00 Spatial Data Analytics: Introduction
Well hello everybody. Um this is the beginning of the graduate level course geostatistical subsurface modeling. I think in the in the calendar for UT. It's listed as stochastic subsurface modeling or stochastic modeling of the subsurface. Or something like that. Whatever it's called this is what its gonna be and let's just get started with it. This is the first lecture so what we'll do here is we'll have a really quick summary of who I am. There will be a lot of new students in the course about half the students from the Jackson school of geosciences. So let me get into that so. Who am I well. First and foremost my last name is pronounced perch. I have been informed by Ukrainians that I'm not saying it right however it is over a hundred years in Canada. We lost the ability to pronounce our last name. And that's what we've accepted it to be. I'm new I'm new to UT. I just left industry year ago means everything is new. Means course material is new this second time. I've conducted this course. I'm making lots of changes to the class. You'll see that what does that mean. I didn't write it down but what I was thinking was I make mistakes. I don't I'm completely open to advice if there's an opportunity to improve and so forth. I'm always. I'm always interested in that information what I want to help. I've done this. I have about 17 years of experience in consulting teaching leading conducting industrial Rd and statistical modeling reservoir modeling uncertain characterizations so forth. And so everything. I'm teaching. I know how this is actually done in industry. I when I was at Chevron I was had the opportunity to review reservoir model work all over their company all over the world. I had the opportunity to teach it and I had the opportunity opportunity to conduct research in it. I left industry last about a year ago to teach really enjoy teaching so I'm motivated to try to do the best. I can with this class. The other thing too is. I'll say all the time in class that I believe in democratic principles which means that if you have ideas in class you have feedback on improving the learning opportunities or even to manage the course load.
Let me know. I'm always happy to discuss. I'm always happy to improve. I'm very flexible. For instance last term about a couple lectures in my students said boy dr. Burchard be really useful if you were to record all your lectures and so I started recording all my lectures like I am right now. Recorded them outside of class so that there is good sound quality it also provides me the opportunity to really focus on the presentation for the video rather than say half. The conversation is cut out while. I'm interacting with students if I was to record during the classroom time. So this optimizes this idea of providing the best product. I think to support the students but it was feedback saying why don't you record all your lectures. I did it. I put them out on YouTube and I think that was a very good thing to do if there's other ideas I'm always happy to entertain those. I'm available I have an open-door policy. You're one of my students who can drop by my office if the door is closed it probably means. I'm just playing music recording a video. No worries just knock come on in it's okay. I'm always happy to discuss. I am an engineer but my BSC is in mining engineering which is basically where I was from a form of geologic engineering. I had a lot of courses in I took courses in stratigraphy and mineralogy in rock mechanics slope stability the structures. I've been through a lot of Geo even contaminant transports hydrology courses so I have quite a bit of Geo in my background. In addition to that my PhD was in quantitative geology. I had committee members such as Octavia nunu up in University Alberta. Who's well known in sequence tree geography and I spent 13 years in Chevron in earth sciences R&D working with geologists geophysicists on reservoir modeling so the summary of all of that is I speak geo.
I'm also a modeler who strongly believes that we should understand the phenomenon that were modeling and we are modeling. The subsurface we should understand the geology. The other thing. I should say is that I am active and outreach. Social media professional organizations. I'm not saying this to brag but I'm associate editor computers and Geosciences. I was just put on the editorial board of mathematical geosciences both for the International Association mathematical geosciences. So a lot of math a lot of geosciences. I've just been nominated ever put in a chair. Position for the espy data analytics technical section. That's international not one of local groups which is awesome. That's a lot of fun and I authored a book on Twitter I'm the geo stats guy on github. I'm the geo stats guy and on YouTube. All of my lectures are available under do stats guys lectures. So what will you learn during the class. So that's don't worry. I'll never do that again for the remainder of the class. That will not spend a bunch of time talking about myself. I think it's good. Yeah you know who. I am where I'm coming from. I hope it helps. You understand my perspective my motivations and so forth okay. So what are you gonna learn from this class. You're gonna learn about data analytics spatial data analytics subsurface data analytics geostatistics. The theory the methods the practice how we do subsurface modeling. How do we build the workflows in order to accomplish that. We're going to learn what is done. Currently what's established practice. What should be done. What are the gaps in that established practice and what are in fact the best practices. What could be done novel innovative approaches. That would really push it forward. We won't spend too much time on that but I think it's a good idea to provide some of that so that you come in to kind of look for thinking forward about new opportunities how to influence what's gets done how to critically evaluate subsurface modeling work if you were on a project team an asset team and you had to evaluate look at the model.
That's being built. What are the questions you should be asking. What are the diagnostic types of features or components. That you should be looking at how to do it yourself. Maybe you want to at some point. Develop your own workflows well. You should know how to do that yourself and we'll do that in the class. You'll get a chance to be building your own workflows. Getting the job done yourself how to communicate what is done. How do you report to people who may not be knowledgeable in the field of modeling in a way that they will understand and be able to evaluate that work and be able to support their decision making everything we do in subsurface modeling is all about decision support. Communication is critical to make sure we can do that. So how are we going to do that. That's what we want to teach you in this class. How we're going to accomplish them. Well we're going to do some role-playing. I like improv. It's kind of like improv. I guess as if it won't be comedy but I do make jokes sometimes mulch anyway. We will operate the class as a term long role-playing effort where basically we'll have the class as a subsurface project team an asset team. And that's going to be starring myself as the subsurface asset manager and each one of you will be a subsurface modeler on my team. You're working for me on my team and we're working within some type of a like a large company in industry and so if all term long we'll be working on the same project but we will progress it through all the fundamental steps of subsurface moment so the students will have the opportunity to work with realistic synthetic datasets to gain practical experience in fact. I just finished the Python code in order to make those randomly for each one students well comment on that complete a comprehensive reservoir or I should say subsurface modeling project including initial data analysis and cleaning often.
You'll hear that 80% effort is data. Preparation cleaning will provide data. That's pretty clean so we won't have to spend 80 percent of our term doing that. Univariate multivariate and spatial analysis estimation and simulation of reservoir properties and also model checking post-processing. Lots of things we can do with the model when we build models and then to support and make decisions in the presence of uncertainty. The students will be expected to prepare and present project updates as we go along consistent with what. I would have expected when I was a program manager or a project leader within industry. You can expect that type of presentation. I'll help you get that to be able to do that so really how are we going to be able to do that. This is a good question after what we've all talked about. It might seems ambitious as far as what. I've laid out for the goals for this course. Well the way we're going to accomplish. This is that you can have resources available to that are going to help you. There's in-class lectures. They will cover all the necessary theory. The methods the practice the strategies and the workflow designs that will be given to you in lectures now during the lectures also so sorry. The class lectures are going to be recording put on YouTube for review and they'll be Accord it just like this is right now from my office so it's good sound quality and so forth you can use the review. I tend to talk pretty quick so I'm finding my students don't actually have to speed up my lectures two times. Or what in order to listen to me. I think it's more efficient but you can always pause me if you have to. All lectures demos workflows from the undergraduate class are also available on YouTube github and so forth and these are pretty useful because a lot of the basic concepts the theory was taught in the undergrad - in fact I will be leveraging it from it at times not everything but we will go into much more practice than we did an undergrad but a lot of theories there so all of those lectures are available on juice that lectures channel on YouTube.
