EGU WEBINARS: Topographic Analysis Using TopoToolbox in MATLAB And Python - Session 1

Welcome everybody. Um I'm really amazed to see that many participants which is great. Um so welcome everybody and good afternoon to those who are in Europe I guess in Africa. Good evening or good night to those in Asia and good morning to those in in the US. So I think we probably have people from all over the world. um which once again underlines uh how cool it is to to have this webinar um so that people can uh participate uh globally. So this webinar uh is hosted by EGU. So thanks foremost to the European Geoscience Union uh for um hosting us and and giving us the opportunity to give this webinar about topographic analysis using topo toolbox in MATLAB and Python. So just to check in and to see um whether you have actually participated in the right webinar, here are a couple of questions that you might ask yourself. And if you can answer any of those with yes or you can even answer all of them with a yes, you're completely right here. So do you want to develop your skills in topographic analysis? Do you want to learn how to address scientific problems by using Toper Toolbox? Do you want to discover new functionalities in Toper Toolbox 3? in MATLAB and Python. Do you want to learn from examples of advanced terrain analysis? And finally, and perhaps also most important, do you want to know how to contribute to the development of top toolbox? As I said, if you say yes here, uh you are more than welcome and you are completely right. What are you going to see in this webinar? Um well, what we are going to do is that we'll present new functionalities in top of toolbox. We'll show how one can contribute irrespective of programming language and programming skills. And we also want to know from you how you might want to contribute and uh at the end I think this should be also taken as a great opportunity to bring the international geomorphology community together. Now, I've mentioned quite a few times the word we. Here's a small gallery of who has actually instigated and uh who's now leading uh this little webinar. Um the names were just here before, but Boris, there's me, Will, Bastian, and D. Okay. So, what are you going to expect in that webinar which is going to last 2 days. So we have a session today which goes from 4 to 6 PM central European summer time. Um today we will briefly introduce the webinar aims and scopes. That's what I'm doing right now. Uh then we will have an introduction to topolbox in MATLAB. Then we will see how we can do similar stuff or actually the same stuff using Python. And then we will have some time for answering questions and answers. Um I'll get back to that just in a second. And then we will have or we will cover a bit of a more advanced topic which is about modeling nickpoint locations using top toolbox in MATLAB. Um at the end of today's session we will talk about the future of top toolbox. Then on the second day, so tomorrow again from 4 to 6:00 p.m. Central European summer time, we will briefly recap on what we have done today or yesterday in that case. Um and then Boris will give an introduction to graphlet which is a simplified hydrodnamics toolbox in top of toolbox or software in top toolbox. Um Bastia will talk about transact which is also a kind of add-on uh to interlink different landforms and analyze them. Um then we'll will show um how to contribute to top toolbox. So give an introduction how you can actually do that and then again we will have a discussion and then finally conclude that webinar. So this is what you're going to expect in the next two days. So this webinar is about topographic analysis and know one of the main questions is why do we actually want to do terrain analysis and why and and what is the role of toper toolbox here. Well very simple and very general we could say that we want to understand how the earth system works or even that of other planets and topography gives us many clues about it. So for example, think of the dynamics of tectonic faults or perhaps sea level changes. These changes uh leave an imprint on the river network and in topography. That's something that we can infer then from river networks data. Think of landslides and the occurrence of landslides. Topography is a first order yeah predictor to or that helps us to predict where landslides occur. So in the end also many models and simulations rely on topographic data uh think of hydrodnamic models, hydraological models um and some of these models require pre-processing. All that is done using topographic software. The other incentive to work with terrain analysis is that terrain data has become really ab abundant in the last years. Not everywhere but in many countries of the world we now have lighter data for example and we have never had as much and as detailed topographic data as we have today. So there's a good incentive to harness that data to learn more about how the earth system works. Now top toolbox is a software that should support that process. Okay. And perhaps some of you might ask, well, we have GIS for that. And of course, there are a lot of, you know, yeah, overlappingings between what topo toolbox can do as a terrain analysis software and what you can do in GIS. But I think one of the main distinctions is that top toolbox is centered around a model based approach to the analysis of DMS. So we think in terms of or often things in terms of models like the stream power incision model um that guide our analysis and that guide our statistical or machine learning workflows that we use. Um top of toolbox is also at least if you can do some programming and I hope you can um otherwise I encourage you to learn that um is for prototyping new algorithms and workflows. So you're in a relatively easy space, let's say, where you can develop your own stuff and try it out. Finally, toolbox is also a nice way to visualize DMs and uh any related results that have something to do with terrain analysis. So for these purposes, Topo Toolbox has served the community as a MATLAB based software for now for uh 15 years. That's quite a while. So, we're quite proud to now introduce um let's say the the latest version, Topo Toolbox 3. Okay. Um so we were lucky to receive funding from the German Research Foundation which I really like to acknowledge here also uh and and thank you for that uh for that funding um for conducting a a project which was called topper toolbox 3 improving the quality and reuse of research software for terrain analysis. Um the main aims of the project were to bring Topopper T toolbox to a stage where it is fair. Okay, fair in the software sense. Um so following the fair principles, it should be findable, it should be accessible, interoperable and reusable and well one of the main aims was here with to make it also available in other programming languages like Python and R. and also to more involve the community in developing and maintaining topper toolbox um to make Topo Toolbox um a sustainable software for the decade to come. Okay. So for top toolbox users who have used the MATLAB version 2 uh which has been around for now 10 years or so even longer um if you have already switched to top toolbox 3 you might not notice many differences. Okay, that exactly was our aim to keep that backward compatibility. But you might have noticed or not um that there had been happening a lot under the hood. Okay. So, Will Kier who's here today also um is uh was or put a lot of effort into you know rewriting parts of the core algorithms um as an own C library which is called liptoper toolbox. Okay. You see that in the center of that figure here and top toolbox 3 the matlab version basically interfaces to that library uh using the max functionalities so which allows to run uh C functions from within uh top toolbox. The same is basically happening in Pytop toolbox which is the Python version um which yeah already attains a yeah high maturity in terms of what it can also deal or have uh in terms of functionalities like the metup version. Um it also has a binding to lip topper toolbox but also of course interfaces to rest or io sci numpy and other libraries and in fact developing that core uh makes it also possible to create bindings to QGIS or uh R. Okay. So whatever your favorite programming language is or your favorite GIS there are possibilities now to use uh toolbox. Now the challenges when we did that is of course that you know a lot of algorithms exist already a lot of libraries exist um a lot of packages in Python that do some of the stuff that we are doing but you know we always were walking on the knives edge between reinventing the wheel okay no we don't want that but on the other hand avoiding that we rely too much on existing um Yeah. Uh software architecture that we cannot get really control of. And if some of these software might break, then it might tear down uh all the other packages that rely on it. And that's something that has to do with uh software sustainability to make sure that you're only dependent on other software packages to some um reasonable degree. The other challenge is community involvement. Okay, many of you are users, many of you code. Um, but I don't know how many of you have actually thought about contributing to their code uh to top toolbox. So far, Toolbox was developed and maintained by a rather small team. Okay, but as you can see, if you have, you know, the same tools in different languages and so forth, it is a hard work also uh to to do that. And uh so we hope for more community involvement um because it is required FTT shell exist in other programming languages. But of course there's also involvement possible when it comes to documentation to testing stuff like a book project. Um and we offered a couple of opportunities in the last month and two years. A hackathon. We offered courses at conferences. uh an online workshop webinar as we have it today or uh something like top toolbox gallery which I'd like to particularly mention. The other uh thing that we wanted to do is also to provide a space where people can actually link their software or simulations