection to predict Volve production rate

We are also coming up with a correlation of logs to further show the use of software.

Geologists are really good at rotating a 3D volume in their head. If you can make the shift from spatial coordinate systems to other abstract principal components, it’s a superpower for navigating complex data.

Specific workflows that directly apply:

Data mining (who likes transcribing tables?)

Petrophysics

Depositional systems (examples in the work of Zoltan Sylvester at UT and Zane Jobe at CO Mines [also Pete Burgess at Liverpool])

Image analysis to scale up core, outcrop, and thin-section work

Most geologic careers have a significant amount of reporting that few enjoy. Automated the boring stuff so you can spend more time on the fun stuff!

A short review of cognitive/behavioral psychology (Kahneman) will leave a geologist unnerved about the reliability of their interpretations. Incorporating the modern algorithmic toolkit creates an opportunity to amplify interpretive workflows either to achieve consistent outputs, consistently calibrated interpretations, or probabilistic interpretations from integrating varied data filters or interpretive approaches.

One article I am fond of from my blog is this one: Be a geoscience and data science detective. I like to take apart papers and go a bit beyond. Without having learned a bit of coding I could never do that.

Interesting question. Just to be clear, I am a geologist by training and background, not a geophysicst, but writing code is still a key part of my research and has been for 30 years. The animation [below] is an example I am working on currently to create virtual mapping areas.

I had a coding class myself as a geology undergraduate in the late 1980s, so this issue is not a new one. Geologists have been writing code for decades and the number doing it is increasing now rapidly. I have also worked in industry in the past from 2002-2010, and being able to write code in my industry job in the energy sector was absolutely central to success of the teams I worked in. So coding has been a useful skill for geologists for a long time, but now more than ever. My advice to you all is that you should embrace the opportunity to learn this skill if it is new to you and get ready to use it frequently throughout your careers in Earth Science.

Hopefully this insight from professional geoscientists will be good inspiration and help find the motivation to continue learning about computing. It’s by no means an easy subject and requires approaching problems with a completely different mindset. That is difficult to achieve and requires practice to become second nature. But it can be done!

By the way, if you are interested in geoscience and coding/technology at any level, the Software Underground is the best place to find your peers! They are a registered non-profit and a legit professional society, the likes of which you won’t find anywhere is the geoscience space. Signing up for the Slack with over 4000 members is free and you can also join as a paying member to help support the community.

Finally, I want to extend my heartfelt gratitude to the Software Underground community for providing all of the answers above and always being keen to help!

Update (2022-03-18): Add testimonial contributed by Peter Burgess (including an awesome 3D model animation).