There's I think 40 or so lectures there and are available to you and on github there are a lot of workflows available to you. You can download in fact many of the critical steps that we have for this term long project. I've already put together sample workflows with generic datasets in order to accomplish a lot of these steps. So this will be there to support you to help you put together your own custom workflows with your own data. Why learn about geostatistical subsurface modeling. You may have got to this point right now and you're having this realization of why would. I spend this time why would I spend all this effort doing this working on these workflows play this role-playing exercise in this course in order to learn about subsurface modeling and so level one you want a basic understanding of subsurface model most reservoir assets surface assets subsurface teams are working with three-dimensional models of the subsurface for any major capital project. That's a given. There's going to be three dimensional modeling if you're going to work with the subsurface you're going to be exposed to stochastic reservoir models and if that's happening within your subsurface asset team what generally happens is they'll be at the center there will be the model construction of all this information integrated into it. It's essential for you to understand how these models are built to be cognisant to be aware of how they're being built. What's the next level to would be. You want to improve communication. Once again the reservoir modelers going to sit in the middle of the subsurface team. I'm not suggesting that they're the most important member of the subsurface asset team but they are in the middle the air interfacing with all of the other disciplines in the subsurface team and they're integrating all that available engine geologic and geophysical information into the subsurface model.
And so the best thing to do is to improve your capabilities in modeling so you can improve your communication and your integration of your concepts your information into the subsurface asset team. You'll have better communication on the team that's all. Let's just say you have better communication across the team. If you understand what's going on. At the moment level 3 would be where we now maximize our impact now. Reservoir models are used directly to support the wall to calculate the forecast and other types of results from the subsurface that are then used to support decision-making if you want to best integrate your knowledge into the subsurface model you should understand about how rows bar modeling gets done so you can make sure your engineering your geophysical your geologic information gets into these subsurface. If your expertise does not impact the model that's the worst case scenario. You may not impact the development decision. The exploration decision whatever decision it is and then you don't have impact we all want to have impact when we're out there. Working level 4 level 4 is where you go over to the dark side you decide to become a reservoir mother or subsurface mother in my experience the vast majority of those who were participating in building these subsurface models were in fact engineers and geo scientists who learned on the job. If you're interested in that path this course will be a very valuable opportunity for you to jump forward to spring forward in the knowledge required in order to do that to gain the theoretical knowledge to learn about the best best practice so that you can be a good modern black box. Uninformed reservoir modeling or subsurface modeling is dangerous. It results often in bad forecasts and that results in bad decision-making in this class you'll learn enough about what's going on there under the hood to critically evaluate improve reservoir models to be a better subsurface or reservoir model.
So how are we going to sign grades within this. Course there's no quizzes. There's no examinations the idea here is I want to incentivize the performance. That would help you succeed within industry in industry. You will succeed when you're able to integrate concepts you're able to provide professional updates that provide your leadership with the information that they need in order to make decisions and you yourself support and help make those decisions so. I will base the grade completely on project. Updates which you'll see later are oral and written a completed class project which is going to be a compilation of all of those updates into a final report and also class participation the questions you ask of other groups welder or other individuals as they're presenting and how you participate say participate in class like that okay textbook. There's no assigned textbook within this class. Although there are books that could help you in this class now. I could also make my lecture notes available from the undergraduate course. That could be helpful. I'll probably upload them and make them available to you. The lectures are available to you my book with Clayton Deutsch. I think would be useful. I've asked the library to stock a couple extras of these. This book specifically the library over in petroleum engineering. I could reach out to geo systems and ask for the same thing but that book is also available of course for purchase. I think if I was to recommend a single book that would be the one I'd recommend if you want to learn about the methods behind geostatistics kind of more in detail the geostatistical software library user's guide is a very good deutsch in journal. And of course. Larry Lake one of our full professors here in my department wrote a very good book. Statistics for patrolman engineers and geoscientists that maybe also use well to you now. If you want to hear about my thoughts around other books that could be a user of interest to you.
Well here's a list right here. Of course. I put the Jews statistical library users guide up top. There's also an applied geostatistics with s gems which is a open source soft side of software for jews. Statistics developed at Stanford University Romney Buescher Buescher and will provide a users guide for that. There's an introduction applied geostatistics which is the book one ed MO known as the book from Ed in MO Isaacson. Serve a semester. Rostova very readable very readable. And there's a variety of other books right here listed if anyone's interested that course have these books in my office if you want to check them out before you pick them up. What are the subsurface modeling steps that we will cover in this this course well our chapters of this of this course will follow and the updates will follow these mean steps within a a subsurface modeling workflow now we'll start with some prerequisites up front. We'll spend a couple lectures and we'll get into some practical things around maybe Python or around fundamental probability theory that I think is essential to get you started but after that will be very much focused on the steps of subsurface modeling data preparation univariate multivariate analysis spatial analysis estimation and trend modeling stochastic simulation uncertainty analysis. How do you check your models. And then how do you make decisions in the presence of uncertainty with your models for each one of these chapters there will be lectures and demos there will be there will be examples of how we do this for subsurface inference modeling. And then there'll be an opportunity for you to complete an update towards the final project and provide an update with documentation and presentation of that work. Now let me just make a couple more comments about the project. What's the motivation. What's the idea well. The purpose of the project provides you of hands-on experience to teach you the theory and then be able to thoughtfully empirically.
It's experiment around. Apply this theory to realistic subsurface modeling problems the deliverables from it well you'll provide regular updates documentation and in-class presentations. I want you to gain from that experience with professional presentations and documentation the style. The skills also one you gain opportunities to learn about critical review of a subsurface project. So when you're not presenting you have the opportunity to ask questions clarification questions that can in fact be quite challenging which will help everybody come to a better solution. That's all about being a better observer a better better at critical review of models and I know in my professional career I encountered a lot of people who did a very good job of asking those very good questions finding those critical gaps and helping the project improve at the end and I faced that a lot and it made me better at what. I do. Provide ample opportunity to demonstrate your knowledge tips for success work to demonstrate. Your knowledge of the theory of the concepts don't just paste figures into a PowerPoint. Think about what they mean. Think about how to interpret them be very efficient with that ensure that you follow the guidelines and address topics provided in advance don't provide an update. That doesn't actually address the fundamental technical questions or challenges that were given to you at the assignment at the time of the assignment and for that update presentations and documentation must follow logical flow. They can't just be floating in the air. They they have to have be defendable build-up you have to explain assumptions limitations and other considerations if you do that you're going to do very well in this class or the project. I have already generated multiple synthetic truth datasets and they are based on a simple two dimensional multivariate spatial problem it's one kilometer per one kilometer. Here's and there's a variety of different attributes there's facies sandstone and shale there's porosity there's permeability there's a seismic property which is acoustic impedance that you also will be able to work with.