tools to okay and that is something that has happened over the last years. So we have created even though we haven't made available yet TTLM 3D so a landscape evolution model um built on top of toolbox uh ingesting 3D lithologies. There is also TT mini which is a um landslide a dynamic landslide model written by Stefan Haggarden and which we provided an interface now to there's graphlet uh Boris will talk about that more tomorrow uh which is a hydrodnamic model and there's a tool called bankful mapper created by Michaela deiavo uh from the university of Roma wand um who or which provides tools for deter determining the bankful discharge in uh or based on uh higher resolution DMs. So I can only encourage you check out our GitHub organization to learn more about these tools and to get an overview. But I know that it can be a bit you know intimidating sometimes to see all the stuff there. Uh so uh we provide guidance here. So the other thing that I mentioned is the top toolbox gallery and that is something where I hope that people will contribute to in the future. Again we will talk about that later on. Um it gives you an opportunity to showcase your own code um online together with a lot of other examples and people can learn from that and you can showcase your study. Okay. So this is just a very brief introduction at the beginning. Um now before we head on we would like to know a little bit more about you and how you use Topper Toolbox and to do that we have prepared a small survey and so if you have your smartphone with you just next to you uh please scan um this um QR code and you will be brought to um a Google survey. Um, alternatively, D perhaps you can also if you have the link with you because it's not the link is not on here but perhaps you can post the link in the chat. Uh, so for people who don't have um a smartphone just at the fingertips. Okay. So we can can go ahead. Um, I think it doesn't take much more than a minute or a minute, minute and 30 seconds and we're waiting for your answers to come in and I'll just continue talking during real waiting. Okay, I think all of you might have scanned the QR code. If not, let me know. Okay, so I'm going to go to the next slide. Um and hoping that you have all scanned the QR code and were able to fill in the survey. Okay. >> The link is also in the chat so you can continue I think. >> Okay. >> Right. >> Okay. So um what up next? Uh the next thing is uh introducing you to toper toolbox in MATLAB. Okay. And um while we are doing that uh just a couple of notes. Uh perhaps I should have said that earlier but please make sure that while we're doing the presentations um um that the microphone is muted. I think we have agreed on rather that uh the microphones shall be muted like all the time because um this is a webinar and which will be recorded and will be made available on YouTube. Um so um if you have questions please write them in the chat. Uh you can do that anytime. Um and there will be a dedicated Q&A session um in the midst of this session today where we will address your questions then we will moderate those um a little bit so that it remaining exciting and not too much overlap within the questions. Um okay okay so um the exercise or the let's say modeling uh programming uh demonstration let's say takes us to the Rome plateau on the Colorado plateau in in the US um and what we are going to show is how we can quantify rates of landscape rejuvenation by flu wheel incision and nickpoint migration in response to base level drop. Okay. So, we are going here to that area that is marked here with that um white rectangle. And what you can see is basically a landscape that is heavily insized by canyons. And these canyons, you know, they um yeah, inhabit not the entire basin as you can see, but you can see that the basin consists also to a big part uh of some old remnant paleo surface. And where those rivers starting on the paleio surface basically go into the canyons that's where nick points are located. So nick points are convexities along the river profile. And we will take those convexities or nick points to study the dynamics of this incisional process that has basically rejuvenated that landscape. Okay. The demonstration will guide you through you know the use of topper toolbox and what we plan to do is in particular in that first phase to focus on three numerical objects that are part of top toolbox that's the grid object the flow object and the stream object at a later stage we will also talk about PPS which is about point patterns on stream networks okay now let me switch my um window. Okay, let me just so Okay, so you should all see um Matllet um and yeah, before we start actually let me say a few words on how to actually well download um Tupper toolbox and how to install it. Okay, um it has never been as easy as it is now. Okay, so where you should go is GitHuber toolbox. So the topo toolbox organization which is there to you know to organize all the different repositories uh within um the topo toolbox realm let's say um to install top toolbox the matlip version go on top ofbox 3 which then brings you to the repository and once you're here go on release we see that the most recent release is pretty young just reflecting that we did a couple of changes right before the webinar and then you can go down and then choose uh your operating system whether you have Linux, a Mac or Windows and there's also a version that is means um no lip tt which just means that there's no lip topper toolbox with it. This is something that you might want to use if you are using topper toolbox in MATLAB online because the other ones won't work with MATLAB online. Okay, once you're here, download your the toolbox of your choice and then you know you can directly open it. I won't do that. Um but it will be opened in MATLAB and then it's installed and there's nothing else you need to do. Okay, once you have done that you can check whether you know whether you have uh or whether everything is uh correctly installed um by just typing grid objects. Okay. And you can see okay since you have some order completion here it seems to be available. Good. If that is the case great uh you can also check has lip topper toolbox and if it says true which it should then that also tells you well even the binaries have been successfully installed um or they are available um and so we can get started. Good. Okay. So what I'm going to do now is to to walk you through the code um and give my explanations from time to time. I will also just live code uh not all of it is live coded simply because it would take too much time to do that uh for all of the functions in particular when it comes to plotting. Okay. So first of all your analysis probably starts with a DM uh that you need to download from somewhere. Okay. um that somewhere is best expressed in terms of coordinates. Um so for example like these four coordinates from east uh west to east and from south to north um which can be then put into the function which is called read openo. So what read openo does is it uses the API of open topography to enable downloading all kinds of different DMs. Here we chose the DM type COP 30, but you can also use other ones. And if you remove that, you know, you get a list of all the different DMs that you can take. COP 30 is a fairly good DM um which we will use here. Okay. Now, we'll run that code. And what you see here then is a message that either tells you that now your data is downloading or like in my case uh well that the downloaded DM is not in protected coordinate system and so forth. What has happened here is that I have downloaded that data before also to make sure that it's there and it's in the cage in a top toolbox cage. Uh so if you run that code again it will just take um the DM from your hard drive rather than downloading it again. Your message will be a little bit different and also what you need if you run that is an API key for open topography u otherwise um you will get an error but there will be a message that tells you enter the API key here um to go forward. Now what we have is DM that's a grid object. Okay. Now this DM comes in geographic coordinates. Okay. That's how open topography distributes its data. Um but for most of the analysis that we want to do uh we will rather need um a DM in a projected coordinate system. So we are going to do that using for example the function repro to UTM. Okay. So that's a fairly easy function which just requires the downloaded DEM and the only thing you need to yeah provide is um that the output resolution 30 meter in that case um and it will then project your data into a uh universal transverse macado projection. Okay, so this works. And if you plot your DM, I'm going to do that here using the function image SCHS, a function which is quite nice to, you know, to plot a hillshade. It just means image and hill shading, a scaled image and hill shading. Um, you can see we have our DM here. Um, the only thing that bothers a little bit is these wide edges here. um which come uh by the transformation into another coordinate system. We can um remove those using a command uh which is called largest inscribed grid. Okay. So what the function does it tries to find the largest grid within your uh DM that doesn't have any missing values. Those are the these white areas. So let's do this again and then we can plot our data image ss and the result will be this one here. Okay. So we have removed the edges and we have um a DM which is in a right coordinate system to go on. The next step is then having that DM to derive the stream network and to do that uh we need two steps. The first one is to derive a flow object. That flow object is derived from the DM itself. Um and contains all the information required uh to infer flow directions. Okay. By default, the flow direction algorithm used here is a single flow direction or D8 flow direction meaning that we have always converging flows no diverging flows. Okay. So FD you can see it in your workspace that you now have FD as an additional variable which is called a flow object. It's basically the flow direction grid