EGU WEBINARS: Topographic Analysis Using TopoToolbox in MATLAB and Python - Session 2

So once again I think still people are dropping in um but nevertheless let us start again. you are here in the topographic analysis uh webinar organized by the EGU um and you have well I guess most of you have already attended uh yesterday's session um and why this session was mainly you know introducing to top toolbox and its latest developments uh and then we had also um yeah a a coding session comparing basically uh the latest developments in MATLAB uh with those in Python and how they are implemented in in top toolbox and then we went on with a more advanced um topic when it uh was about nickpoint location modeling using top toolbox and we also had a lot of time to discuss things so I think that was a very uh good um um day and I hope you enjoyed did as much as we did. Um so today the second day again from we have two hours um and within these two hours uh we will uh mainly deal with uh applications in top toolbox. One is graphlet a simplified hydrodnamics tool in topo toolbox. Um and then there's transact another tool uh which is about analyzing landscapes and the interconnected uh landforms. Uh we will also talk about uh contribution workflows. How can you actually contribute to topo toolbox development um and we will again conclude with a discussion and conclusion. uh we will do it like uh yesterday that um that you should use the chat to to ask you questions. So please ask questions uh during the presentations just using the the chat functionality um and we will then have ample of time to uh yeah to cover your questions in the discussion and also have the time to to moderate those. Um before we finally start, I would just uh give a short um you know call out to uh a latest preprint that just appeared last week. Um uh it's the it's yeah it deals with our uh latest work that we did uh Topo Toolbox 3 the developments that we implemented in our new software um in particular testing frameworks um but also the whole continuous integration continuous development stuff u that as a user you probably don't have to care that much about but that's exactly that's why we are doing it so that no one else has to care about it. Okay. Um without further ado, um I would like to head over to um or get the mic over to Boris uh who will talk about graphlet. So I'm unsharing my screen I guess uh and you will take over. >> Thanks a lot Volvang. I will I will take over. Let me just share my screen uh very quickly. And there we go. You should be able to see now my my slides. Uh so yeah I will be talking about graph floods. This is the work I've been doing in the past couple of years and it's becoming my main tool and I've implemented the final version in in TooBo toolbox uh because it's very suitable for topographic analysis. I will be first introducing graph flood why did we come up with uh with that algorithm uh what is it used for? Um then I will very quickly talk about how it works in term of of of numerical methods. what are what's in the pipeline for the future development and the most interesting part we'll do a bit of live coding at the end so graphled is a 2D hydrodnamics model basically it takes a topography um you put some water on it via precipitation rates and it computes the fields uh of flow depth and flow velocity in 2D the motivations behind developing graph flood comes from the rise of high resolution DMS uh we switched basically in the last decade or so from a referential where we can roughly have the shape of the fluial valley from which we can calculate the fluial slope but not much more than slope and range area to a referential where we can get that's what you can see on the right um systematically we can measure the width of the river we can see the small gullies we can see um a lot more informations and we can access it more and more systematically thanks first to LAR nationwide LAR survey like in the US, in New Zealand, in France, in Germany. This data is now roughly available for the whole country or even better thanks to satellite photoggramometry. We have satellites like the PlaySt. But it also brings some challenges because our methods in geomorphology which are mostly based on dage area when we have a high resolution DM metric scale that's what you can see at the screen here d area fails to capture the whole geometry of the rivers because even with a multiple flow algorithm we are still flowing water on the topography but rivers are now multipixels and the alternative to drain area is to use flood models and if you take is a flood model that are designed for risk analysis like hecas for example a flood computing a flood for 1.5 km basin uh six physical day in real life of flood takes about 10 hours of compute time. So if as a geomorphologist you're interested by montain range scale basin scale it's a bit limited and that's where graph floods comes in it calculates the steady state flow depth and velocity and it does an approximation of it. So what does that mean? That mean that if you have a topography and any way to quantify your water inputs for example with a predition precipitation rate it directly calculates the corresponding equilibrated flow field. You can think about it of calculating the peak discharge of a flood and propagating it instantaneously on the whole landscape and it makes an approximation meaning that it takes just a couple of seconds, minute, maybe hours if you have a very big DM uh to compute. What can we do with it? We can calculate flow extent. For example, you what you can see on the left figure each color correspond to the area flooded for a given precipitation rate. So you can do a bit of simplified flood analysis before running the very big model like heas. On the middle figure we've been computing flow width systematically for basin and you can do that across different precipitation rate to test for the effect of climate forcing for example but you can also go a bit further for example we've been doing a bit of classifications using that data. Basically we use the hydraulic slope and the discharge. This is uh on the x-axis the discharge to have some sort of hydraulic slope area plot. And we use that data to segment our landscape in fluial domain that's what you can see in blue collivial domain that's what you can see in yellow or even hill slope that's what you can see in red to give you a more sophisticated example um just just for the example I'm um tectonic geomorphologist so I'm interested into um identifying tectonic signals in the topography so this basin is a 10 km wide basin in the east in the west coast uh of the us in California where we have LAR data. Um, and it has been subjected to recent tectonic activation. So there's recent tectonic uplift over there. And I've been calculating for the main river. That's what you can see in red on the map. Two metrics with graph flood the sheer stress. This is basically the erosion power of the river and the river width for a given precipitation rate. And we can identify different type of signals that were not visible only with slope and drainage area. For example, we have some nick points here that are expressed by an increase of she stress. That's the the typical nick point where you have an increase of slope. We also have wit nickpoint where you have a narrowing of the river locally and a slight increase of ch stress. um in the literature that usually shows that the river is responding to a recent input of sediments and we have more combined um signals where you have an increase of sheer stress near my headwater combined with a narrowing of the river which is a bit more consistent with um tectonic signals as we as we see them usually. So how does graph flood work? It's an iterative algorithm. The first step is to fill the topography with water. If you have a lake and an infinite amount of time, uh you will eventually fill it with water. Then you calculate what I call the discharge input. We calculate the drain area. So with topo toolbox and we multiply by the precipitation rate that gives us ultimately the total amount of water we can use to flood our landscape. And then locally for each pixel we calculate the discharge output. We use Manning's equations to compute how much water can my cell store. And then if we have more water coming in than out, I add some water. The opposite. If I have more water coming out than in, and I will iterate for hundreds of iteration between these three step until I reach flow depths, field of flow depth and velocity that's totally equilibrated on my nonscape. The key point to address before I move to uh the the demo are how to handle bundary conditions. Where and how do I tell the code I'm inputting or outputting water and checking that my model has reached convergence and that I can stop iterating. Oh, just a last word about implementation. Um most of it is coded in liptop toolbox because it's quite a computationally demanding algorithm. The main interface at the moment I'm still working on it is in Pytopo toolbox. That's where I will um implement the the the late um cutting edge improvement I will do on the algorithm and we are working now with will and wg going to have a mat lab and our version there's a matlab version working but it's it's a bit less sophisticated than the python one but it will come soon and just for the future I'm working now on GPU version of this algorithm and on the main um algorithm we find in in geommorphology and that will come hopefully in 2026 six in a module called Pyas flow which is also part of the lip of the toolbox uh GitHub and and and all right let me share the live coding window now and we will get to it >> just briefly Boris um we say that anyone who has questions is welcome to post these questions in the chat and we will address them later on. Yeah. And also feel free to contact us on um by email or on the topograph on the GitHub discussion page um after this workshop if you have some questions that arise. All right. So you should be able to see my screen now. Um this is a Jupiter notebook. So I will be using the Pytopo Toolbox version that will presented yesterday. This notebook will be shared I guess on the gallery or or within the the top toolbox and you can follow it uh without this talk initially. That's why I won't read everything that's in in the cells because I assume that some people will want to follow it as a tutorial offline. So I'll just assume you have installed too. We will just import it. I will also use u numpy, scypi and scikit imageh which are classic scientific computational tool uh in uh in the python ecosystem. And maybe something that you may not be um used to is that I'm using IPMPL. This is a way to tell Jupiter that I need interactive figures. It's not something that's mandatory, but now you know. All right. Uh let's go. I will start with a quick start. Basically, you have a DM um the classic DM