And that's exhaustively sampled at all locations. The other ones of the the other properties of interests are going to be sampled only at limited locations within that one kilometer by one kilometer. Now there are multiple truth data sets and sample sets for every one of the students or we may use groups in the class. We'll see how that goes and every single thing has been randomized. So what you'll find. Is that each one of these truth. Data sets and in fact that the sample sets from each one of them are randomized. So they'll be very different from each other software. What software should you use. Now we could not have hands-on practice an opportunity to build your own workflows without using the computer. We have to do something with the computer to get that down. The point is I refer you back to one of my first slides where I said I am Democratic democratic or a believer in democratic principles. And so you're free to use any software you would like to complete the term long project as long as you meet. The goals provide me with the summaries and that are going to be discussed in more detail just pretty quickly here then. It's all good. So what are some of options that you could use for doing this. Well you could use open source statistical and geo statistical software. The most simple approach would be just to use gia slot now gia slide is the standard geostatistical library. And so it you could just run the individual algorithm so you could build a workflow with dot bat files that would run each one of those methods and sequence you can produce a series of outputs which could be output files data files and models or post group files for your summaries and your reporting you could do all of that in. Jesus life and in fact in fall of 2017 most of the groups did in fact complete this course just doing it like that and I can provide examples workflows and so forth to help people do that you could use our in our studio there are stats and G stats.
Packages are actually pretty good. It's a quite robust. It works quite well and I could also provide example workflows. I put them together for my undergrad course and so those could be available to YouTube you could do. Python PI G asleep ideas live as a package it is a more of a formal wrapper of the GS slide in Python. It's got its own kind of design. I've used it. I've developed example workflows. The package is not currently current it's maintained but it's not quite current which means many students may run into problems trying to install it. You may have to go back to a previous version of Python because I found it became bit of a headache in order to get that all working but you could also do that I have developed workflows using pi GS live. That can help you get started there. I put this in red. This is my recommended path board. Python geo stands PI. Jia slide. What we do here is. I'll give some more information about it right away but basically it's my own I've coded up geo stats. Pi is kind of a very simple wrapper of the gist. Live with reimplementation of a lot of the kind of simple numerix or visualization. And that way you can build the entire workflow in python. It's quite clean. It's easy to use and all you need is the examples you could work with the functions that I have induced. That's pi plus the executables. Now another option would be to work with more commercial type software such as patrol if you have access to patrol in our department. There's an educational license students can get access to Patrol. It is commonly used in industry for building subsurface models. So it it wouldn't maybe make sense to use something like patrol. I'm not going to expect that for this class. I'll make other comments about software right away here. There's also course other commercial software with appropriate licenses that you could use and so.
I list those here. I am completely open to whatever software. You're comfortable using now. A couple more comments around software encoding. This is not a software coding class so people have asked me. Why don't you just teach patrol and that we could do that. Patrol has some very nice tutorials. We could try to teach that within a class. I don't think that that has high utility. The reason being is that of course patrol is a highly specialized software suite. It's um receives regular updates it changes and so what you'll find is that there's so many details to patrol if you don't use it continuously you start to lose it pretty quickly and so you probably buy time you finish and leave. UT you may not be current anymore. I think we're better served if we teach you the concepts for deeper understanding of how do you build workflows and for subsurface modeling and then you're able to when you get into a company whatever software they're using you'll understand those concepts which will be general in fact it's amazing even the basic building blocks of what we did with. Jesus live from like 20 years ago is exactly what people do in patrol. It's just more fancy with the gooeys and databases and so forth. But it's all the basically the same thing if you understand those deeper understanding of those concepts you'll be able to use whatever software is available. Come up to speed with basic get up to speed how it works and be able to put that together now. Workflows could be developed using very simple geus live. That's all comment on that. So no coating would be required at all no event software would be required at all that would be possible. Python and our approaches you use available packages with very basic code. You don't have to do really deep. Python coding in order to do this. It's more about using the packages and putting workflows together in Python so if people are interested I have a variety of different demonstrations available for basics of Python if a bunch of people are going to use Python in this course I could run a couple of the lectures just on basics of building work workflows in Python regardless I've got lots of example workflows for people to start with and I'm always happy to do more to support people pythons powerful a basic ability to use Python probably looks pretty good on your CV so I encourage that now.
I want to emphasize this point right here. This class could be completed with very basic scripting and geus libe use of. Jesus like programs so one option you just used dot bat files which is basically with an example which shown right here I in fact use dot bat files that build the truth model models back in 2017 not Python at that point. And so you could. You could have all kinds of looping and conditional statements you could do a lot of things in bat files and you could be able to build your workflow up and it would not require any coding at all so the reason. I'm saying all of this is. This course is completely accessible to any geo scientist or engineer even somebody who showed up here today and said I absolutely refused code and I don't want to learn complicated software. Then you could do it this way. You could have complete success in this course you would not be at a disadvantage as far as the result from this class. But if you're not already convinced please provide me with the opportunity to show you two slides and on those two slides. I would like to try to convince you that. In fact coding would be of great benefit to you and so here. Is that some arguments for why coding should be something that all geo scientists and engineers should learn transparency. No compiler accepts hand-waving coding forces you to to lay your logic bare for anybody any scientist or engineer to review and look at your logical flow of your workflow. It's code there it is. We're pretty reproduce ability you run it get an answer hand it over. They run it to get the same answer one of the main concepts behind the scientific method.
Coding is pretty handy for that. Quantification programs need numbers in order to run. You need to feed the program. You're going to discover new ways to look at the world with quantitative methods you will learn new things open source leverage the world of brilliance. Check out the packages. It's amazing what's available to you. How many workflows and packages are being shared freely all over the place very easy to install all kinds of repositories for it. And you're able to run and try something new. I remember one day. I learned about the Hurst coefficient. I thought that'd be really interesting. The tried out for time series analysis and within minutes. I was calculating the Hurst coefficient. I downloaded one package. I was going. It was very great break down barriers. Don't come up with your engineering. Geoscience solution and throw it over. The fence said that the table with the developers the people who are taking your knowledge and turning it into a tool sit at the table with them and share more of your subject matter expertise get a better product. If you understand something about what they have to do to code it deployment come up with something great code it up share it with others and multiply the impact leverage performance metrics or altruism. It doesn't matter. Your good work is going to benefit many others and in the end you're probably going to get recognized for it to efficiency minimize. The boring parts of your job doesn't matter how scientific we are. There's going to be boring parts of your job if you can build a scripts set of scripts some EPIK code. That automates the common tasks. Things that are boring. You spend more time doing engineering insights always time to do it again. How many times did you only do something once. Probably takes two to four times as long to script to code to automate a workflow. It's usually worth it. Turns out you'll have to often go back and redo it over and over again be like us.
I hope this doesn't sound too arrogant. It will change you. Users will always feel limited programmers. They truly harness the power of their applications in their hardware. They feel empowered. And so. That's why you might want to consider coding. Just in case you had not already been convinced. There's great benefit in coding. I have had coding in my entire life. In fact it seems all the way from when I was very young. Coding on cassette tapes on my tee in the TI 99 for eighth and my Amiga 1000 my not above my newspaper route my Amiga. Oh shucks shucks I went to Amiga 2000 after that upgraded and then of course to various different pcs and so forth but coding always provided me with great benefit no matter what I was doing what I was working on it was great now. I want to just give a couple of caveats. These people will hear me talk about coding and they may make some assumptions now first of all. I'm not arrogant about it. Any type of scripting any simple type of coding workflow automation matched appropriate to your working. Environment is great. It's all great. We don't all need to be C++ experts I spent probably 13 years working with C++. I'm sure I'm not an expert in it and I'm sure there's some C++ experts who've looked at my code and would agree with me but anything is good is very flexible. I respect the experience component of Geoscience engineering expertise. This is really beyond coding. This is the idea that is essential to good workflow logic development to best use the data we need to really be focused on the geo science and engineering coatings. Huell to assist us in doing a better job as a geoscientist engineer. Now some expert judgment war remains subjective when I say quantification and so forth. I'm not suggesting that we're replacing expertise with a machine. I'm not advocating for that in talking about the value of coding. Alright my caveat sigh think. I protected myself there it sincerely I how I feel about that.