that you would have in any GIS but it has a better way to store the flow directions and so it's much better and faster that to use this one. Based on the flow object you can derive a stream object. So what is a stream object? Basically it's just a sub network of the flow object. So meaning that well we just want to have the flow network for those parts of the landscape where you might have constantly flowing water for example and that just happens or permanently uh flowing water and that happens let's say where the um upstream area exceeds a certain threshold. In that case we can just say if it exceeds for example a thousand pixels that flow together that's where basically streams begin. Okay. So we are going to plot this stream network. I'm going to going to do that in the command window. Okay. So we have a nice representation of uh the river network in our area. Now let me go back. What we want to do and what we want to focus on is that catchment here in the middle. Okay, I hope you can see my mouse uh that shows it. And basically, you know, from from that point on, we're not that much interested in that part here. That's the Colorado River that is also not completely here covered by that. Um the we're just interested in in this basin. Okay. to do that there are different approaches but I'm going to show a little or one here which is actually quite uh cool because it's completely programmatically uh I will also provide code that shows how to do that more interactively but we can just the first thing that we are going to do is to extract the large com connected component which means we are extracting the largest drainage basin and then in the second step well let's do that and I give you an intermediate result so that you can see what has happened. Okay, now all the small basins have been removed and we still have that one and then we just want to have that one here. Okay, upstream of that junction. Trick that we are using here is to identify that part of the network that is somehow connected to the edges. So where some of the streams are just you know have upstream areas that are directly connected to the edges. And so this is what we are going to do here. um remove edge effects like this. Uhhuh. Okay. Now, we already have this stream network here, but a bunch of other ones which are probably not connected to the edges. The only thing we just need to do now is to run K largest concoms again. Okay. And here we go. So, that's our network that we want to look at. Um as I said before, there are other ways to approach this problem. Um but um for the reason of times, we I won't do it here. This is also not a method that works with every DM for sure, but it should show you that by combining different tools, for example, also the function modify that you have a lot of possibilities to programmatically uh modify your stream network. Now we can we create a map. Okay. And to see whether we have done everything right, um me just um I have always this this bar in front of me. Okay. So, so we identify in the next step the drainage basins based on our stream network that we identified and then there's some additional plotting stuff uh that together then gives you a relatively quick map. The next thing that we are interested in is to plot the river profile. This can be done using the command plot D set. So we take our stream network and we take elevations which are not stored in the stream network itself. Uh but we take those from the DM and see our flow or our uh representation of the uh distance versus elevation. Okay, great. And what we can see is okay we have a main trunk river that you know once it goes into its tributaries these tributaries go pretty steeply up but then they transition into more flatter sections or less inclined sections uh which basically represent these uh paleo surfaces or these rel surfaces. And you can see that those relict surfaces are basically separated from uh those canyons uh by nickpoints. Can see those here. We could identify them by hand. But what I'm going to show next is how to use the so-called nickpoint finder. Okay, I'm going to use it like this. Nickpoint binder takes the s the DM and the stream object and then uh what we need to define is a tolerance. This tolerance can be set fairly high here 200 m because those channels are actually so steep in size by 500 to 700 m that this is a we can really have a fairly large tolerance. In other instances, you should choose a smaller one because otherwise you don't get any nick points. And we just say split is false because otherwise um it only makes sense to make split uh is true if you use your um if you have the parallel computing toolbox and if you have multiple basins. Okay, so you get an idea how that algorithm works. Now it is already finished um and it has identified 36 nick points. Let me close this again and we can plot the results. That's what it is. Okay, so we have our river profile with the identified nickpoints and these nickpoints are here also shown in map view um showing yeah where we have our nickpoints and I think we quite reliably detected those. Okay, the next step that we want to do and I yeah need to come to an end uh just in a second at least for that first part is Kai analysis. When we analyze uh nickpick points and I'm going to talk about that in more detail later on. Kai analysis is extremely helpful in identifying the mechanism behind nickpoint formation. Um here I'm just showing how to compute it and I will give a few more sentences or provide a few more sentences later on about what is actually done by that function. So kai transform um s and a what is a? Well we need to compute it before we have it here. A is upstream area a equals upstream area which is calculated by the function flow act which is flow accumulation and then we do our kite transformation um providing some additional input here n ratio of 0.45 45 as I said before I'm going to say a few more words about this later on and a Z which is a reference area which is often set to 1 million so 1 million cubic uh square meters okay the only thing that we now need to do is to plot it to receive something which is called the kai map okay and what I'm going to do to make it look nicer is to have a underlying uh hillshade color map. Yeah, I just say it should just be white. Okay, so it just has just shades of gray. Um, and in that case, we don't need a color bar. Hold on to make sure that you're plotting on the same axis that we had before. Let me just get a figure before. And then plot C. And plot C is a function which plots the map layout of a stream network, but gives some additional coloring. And in that case we are going to color the network uh using the kai values. Finally we add a color bar and say okay it's label string should be kai and kai has the unit meter. Okay let's run this part. And here we go. Here we have our first kmat. The only thing that is missing is the nick points itself. So we add those as well. plot kpx and kp y. So this is what is contained in this nickpoint uh data structure returned by nickpoint finder um okay which just means we have circles and uh which are plotted in black. Now let's run it again and here we go. So this is from my side so far and I will hand over to will show exactly the same thing and how it works in Python. Okay, I'm stopping here and hand over to you, Will. >> All right. Um, so yes, hello everyone. I'm Will. Uh, and I will be going through basically the exact same example that Wolffgun just showed. Um, but using PI Topa Toolbox, which is our Py new Python implementation of the Top Tool toolbox interface. Um, we you've heard us mention the Tupper toolbox gallery a few times. So, um, I hope you can see that my screen here. This is our our gallery web page. uh topple toolbox.github.iogallery. I believe I posted the link in the chat earlier. Um and you can find a bunch of examples that we've created for doing um using tpple toolbox in both mat lab and python here including the um mat lab live script that wolf gang just presented this fluial nitpick points on brain plateau mat lab version and there's also this python jupitter notebook which is what I will go through um now but uh first I will show uh the process by which you can acquire pyle toolbox um it's a little bit more complicated than than mat lab partly because there are a few um different ways you can install packages in python. Um I'll show you a simple way that uh we know works. Um if you are a python user and you have other ways of installing packages and feel free to try those out if your usual way of installing pytopics doesn't work uh please let me know um open an issue on the GitHub repository. We'll talk more about how all of that works later. Um, and we'll uh try and fix it so that it does work. We've had issues for instance with um installing packages with Anaconda particularly on Windows machines in the past. Um, so if you're an Anaconda user and you have a Windows machine, uh, please try it out and let me know how that works. Um, so now you should see a terminal window. um uh here uh so I have already prepared a um virtual environment um a Python virtual environment for this webinar. So I'm going to go ahead and activate that one. Um this has some of the the uh other dependencies that we need um already installed on it. So we don't take too much time doing that. And I'm just going to install the latest version of top toolbox using pip um pip install top ofbox. And this will work um where you it'll go through uh install uh bunch of packages that are required by top of Toolbox um to function and you see toolbox version 0.0.12 is the most recently released um one and you'll need that to run the um ren plateau uh Jupyter notebook because it's got some uh new functionality that we recently added there. So now having that um installed uh I can open my Jupyter notebook um interface. Uh as I said I've already installed Jupyter notebook in this environment. Um you would have to install that separately or have a um you know copy of it somewhere. Um so uh you can see I've uh put this down. Hopefully you can still see my um Jupyter notebook