and you want to calculate the flow depth for a given precipitation rate. I will be using uh the Green River DM which is built into topo toolbox. So, if you want to try it, you don't have to um um you don't have to download a new DM and I'm cropping it to a small area to make sure it can run live because the algorithm can take quickly a few minutes, tens of minutes, or even hours to run. if you have a massive DN with um tens of millions of sales and just to show you I will um the the DM is most small basin in a low red area in I think it's in Utah in the in the US and I chose it because it has very well definfined uh fusial form so it will be very nice to see if we can capture them with graph flood very much like the grid object the flow object the stream object in top toolbox graph has its own uh graph object which I named GF object effect because probably graphford was a bit too uh long to to write all the time and to construct it you just need to give the grid object which is the topography on which you want to drop flow as you can see here I'm creating it and you can add some options uh optional uh data but it's usually good to have them P is my precipitation rate you can either give a scalar to have a constant precipitation rate on every pixel or you can give a 2D uh array of DM size if you want specially varying uh precipitation and a manning coefficient for the manning coefficient I don't have time to go through the physics but it's the empirical coefficient that you will use in any flooding model um it's what we call the friction coefficient unfortunately we have to choose a value for it so there are some a lot of literature on on that value here I chose 0.05 05 which is a fairly standard value for open rivers. Once we have constructed our uh graph object um we need to run the model for a number of iteration because as I told you before we are gradually creating a field of flow depth until it stabilizes. Here I will run it for 100 iteration. So while it run it will take um oh yeah of course I forgot to run the cell before. While it run it will take something like 10 seconds. I'll just um state that I'm using a very high precipitation rate 100 mm per hour which I convert here into meta per second. Uh but it's because we have a very small area. So if I want nice rivers and to show you quickly how those rivers work I'm using a a very high precipitation rate. The model also needs uh the number of iteration on which you want to run it and a numerical time step. It's a steady state model. So the time step does not have physical meaning but basically the higher the time step the faster you reach convergence theum the higher number of iteration you run the faster as well the more you reach the convergence but if you have a high time step you can destabilize the model it's a bit of the issue with those physics uh flooding model here I've been choosing a nice balance which usually work for um data with a relatively low relief on LAR DDMs and then once the model has It has finished to run for a long time. We can simply access the field of flow depths in the graph flood object with GFO.HW flow height. Um and we can see that basically we've run the basics of graph flood. We can access at any point the field of flow depths to see if our river is wide or diverging or converging. And that's one of the main strengths of um using floods to describe rivers in the topography. Even if your flow is diverging, like you can see here, it's branching into two branches, you can still describe your reverse even if it's like 10 or 20 pixels. So, all right, what can we do with that? The first thing I'd like to show is to compute shear stress. Shear stress is the force the river is applying on the landscape. Whatever erosion or sediment transport model you're using, it starts with calculating shear stress. Even the stream power low even if we use a lot of different assumption on range area on flow uh steady uh and uniform flow and so on at the start of it it's just fresh. So this is really the most basic way to describe erosion in your landscape and the most basic way to calculate it with craft flood is to use uh this form this very standard formula g h and um hydraulic slope which is fairly easy and uh rapid to compute with the function compute and if I plot it it will simply give me a clue to uh towards where I'm focusing erosion power in my landscape. For example, here I'm focusing erosion uh at that place and if I have a drop of your stress, I can start interpreting that I will deposit the sediment I've been on training. All right. Another thing we can get with graph flood is getting the flow velocity. It can be important if you're tracing for example a pollutant in your river uh suspended sediment and so on. And uh the flow velocity will be what you use to transport and advect information in your um on your landscape. Again, this is fairly standard because this is something that's directly simulated by the model. So, we just have to get the grid object of u which is standard um uh notation for flow velocity and you can plot it to identify where your flow is accelerating or where your flow is decelerating. These two metrics just stress and flow velocity are really the let's say basic first order metrics we can get out of flow. But we can go a bit further and start interpreting our landscape. For the sake of time, I won't go through the very uh the detail of each and every line of code. But what I'm doing here is getting from the model my volutric discharge that's skewing that I get here and my hydraulic slope. That's the slope of my topography plus the uh flow depth I have on top. And I'm using this information as an alternative to um slope area plot. um slope area plot in geomorphology is what define KSN what defined sky it's really behind everything we are using to quantify erosion and using this hydraulic version we have a bit more information for example we can see more breaking slope into this slope area plot and we can use this breakin slope for example here um I use 1.7 and minus 1.7 and minus 1.4 for as a threshold that I identified. I think it belongs to one of those uh breaking slope. And if I identify my data where I'm um where I have sorry uh I'm respecting the threshold, it allows me to identify where I have uh my fusial domain. That's what I was saying before to classify my landscape and that can identify in that using this kind of information a 2D river network. One thing I forgot to comment, sorry, is uh that it can require a bit of extra work to clean. For example, if I use the row data, you can see that my 2D network is a bit um patchy with a lot of noise. So, if I want something cleaner here, I've been using a sci domain, sorry, scikit image, uh Python tool called binary closing. It's just a way to tell the code, I want my 2D field to be as continuous as possible, so clean it a bit. And um I end up with a 2D network and 2D mask of topography. But we can go even further. For example, can we use this 2D mask of top of reverse sorry to calculate channel widths because essentially this is my fluial domain. If you want the full example because again for the sake of time um it's a lot of step to to to to explain but the example from the the paper volgon just show uh is available in the gallery of to box and is developing a bit more in details how to calculate that width but the step are really straightforward. I'm cleaning the channel mask that's what I've just been showing with binary closing. I extract the center line of uh that mask using something called a skeletonized algorithm. Skeletonized algorithm is just a way to collapse this 2D mask into a network of line. And then I'm calculating the distance the eucidian distance from the um border to that center line and that gives me if I uh double it the u width of my river. So if I run that code you can see it's not hundreds of line of code. is really just maybe 10 lines of code uh excluding visualization. You can see that it has been doing what I said. Um if I zoom on an area, the yellow you can see is the mask we've been quantifying a bit before. The line of dots is the center line that I've been calculating with the skeletonized algorithm and the color is my river wits calculated uh with this algorithm. You can see that we can quickly identify the area where my river is widening. I can then uh use some algorithm like a KD3 uh KD3 sorry to related to the real uh river network that I would have been calculating with toolbox and uh cross that data with any other stream object uh analysis that uh that I would have been doing before. All right. Um yes, I'm done with the quick start. Sorry, I was just shaking. And the rest of this tutorial will be slightly more boring. But this is what I said before. There are two aspects that are crucial with Scruff flood. The first one is checking the convergence. Have I been running enough iteration? Can I trust the field of flow depths that I've been creating? Um and to do so there are not there are many different ways. I'm still trying to find the perfect way that I will implement at some point in graph flood to say to to have an automated way to run to convergence. But the safest by far is to use um to to monitor sorry the increase of my flow depths through time or through the iterations. Once the latter has stabilized I can consider that my algorithm has been running to convergence. So to do so I restart from scratch to create my graphlet object and I will be saving every 10 iteration my field of flow depth. So this will take a bit of time to run about 10 seconds uh like we used to to do. I will plot in a when it has um run a different step. You can see that my after my first 10 iteration I have a very faint reverse and slowly through time it starts to accumulate more water up to a point between my two last iteration where it's not moving a lot anymore. So what did I do to check that if I' that I have uh reached convergence? I'm simply running uh I'm simply calculating between each field of flow depth that I've been saving the 19th percentile of the H increment. This is in log scale and what it tells me is that the 19th percentile say that I've been adding 3 mm of water after 10 iteration um in between every 10 iteration and I'm monitoring that it's decreasing through time up to a value where basically not much is happening anymore. Here for example I'm only adding a few like less than a millimeter every 10 iteration to my field of flow depth. So I can consider that I've reached something close to equilibrium. I can run my model for 10 for 100 more iteration 10 10 times more and keep monitoring and uh once it has run I can still monitor my increase of foot depths and you can see that through time it will store and slowly converge towards zero but at that point I don't care anymore because I'm not building anything anymore I'm just refining a few millimeters here and here and there right