So my recommendation for workflow construction once again. I said I had democratic values that I felt that people should be able to build the workloads however they like I would recommend the use of open source methods with workflows in Python I think Python. There's a lot of great resources right now. I would recommend leveraging the strengths of the geostatistical library very robust very well documented example parameters. Everything's available so and I would recommend if you're working in Python to work in Jupiter notebooks so geo stats pie is a set of Python functions for most of the required workflow steps that support you in doing that. It includes visualization. The light numerix have also been reimplemented. There are more heavy numerix. The kind of heavier programs like simulation spatial estimation and so forth i have used the wrappers kind of wrappers in order to run the gist live executables and so the wrapping method is very simple it writes out a parameter file. The data that's required by the geo slide program runs the geo. Slive executable and of reads in the results very simple so this is the method I suggest that you work with the Python geo stats pi. That is wrapping and reimplemented base. Egeus live if any of these steps of anything that you're doing seems ownerís at any time. Then let me know all says none of this should be terrible now let me make a couple of comments about some of those components of my recommendation. Jupiter notebooks it really is it's super cool it's a web-based application where you have a set of blocks block block block and so each one of the blocks can include code documentation or results and so this is documentation this is code documentation code result and so forth you can also work with a variety of kernels in fact when I say Jupiter it does not mean that you have to work with Python you could actually download and work with an our kernel SC kernel JavaScript kernel or any other kernel and you could use Jupiter notebooks with those methods in fact I have a bunch of workflows examples.
Interpreter notebooks with our kernel because a lot of my students preferred the Jupiter notebook to our studio and they like that now. Once you're in the business of doing trip or notebooks there's so many cool things you can do. You can produce very professional-looking workflows. The documentation is in markdown. So you can actually put equations links you can make it look very cool you can in fact use containers and run it on long online docker. You could put this entire workflow online and other people could run it in a container so they don't have to set up the environment they can just run it online experiment and immediately see results. So it's all very powerful now if you're going to work with Python you don't need to know a lot you don't have to be really in-depth in the Python but you want to be able to work with the basic elements the basic objects that are created as part of a subsurface modeling workflow and what are those basic objects. Well they would include gridded data like seismic data and they would include tabulated data like the well data. And so you'd want to learn about numpy which is package that's very good for doing all kinds of numerics but also for working with arrays of data like a gridded data set pandas package provides excellent methodologies the data frames for working with tables of data like we'd help from the well data now. I have tutorials available for the basics or the things that you need to know for arrays gridded data for data frames well data in order to build subsurface modeling work. I've made these tutorials I put them on github and so they are. Jupiter notebooks with many of the steps that you would need in order to build a subsurface modeling workflow. Now if there's an interest from the students in the class we could of course cover this in class. We could spend a couple lectures if most students are working in.
Python in order to do this we could even do a lecture outside of class that people are interested in that but what is geo stats pi. Well it's something. I've been coding. It's a set of functions in. Python these functions are the ones you need for most of the steps in building a subsurface model now if they were visualization or light numeric Syre coded them in Python and so there will be functions available to you like location map all I did was use matplotlib and the standard geostatistical library types of parameters and so people can quickly build location maps if it's something like producing a Vera Graham a very ground function plot I have reimplemented Varg plot indirectly in Python. So people can do that. And that's just using that plot lib again and so those are two examples right there. Heavy numerix such as spatial estimation simulation. I wrapped as I mentioned before the GS slide wrapping is very simple as I mentioned before it writes out the parameter file it writes up the data I have a bunch of functions included in geo stats PI that are able to convert between the geo es us. EAS format and in fact the format of say data frames or India arrays so you can move back and forth between geo slide and Python with your datasets and so all of that is available to you. I've written this. I am constantly adding to it. I put it on on github and so we can even augment it to help support the students with maybe some of the more tricky steps within the project. What's expected from you you're going to work on a term long project. Updates will be due every two weeks. The updates will include oral presentation on the day that they're due only about five minutes a very brief very concise. And then we'll have two minutes of questions and we move to the next group. PowerPoint template is going to be is provided to you on canvas written update. This is very very concise. Here's a template right here. You'll see that it works out to be maybe two pages and total width figures.
It's brief it's concise should not be onerous use efficient creative figures to communicate. Don't just provide us with a document or me with a document that's going to be just pages and pages of figures. Think about how you can be very efficient with combining figures communicating very effectively efficiently and. I also have a word document on canvas for you. Here's the PowerPoint presentation update template very simple. It's just the summary executive summary workflow results discussion conclusions. It should be literally 5 slides in addition to the title slide the final project. It's not that bad it's going to be a compilation of all of the updates with the requested revisions that there were visions that I requested and there will be from based on the discussion from the class and based on my input. There will be chances for you to improve on your work. Those will be expected the documentation a bit enhanced. And then there'll be an additional challenge of a decision making exercise that we will put into it and you'll be expected to complete prior to handing in the final project academic dishonesty let me say we are simulating a corporate environment. We all win together. Collaboration is completely encouraged. Consult with each other. I want you to do that. I want you older to learn together as a group. It's a real-world type of situation. We have a network we're connected to each other you might be modeling in 10. G's and you might talk to somebody else in the world. Who's modeling another carbonate reservoir or you might be modeling a Jack and saint-malo and then you're talking to somebody else who is modeling over in block 14 and Angola you if you you would be talking to each other. Each of your data sets are completely different. The truth models are very different. So if you won't be able to help each other with accuracy but accurate he's not the not the challenge here right to build defendable workflows that use best practice is the Challenger but each update final project must be unique work of each student.
If we have earth the class is large we may decide to form groups in which case will be the unique work of the group assignments hardcopy of the update documents or will be due at the start of the class. Presentations are given and graded in class. Contact me if you're going to be absent or if you have any other type of emergency late. Updates and presentations will be accepted with reasonable excuses. Illness family emergencies interviews conferences. I support the fact that you have many important other things you're working on as far as in grad students in some cases undergrads and I get that disabilities of course if you have any accommodations come see me. I am all about supporting and helping anybody I want to include. I want to ensure that everybody is comfortable and able to enjoy and benefit from this class. Classroom environment pause a positive professional conduct. I think that's a great way to describe it. We're simulating a subsurface asset team. Meeting doesn't mean we're all like super gentle on each other. It doesn't mean we don't ask hard questions. In fact. The general tone should be searching and probing questions but it should be very much like we're trying to encourage each other we're trying to maximize each other's impact on the project. We're trying to be collaborative. We're trying to win together. Hard questions can help everyone learn. That can help us win together. If somebody's done something that you think they have an opportunity to improve. Make the statement. It's totally cool. That's all good. We did that inside of industry actively participating in supporting a dialogue be enthusiastic avoid negative behaviors anything that's disruptive negative. Disrespectful comments of course would not be okay. Anything that's unfocused. No use of electronics other than those needed to participate in the class. Cellphones will not be out laptops if they're out only if they're used to say follow along a lecture and so forth now for any other type use checking the news sports scores or anything like that personal attacks of course would not be appropriate at any time.