window here. um let me know in the chat if anything goes wrong there. Um this is my directory with all of the uh Toolbox code in it. So I'm going to go into our gallery repository. Um so uh excuse me double click in this uh runto. You can see the path here. Um and this is where you can find the the notebooks within the gallery repository if you're looking for them. Um, and this Rome Plateau one is the one that we have published on the gallery web page. Um, I've created this other one that has some of the code taken out so that I can, um, type it in for you here. Um, so, uh, if you're not familiar, this is the Jupyter notebook interface. Um, uh, as with the Mat Lab live scripts, it it's kind of a computational notebook. It allows you to, um, run code text as well. Um so the first thing we need to do when we start using PI top toolbox is to um import some packages uh including top toolbox itself. Um so in Python we do import tutorial box. Um I usually call it TT3 because top toolbox is a bit much to type um over and over again. Uh and then I'm also going to import excuse me uh map plot lib pipot uh map plot lib is included with the top of toolbox. It's a dependency of top toolbox. It'll get installed when you um install the python toolbox. Um and we use this for plotting um the kind of default plotting uh system within top toolbox basically net plot web. Um we're going to import uh some more stuff from net plot web. So we can do some uh hill shading as we saw before. Plotting in Python as we'll see is a little bit more complicated than it is in mat lab. We'll use numpy which again is a dependency top toolbox um for array computing and raster io is what we use um by default to load dems. Um so it's also included as a dependency and we're going to import the coordinate reference systems from there because we'll use that to do the reprojection later. Um so these are kind of the packages that I import when I start running top of toolbox. Um you can do everything um more or less with just the top of toolbox import. Um and you can use the top toolbox functions for plotting and everything but to make nicer looking plots or to do some uh more sophisticated analysis you might need some other um of our dependent packages available to you in the um your environment. So, as before, we're going to load this um DEM. I'm going to uh probably not talk so much about the scientific parts of this analysis and more about the differences that you'll encounter when using Python top toolbox in Python from using mat lab. Um and some quirks of our implementation, but uh because every all the analysis that we're going to present is basically the same as what we did before. Um so, we're going to load this from the open topography API. Um the first difference that we'll encounter between Python and Mat Lab is that the function we use to do that is called load open topography not read open top of so a lot of the functions in the Python version have different names for various reasons. Um uh it's something you will have to figure out. Um you can look at the documentation for instance that has listings of all these functions and um hopefully that'll figure it out. That'll help them figure it out. also the examples um uh of that are provided in the gallery are good resource for learning what all the the functions might be called in um Python and then we'll supply instead of the extent um vector uh we'll supply the coordinates as keyword arguments with um this right okay now we have the dem type is 30 So otherwise it behaves pretty similar to um the mat lab version of read open. So we'll load that and um as before this uh as wolf going showed the um I've already loaded this DM from the API before so it's cached on my computer. Um if you don't have it cached and you have an API key this will download it and it might take a few seconds to download. Um we will uh do our reprojection um as before and for this we use a function it's just called repro and we give it a coordinate system um specification. So the easiest way to do this is from um using the CRS uh object from raster io and then here we'll just give it the EPSG um code for the correct UTM zone which is in 12 north um the mat lab reproje to UTM has some logic that uh figures that out for you but that's something we haven't implemented in Python yet and specify resolution and then we can visualize this DM with plot HS. So this plot HS method is the um equivalent of CHS from MATLAB. It'll plot a cool shaded raster with the color map um of it. You'll notice here that uh the methods on grid objects are called by um taking the name of the grid object dot the method. This is Python syntax which you may be familiar with. Um whereas in mat lab you pass the grid object as the first argument to all of these things. So most topics functions are methods on one of our objects, a grid object, a flow object or stream object. Um in mat lab the first argument of all those functions is that object and that's the one that you call um the method on in this um is just a good way to so we visualize it and there we go. So this is our DM unloaded um and we see our the missing data along the edges that we get from the loop projection. Now in mat lab we use largest inscribed grid to um uh fix that. But you'll notice if you try largest grid that we don't have that implemented in Python yet. So there's lots and lots of functions in the mat lab version of top toolbox. Not all of them have been completely ported over to Python yet. Um this is one of those. it's a little complicated to um get right. So, it hasn't been a high priority. Um there is an issue on GitHub and there's a link in the notebook that you can use to um check that out. If you're interested in contributing, that's a great place to start. You can look at the MATLAB version and try and port it over to Python for us. Um but so instead, I'll just use uh our crop function um prop. And I've gone through and found which pixel indices are the correct ones here. So you would have to do this manually. Um otherwise kind of looking at the DEM. Um this is it's a little annoying but it is what it is at the moment. As I said if you're interested in contributing at large disc but we'll crop it. And there we are. I cropped the um with nodes and data. All right. So having done the uh prepared the DEM for analysis we can now identify our stream network. This is pretty similar to how it's done in mat lab. We have a um a flow object constructor that would take a DEM. Note that instead of being called flow obtain is one of the naming differences. Um and uh that will create our flow directions object there. And then we can be our stream object the same way the passing flow object and the threshold of passing. So that will do our flow routing and um identify the the streams that are considered to have kind of perate flow in this um system. Now we can plot um the stream network on our DEM as we did before. Uh, as I said before, plotting in in Pytopex is a little more complicated. It's a lot more explicit usually than what it is in in mat lab where you're able to just kind of call the plot methods. Um, you kind have to interact a little bit more with mplot lib to use correctly. So I'll create a a subplot. This is using the piplot function subplots. So create a a plotting window and then we'll plot into that so we can get the hillshade grass plotted. um into that axis we pass the axe the plumbing access as the ax implement there and then we stream on top of that. Um there this is our stream object uh as identified um by our constructor. Uh now we want to select our parachute creek catchment from the center of this DEM and we'll do this exactly the same way as we did in uh mat lab using the uh k logistic components method which is a method on stream objects. propended with the S do a components. We then will do remove edge effects with the flow directions object flow object there and then select the um low disconnected component um outputs remains in that. We'll run that and now we will plot our nice little plot of the parachute creek drainage basin. basically identify the drainage basins using the drainage basins method fill object. And instead of passing the um stream object, we have this outlets argument which requires you to do a little bit of um array manipulation to figure out where the outlets are. I'll use the stream POI function here to select the outlets of the single basin in our streamlin. And then this we index into the stream array to convert that um the index that this function returns into an index into our uh DM as a um which is a little complicated but uh you you'll get used to it as you start playing this. But we created drainage basins. And now we will plot our um the drainage basins as a hillshade. Um so here this may be a little confusing, but we're going to plot the hillshade. The the data that we're going to plot is the drainage basin. So that's why it's the the one that's called um on the method. And we pass the DM as the L of argument to actually do the full shading rather than to the axis. Um and uh this is just to set up a nice color map for that show. Um plot the stream object on top of that. And in the second axis we'll plot the screen profiles with the plot dz. Make sure that basically the same as it is now work. Yes. Um so there we have our drainage basin with the stream network identified and the stream profile. Um you can play around this plot to adjust the um the layout a bit more. Um this is something that you uh at the moment kind of have to do a little bit on your own in Python. There we go. Our nice stream profiles. We can see our nick points up at the upper reaches of our uh streams. So now we will go to identifying nickpoints and once again this is pretty similar to um mat lab we use the nickpoint finder which is a method on stream object give it the BM and the tolerance um in this case we're using 100 meters the same as we used in the m um and yes uh the the output of the npoint finder is a little different in Python so it's it's a array that tells you where the nickpoints are rather than an array that contains a bunch of information