so this is the first um aspect to check the second one is to make sure that we manage bund conditions. What I've been showing you so far is really simple. I have my DMs. My water can flow everywhere and will leave from the moment it reach a border of my DM. So my first row, last row, first column, last column. But some cases are a bit more convoluted. For example, let's say that I want on my landscape my water to enter by the side of my river of my river reach uh at the bottom of my DM. I don't want my water to leave the DM until it reaches the output of this rivers. So what I did here is plotting the hill shade of the DM without the keyword extent. So basically the um X and Y axis are not the X and Y coordinate anymore but they are the number of row and columns which allows me to identify the small areas where I want to input water or to leave water which is the indices I've been um uh identifying here and if I now look um visualize sorry my field of bundary conditions I can see that in light blue this is where my my water can flow on my DM um on my wall I have dark blue which means uh that the value is zero and that my water cannot cross this bundary and on the top right which is where I want my water to leave I have a small gate of value three of red values here where I told the model my value can leave through these pixels right that was a bit confused and boring but what does it do in real life I will run my model again using this new data and sorry I forgot to run it for 100 iteration and not one iteration and it will uh compute that's why just a couple of seconds and it will compute the flow depths just for that reverse because I've been inputting river um um precipitation just here and monitoring it forcing it to leave at the end of my rivers and if I do that again with a 10-fold increase in precipitation rate which will take again a couple of seconds you can see that I can quickly test different peak discharge for a river reach and no need to compute it for the entire landscape. This is what we can see here. This is my last run with a 50 uh cubic meter of water inputting from the south and leaving from the north. And same thing on the left with 5 cubic meters. All right. Um I have a few minutes left because I probably spoke a bit too fast. So feel free to contact me if you have any more question or to contact me through topo toolbox through the chat if you have more question or through my email which I will write right now on the screen and I'm uh planning to really use uh graph flood a lot more in in my current research. So I'm planning to add more and more features either in the CPU version or the GPU version. For example, everything I've been showing with the wits, everything I've been showing with comparing or mentioning comparing with um the stream object information or automating the convergence. It's something I'm actively working on uh at this time. So in the future pull request, I will I will add a bit more information. All right, I will stop my rambling and give back the microphone to uh to Volvong. Thanks a lot for listening. >> Thanks a lot, Boris. That was very insightful and uh exciting to see where graph flood has gotten to right quite a and seeing it how quickly it evaluates in a yeah for quite a I mean it was that was not a very big DM but even for such size DM it's amazing how fast um you get a 2D water surface that's really great uh perhaps one thing that I'd like to add here um because you have uh you have given in your mail address, which is fine. When people want to contact you by mail, that's good. But I would still want to encourage you to, you know, to to use the discussion uh yeah tools on on git. Um because then everyone can learn from these discussion also and and see what's going on. Uh that's better and more open and more fair um to our research community. Okay. If I may just add, it's also uh good because uh we're not always available to answer and other users might have an answer to your question. U so it's for the sake of time. >> Exactly. There might be other users who might actually be able to to answer and we don't have to do that. That would be a dream. [laughter] [gasps] Okay. Um there are already a couple of questions online. Uh but we were saying that uh we will cover these at the end right D. >> Uh yes except maybe for the last uh question here uh cuz Boris you have this uh Jupiter notebook and I guess that you will share it right? >> Yeah definitely. Um I may need to chat with VGO and Will to know if we share it through the Gary in Topo toolbox or any other way but it will be shared. Um >> sure. Yeah, I think the gallery is the preferred option nowadays. And so that already answers one question. Okay, thanks Wolf Gang. Back to you. >> Cool. Okay, no, not back to [clears throat] me. I'll pass directly on to Bastion. Um, go ahead. >> Thank you. So, we just share my screen. Okay. Okay. So, hello everyone. So, today we talk about transect. So a toolbox I have developed within topo toolbox environment and well so the the aim of this toolbox was to uh try to connect the different gemophilic processes that are usually so analyzed separately as a river profile is slopes or flute plane and particularly to try to to understand how these processes can qu together across the landscape. And so here I will just show you a few slides explaining the main principle of this toolbox and then I will run some life coding to to shows the main application we can make with translates. So the motivation for developing this toolbox come from my PhD work on the landscape version of the V mountains and there the the in this area. to the eastern side of the massive is geomorphologically very dynamic and undergoing transient transient processes as you can see here in this catchment where we have so different waves of migrating points and this is a systematic signal we we can observe all along the range and uh we can also observe an adjustment of the east slopes to this nickpot migration with an increasing in slopes just downstream the neck point. So at this point we wanted to find a way to quantify so this lateral adjustment of the valley to the vertical incision of the of the knee point and so starting from this observation. Uh so we wanted to build a tool that don't treat so the river profile all the all the slopes as separate objects but instead so transct connect them by extracting matrix along transexs that run from the channel to the device. So the two box works into two complimentary mode. uh we have the geometric one that is based on minimal distance that generates a subperpendicular uh transct all along. So a baseline can be of any type. So here we choose the trunk channel but in this mode so we can observe that in curved ridges the the transet can converge into one point forming this type of fan structure. So in this mode I've implemented an algorithm that allowed me to spread those fans in order to ensure a more even coverage of the valley but also some processes would be better linked for the natural pathways of water or or sediment and that's why so transact also include the flow routting mode uh that follows the the topographic gradient that show better representation of how material move through the landscape. Okay. So once the transexs are traced, so the next step is to connect them to to other geomorphic uh features. So for instance we can extract fl plane with graph flood and connect the the resulted plane mask with the transact in order to measure the the width or we can also connect the transct with um the drainage divide in order to capture the full valley geometry. And so here's an example of the type of output we can have with transact. So in this case I have run to the geometric mode and stop the transect to the divide in order to construct the valid geometry and [snorts] from that we can extract all type of information along the the transct illustrated here by the black lines. So we can extract the value width value height or the the e slope angles and uh as all those transects are referenced along a common baseline. So here's the river we can relate all those valley matrix to the fia matrix as here the stiffness index but also to the flid. So now I will run the demo to to bit show you how we can collect this type of information and how we can manipulate uh through manipulate transact. So let's share uh my MATLAB window. Okay. So for this demo so uh I will surely made it available to to put toolbox gallery in the upcoming days but uh but well so this demo will be run in the catchment I've just showed you in the for montain and it will be divided into two parts each focusing in one chopic domain. So the first one will be on the valley geometry uh and then we'll focus on flute planes and particularly how we can relate to the plane width extracted with craft flood to flu matrix and is slope gradient. So before starting so I have just premputed some topbox objects in order to save some some computing uh time. So here has the premputed object. So I have already prep-processed the DM. I have already computed to the flow objects the flow accumulation the stream network and the drainage divide and I have so the the weather head resters that gives the MATLAB version of graph. Uh okay so let's go with the first part. So to reconstruct the valid geometry from geometric transct. So the first step here is to set a baseline a baseline and for that I will use the modify uh function of uh of to put a box and I will select a ridge that uh uh that I will uh from which I will trace my transit. So okay so just close the the the figure. So this is my selected ridge and then I will do the same but to select a portion of my trainage divide that will set so the the outline of my value. So the function works similarly than the modify function of stream object. Let's select this and this is so we have our baseline and we have our val outline and now before running transex so we have to set two parameters so the the width there will be that will correspond to the length of the transacts and the number of iteration uh will make within transact so and to uh and to estimate the width so instead of choosing it manually we can estimate it from the size of the Austream D basin. So let's let's do it. So I will use uh the the uh to put toolbox function D basin in order to extract um the the basin mask. So let's do that with uh the outlet of my selected bridge and then I will convert um this into a mask. So doing logical z up. So this give us our basin mask. Then we'll set the iteration to three. So so basically the iteration is to set how much intermediate outline. So the transact we have to pass through during the uh during the the tracing and then I have so a small custom function that is called max width that will um determine so the the optimal width for my transect run. Oh miss here and up. So I have my W and I will display what I obtain. So this is the