Presentations must not run over time. And that's part of professional conduct to imagine everybody in your audience will have a lot of responsibilities and competing pressures on their time. They will lose interest. And what you're saying if they realize you're running late and they're uncertain about when they'll be able to escape. Some personality is fine but be concisely. Be on message or this class. We're going to teach the reticle in practical reservoir modeling and we're going to do it with a lot of hands-on experience and opportunity to simulate what it's like with in industry or than a company to be conducting and subsurface modeling exercise and so now it begins of course anybody who's watching or listening along here at any time you're welcome to send me questions. I am Michael perch I am a associate professor at University of Texas at Austin I am also the geo stats guy on Twitter and I have all my lectures juiced at sky lectures on YouTube and a lot of my workflows and methodologies code are available on github under geo stats guy - alright thank you.
Let me know. I'm always happy to discuss. I'm always happy to improve. I'm very flexible. For instance last term about a couple lectures in my students said boy dr. Burchard be really useful if you were to record all your lectures and so I started recording all my lectures like I am right now. Recorded them outside of class so that there is good sound quality it also provides me the opportunity to really focus on the presentation for the video rather than say half. The conversation is cut out while. I'm interacting with students if I was to record during the classroom time. So this optimizes this idea of providing the best product. I think to support the students but it was feedback saying why don't you record all your lectures. I did it. I put them out on YouTube and I think that was a very good thing to do if there's other ideas I'm always happy to entertain those. I'm available I have an open-door policy. You're one of my students who can drop by my office if the door is closed it probably means. I'm just playing music recording a video. No worries just knock come on in it's okay. I'm always happy to discuss. I am an engineer but my BSC is in mining engineering which is basically where I was from a form of geologic engineering. I had a lot of courses in I took courses in stratigraphy and mineralogy in rock mechanics slope stability the structures. I've been through a lot of Geo even contaminant transports hydrology courses so I have quite a bit of Geo in my background. In addition to that my PhD was in quantitative geology. I had committee members such as Octavia nunu up in University Alberta. Who's well known in sequence tree geography and I spent 13 years in Chevron in earth sciences R&D working with geologists geophysicists on reservoir modeling so the summary of all of that is I speak geo.
I'm also a modeler who strongly believes that we should understand the phenomenon that were modeling and we are modeling. The subsurface we should understand the geology. The other thing. I should say is that I am active and outreach. Social media professional organizations. I'm not saying this to brag but I'm associate editor computers and Geosciences. I was just put on the editorial board of mathematical geosciences both for the International Association mathematical geosciences. So a lot of math a lot of geosciences. I've just been nominated ever put in a chair. Position for the espy data analytics technical section. That's international not one of local groups which is awesome. That's a lot of fun and I authored a book on Twitter I'm the geo stats guy on github. I'm the geo stats guy and on YouTube. All of my lectures are available under do stats guys lectures. So what will you learn during the class. So that's don't worry. I'll never do that again for the remainder of the class. That will not spend a bunch of time talking about myself. I think it's good. Yeah you know who. I am where I'm coming from. I hope it helps. You understand my perspective my motivations and so forth okay. So what are you gonna learn from this class. You're gonna learn about data analytics spatial data analytics subsurface data analytics geostatistics. The theory the methods the practice how we do subsurface modeling. How do we build the workflows in order to accomplish that. We're going to learn what is done. Currently what's established practice. What should be done. What are the gaps in that established practice and what are in fact the best practices. What could be done novel innovative approaches. That would really push it forward. We won't spend too much time on that but I think it's a good idea to provide some of that so that you come in to kind of look for thinking forward about new opportunities how to influence what's gets done how to critically evaluate subsurface modeling work if you were on a project team an asset team and you had to evaluate look at the model.
That's being built. What are the questions you should be asking. What are the diagnostic types of features or components. That you should be looking at how to do it yourself. Maybe you want to at some point. Develop your own workflows well. You should know how to do that yourself and we'll do that in the class. You'll get a chance to be building your own workflows. Getting the job done yourself how to communicate what is done. How do you report to people who may not be knowledgeable in the field of modeling in a way that they will understand and be able to evaluate that work and be able to support their decision making everything we do in subsurface modeling is all about decision support. Communication is critical to make sure we can do that. So how are we going to do that. That's what we want to teach you in this class. How we're going to accomplish them. Well we're going to do some role-playing. I like improv. It's kind of like improv. I guess as if it won't be comedy but I do make jokes sometimes mulch anyway. We will operate the class as a term long role-playing effort where basically we'll have the class as a subsurface project team an asset team. And that's going to be starring myself as the subsurface asset manager and each one of you will be a subsurface modeler on my team. You're working for me on my team and we're working within some type of a like a large company in industry and so if all term long we'll be working on the same project but we will progress it through all the fundamental steps of subsurface moment so the students will have the opportunity to work with realistic synthetic datasets to gain practical experience in fact. I just finished the Python code in order to make those randomly for each one students well comment on that complete a comprehensive reservoir or I should say subsurface modeling project including initial data analysis and cleaning often.
You'll hear that 80% effort is data. Preparation cleaning will provide data. That's pretty clean so we won't have to spend 80 percent of our term doing that. Univariate multivariate and spatial analysis estimation and simulation of reservoir properties and also model checking post-processing. Lots of things we can do with the model when we build models and then to support and make decisions in the presence of uncertainty. The students will be expected to prepare and present project updates as we go along consistent with what. I would have expected when I was a program manager or a project leader within industry. You can expect that type of presentation. I'll help you get that to be able to do that so really how are we going to be able to do that. This is a good question after what we've all talked about. It might seems ambitious as far as what. I've laid out for the goals for this course. Well the way we're going to accomplish. This is that you can have resources available to that are going to help you. There's in-class lectures. They will cover all the necessary theory. The methods the practice the strategies and the workflow designs that will be given to you in lectures now during the lectures also so sorry. The class lectures are going to be recording put on YouTube for review and they'll be Accord it just like this is right now from my office so it's good sound quality and so forth you can use the review. I tend to talk pretty quick so I'm finding my students don't actually have to speed up my lectures two times. Or what in order to listen to me. I think it's more efficient but you can always pause me if you have to. All lectures demos workflows from the undergraduate class are also available on YouTube github and so forth and these are pretty useful because a lot of the basic concepts the theory was taught in the undergrad - in fact I will be leveraging it from it at times not everything but we will go into much more practice than we did an undergrad but a lot of theories there so all of those lectures are available on juice that lectures channel on YouTube.
There's I think 40 or so lectures there and are available to you and on github there are a lot of workflows available to you. You can download in fact many of the critical steps that we have for this term long project. I've already put together sample workflows with generic datasets in order to accomplish a lot of these steps. So this will be there to support you to help you put together your own custom workflows with your own data. Why learn about geostatistical subsurface modeling. You may have got to this point right now and you're having this realization of why would. I spend this time why would I spend all this effort doing this working on these workflows play this role-playing exercise in this course in order to learn about subsurface modeling and so level one you want a basic understanding of subsurface model most reservoir assets surface assets subsurface teams are working with three-dimensional models of the subsurface for any major capital project. That's a given. There's going to be three dimensional modeling if you're going to work with the subsurface you're going to be exposed to stochastic reservoir models and if that's happening within your subsurface asset team what generally happens is they'll be at the center there will be the model construction of all this information integrated into it. It's essential for you to understand how these models are built to be cognisant to be aware of how they're being built. What's the next level to would be. You want to improve communication. Once again the reservoir modelers going to sit in the middle of the subsurface team. I'm not suggesting that they're the most important member of the subsurface asset team but they are in the middle the air interfacing with all of the other disciplines in the subsurface team and they're integrating all that available engine geologic and geophysical information into the subsurface model.