about nip points like their coordinates and things. Um so you have to obtain the um information that you're interested in plotting separately. So here we're going to identify the nickpoints that just tells us the locations um within the stream network. And then here um we'll create an array of the elevations of those nickpoints. So we use the easy get nal function to extract the elevations from the DEM and then um index by the the nick points to pull out the elevations just at those uh nickpoint locations and little plot. I am going to to save time copy and paste this plot from the direct notebook here is very similar to what we just did before. Um we have the hillshade of the DEM uh the plot of the stream network and here we plot the um x and the the nick points but using the x and y coordinate vectors of the stream object. Sx is the x values and sy is the y values of those um every node in the stream network. And so we index by the nick points to um pick out which ones of those are nick points. And then we'll do a string profile again very similarly plotting the um elevations that we extracted in the previous step against the um distance of the upstream distance of the stream network again uh index by the midpoint values there and there we go we've identified all our nickpoints there both on the map link in the stream profile and then finally uh we will do our kai analysis here um again the analysis is very similar to the the mat lab web version. Um we use the flow accumulation function rather than flow. This is another rename and the kai transform giving the flow accumulation the reference area and the mn ratio um assumed to be zero that will compute the kai transform and then uh plotting. So at the moment we don't have a um stream object.plot plot C function which allows you to plot the um uh stream network uh with a color. Um so here I just do a a scatter plot of the um x and y coordinates of the stream object colored by that kai transform value. uh if you're at a large enough scale and you can't really see the dots and they they overlap and it it looks all right. Um here there are more complicated ways if you want to do something more similar to what plots does um in mat lab but there we go our kai map of the parachute creek catchment done in pi top toolbox. So I think that concludes my presentation of this segment. Um, so I thought sharing my screen. >> Thanks, Will. Wolf gang, do you want me to just continue right away? >> That's f. >> Okay, great. Uh so we are more a little more than an hour into our twohour session and uh we've seen examples now and as was written or as I have written in the chat we also um have now something like 25 minutes maybe for question and answers and instead of uh you asking the questions personally we opted for uh typing them into the uh chat window and there are few questions in the chat window so far. I noted them down but then we have a bunch of questions that were typed into the questionnaire that we sent around initially and now I have them on my screen and somewhat sorted. So we want to address these questions one after the other and see how far we get and um right so I will just start away and ask uh Wolf Gang and Will um one user said that um they use Tupo Toolbox 2 and they just started with Tupo Toolbox 3 and is there something to be aware of regarding old files that were built with Tupo Toolbox 2? >> H well I think I can comment on that. Um if you have files that contain let's say uh grid objects and so m files uh containing grid objects or flow objects um these might not be read by uh if you you know if you have top toolbox 3 on your uh path in in matlib because the structure has slightly changed. It's a minor change in fact uh of how we handle transformations. Um it's also a change that you know was necessary to comply with um changes in the mapping toolbox and new functions here. Um so you might face some issues here. Um so best thing is probably to export your DEM from top of toolbox 2 uh to a grid object uh to sorry to a TIFF and then read the TIFF again. Um that should be all right. The other thing is that you know if you if your implementation relies on liptoper toolbox the top toolbox or top toolbox 3 relies on lip topper toolbox. So there might be minor differences in the flow network that it might be somewhat a little bit different than uh the ones obtained in TT2. Um, this is what you might consider, but otherwise I think everything should work as it did in top of Toolbox 2. If it doesn't, let me know. >> Yeah. Yeah. I guess there's not much to add maybe from Will's side or is there? >> Um, no. Yeah. I mean, the the the details about the the liptop toolbox flow network differences are interesting, but uh I think rather technical and not of you, you can find me and talk to me if you're curious. >> Great. Thanks. Uh also um when there are specific questions uh we have the GitHub uh organization for the tupal toolbox and there also uh there are possibilities to ask questions there and to get help right will do you want to say something about this? >> Yes. Yeah. Um so uh there are many ways to uh contact us uh through GitHub which is where most of our um organization for the project is is happening at the moment. So um you can if you have pro specific problems with the software the best way to um you file a bug report essentially is to open an issue on a repository I'll put a link in the chat um to the issues on pi topper toolbox bugs for instance um so uh you know if you have a problem you want to report there if you're thinking more about if you have more of a a general question um about you know top toolbox especially a more scientific questions, you know, what can I do with top of toolbox? We have a discussion forum um which I will also post a link to that's also run through GitHub that is an excellent place to um ask those kind of questions. >> Okay, perfect. Well, then there was another question asked early on and uh the question was about liptop toolbox and even though it's somewhat specific most questions are relatively specific so I'll ask it anyway. Uh does lip top toolbox currently support scale dependent analysis of river platform geometry such as multiscale senosity or identification of characteristic meander wavelengths. Wolf gang will so what can you tell us? >> The short answer is no. Um but it it could there's no reason why it couldn't. Um, and that you know if that's something you're really interested in like uh you know get in touch uh and we can see what kinds of things we might need to do to uh make that available to you. >> Good. Yes, there was also the question or the comment of a person who wants to integrate toolbox with cascade and uh maybe I asked back the question via the chat. uh that person um could you please specify which cascade because I looked it up and there is a cascade landscape evolution model which was developed by Jean Brown. I think that was developed in Fortrron and I also noted there is a coastal erosion model. So maybe you can specify that question and then we at the moment move on to the next question which is about nickpoint detection. Wolf Gang how to estimate the tolerance for nickpoint detection. If it is too high, not all nickpoints are found that were identified in the field. But if the tolerance is too low, too many nick points which uh do not exist in reality. So how to identify the best tolerance value? >> Yes, that's a a good question um that I've received also before. I've actually covered that also in the top toolbox blog uh which is still active and which continues to be active. Um in short um what you need to consider is what is the error in your DEM. Okay. And so the nickpoint tolerance should be at least larger than the kind of average error or actually more like the maximum error that you see in your DM. And then you can do that using a couple of tools. It will be too much uh detail here but you as I said before you will find the description in the top of toolbox blog um I can also send the link later on um uh which describes this like you're creating a band uh around an arrow band around your DM indicating let's say the 10th and the 19th percentile and you know the tolerance should at least be as big as that band or even larger and that uh makes sure that you know you don't just map a lot of uh errors in your uh DM, but rather get the signal behind the noise. >> Okay, great. Uh yeah, or you also mentioned the the block uh which still exists. Um but I think and correct me if I'm wrong, Will, some of the blog uh the posts that you have there are now part of the gallery, right? >> Yes, that's true. Um, and uh, you know, I think we're kind of slowly migrating a lot of the um, you know, more substantive posts from the blog over to the gallery. Um, the blog I I assume will continue to exist um, you know, going into the future. We may we may move some stuff around or or find another hosting solution for that at some point. >> Okay, good. Thanks. So, I have another question um, which is directed maybe towards Boris and Wolf Gang. It's about the feasibility of DMs with different resolutions to use add-ons such as bankful mapper and graph flood especially if these are lighter derived. No sorry especially if lighter data is not available. So how useful is graph flood if there's no lighter data >> right I may answer that one if you can hear me >> yes >> perfect so it's roughly the same answer that uh vulgar point it depends what you are trying technically speaking graph flood works on any type of of resolution and in top paper that just came out as a discussion paper we tried it on a copenic 30 m so it definitely work but your river needs to be wide enough to be described by uh by the model and then there might be a bit of uh physical parameter tuning like friction coefficient and everything. We will talk more about it tomorrow. But yeah, it can work with any resolution as long as you can see the object you're trying to measure. >> Uh W do you want to add anything on bankful mapper or is that um >> well I can only say that here you also need a good representation of the river and its banks