gen the upstream gen basin from the outlet of my selected reach and we can observe that we have here's the outline with that will take into consideration to the entire genes basing. So now we have everything to run transact that will just [clears throat] run. So I just have to input the DM or selected reach uh the width the number of iteration and verbose to display um the the progress. So if I don't work okay so that's done. So now I will show you what obtain. So these are our row geometric transect. So we just zoom a bit more to better show the how the transact were traced. Okay. So we have here so the the transexs that are tra. So we can observe that those transact doesn't cross uh don't cross each other and they they cover like the interior areas that we considered and uh so here are the number of iteration that transact with trace. So more iteration input more it will respect the the baseline geometry uh as if you only take like one uh one iteration and to have a tendency to to smooth as the baseline geometry. So once we have traced our transex so the next step will be to connect them with our uh basin not line. So let's do that and ah yes and we can enforce the structure of transect so it's a class object as we can have with grid object etc uh that give different uh information where we have so our base more the information of of the baseline so the coordinate or if it's a stream object or divide object etc. The con will are the the path that just run from the baseline to the uh intermediate outline. So in this case they just pass with with four nodes and int there are the interpolated path where we have so all the nodes all the pixel between the uh the different outline. So now we want to pair those geometric uh transact with a selected divide. So for that just I have to call a function called pairing where I will input my transect and my selected divide object and I will extract the stats that calculate um the width the eight and the and the the slope along those transects. So it should took [clears throat] a few seconds. Okay. So note let's display our results. So now we can observe that we we have our our transacts that efficiently stopped as the trans divide. So the transact only extract matrix within the the valley boundary. Okay. And the next step will be to try to look as well to the foot plane width. So for that I already computed so graph water 8 water hat restor. So I've used the MATLAB version of CL flood and to extract a mask. Uh so me I've used more direct method than than BIS but I think uh it was a good way to to maybe show the F domain as you showcase BIS. But uh well in this demonstration we just set a simple threshold above which we consider it to be a full plane cells. So here we have our um our plane mask if we can zoom. So it's quite well it's more try and error to to set the the good uh threshold but we can observe that it doesn't go very it doesn't go to the his slopes and cover well the uh the fruit plane and now we do the the same uh we call the same function uh where we pair so our transect with the footpen mask and we extract the stats in order to obtain the width of the footprint. Okay. So now let's see the results. So now we have our transect that stops efficiently on to the to the foot plane mask. So basically stops until it encounters a false value in the in the foot plane mask. So now we have everything for this first part and uh so we'll we try to to to trace so along our selected reach all the value matrix and the plane width. So I've just premputed the stiffness index of the baseline here. We not enter to the detail of the plot but it's quite straightforward. And so this is the type of result we can obtain from this analysis where we have so different metrics that are plotted along the selected reach that is here the the trunk channel that I've colored based on its stiffness index. Uh so here in the upper left you can see the two elevation of the of the drain divide. So I think red is the left side looking upstream and the blue color is the red right side. So we can also extract the his slope angle. So in the same principle here you have the the two sides in black you have the mean value between the two side. Then we can estimate the valley configment by doing the valley width divided by the the valley height and uh we can also extract the footprint width along the the the channel. Uh so here we can observe um a global trend between all those uh landforms where we have a less slightly steepening of the east slopes uh higher confined valley and an foot plane towards the neck point uh that showcase this um this river. And well we can also observe here that um that the foot planes varies from uh for almost three orders of magnitude from few t of mters to almost 1 kilometer. So we make an interesting case and to compare it to to fluial matrix and is gradient to see how this range of len evolve uh with those matrix and for the gradient. So in this part instead of extracting it geometrically I would do it along so the the flow path so to which is which are more possible pathway for sediment transfer from isoprop to channel. So this mode is more straightforward than the geometric mode as um you I have to put them [clears throat] the selected reach as we don't need to select to define a width or an iteration as it will simply compute the flow object from to put a box and follow it. So method flow and go to true to show. So it's much faster than the geometric mode. Okay. And let's plot the results. So here we have so so a flow path that is uh so extracted. So from each uh baseline nodes and here so the the dot correspond to the to the end of those those flow of those flow path that correspond to the valley head. But as we don't want to extract matrix all along the basin outline now we'll pair those flow path to the first basin outline note say encounter so we take so tsl so we pair our new flow path transect with our selected t then we extract the stats among them they is done. We know our flow pass stops as the first train activate the encounter and um now we can extract the esop gradient along each of those flow. Uh okay so now I will just um extract some previous metrics I want to relate with my flu plane width. So firstly I will extract some my free plan that correspond to the to the to the stats of my um TF transect then I will extract my train area along my uh my select reach. So it's get null uh message reach and my full accumulation grid that I've already computed that we will multiply by my cell size of my DM to in order to get train area and uh although there's a small specificity to this version of transct is because uh well I have to to do some premputation of my baseline. So, so the baseline of my transact is not is not exactly the the stream object I've inputed. I may have some notes that are deleted, but uh you can refine. So, this uh arrays that allow me to reconnect to the baseline of the transact to the stream object I have input uh in my transact uh run. So, this give us my trainage area. Then I can uh extract my rever gradient like so. So in the same way I extract like connect them to the original stream object and then I have a case that correspond to so the case. So we just recmp computed here. I will put the concavity of the open 45 which is standard uh smoothing factor and I will link it back to uh oh no I don't need strictly to the uh stream object and I will uh extract my uh I have to base and uh the last one will be my slow gradient where I will directly use so this flow based transact I call ths and for that so I take the the mean value of the slope of both sides of my uh transact object okay I just run that and now I can plot the the result so pw plot is just a custom function I've made to to make this polo fit with so average bins beans average. Uh so here I will not enter. So here I will not enter in too much details to the interpretation of the this results but uh basically these two application I was to show how can we uh provide so interconnected analysis along unified framework that is transact and we observed here also in this simple case how they they can co-eolve together in the case of catch ment that is undergoing strong adjustment to Nickpoint migration. So thank you. I've done for my part and now I will let we'll hand over to Will I think. So yeah, thank you. >> Well, thank you um Bastian. That's a a really nice example that shows you know how to to use all the different tools that are available like graph divide object. So divide object is one that we haven't mentioned so far but is which is also you know part of the toba toolbox distribution um and then build your own tools basically and and transact basically builds on this and adds new stuff and so I guess you know if you have done that and you are at a point where you need to decide is it something that I want to contribute to the community is it something you know that other people might well value well you use in their own studies Uh, of course that also entails a little bit of work. We can discuss that later on. Um, but before we go on to the discussion and once again I I want to encourage you if you have questions ask them in the in the chat. um to Bastiana to Boris to us um and so but before we do that um well let's go on uh with the next talk by by Will who will show the workflow that you know Bastion might encounter now once he wants to contribute and and make transact available to everyone. So over to you Will. >> Uh yes hello everyone again. Um so I will start I think by sharing my screen uh screen and start with um a slide that's from the beginning. Let me go to the correct slide here. Okay. Um so I mentioned this kind of a little bit briefly at the end of uh yesterday's webinar but um you know that we really want people to um think of Top of Toolbox as a a sorry um all right think of Topa Toolbox as a a community project um and one that's driven by the the needs of the community and that community members can actively participate in. Um so we really want to get you know feedback from you all about what kinds of things you want to see in top of toolbox you know how you like it um to participate in different ways. We have a few different um ways listed here that are good ways to get started contributing um and you know I kind of walk through these a little bit um here. So um I I've included some links as well here uh that you can visit but I'll I'll show off these different websites here um now. So hopefully you all can see my uh web browser here. So this is the uh topa toolbox um organization page on GitHub. So we have a a GitHub organization now that coordinates all of the development of topa toolbox packages since we now have the mat lab package. We have the Python package. We have lib topa toolbox. We have our R package and many other things. We we need all a centralized place to put all of that stuff. So that's in the top of