And so the best thing to do is to improve your capabilities in modeling so you can improve your communication and your integration of your concepts your information into the subsurface asset team. You'll have better communication on the team that's all. Let's just say you have better communication across the team. If you understand what's going on. At the moment level 3 would be where we now maximize our impact now. Reservoir models are used directly to support the wall to calculate the forecast and other types of results from the subsurface that are then used to support decision-making if you want to best integrate your knowledge into the subsurface model you should understand about how rows bar modeling gets done so you can make sure your engineering your geophysical your geologic information gets into these subsurface. If your expertise does not impact the model that's the worst case scenario. You may not impact the development decision. The exploration decision whatever decision it is and then you don't have impact we all want to have impact when we're out there. Working level 4 level 4 is where you go over to the dark side you decide to become a reservoir mother or subsurface mother in my experience the vast majority of those who were participating in building these subsurface models were in fact engineers and geo scientists who learned on the job. If you're interested in that path this course will be a very valuable opportunity for you to jump forward to spring forward in the knowledge required in order to do that to gain the theoretical knowledge to learn about the best best practice so that you can be a good modern black box. Uninformed reservoir modeling or subsurface modeling is dangerous. It results often in bad forecasts and that results in bad decision-making in this class you'll learn enough about what's going on there under the hood to critically evaluate improve reservoir models to be a better subsurface or reservoir model.
So how are we going to sign grades within this. Course there's no quizzes. There's no examinations the idea here is I want to incentivize the performance. That would help you succeed within industry in industry. You will succeed when you're able to integrate concepts you're able to provide professional updates that provide your leadership with the information that they need in order to make decisions and you yourself support and help make those decisions so. I will base the grade completely on project. Updates which you'll see later are oral and written a completed class project which is going to be a compilation of all of those updates into a final report and also class participation the questions you ask of other groups welder or other individuals as they're presenting and how you participate say participate in class like that okay textbook. There's no assigned textbook within this class. Although there are books that could help you in this class now. I could also make my lecture notes available from the undergraduate course. That could be helpful. I'll probably upload them and make them available to you. The lectures are available to you my book with Clayton Deutsch. I think would be useful. I've asked the library to stock a couple extras of these. This book specifically the library over in petroleum engineering. I could reach out to geo systems and ask for the same thing but that book is also available of course for purchase. I think if I was to recommend a single book that would be the one I'd recommend if you want to learn about the methods behind geostatistics kind of more in detail the geostatistical software library user's guide is a very good deutsch in journal. And of course. Larry Lake one of our full professors here in my department wrote a very good book. Statistics for patrolman engineers and geoscientists that maybe also use well to you now. If you want to hear about my thoughts around other books that could be a user of interest to you.
Well here's a list right here. Of course. I put the Jews statistical library users guide up top. There's also an applied geostatistics with s gems which is a open source soft side of software for jews. Statistics developed at Stanford University Romney Buescher Buescher and will provide a users guide for that. There's an introduction applied geostatistics which is the book one ed MO known as the book from Ed in MO Isaacson. Serve a semester. Rostova very readable very readable. And there's a variety of other books right here listed if anyone's interested that course have these books in my office if you want to check them out before you pick them up. What are the subsurface modeling steps that we will cover in this this course well our chapters of this of this course will follow and the updates will follow these mean steps within a a subsurface modeling workflow now we'll start with some prerequisites up front. We'll spend a couple lectures and we'll get into some practical things around maybe Python or around fundamental probability theory that I think is essential to get you started but after that will be very much focused on the steps of subsurface modeling data preparation univariate multivariate analysis spatial analysis estimation and trend modeling stochastic simulation uncertainty analysis. How do you check your models. And then how do you make decisions in the presence of uncertainty with your models for each one of these chapters there will be lectures and demos there will be there will be examples of how we do this for subsurface inference modeling. And then there'll be an opportunity for you to complete an update towards the final project and provide an update with documentation and presentation of that work. Now let me just make a couple more comments about the project. What's the motivation. What's the idea well. The purpose of the project provides you of hands-on experience to teach you the theory and then be able to thoughtfully empirically.
It's experiment around. Apply this theory to realistic subsurface modeling problems the deliverables from it well you'll provide regular updates documentation and in-class presentations. I want you to gain from that experience with professional presentations and documentation the style. The skills also one you gain opportunities to learn about critical review of a subsurface project. So when you're not presenting you have the opportunity to ask questions clarification questions that can in fact be quite challenging which will help everybody come to a better solution. That's all about being a better observer a better better at critical review of models and I know in my professional career I encountered a lot of people who did a very good job of asking those very good questions finding those critical gaps and helping the project improve at the end and I faced that a lot and it made me better at what. I do. Provide ample opportunity to demonstrate your knowledge tips for success work to demonstrate. Your knowledge of the theory of the concepts don't just paste figures into a PowerPoint. Think about what they mean. Think about how to interpret them be very efficient with that ensure that you follow the guidelines and address topics provided in advance don't provide an update. That doesn't actually address the fundamental technical questions or challenges that were given to you at the assignment at the time of the assignment and for that update presentations and documentation must follow logical flow. They can't just be floating in the air. They they have to have be defendable build-up you have to explain assumptions limitations and other considerations if you do that you're going to do very well in this class or the project. I have already generated multiple synthetic truth datasets and they are based on a simple two dimensional multivariate spatial problem it's one kilometer per one kilometer. Here's and there's a variety of different attributes there's facies sandstone and shale there's porosity there's permeability there's a seismic property which is acoustic impedance that you also will be able to work with.
And that's exhaustively sampled at all locations. The other ones of the the other properties of interests are going to be sampled only at limited locations within that one kilometer by one kilometer. Now there are multiple truth data sets and sample sets for every one of the students or we may use groups in the class. We'll see how that goes and every single thing has been randomized. So what you'll find. Is that each one of these truth. Data sets and in fact that the sample sets from each one of them are randomized. So they'll be very different from each other software. What software should you use. Now we could not have hands-on practice an opportunity to build your own workflows without using the computer. We have to do something with the computer to get that down. The point is I refer you back to one of my first slides where I said I am Democratic democratic or a believer in democratic principles. And so you're free to use any software you would like to complete the term long project as long as you meet. The goals provide me with the summaries and that are going to be discussed in more detail just pretty quickly here then. It's all good. So what are some of options that you could use for doing this. Well you could use open source statistical and geo statistical software. The most simple approach would be just to use gia slot now gia slide is the standard geostatistical library. And so it you could just run the individual algorithm so you could build a workflow with dot bat files that would run each one of those methods and sequence you can produce a series of outputs which could be output files data files and models or post group files for your summaries and your reporting you could do all of that in. Jesus life and in fact in fall of 2017 most of the groups did in fact complete this course just doing it like that and I can provide examples workflows and so forth to help people do that you could use our in our studio there are stats and G stats.