uh and which is probably hardly feasible if your DMS is too coarse. Of course, you perhaps you don't need LAR. Um perhaps a 5 m or 8 m resolution model might be good enough, but I can barely imagine that it works with uh the common global DMs. >> Okay, good. Thanks. Uh Wolff, I have another question from you for you. One user would like to know about creating DMs of differences and tools for mometric analysis. I guess in connection to those DMs of differences are there specific functions or functionalities currently because I guess DMs of difference require precise co-registration which is not necessarily always the case. >> Yes, exactly. So there is no let's say dedicated function to or set of functions that let you work with DMs of difference. This is might be a nice add-on perhaps in the future. If you have some good ideas for for what to implement in in top toolbox in whatever language, go ahead please. What can be done of course is just to substract one DM from the other one and to make sure that the underlying geometry is basically so of two DMs is actually the same. Um so interpolating one or resampling one DM to the exact geometry. So same extent and same um pixel centers let's say uh this is uh no big deal using the function resample or repro if they or project if they they come in different uh coordinate systems >> okay good >> well there is another somewhat longer question that I took on myself now it's about many studies using kayen channel steepness index as indicators of landscape at this equilibrium and erosion potential And then how well do these topographic metrics correlate with basin wide erosion rates derived from tenderillum and can they be used as reliable proxies for long-term denodation across different tectonic and climatic settings. I would say this is a scientific question which is very interesting but it's very hard to to answer in the scope of this uh the seminar but in fact I mean this is something where many papers have been published and it's probably useful to get into the details of these sightspecific studies but overall I think there is good evidence that there is some correlation but um it's it's hard to make um you know blunt statements whether this works everywhere and for every setting. So, uh, Wolf Gang, how do you choose a sensible value for the threshold on flow accumulation when creating a stream network, a stream object? >> That's a good question. It's, uh, basically where do channels begin, right? And I have no good answer to that except that, you know, it's often a trial and error, just seeing whether, you know, it works for you. Um, and of course, um, it depends on the DM resolution. um and what kind of channels you want to capture. Um if it's if it's just the main rivers, of course, you choose a larger um cutoff. Um and if you want to go up to the divide, you choose a very low uh cutoff. But at the um you need to be cautious to not map the hill slopes, but there might be cases where you actually want to do that. So it's up to you. It's a trial and error process. there is no good answer and it depends on the climatic setting in which you are and probably also uh even the uh lithology. Yeah, maybe I can add to that. Um that when you plot slope against area, drainage area um from theory we often think that in on hill slopes we have either a constant or even declining slope angles towards smaller drainage areas and only when you have this rollover and then uh approaching the area where slope and area is a linear plot on a lo lot lot then um you are in a in a fluial domain. Uh however this is not always easy to apply to course resolution DMs. But another thing that I think nowadays can be easily done is when you derive a stream object and export it for Google Earth and then plot it in Google Earth, you often have very high resolution uh satellite or aerial images and you can zoom in and you can figure out really whether the streams that you have derived using a certain certain threshold coincide with streams in high resolution satellite imagery or if it's really already on the hill slope. So this is maybe something uh to to do. But it's clear that in like aid settings and weak rock types, you sometimes have drainage areas that are very very small and that are hard to see even in uh satellite imagery. Okay, good. Wolf gang, another question if we have time. Yes. Uh how sensitive are topo toolbox derived KSN values to DEM resolution and stream extraction thresholds particularly in steep Himalayan terrain? >> Mhm. Yeah. um they are very sensitive I would say um but of course it's it's the errors and but but also the natural variability that um is um you know the cause of this uh yeah sensitivity. So um talking about errors um if you work in the high Himalayan region you will have very steep um hill slopes uh which cause that you have a lot of errors along the Talvig um so and your uh you know your river profile will have a lot of wiggles that um you might want to uh filter out. There are some dedicated functions for that like the constraint regular regularized smoothing function CRS or quant. Um these are some functions that are probably helpful to you know to get rid of all these bumps. Um but nevertheless um yeah you will need some additional smoothing to get some meaningful signal. Um why is that the case? Because even if you have um if you would have a perfect representation of the river elevations at a higher resolution, let's say you have lighter data, then you would see some riffle pool or um yeah step pool sequences that actually show you also high variability in river gradients which are the basis to calculate um ken. And even here you might want to integrate over a larger distance. So even here smoothing is required. So both to get rid of some of the natural variability as well as getting rid of DM errors um you will need smoothing. Um and again there is no good answer to say how much of smoothing you need. It's something that you will need to find out by trial and error. >> Yeah maybe I can also add uh I think the question is always also related to how you intend to interpret these values. um measuring something from a DEM and then even calculating a metric like channel and steepness or something is in itself possible anywhere and can be done for different resolution etc. But then again uh the question is how do you intend to use it? Is it for relative comparison among different regions within 1DM? Is it for comparison with let's say tambber data or making inferences about steady state badrock river incision or whatever. So there are many nuances to this and I think this is uh important to keep in mind but there are papers that are dealing with these kind of things and I remember that cam wobbis in 2006 I think had a GSA special paper where they also talked about different resolution maybe not but smoothing etc. So there is some some um published work out there that can be accessed. Maybe another question for Wolf Gang even though um okay so assume like many of the Tobo toolbox functions are um using also the big hunga catchment 30 m resolution DM as as examples but people may come and have their own generated DMs from lighter or photoggramometry and is there anything to keep in mind when using these DMs along with the topo toolbox so what I'm thinking about it. I could imagine that the georreerencing may not exist or maybe um somehow stitched together rather than uh adhering to a certain format or something. Yeah, I mean yes that can happen if you have a if you don't have any coordinate system at all in your let's say uh lighter data even though I think that mostly they have there might be some unexpected errors that might occur since we haven't we have focused more on you know putting functions together that that forward all the geographic information and uh with it or projection information. Uh we have I think we don't really have good test data where that is done even though they I'm not entirely sure but um the other thing that um of course you might need to consider is if you have I mean you should bring your own data obviously uh that lighter data can become quite large um in terms of memory requirements uh just as a rule of thumb here I mean um you can take as big DMs as you nearly want uh as long they as they fit into your main memory. Um so on my 60 GBTE ROM machine I can work with DMs up to 30,000 by 30,000 rows times columns. Okay, so that that's about the the size of the DMs that you can work without. You can even go beyond that. Um but it can get quite slow then. Um yeah >> right uh I think this also pertains to um this question with lighter or structure for motion derives. I could imagine or we once had I think somebody who was using an analog model like a sandbox model scanning so producing a DEM of that but without really any ge referencing and so there it is possible to use just the x and y coordinates uh to put together a grid object with the dm uh sorry the elevations in a matrix but and I think you can also derive flow object stream object and all these kind of things but at some point there might issues with some more specialized functions that um are exporting something or use basically the projection information if that is not available. Okay. Um so there was also more like an internal question of what is the stage of development uh of the Python version with respect to the MATLAB version and I heard that there's also an R version. >> Yeah. Um so I can I could talk through that. you've seen um in the the pi top toolbox demonstration here um we have a fair amount of stuff uh implemented in Python that's also available in mat lab it's not uh anywhere near everything um and actually there's quite a lot that hasn't been implemented um partly because a lot of the stuff in mat lab is either very specialized or um uh is designed to do things that we don't really need to do in python you know a lot of the functions have a lot of built and plotting um mechanisms and stuff that I think um at least the plan for now is just to kind of push that off to Mattplot Lib in Python and say like you know you're going to create your own plot