Toolbox organization. It's github.com/top toolbox. Um and you can find links to all of the other resources. Um from there, though it's a little um can be a little intimidating to navigate at first. Um we do also have a website. Um the link to it is here on the organization page top ofolbox.github.io um which you can visit and um this is mostly again links to other um resources. So we can you can visit the mat lab repository the python documentation which has its own website. There's a link to the gallery which I'll go into in more detail um later. And there are some contribution guidelines here which um provide kind of highle overview of how you might go about um contributing code to top toolbox. Um so I'll show that here. This is the um the contribution guidelines. So it explains kind of how we work through GitHub in the top of toolbox organization um and goes into more detail about what kinds of things are in the different repositories um as well. Um so the um >> right um I think the we'll start off from perhaps the easiest way to get started uh contributing to top toolbox is through our discussion forum which again I mentioned yesterday. Um that again is accessible from our organization page. You can go over to the the right side here and view the discussions and we have some you know discussion topics. Um, this has not been particularly active, but we're trying to encourage this as a place to to start having conversations about Topa Toolbox. Um, we'd really encourage you particularly to introduce yourself in this thread um so we can learn more about um our audience here and what you guys are working on and um what kinds of things you're interested in um as well. So, I encourage you to go check that out um and you know, feel free to add a reply to this thread there. But if you have any any general questions about top of toolbox about the um I would say you know maybe like the more scientific questions about you know how do I use this to do a particular analysis that I'm interested in this is a really good place for that. Um if you have uh more specific questions um or problems that you've encountered using the software, we have um the uh individual software repositories. Um and there are a bunch of these in um the top of Toolbox organization. Um but you'll see the top toolbox 3 here is the mat lab package for the third version of topa toolbox. We have pi topa toolbox our python package lib topa toolbox our C library and and a few others. Um but if I go to say the Pytopa Toolbox repository here um we'll see as with all um GitHub repositories uh you can find issues and this is where you can report u problems that you run into with Top of Toolbox. So um go to that issues tab. If you click new issue you'll be given a little form. You can say you know I've run into this problem running this function. It didn't behave the way I expected or I hit an error. um and one of us will uh get back to you uh ide pretty quickly and and try and sort out your problem there. Um we also use the issues as um kind of public to-do list. So um if you are looking for ways to contribute to top toolbox, feel free to feel free to scroll through these issues and see what kinds of things we um could possibly need help with. You know there are some functions we have these help wanted tags which are used more in some other repositories than in um this pytopa toolbox one for instance but um for instance we have uh let's say the grid object inpaint nanss function so this is a function that exists in the mat lab version that we need um for some example uh things I'll talk about that a little bit later as well um but it's a a function that exists in mat lab that we would like to maybe at some point include in pyop toolbox so if that interests you, um, you know, you are totally welcome to, um, you know, take a crack at some of these issues as well. Uh, but like I said, uh, this is a good way to communicate for bug reports, u problems, feature requests, if there's something that you really are interested in seeing in any of our repositories. Um, uh, communicate through the issues is a great way to do that. Um and don't worry too much about opening issues on the wrong repository or something. If you uh you know have a Python question or Python issue that you open the issue on the mat lab repository, we'll sort that out and uh we can move that around. So so you know it's better to get in touch with us um through any means possible than to uh worry too much about making sure you're in the right place to to do that. If you ask a software question in the discussions, that's totally fine as well. Um, so yeah, those are the kind of the the easy ways to um get in touch with us. Uh, the next thing that I want to share, we've talked about this a few times, we mentioned it several um, examples, the Tapa Toolbox gallery um that we've established to share examples of using Topa Toolbox. So um you can find here um mat lab live scripts and jupyter notebooks that implement various examples. Um the front page just kind of has a list of all of them. Um you can also sort by python and mat lab using the the um tabs at the top. But for instance you can find the um the ran plateau nickpoints example that we discussed yesterday in mat lab there. The python one is down here. Um you can find the uh graph flood a graphite example that Boris implemented here. Um so yes uh this is a great way to get get started contributing. So if you've done something with topo toolbox and you would like to contribute it to the gallery um that's a great way we can you know share that example with the community um get uh it in front of other users of top toolbox who might be interested in what you've done. Um, and really all you have to do is create a live script um, if you're using mat lab or in Jupyter notebook if you're using Python and um, submit that to us. Um, if you're uh, we'll talk a little bit about the the pull request workflow through GitHub that we use to um, uh, update our code. Uh, you can make a poll request to the top toolbox gallery repository which is this one here. Um, but if uh you're not a confident Git user yet um and you have a G you want to share with us, you can email us a live script and we'll um sort out how to put it on the gallery. Um and we'd really like to see more examples from the community. Um we have mostly the examples there are things that we've created um to demonstrate various aspects of top toolbox, but we really want to see what you guys are doing with top toolbox. So um you know don't worry if it's uh you know not not the most sophisticated or cutting edge example you know um it could definitely be helpful to someone just to getting started um uh using the software to see what other people are doing there. Um so the uh if if you are motivated to contribute code to Tupper toolbox now um we have a workflow which I as I said is is documented in the contribution guidelines here um in our contributing via pull requests here. Uh this is is a pretty standard workflow for using git and GitHub. So um if you're familiar with that from another project then uh it it should be pretty um uh familiar in the top of toolbox context. Um if you're not familiar with git and github there are plenty of great resources online. I think there's some there links at the beginning here to the github documentation and this um git uh tutorial from organization called the touring way that's very good um for getting started with git and github. Um, and you're totally welcome to ask me um questions about using Git and interacting with us um via GitHub if uh you're running into problems um in any way. Um but uh the basic idea is that you would go to say uh let's go to the Pytopa toolbox repository. Um so this is topa toolbox/pytopa toolbox um which is the the main organization repository. you uh don't have permission to change anything in the official repository. Um that can only be done by one of the um uh maintainers of the organization. So don't worry about messing anything up. Uh there's no way you can mess up anything um for other users. Uh if you're just experimenting or something um or you know you're new to get git and um you're afraid that you're gonna uh you know delete someone's code, it's not going to be possible to do that. Um there's checks in place to make sure that um that kind of thing uh doesn't happen. But so we're in the top of toolbox repository. You would fork this repository using this little fork button. Um assuming you have a GitHub account account here. Um and in fork it pops up this little um window to create a new fork. I already have a fork of um Pytopa Toolbox. So I'm not going to be able to do that. But we can go visit my fork here. And my fork is just a copy of uh Pytopa toolbox that I keep for myself. Um and on GitHub um and so then what I have done is um uh once you've forked it the the best thing to do then is to obtain this copy of Pytoper Toolbox or whichever repository you're using the source code on your computer. Um so you can go to this little code window here. Um and it provides you various options for the URLs for downloading the um the code from your fork. Um so uh you can use that. So um for instance, I would go copy this SSH line here and then um oh I'm going to need to open a new terminal. um terminal uh and then I you know uh you can run a git command get clone and use that um URL to get your uh copy of Pytopa to Toolbox. Um so now I'm I'm in the Pyopa Toolbox repository. Um you know I'm using Linux here. things may be slightly different if you're working on um a Mac or Windows machine, but um I'm in Pytopa to Toolbox and you can see uh yeah, I have a this main branch. Um I I'm not going to go into like a full git tutorial here because it's going to take too long. But um uh basically now I have a copy of Py to Toolbox on my computer. Um, I have all of our code. Um, you know, it'll take some familiarizing with the repository structure. The code for Python package is in the source uh directory. We also have tests in our test directory. Um, and docs in the docks directory. So, um, we could go visit our source and pop up toolbox here. And you can see we have a file for each of our objects. Basically we have the flow object, the stream object, the grid object um and then some the graph object and uh graphled functions um as well and then some other auxiliary functions there. So you make a change to one of these files and then um uh you commit that change in git parlance, push that change to GitHub um which again you know if you're not familiar with these terms you'll learn as you get used to GitHub and then we make a pull request to the um original repository. So again um in the interest of time I'm not going to do that but I will show you uh some of the poll requests that we've had. Um, so, uh, as you can see, most of these are from me, um, as I've been doing things recently to clean stuff up for