Packages are actually pretty good. It's a quite robust. It works quite well and I could also provide example workflows. I put them together for my undergrad course and so those could be available to YouTube you could do. Python PI G asleep ideas live as a package it is a more of a formal wrapper of the GS slide in Python. It's got its own kind of design. I've used it. I've developed example workflows. The package is not currently current it's maintained but it's not quite current which means many students may run into problems trying to install it. You may have to go back to a previous version of Python because I found it became bit of a headache in order to get that all working but you could also do that I have developed workflows using pi GS live. That can help you get started there. I put this in red. This is my recommended path board. Python geo stands PI. Jia slide. What we do here is. I'll give some more information about it right away but basically it's my own I've coded up geo stats. Pi is kind of a very simple wrapper of the gist. Live with reimplementation of a lot of the kind of simple numerix or visualization. And that way you can build the entire workflow in python. It's quite clean. It's easy to use and all you need is the examples you could work with the functions that I have induced. That's pi plus the executables. Now another option would be to work with more commercial type software such as patrol if you have access to patrol in our department. There's an educational license students can get access to Patrol. It is commonly used in industry for building subsurface models. So it it wouldn't maybe make sense to use something like patrol. I'm not going to expect that for this class. I'll make other comments about software right away here. There's also course other commercial software with appropriate licenses that you could use and so.
I list those here. I am completely open to whatever software. You're comfortable using now. A couple more comments around software encoding. This is not a software coding class so people have asked me. Why don't you just teach patrol and that we could do that. Patrol has some very nice tutorials. We could try to teach that within a class. I don't think that that has high utility. The reason being is that of course patrol is a highly specialized software suite. It's um receives regular updates it changes and so what you'll find is that there's so many details to patrol if you don't use it continuously you start to lose it pretty quickly and so you probably buy time you finish and leave. UT you may not be current anymore. I think we're better served if we teach you the concepts for deeper understanding of how do you build workflows and for subsurface modeling and then you're able to when you get into a company whatever software they're using you'll understand those concepts which will be general in fact it's amazing even the basic building blocks of what we did with. Jesus live from like 20 years ago is exactly what people do in patrol. It's just more fancy with the gooeys and databases and so forth. But it's all the basically the same thing if you understand those deeper understanding of those concepts you'll be able to use whatever software is available. Come up to speed with basic get up to speed how it works and be able to put that together now. Workflows could be developed using very simple geus live. That's all comment on that. So no coating would be required at all no event software would be required at all that would be possible. Python and our approaches you use available packages with very basic code. You don't have to do really deep. Python coding in order to do this. It's more about using the packages and putting workflows together in Python so if people are interested I have a variety of different demonstrations available for basics of Python if a bunch of people are going to use Python in this course I could run a couple of the lectures just on basics of building work workflows in Python regardless I've got lots of example workflows for people to start with and I'm always happy to do more to support people pythons powerful a basic ability to use Python probably looks pretty good on your CV so I encourage that now.
I want to emphasize this point right here. This class could be completed with very basic scripting and geus libe use of. Jesus like programs so one option you just used dot bat files which is basically with an example which shown right here I in fact use dot bat files that build the truth model models back in 2017 not Python at that point. And so you could. You could have all kinds of looping and conditional statements you could do a lot of things in bat files and you could be able to build your workflow up and it would not require any coding at all so the reason. I'm saying all of this is. This course is completely accessible to any geo scientist or engineer even somebody who showed up here today and said I absolutely refused code and I don't want to learn complicated software. Then you could do it this way. You could have complete success in this course you would not be at a disadvantage as far as the result from this class. But if you're not already convinced please provide me with the opportunity to show you two slides and on those two slides. I would like to try to convince you that. In fact coding would be of great benefit to you and so here. Is that some arguments for why coding should be something that all geo scientists and engineers should learn transparency. No compiler accepts hand-waving coding forces you to to lay your logic bare for anybody any scientist or engineer to review and look at your logical flow of your workflow. It's code there it is. We're pretty reproduce ability you run it get an answer hand it over. They run it to get the same answer one of the main concepts behind the scientific method.
Coding is pretty handy for that. Quantification programs need numbers in order to run. You need to feed the program. You're going to discover new ways to look at the world with quantitative methods you will learn new things open source leverage the world of brilliance. Check out the packages. It's amazing what's available to you. How many workflows and packages are being shared freely all over the place very easy to install all kinds of repositories for it. And you're able to run and try something new. I remember one day. I learned about the Hurst coefficient. I thought that'd be really interesting. The tried out for time series analysis and within minutes. I was calculating the Hurst coefficient. I downloaded one package. I was going. It was very great break down barriers. Don't come up with your engineering. Geoscience solution and throw it over. The fence said that the table with the developers the people who are taking your knowledge and turning it into a tool sit at the table with them and share more of your subject matter expertise get a better product. If you understand something about what they have to do to code it deployment come up with something great code it up share it with others and multiply the impact leverage performance metrics or altruism. It doesn't matter. Your good work is going to benefit many others and in the end you're probably going to get recognized for it to efficiency minimize. The boring parts of your job doesn't matter how scientific we are. There's going to be boring parts of your job if you can build a scripts set of scripts some EPIK code. That automates the common tasks. Things that are boring. You spend more time doing engineering insights always time to do it again. How many times did you only do something once. Probably takes two to four times as long to script to code to automate a workflow. It's usually worth it. Turns out you'll have to often go back and redo it over and over again be like us.
I hope this doesn't sound too arrogant. It will change you. Users will always feel limited programmers. They truly harness the power of their applications in their hardware. They feel empowered. And so. That's why you might want to consider coding. Just in case you had not already been convinced. There's great benefit in coding. I have had coding in my entire life. In fact it seems all the way from when I was very young. Coding on cassette tapes on my tee in the TI 99 for eighth and my Amiga 1000 my not above my newspaper route my Amiga. Oh shucks shucks I went to Amiga 2000 after that upgraded and then of course to various different pcs and so forth but coding always provided me with great benefit no matter what I was doing what I was working on it was great now. I want to just give a couple of caveats. These people will hear me talk about coding and they may make some assumptions now first of all. I'm not arrogant about it. Any type of scripting any simple type of coding workflow automation matched appropriate to your working. Environment is great. It's all great. We don't all need to be C++ experts I spent probably 13 years working with C++. I'm sure I'm not an expert in it and I'm sure there's some C++ experts who've looked at my code and would agree with me but anything is good is very flexible. I respect the experience component of Geoscience engineering expertise. This is really beyond coding. This is the idea that is essential to good workflow logic development to best use the data we need to really be focused on the geo science and engineering coatings. Huell to assist us in doing a better job as a geoscientist engineer. Now some expert judgment war remains subjective when I say quantification and so forth. I'm not suggesting that we're replacing expertise with a machine. I'm not advocating for that in talking about the value of coding. Alright my caveat sigh think. I protected myself there it sincerely I how I feel about that.