um and add all the information and stuff like that um there. So uh uh it you know you you could try and measure like which functions have have been implemented in both mat lab and python um uh but I don't think that's a very useful um number to to tell you um simply because I don't think we're aiming for a onetoone feature parity there. Um you could do a lot of obviously do the the find the flow direction the flow object and stream object and do a lot of the um stream analysis uh as you've seen here um we have things like Kai we have um the smoothing of stream profiles has all been implemented we had a um very industrious student last summer who um got through a lot of that stuff um the I believe there's a KSN implementation if there isn't there should be um I think we did did some work on that as well last summer. Um so so a lot of the main uh tools are available in Python. If there's something that is available in mat lab and you want to try the Python version out and um it's not yet available there uh do open an issue about that because that's you know um good feedback to know what kinds of things we should be implementing. Um >> yes >> there is also the R version. The R version is much less um further along than the Python one. Um it works. You can load DEMs. You can construct flow objects. I don't think you can properly construct stream objects actually though. Um you could do some analyses like uh compute gradients and access topography. Um I know we have available there. Um so that's definitely a work in progress. Um we would really love more people who like use our um to do actual research to help let us know what that package should look like a little bit more. Um so if if that's you uh please get in touch. We we would love to hear from you and um know what your needs are there so that can help us uh design that a little bit better. >> Perfect. Well, thanks Will. Um before we end this question and answer session uh I just took a quick look at the evaluation of the questionnaire and maybe just could to connect to what will just said is uh I've seen that there are among the different questions that we're that we have asked that was like 40% or more than 40% would like to contribute uh to toolbox but they are not sure how they can then there uh uh onethird of the people are using Python most of the time and then 70% of the attendees are graduate student or postto researchers. So I think there is basically uh the people that are familiar with Python and that want to contribute and that maybe have the time to do so. So that so that is an opportunity to get involved and I think will will u finish this today's session here uh with a quick overview of how this can be done um but I suggest that at the moment we now move on Wolf Gang is that correct to to the second example that you wanted to to present. Yes, exactly. So, uh just one more note. Um if your question was not answered, uh please go to the discussion forum on on GitHub, uh and and we will try what we can do to to answer your questions. Okay. Um Okay. >> And we also have I think we also have time tomorrow for more discussion uh which will obviously pick up content from tomorrow, but it can also pick up content from today. So if there was uh a question that arose arises during the transition from today to tomorrow then we can also pick up that question there. Exactly. Okay. So we are going to continue with um a little bit of of MATLAB coding. Okay. And um now in our previous steps uh we um yeah basically looked at um the the parachute creek at the or on the Rome plateau. And I said a little bit about you know the that this was actually produced by an incisional wave uh cutting through this uh yeah paleo relief because why? because the Colorado River basically insized and that incision commenced 8 million years ago and so this landscape that we see has seen a rejuvenation phase of about 8 million years. Um now this is also shown here in that figure. It comes from a publication that Durk and I uh were working on a couple of years back which was published in geology. Here we uh took a uh yeah bit of a more uh detailed look at the distribution of nickpoints. And uh what we found was that not only you know we could predict where nickpoints should be but also that some of them migrated much faster and much farther than we would expect. This is here in the east fork uh basin and that was due to some uh yeah divide migration or that's what we inferred from that data. We are not going that deep today into the data. Uh but what we are going to do is um to aim to you know to model the distribution of nick points uh and even to simulate it. Okay. Uh here to the right you can also see just a photo so that we just don't have uh just digital data um but that you have an impression how these canyons in the uh RTO basically look like. So these are really massive. Um okay to to get an idea of what I'm doing, I assume that you have heard about the um stream power incision model before because it's the the model of our choice that um and actually probably also the only model that can actually be used to to do what uh we are doing here besides some some other stream power based models. But it's um well what it says is that the river elevations change over time or the rate at which they change um is a function of uplift on the one hand side and on the other hand uh due to incision. So we're just looking at rivers here not at the hill slopes. Okay. And incision is a function of um yeah upstream area which whereas as a function of x distance in that case um to the exponent of m and then there is also um incision is also controlled by slope itself. So the river gradient det over dx and also this one here has often an exponent which is often set to one. Um now the only thing that I want to mention here is if you set this to zero then you are in a steady state um river profile that doesn't mean that you have uplift or so but of course you have uplift but it's perfectly balanced by incision and if you have that uh state um so a kind of steady state um that means that your slope is actually a function of upstream area to the exponent of M / M the MN ratio or theta um concavity index also known as that. Um so but your your river slope is basically a function of a and then it's also modulated by the uplift rate. The higher uplift rate the steeper it becomes. Um and also K and K is some kind of bulk parameter that um yeah is yeah quantifies the uh incision efficiency. Um and so the higher the efficiency uh is the less deep will rivers become because the river just is really able to insize into the bedrock. Okay, this is a common equation that you have probably seen before. uh it leads us to a nice transformation which is called uh the kai transformation. So taking this equation up here and then integrating it in upstream direction gives you the what is called the kai transformation. Um so what does it do? Well, it takes this second uh term on the right hand side uh and we integrate it um in upstream direction and yeah that has the nice property that it linearizes the river profile. Okay, in that case we have x not the distance but uh sorry kai not the distance on the x-axis and yeah what we actually have is just a linear model where set so when you integrate the gradient in upstream direction well you actually get just elevation um you get some integration constant set zero and then you have actually a linear model where the gradient is determined by K or sorry the gradient is determined by U the uplift ratio over K. Okay, as I said the good thing is that it linearizes our river profile. But there's one more thing that it does um is that you know when you when we think of uh the upstream migration of nick points in our river profile um you know it could happen like it's shown here. you have a sudden base level drop of about 50 m and then this yeah this drop basically migrates upstream and you have a series of nick points that once they enter into tributaries uh will slow down. Um now if we do the same in kai space all of these nick points will cluster at one point in kai or they will have a very narrow range in which when in which uh in which in it will they will cluster. Okay. So here this is just after might have uh you have a a level lowering and here uh all these nick points are all these river lines basically collapse to one line and all these nick points are in one point. Um so if we assume that n equals 1 same erodability and uplift and climate and so forth then the time required for nick points to travel to their location is k times uh 1 / k. So in tow is basically time um that is required and which is directly related uh to k. Okay. So this is the idea that we will take um as I said before what we know is that um incision started about one 8 million years ago and we will take this information to retrieve stream power parameters um and then which allows us to to model nickpoint evolution. Okay, let me go back to um my MATLAB window. Okay, here we are. So that's where we stopped before. Okay, that's when we have our kai map together with all those uh points that indicate the location of nickpick points. So what we are going to do now is to learn a bit about um how we can basically approach this using uh pps using uh pps and pps is a numeric class in in top toolbox which is called or which is which stands for point pattern on stream networks. Um it brings a lot of ideas together from spatial statistics and geommorphometry. Um and is basically just about you know the combined analysis of streams and points and these points could be anything beaver dams or uh landslide locations whatsoever. Um but in the thing is they are attached to the stream network and so they are also used to analyze um point uh nickpoints. So we initiate or construct such a PPS object using our stream network. Okay. And then um we also need to provide our uh yeah nickpoint locations and these are in our structure here structure area KP which is and this field. Okay, this sounds probably a bit complicated but this script. So you could also add some coordinates X and Y coordinates but EX grid is