our webinar. Um, but, uh, we, the poll request is just a request to, uh, contribute your changes from your, uh, repository into Tupper toolbox. Um, so, you know, I give a little description of what's in here and, um, we see uh, what's going on. the uh when you contribute the the pull request, it'll it'll run these um tests automatically um here. So there a bunch of tests in each repository that check to make sure that the the software is working correctly. Um and you want all of these to be green um before uh we merge your contribution into the main um repository. And so, um, I will work with you if you know you run into errors to figure out what's wrong with what you've got and, um, make sure that everything looks good. And then we merge that pull request. Um, this has been merged. And now this, in this case, this the row hat function for our PPS um, object which I've recently been implementing is now available in Pytopo Toolbox. Um, so, uh, yes, like I said, this is a a relatively standard way of contributing, um, via pull requests on GitHub. So, a lot of the tutorials that you find out there for, um, you know, how to contribute to open source software projects um, will apply in this case. Um, I don't expect you to, you know, be able to do this from this description that I've just given to you. Um, but that's, uh, um, you know, as I said, let me know if you run into any trouble with this this process. Um we really want to make sure that everybody who wants to can contribute code if they wish. Um as I said uh you know if you're looking for a project you can check the issues um in the different repositories. This is the pi topa toolbox one. I'll show you. I know lib topa toolbox has a fair number of um to-dos in its issue page. Um so if you're a confident C programmer um we'd love to have some contributions here for um different functions that uh could benefit from the libtop toolbox implementation there. Um so yes that is the code contribution process um for code to uh that you want to contribute to a topa toolbox package. Um if on the other hand you're um developing something that maybe is a bit more specialized or that integrates a bunch of different packages together um there you might want to keep that code in a separate um package from the main PI top of toolbox. You know I think if you if you can narrow your implementation down to a single function or something that would be a good candidate for inclusion in one of the top toolbox packages. if you have a bunch of different functions or some additional data structures and in particular if you depend on other packages if you're interfacing between top of toolbox and something else um it might be a good idea to maintain a separate package for your software um and we have a few examples of that um here we just saw Bastion's transsect um demo which I think I opened um here but yeah so uh Bastion has this the transect code um in its own git repository on on GitHub um that he maintains. Um we also uh we are happy to host packages through the top of toolbox organization. Um if you like the a good example of this is the bankful mapper which I think wolffko mentioned a few times yesterday contributed by Michaela deiaro. Um and so this is it again its own mat lab toolbox. Um and we have it set up uh so that it works with our um systems. there's a release uh mechanism so we can make releases of the toolbox for instance um that people can download um and so if if you uh you know would like to include packages in the topics organization we are happy to um uh and you know talk to us about that and we can um discuss um that the benefit of that is mostly that um you know maybe more people can find it um because it's going through our um as available on our organization page and you know we can work with you to set up some things like the automated release process um for example uh but you don't have to if you if you want to maintain your own GitHub repository that contains code and just links to the top of toolbox packages that's totally fine as well that's a great way to do it in that case we'd still love to know about it you know get in touch um through the discussions or you know submit an example to the top ofbox gallery um those all sound great uh those would all be great ways to um you know help inform us and ultimately the topics community about what you are doing. Um so uh with that I think I'll um you know end my little discussion of the contribution workflow there. Like I said get in touch if you have any questions um and yeah we hope to see you in our various um you know spaces uh in the near future. >> Okay great. Thank you, William. Um, I think we transition over now to the discussion uh part of this session and we have a couple of questions that were posed through the chat. Um, and I have put them here in a separate sheet for myself so I can overlook them. And these relate to either graph flood to the transact software or they are more general in questions. And I would now go through some of these questions or basically all of them. And then since there are not that many uh I also added some questions that we think are interesting for either you guys or for potential um people that would later on uh look at the recording of this when they want to get involved. So one question was uh could and I start with graphlut could graph be used useful for investigating how topographic perturbations such as those associated with falting influence river hydraulics and channel adjustment. So a relatively general question to you Boris. >> Um thanks a lot for the question. Yeah uh it can basically that was one of the idea of developing graph flood. We know that when you have a fault or a perturbation, you change the slope of your river but also the width or eventually the shear stress or the things like that and graph is there to measure them. Uh so if you think the width of your river has been affected by the fault or any other perturbations then you can measure it with graphlet. >> Great. Maybe uh I continue with a question that goes in a similar direction asking whether graph flood could be used in an inverse modeling framework where observed valley geometry is used to constrain the shear stress distribution and erosion parameters needed to reproduce the modern landscape. More specifically, could this approach help to calibrate stream power model parameters and explain spatial variations in incision rates that are not captured by channel steepness alone? Yeah. So, um it will require a bit of work uh because you would you could consider making it for example a a landscape evolution model by using graph flood calculating share stress and using it to apply erosion and so on. But it's a bit more work because there's um there are there's a difference between describing the field of sheer stress we have on on a landscape. That's very first order. You just need to think about precipitation rate and friction and so on. But converting that share stress into erosion or sediment transport is a bit more difficult because you need to assume some information for example about grain size about how you transfer uh this chest stress into incision is my incision linked to um sediment abrasion is my incision linked to plucking and so on. So it is definitely possible but I would say that graph flood in that case is just the first step and then you can use it with whichever the physical law to to invert. >> Okay cool. Yeah, I guess the speed of the calculations make it suitable for inversion schemes in general. And when you say that you're developing something based on GPUs that would be even uh favorable more more favorable for that. >> That's one of the idea. >> Exactly. So you had uh Wolfkang had a question how to choose a suitable value for the time step in graph flood and how does it depend on cell size of your DM? >> So theoretically it's linked to what we call the CFL conditions the Kuran Fredish levy conditions. It's a physical way that say that if your velocity uh of your floor for a given pixel is too fast and if it can goes in one time step belong the size of your um of your pixel the model will be unstable. So basically there's a physical way to calculate the maximum stable time step. uh it's a flow velocity divided by uh dx or something like that and it works relatively well for graph flood but because we decouple the physical flow from the propagation of drench area it's not the ultimate answer so long story short trial and errors is the best way uh to get the best time step for your given landscape >> okay as often the other question I have down here is are there plans to use graphlut in a landscape evolution model >> there are plans uh there were plans a lot before when I started to develop graphlet it's tricky that's uh the main issue it's tricky because um our flow adapts instantly to any change of uh of sheer stress of erosion so you quickly start to develop some over steepening and once you have an over steeping steepening you destabilize the model you converge of all your flow and everything crashes so Why did I ported graphlet to GPU and to gain more speed? It's one of the main reason is to try and create a landscape evolution model. So hopefully it will come as soon as possible. But uh yeah, if you have some ideas, if you if you want to develop your your landscape evolution model using graph as an hydraulic uh uh engine, you're more than welcome to to try and contact me and share your results if you manage to do it. So hopefully as soon as possible. >> Okay, good. Well, thanks Boris. Um, uh, I will now transition over to some questions related to transact. Also, one that involves transact and graph flood and, uh, but it's possible to continue asking questions via the chat even those that address graphlet. So, there was one question um, relating is there a specific threshold used to identify channel width? Uh, I imagine the result highly depends on DEM resolution, but is there a recommended minimum resolution or grid size needed to obtain reliable channel width estimates? >> Okay. Uh, Bastian, I don't know if you have a >> you would be better to to answer this question. So what I've been observing but it's an empirical observation on the the site I've been studying with graph is that you need at least three four five pixel wide channel to get a relevant uh width calculation below that uh you start to see a lot of artifacts um so or or like you start to measure channel that don't really exist or are sub pixel so basically if you are interested in a very big river like the Danube tube you can get a 25 m 10 m DM and you will still get a reasonable estimate of of the width because the river itself is like 50 m 100 200 m wide but if you are in the Alps and you are you you are looking at small river small stream it's better to have a metric resolution >> it depend on the object you're trying to measure I'd say that below two pixels