So my recommendation for workflow construction once again. I said I had democratic values that I felt that people should be able to build the workloads however they like I would recommend the use of open source methods with workflows in Python I think Python. There's a lot of great resources right now. I would recommend leveraging the strengths of the geostatistical library very robust very well documented example parameters. Everything's available so and I would recommend if you're working in Python to work in Jupiter notebooks so geo stats pie is a set of Python functions for most of the required workflow steps that support you in doing that. It includes visualization. The light numerix have also been reimplemented. There are more heavy numerix. The kind of heavier programs like simulation spatial estimation and so forth i have used the wrappers kind of wrappers in order to run the gist live executables and so the wrapping method is very simple it writes out a parameter file. The data that's required by the geo slide program runs the geo. Slive executable and of reads in the results very simple so this is the method I suggest that you work with the Python geo stats pi. That is wrapping and reimplemented base. Egeus live if any of these steps of anything that you're doing seems ownerís at any time. Then let me know all says none of this should be terrible now let me make a couple of comments about some of those components of my recommendation. Jupiter notebooks it really is it's super cool it's a web-based application where you have a set of blocks block block block and so each one of the blocks can include code documentation or results and so this is documentation this is code documentation code result and so forth you can also work with a variety of kernels in fact when I say Jupiter it does not mean that you have to work with Python you could actually download and work with an our kernel SC kernel JavaScript kernel or any other kernel and you could use Jupiter notebooks with those methods in fact I have a bunch of workflows examples.
Interpreter notebooks with our kernel because a lot of my students preferred the Jupiter notebook to our studio and they like that now. Once you're in the business of doing trip or notebooks there's so many cool things you can do. You can produce very professional-looking workflows. The documentation is in markdown. So you can actually put equations links you can make it look very cool you can in fact use containers and run it on long online docker. You could put this entire workflow online and other people could run it in a container so they don't have to set up the environment they can just run it online experiment and immediately see results. So it's all very powerful now if you're going to work with Python you don't need to know a lot you don't have to be really in-depth in the Python but you want to be able to work with the basic elements the basic objects that are created as part of a subsurface modeling workflow and what are those basic objects. Well they would include gridded data like seismic data and they would include tabulated data like the well data. And so you'd want to learn about numpy which is package that's very good for doing all kinds of numerics but also for working with arrays of data like a gridded data set pandas package provides excellent methodologies the data frames for working with tables of data like we'd help from the well data now. I have tutorials available for the basics or the things that you need to know for arrays gridded data for data frames well data in order to build subsurface modeling work. I've made these tutorials I put them on github and so they are. Jupiter notebooks with many of the steps that you would need in order to build a subsurface modeling workflow. Now if there's an interest from the students in the class we could of course cover this in class. We could spend a couple lectures if most students are working in.
Python in order to do this we could even do a lecture outside of class that people are interested in that but what is geo stats pi. Well it's something. I've been coding. It's a set of functions in. Python these functions are the ones you need for most of the steps in building a subsurface model now if they were visualization or light numeric Syre coded them in Python and so there will be functions available to you like location map all I did was use matplotlib and the standard geostatistical library types of parameters and so people can quickly build location maps if it's something like producing a Vera Graham a very ground function plot I have reimplemented Varg plot indirectly in Python. So people can do that. And that's just using that plot lib again and so those are two examples right there. Heavy numerix such as spatial estimation simulation. I wrapped as I mentioned before the GS slide wrapping is very simple as I mentioned before it writes out the parameter file it writes up the data I have a bunch of functions included in geo stats PI that are able to convert between the geo es us. EAS format and in fact the format of say data frames or India arrays so you can move back and forth between geo slide and Python with your datasets and so all of that is available to you. I've written this. I am constantly adding to it. I put it on on github and so we can even augment it to help support the students with maybe some of the more tricky steps within the project. What's expected from you you're going to work on a term long project. Updates will be due every two weeks. The updates will include oral presentation on the day that they're due only about five minutes a very brief very concise. And then we'll have two minutes of questions and we move to the next group. PowerPoint template is going to be is provided to you on canvas written update. This is very very concise. Here's a template right here. You'll see that it works out to be maybe two pages and total width figures.
It's brief it's concise should not be onerous use efficient creative figures to communicate. Don't just provide us with a document or me with a document that's going to be just pages and pages of figures. Think about how you can be very efficient with combining figures communicating very effectively efficiently and. I also have a word document on canvas for you. Here's the PowerPoint presentation update template very simple. It's just the summary executive summary workflow results discussion conclusions. It should be literally 5 slides in addition to the title slide the final project. It's not that bad it's going to be a compilation of all of the updates with the requested revisions that there were visions that I requested and there will be from based on the discussion from the class and based on my input. There will be chances for you to improve on your work. Those will be expected the documentation a bit enhanced. And then there'll be an additional challenge of a decision making exercise that we will put into it and you'll be expected to complete prior to handing in the final project academic dishonesty let me say we are simulating a corporate environment. We all win together. Collaboration is completely encouraged. Consult with each other. I want you to do that. I want you older to learn together as a group. It's a real-world type of situation. We have a network we're connected to each other you might be modeling in 10. G's and you might talk to somebody else in the world. Who's modeling another carbonate reservoir or you might be modeling a Jack and saint-malo and then you're talking to somebody else who is modeling over in block 14 and Angola you if you you would be talking to each other. Each of your data sets are completely different. The truth models are very different. So if you won't be able to help each other with accuracy but accurate he's not the not the challenge here right to build defendable workflows that use best practice is the Challenger but each update final project must be unique work of each student.
If we have earth the class is large we may decide to form groups in which case will be the unique work of the group assignments hardcopy of the update documents or will be due at the start of the class. Presentations are given and graded in class. Contact me if you're going to be absent or if you have any other type of emergency late. Updates and presentations will be accepted with reasonable excuses. Illness family emergencies interviews conferences. I support the fact that you have many important other things you're working on as far as in grad students in some cases undergrads and I get that disabilities of course if you have any accommodations come see me. I am all about supporting and helping anybody I want to include. I want to ensure that everybody is comfortable and able to enjoy and benefit from this class. Classroom environment pause a positive professional conduct. I think that's a great way to describe it. We're simulating a subsurface asset team. Meeting doesn't mean we're all like super gentle on each other. It doesn't mean we don't ask hard questions. In fact. The general tone should be searching and probing questions but it should be very much like we're trying to encourage each other we're trying to maximize each other's impact on the project. We're trying to be collaborative. We're trying to win together. Hard questions can help everyone learn. That can help us win together. If somebody's done something that you think they have an opportunity to improve. Make the statement. It's totally cool. That's all good. We did that inside of industry actively participating in supporting a dialogue be enthusiastic avoid negative behaviors anything that's disruptive negative. Disrespectful comments of course would not be okay. Anything that's unfocused. No use of electronics other than those needed to participate in the class. Cellphones will not be out laptops if they're out only if they're used to say follow along a lecture and so forth now for any other type use checking the news sports scores or anything like that personal attacks of course would not be appropriate at any time.
Presentations must not run over time. And that's part of professional conduct to imagine everybody in your audience will have a lot of responsibilities and competing pressures on their time. They will lose interest. And what you're saying if they realize you're running late and they're uncertain about when they'll be able to escape. Some personality is fine but be concisely. Be on message or this class. We're going to teach the reticle in practical reservoir modeling and we're going to do it with a lot of hands-on experience and opportunity to simulate what it's like with in industry or than a company to be conducting and subsurface modeling exercise and so now it begins of course anybody who's watching or listening along here at any time you're welcome to send me questions. I am Michael perch I am a associate professor at University of Texas at Austin I am also the geo stats guy on Twitter and I have all my lectures juiced at sky lectures on YouTube and a lot of my workflows and methodologies code are available on github under geo stats guy - alright thank you.