uh just a linear index into the DEM. Okay. So we have the locations here of the nick points and one thing that we could add also to our PPS class is actually the elevations. So we can just input our DM from which our stream networks was uh derived and so we have everything together in one class and u will also implemented that in in Python and here it's just a some kind of data class that we need. Okay. Oops. What did I do wrong? It's not ex what is it then? H it's PP the planer point pattern. So apologies. Okay. So you see we have a new object here a PPS object and you know just stuff that you do is like plotting and this is just a a simple plot of the stream network together with the with those dots. Okay, it makes things easier first of all to have that all together in in in one file and we can also use for example plot D set in that case we can just input P because it already has the elevations here. Um and here we go. Okay. So, uh PPS makes things easier. Okay. Now, there's one particular function that uh is very nice um because it um it uh looks uh at the distribution of um the points along the rub network. So it's um it is called rowut and rowut is just a function you know uh that or an approach to test whether a point pattern is dependent on some possible co-variant. Okay speaking in in statistics terms um I'm going to do this here like this that I'm going to create a new figure. I'm creating a plot of the d versus set. So a profile with the points and then I'm calling row hut. Um and let's do this. Okay. So what do we see? Well, we first of all see the river profile again with all the nick points. Um and then we have that black thick line surrounded by by some bootstrap uh bootstrapping intervals which well the values of which are related to the second axis here. So it gives us a row x. So it gives us the density of points as a function of the distance and you see it attains relatively high values where you have a lot of points and then there are low values where you have a few points. So it's a bit like a kernel density estimator. The only thing that it does it accounts for the number of you know how often this value actually occurs along the ribbon network. So you have a lot of channels here. So that's why of course you have also a lot of points here. Um now obviously our data spreads over quite a distance from 10 kilometers to up to 35 kilometers upstream from the confluence. Um now let's do this again and uh with some slight adjustments because we are now changing um the distance on the x-axis to kai. Okay. Instead of plotting the normal ukidian distance, we are going to take the um distance that we have in the in the variable C which is kai. Okay, here we need to say okay, we have a certain co-variant and that is C. Let's do it again. And what you now see is that we we plot uh the rip profiles in kspace. You can see that they all collapse more or less to one line um or at least to a relatively narrow band. Um and what it also shows is that all our nick points basically cluster now in a relatively narrow range of kai something between you know 2,000 and 3,500 mters but it's uh in in kai units. Okay, this already gives us a a strong indication for the fact that all these nickpoints emanated from one common point the outlet in that case. Okay. So the nickpoint started migration from here. This is what when they cluster really narrow the narrow range uh this is a good indication for that um or to to infer that that mechanism behind it. Okay. Um what else can we learn from the distribution of C of of the nick points? Um there's one nice function that I want to briefly show. Uh it's called um you know uh MN optim KP. Okay, we sometimes have to work on those uh function names, but actually it's quite easy. It just says okay, we are we want to find the M and uh ratio of the stream power model by optimization using our nickpoints M and optim KP. It's a PPS function. So it takes P um but it also takes FD um our flow object and then there are a lot of options and you can you're welcome to to read through these options um but here I will just take the let's say the best method it's relying it's called devian um and it allows us to set a couple of other you know parameters in that case it allows us to to include uh in our optimization scheme. The results you will see what comes out in a second. But some more stuff and for example that the onset of incision started 8 million years ago. And we might even say okay uh this value is you know 8 million years uh crisp probably not. Uh there's some variability associated with that onset. uh and we can set this one to perhaps a million years plus minus a million years. Okay. We will then take a look at what is inside. Okay. Um so let's run that section. Okay. So it returns a structure array um with optimized values and a lot of other stuff. Okay. So first of all you will recognize that immediately 0.44. Okay. So that's the optimal MN ratio that minimizes the scatter in our nickpoint distribution. Um we have some uh intensity value. I will show that in a second. Uh and there's also an in we have inferred the K value also from this uh approach. It's fairly low 8.2 10 to the minus 7. Okay. it has some uh associated um uncertainty with it. This is more or less directly related to that sigma t that we have provided here. it would be zero more no not zero but nearly zero uh if we don't have any variability here um and tow for example now tow gives you also a node attribute list uh which um gives you for every location in the river network the age when that nickpoint passed by okay and it even has some uncertainty associated with that so it means that if you know if you don't have a very narrow range of uh nick points in in kpace uh at the beginning then of course the um uncertainties will go up here quite tremendously. Okay, let's just plot what we have here. I'm going to do that quickly because we're running out of time. Um plot the set p distance is and this time we don't plot pi but we plot toao. And we can add some color data to the um to the points so that they are better visible. Um yes, and we plot everything on the right axis as well. And so we plot the intensity. Well, you'll see in a second what it does. Um and the distance again is towel. We add a different labeling. Um and this is as function of uh cut. No of tow actually. So okay. So here we go again. This is now slightly different but only slightly because the M& ratio is a bit different 0.44 instead of 0.45. Okay. But also this mathematical model that basically describes the distribution. So it's a lo linear model that describes the distribution of these points. Now with that model um we can now run a nice simulation. I will not write it all down here but um but it's something that you can uh then also run on your desktop. It's just what it does. Okay, we want to predict where the nick points are at the onset at zero and then in steps of 10,000 years up to 10 million years. So we're going even in the future um not only to 8 million years uh and then we are going to plot all that. This is the top toolbox function. The rest is just pure MATLAB. Okay, let's run this section. It should actually plot something, but it doesn't want to. Let's see. Ah, it's up here. Okay, the good thing is that in when you do that in a MLX file, so in a live script, it will produce a a video that you can also then export to an MP4 or GIF image. So, what I just plotted is and I will run it again. It's how the nickpoints basically start and then once they enter into the tributaries they will slow down until they finish. And you know there is that point about 8 million years ago where basically all points are pretty close to the ones that we also identified in our uh river network. And so it's showing well how nice basically the model is able to hindcast uh where nick points should have been at some time in or at some certain time. Of course provided that our model is correct and the screen power is correct as it is. Okay. So that's it. Um you can read through the through the scripts which are fairly you know fairly short. I hope it's still understandable and uh by looking at the gallery uh online. Thank you, Will. >> Yes. Um so since we only have a few minutes left, I think I will uh skip sharing my my slides. They're not very complicated, just uh really some links and I'll post the links in the chat that I would like you to take a look at. Um but uh that concludes our first day of our webinar. Um, we hope you uh learned something about the new version of Tupper Toolbox. Um, we really want to emphasize that this we want this to be a a community project uh that both serves the needs of the geomorphology community and is driven by the um members of the community. So, we really want you to participate in whatever way you feel uh comfortable doing so. Um, we have many different ways. I'll talk a little bit more tomorrow about ways to get involved in Top of Toolbox. Um, for now though, I'll post a link to our discussion forum which is hosted on GitHub. Um, so uh that's in the the chat now. Um, and you'll find there uh a few things uh including one um place to post some questions if you have any questions about the webinar that you'd like us to address tomorrow. um and uh thread to introduce yourself and tell us what you're working on and what kinds of things you're interested in. Um we'd really like to know, you know, who's in our webinar and um who's in our community and what they're using to walk for. The uh uh as I said, we we'll talk a little bit more about other ways to contribute to Tupperbox tomorrow. Um we also will hear from Boris who will talk about his uh graph flood tool and from Bastion who will talk about his transct tool. Um so we'll go through those first I believe and then have a Q&A period and we'll talk about the contribution um uh workflows and things tomorrow. Um so yeah, we hope to see you there. Um and please yeah, feel free to leave questions in the discussions um and start you know having some conversations there. We'd love to see it. >> Thanks. Yeah, thanks for listening. Hope to see you tomorrow again.