you definitely get nothing below four pixels you get a lot of noise and once Once you you reach four five pixels of uh of of channel width um you're good to measure it continuously. >> Yeah. Yeah. Makes sense. I mean it depends always on the on the wavelength of a landform that we are interested in and channels are typically wavelengths of tens [snorts] of meters or whatever. Okay. The other question transact is not yet available as a repository in the tupo toolbox git organization. Will it become available? >> Uh [clears throat] yeah sure. So yes the toolbox is still in development. I didn't give so much thought to how to share it but if it's better for the community to share directly to put toolbox I think can be a great idea. >> Okay that's that's good to hear. >> Yeah perhaps adding to that uh I I didn't look at the repository before but I've seen that now that it's available via git. Um so there are some uh you know special ways to to wrap your code into a toolbox um which is provided by by Melbourne. You can have a look at the bankful mapper for example um which where we have also done that uh to make it compatible and it in [clears throat] the end it enables uh actually that it's shared also on the file exchange so [clears throat] it's better visible to the MATLAB community. On the other hand, uh you can just download it as a toolbox and it will you know you can install it by doing that and it will even install requirements and in that case these requirements would be top of toolbox. So it's everything is you know handled automatically and the user doesn't need to uh care about any dependencies and and stuff like that. >> Perfect. Um Bastian uh question I have how long did it take you to develop transact? Uh well I started in the start of my PhD so maybe four years uh ago but basically um I've learned how to code in MATLAB and developing this uh this toolbox so it was a great way and to put to box it's very good environment to to start to be involved in coding but it was a continuous development with my skills also in coding but I would say all during all my PhD so four years. >> Okay. But not entirely just the development. You also did research with the right. >> Yeah, it was a project. It evolved also with my PhD funings. So what I needed during my PhD. So yeah, it was great uh a great uh hike. >> Excellent. Uh also question from my side. Do the transact functions that you have developed also contain help text that explain their behavior as well as easily executed examples to play around with the functionality as we know it from the toolbox functions. [clears throat] >> Yes. So I'm almost done with all the documentation as a function. So with text and syntaxes examples uh all I need maybe is to to some notebooks like to to have some application. But yeah, boy, I always have to to improve the the documentation, but uh the majority is done. >> Okay, great. Thank you. Thanks for contributing. Uh now I have a couple of more general questions and um one user asked which Pytopo toolbox metrics are most useful for identifying tectonic controls on river geometry and longitudinal profiles. I guess that addresses the question of which Pytopo toolbox metrics exist at the moment. William, you're muted. What do you >> Yeah. So, um [clears throat] um the basically the the metrics that we have um I'm kind of scrolling through them at the moment that uh you can compute. We uh saw yesterday you can do uh kai analysis with um pitopa toolbox. All of that stuff works um as well as it does in mat lab more or less. We have um also the ksn computation um which you can use. uh we don't have a lot of examples using that in Python yet but it's uh fairly straightforward. Um we have all of the channel um the stream profile smoothing stuff as well which can be useful when um doing those kind of analyses. Um so I'd say those are kind of the main main like metrics that we have that you can compute. >> The point the pattern the point patterns on streams is not yet included. Correct. >> Well so it it is it is as of um yesterday. Um so the I I included a few because I was trying to make um the Python uh version of the nickpoint analysis that Wolfcoin presented yesterday. Um so there there's a little bit of the the the PPS stuff that's um exists in Python now um as a result of that effort. Um so it that's nowhere near um kind of feature complete with mat lab and I think we're going to uh pursue a slightly different um trajectory there because of how it can interface with the python kind of statistical ecosystem. Um but that's that's uh you know in progress if that's something that people are interested in especially that is a good place to get started contributing as well because it could use some work >> right maybe another question along those lines a little bit um uh and that is asking you know there are matlab and python versions and you know one could think of that the target is to produce a python version which is a onetoone copy of the matlab version. Um, is that the goal? And if not, because maybe the the environments in Python and MATLAB have uh strengths and weaknesses that could uh lead in different directions. What maybe um features uh are you know what are better done in Python uh or in MATLAB or is is it all the same or something? >> Yeah. So I think the um uh we have so far implemented pytopa toolbox largely by um taking functionality that's available in mat lab and trying to make it available in python. Um that's partly because that's uh you know what we know people want is the stuff that people use in mat lab's top of toolbox. Um we're not however committed to uh the two packages being identical forever. um you know if people want to see functionality um that's better expressed in Python or you know people the graph flood stuff is maybe a good example where um because Boris is a Python user a lot of that stuff hits Python first and um so it'll uh take um you know some now we need to kind of go the other direction if we want to make that stuff available in mat lab um so you know I can envision that for other tools as well if someone says you know we want to do this particular analysis this and um you know I want to implement it but I'm a python user that might show up in pytopa toolbox and not be available in mat lab. Um the same is true of course for um future developments of the mat lab type of toolbox um as well. Um I would say the the uh one thing that's very different between the two platforms is kind of the um the the way in which they manage the ecosystems a little bit. I'd say um you know mat lab you're kind of expecting mat lab the platform to provide a lot of stuff for you. So for instance like machine learning algorithm statistics um you know like a GPU um framework um that comes from mat lab basically and so we're able to use that um kind of stuff when it's available in mat lab um when it uh arrives in the the the versions that we have um in python a lot of that stuff is developed in in separate libraries by people so you know we depend on a few um kind of core libraries for doing some of these things. Um, but there's lots of other thing libraries available. Boris has been doing a lot of work on um his GPU stuff for instance um that uses other libraries and um packages as well to do that kind of thing. the PPS as I said like um I think my goal for the Python implementation of the PPS is not to provide so many uh functions for fitting models to the PPS um data within Pytopa toolbox but to provide the functionality that allows you to connect that to the statistical packages um available in Python. And I think a similar approach, you know, if the R package got further developed, that would also be a a similar approach there because there's a lot of tools available in R to do more complicated um statistical analysis. Um that we don't want to just, you know, replicate all of that functionality within Pytopa toolbox. We want to let people use the tools that they're used to in the languages that they use to do that kind of analysis. >> Cool. Uh along those lines, I don't know if we covered that point yesterday. Um but there is a link to QGIS right of >> this is something that we um we're very interested in. Um we have a QGIS plugin um that has been created uh by one of our students um Teao uh and it works. You can load it into QGIS and start using QGIS data to do um certain analyses. it's not um very fully fleshed out and for instance it's also not available through the the plug-in manager. So it requires kind of manually installing it at the moment in QGIS. But that's something we're definitely interested in is um you know because we now have this Python version um and QGIS plugins kind of um you can write Python plugins for QGIS pretty easily. So um you know that's an opportunity as well for um you know other users, people who are QGIS users. um and interested in using Tupper toolbox um to get involved in in trying to flush that out a little bit um more because I think that'd be a really good um good integration where you know you can you get all the the power of a GIS system but you can access some of these more advanced um topographic analysis tools that top toolbox provides. >> Right. Yeah. And now I think this is the right way to ask another question I had in mind or maybe some others have in mind but haven't dared to ask them yet because there's so many opportunities to contribute and to write functions and functionalities in Python and MATLAB and R and for QGIS and this and that. It's a lot of work. So um how useful uh are chat GPT and related AI to convert MATLAB functions into Python functions for example? Um that's a good question and one that I have not really experimented too much with. Um you know I think that uh we definitely uh you know if people if people are able to get good results doing that like we're we're happy to take contributions. I know Boris has been using some um uh coding agent kind of stuff to do some of his work um and that has proved um you know moderately successful I think. So, um, you know, have at it. I' be curious to see what people are able to to produce. Um, I don't know if Boris has more he wants to say about how that experience goes. >> Yeah. Um, so when it started to come out, I used it a lot cuz, you know, it was new and everything. Um, now I have a bit of experience with it. It's I'm definitely not an expert and it's definitely not my main tool. you it can produce results uh but you still need to control a minimum what you're doing otherwise it quickly explode into a gigantic pile of code um that becomes unmaintainable but if you manage to save time with it which I did on some occasions and definitely lost a lot of time on other occasions when I tried to damage control um can be good so I'm not an expert in agent coding especially because it's becoming way too expensive for my academic salary. But if you're it's a tool, so if you manage to to to control it, it's nice. If you want to see a nice example of uh uh the mess, but also the time it can save, you can check the an old pull request we I did for the swath tool in Lipto toolbox. Um, in two words for the story, I needed a swath tool to be as like I really needed them quick for a specific project. So I coded them with a nigantic um with clothes if you if you must know and the pull request was was really a mess. So I could use the result but then it it required like probably I don't know a few days of accumulated time to make it um good quality enough for for for to box. Okay, >> this this is a story. You do whatever you want with with that result. But long story short, it's not a magic box. Yeah, it's it's good, but you need to know what you're doing. >> Okay, I think we covered that topic. Now, uh I have two more things that I would like to uh address. And one is assume that I do have some tools available uh that I think might be useful for the community, you know, either Matlab, Python or whatever. and then I would like to contribute them but I need some help to integrate them into the existing architecture and maybe I was also I'm not so an expert on GitHub and so on. Is it possible to get to receive help in order to contribute and who should I contact? >> Yeah, so that is um essentially my job as a research software engineer um working with top toolbox is you know to help you all uh sort out how to do that um kind of thing. So yeah, you know, get in touch with me. Uh um uh I don't know if my email has been posted somewhere um that people can access, but uh you can certainly get in touch through any of the GitHub um platforms. Um if you're able to do that, uh shoot me an email. I think my email is actually in the contribution guidelines. >> Yeah, I think you're also as a first author of the 2.23. >> Yes, indeed. So my my my email is there. You can find my email somewhere. Um >> perfect. and you know let uh shoot me an email and tell me what you're trying to do and I will help you figure that out. Um you know I'm happy to walk through using git with you or u to help you you know edit some code or something um whatever you offer. Thanks. Um, there's also the question [laughter] that was coming from Wolf Gang I guess. Should we still use the comment function on the TT or Tupo to Toolbox blog or do you prefer users to use the Git discussion forum? Wolgang, tell me your opinion. [laughter] >> Tell me your opinion about your own question. Yeah, >> u, yeah, a lot of people still, or not only still, I wouldn't say that, but use the top toolbox. um and it has a comment function um and people ask question there. So uh the frequency at which I get questions you know vise uh it could be at one to two sometimes three per week. Um so it's good to see that uh people visit the the blog and use it as a resource. Um and I like to keep it like that. Um as long as we have not completely switched over to uh let's say a solution that works better. Um I think the discussion forum uh on git works better um or is better. But as you can see and as we we have seen from you know our attempt yesterday also to to instigate you to um to use it. still not like in the day-to-day uh or it's git is not like the day-to-day tool for for many of you. That's my you know what I perhaps take from that. Um and so please use the the topper toolbox comment function you know in in in the on the blog uh if you like if you want to you know if you want to make your question more visible also to others u and easier to link basically to perhaps existing uh solutions that are available on the gallery and stuff like this then please uh don't hesitate to use the discussion forum on the top toolbox GitHub count. So now we're running like both in parallel. Um but the other thing is the blog it's basically just me who's reading that um or um you know looking at the who has admin rights and so I can just uh um work on on this. You are more likely to get a quick reply when you use the discussion on git. And I should I'd also maybe like to say um that you know if if anyone has better ideas for like communication channels that they would like to hear or discuss top of toolbox over you know what kinds of things you know would be helpful. We kind of have an abortive discord server that we tried once but that never really took off because we the developers didn't really use it but if that's something that people are really interested in you know we can start that up again. um you know if an email list would be really useful um you know let us know what ways you would like to engage with the project because we can make those happen um and uh you know we're not we're not 100% committed to using the GitHub discussions for everything um if another platform would be more effective um for people to um participate. >> Yeah. [snorts] Yeah. Good point. Although I think that uh in the interest of also uh time and efficiency, it's also useful to maybe pick one and then stick to that rather than distributing to too many uh because it also requires time to invest in maintaining these >> but that can change and then pick another one or something. Okay. So I think most of the questions that I noted down including all those that I had or we had ourselves uh have been addressed and I would like to maybe finish this part of this discussion session with the last question that was posed here in the chat. Would it be possible to organize such webinars on tectonic geommorphology? And uh I think from my overview of the questions we had, there were quite a few that were addressing um the usefulness of channel steepness, how to comput it, sensitivity to certain thresholds etc. Uh so it's more about the scientific interpretation of these uh these analysis topographic analysis. So would it be possible to organize such webinars on tectonic geomorphology focusing on the interpretation of geomorphic indexes and their relationship with lithology, climate tectonic forcing and seismonics. It will also be highly valuable to cover the linkages among tectonic uplift, erosion rates, sediment production, sediment connectivity and landscape evolution, including how these relationships can be interpreted from topographic and geomorphic data sets. I apologize for reading out quickly, but uh the message is clear. So there seems to be um interest in this and I don't know Wolf Gang um maybe you already have even such uh recordings of previous talks on YouTube or something that address these topics. I don't I'm not sure. >> Yes, there are uh a couple of um lectures that I gave um well some of them are on YouTube some of them relate to tectonic geomorphology. So for example, if you have a bunch of uh basin by denudation rates from cosmogenic nuclides, how to uh go along with that? Um uh these were courses that I gave at the IIT ROI in India. Um so I think they are available on YouTube but I might put them together at some point and uh repost them. Um [clears throat] and otherwise I mean uh it would be interesting to hear from you uh like I mean yes we have mainly talked about techniques right here so and uh the technical issues um we have less or spend less time let's say on the interpretation um and yeah about the scientific uh part of of terrain analysis and from what I said at the beginning one of the main aims of top toolbox is to provide a you know an environment where you can uh experiment and and test models for example and to test hypothesis and that is what we often do as scientists. Um so if you have any ideas or what uh you would want to see in the future um again let us know. I'm not sure whether we can like do everything in uh it's uh it's a lot of work I can tell you to to prepare such workshops or or webinars in uh in particular when it comes to you know carefully in addressing scientific questions as well. Um but I think speaking for the entire team here uh we are definitely open to to do that. Yeah. >> Yes. Maybe just just one point um worth mentioning is the landscapes live which is around since 2020 and has a lot of YouTube recording of scientific talk around tectonic gemophology. It's not only tectonic gemology but very good timing tomorrow it's uh Benjamin Comfort and he will be giving a talk very much around modeling and topographic analysis. So uh feel free to check their website. They have like uh four years of recordings of uh scientific talk. >> Yes, exactly. Um I think the website is a little bit deserted but the YouTube channel uh has all the recordings available and uh and there are plenty of um plenty of talks to to listen to and at the end of the day it's really also the purpose of what we think the gallery is supposed to provide. So examples uh that can be worked through including uh the code including the the figures that are plotted from that etc etc. So these are resources for educational purposes that can be tailored towards specific um applications and anyone who's interested in contributing a gallery can do so. So this is really not that you have to uh deliver um uh code through GitHub that can be implemented in the Dupo toolbox. It can also be uh a script um preferentially including uh text like a MATL life script or a Jupyter notebook but that is more like an educational thing and working through an example. So this is really what the gallery is supposed to do and whenever you have something get in touch with will in order to find out how to to contribute. >> I want to add here that you know it's it's also a way to to increase uh your own visibility right we have on purpose we have uh you know that there's one of the first lines is the author and so forth. So it's something that where you can also you know demonstrate your work to others and uh showcase what you have done in particular um link it perhaps to a publication of yours and your publication will also receive the attention that it deserves. Yeah. >> Excellent. Yeah. I think I'm done with the discussions part now and so the final words I don't know Wolf Gang will >> final words I don't know um I can start [laughter] now first of all I I'd like to thank everyone for uh participating and uh yeah I've quite quite many uh who have now joined this webinar. I hope you found it interesting and you have learned something uh something that you can take along to your own studies and to your own work perhaps even to your own classrooms uh and and forward that and propagate it uh to others. I think that's um that's one thing to do um not sure whether some of you will or do use uh for example top toolbox in teaching also. So perhaps the stuff that we have presenting here uh might be useful for some of your classes too. Um uh on the other hand uh of course many thanks to the EGU for hosting us and for giving us the opportunity to give this webinar and yeah so thanks all and hoping to see you soon again. >> Yeah thank you. >> Thank you. Bye-bye. >> Thanks everyone.