So, thank you for inviting me. It's a great pleasure to be with you, although I hope that I would be in Prague, but I'm not too far away. I'm in the Netherlands currently.

So what I would like to do during this talk, I would like to share with you some of the work that we have done studying spatial temporal changes in topography and using the new technologies that emerged over the past 20 years.

And before I start, let me just acknowledge the fantastic work that our team at the Geo4ALL laboratory has done on this topic, and especially Anna Petrasova and Vasik Petras, who some of you probably know in person.

So let's recall that we had a huge transformation in elevation mapping over the past 20 years, mostly driven by LIDAR, but several other technologies as well. And we now have global DEM, several of them available.

We have many countries and regions that were mapped by airborne LIDAR, leading to one meter resolution, digital elevation models. And most recently, we have many locations where centimeter resolution, digital

elevation models were created using structure from motion from drone imagery. And the good news is that a lot of this data is open, and it's available through excellent repositories, such as open topography.

Some of the US data are available through digital coast or USGS3D nation. So not only do we have new higher resolution data available for many regions, but the data are much more complex than they used to be.

So we are now going beyond the contours representing the bare ground topography towards the full multiple thematic layers representation of multiple elevation surfaces.

So we have digital surface models, which include all the features and vegetation and structures. We can also extract the bare ground, and we can extract thematic layers, such as vegetation canopy surface. So we are talking now about digital canopy models or digital crop models.

And very important are digital urban models that include bare ground with buildings, other urban structures, and also the structure of urban vegetation. So let's just look at the one example where we can apply the techniques developed for geomorphometry for

bare ground digital elevation models to the new types of digital elevation models such as digital canopy models. So on the left, you can see the geomorphons applied to digital surface models and used to extract the individual trees.

On the right, we have a centimeter resolution smooth digital crop surface, which shows the rows of crop and we can extract the

infor the structural information about the surface and link it to the stress in the crop or the nutrient status of the crop. And we can go beyond surfaces, we can take the multiple return dense point

cloud, and we can create a voxel model that represents the structure of the vegetation. So, for example, in this case, Vasek has derived a three dimensional fragmentation index using the point cloud, and you can see the animation slicing through this voxel model, showing this fragmentation index for the for the relatively complex for a structure in this area.

In addition to individual single snapshot digital elevation models, we are now able to collect multi temporal elevation models and study

dynamic topography such as sandy coast, landslides, debris floats, but also urban development, especially in those areas that are rapidly evolving. And of course vegetation growth and removal.

But the working with multi temporal elevation models can be challenging because the data are fragmented in space and time, and we have much more limited historical data compared to, for example, two dimensional land cover data.

But this is changing. And as you can see from this time series of elevation models, we have had maybe every 20 years, one digital elevation model or a contour map available. As we are going over the year 2000 with emergence of LIDAR technology and beyond 2020 we are getting almost annual or even denser elevation data.

So here is an illustration of the fragmentation like a spatial fragmentation that we can see in the time series of LIDAR data from 1997 through 2011.

And what you can see here, these areas that have the these blocky rectangles. Those are the areas that have no elevation data. So, so you can see that the coverage is very fragmented in this location. And of course we

can address this fragmentation by pitching in the existing data, either the closest fast data or the closest future. So that we get an acceptable multi temporal digital elevation model, but we always need to keep in mind that if there

was a short term change in topography, or in this case in the vegetation canopy, it wouldn't be captured by this dynamic model.

So what we have done over the past 10-15 years, we have developed a lot of tools for measuring and analyzing topographic change in GRASS GIS. So there are many modules that allow you to process the LIDAR time series and to go from fragmented point clouds to consistent digital elevation models.

And then to study the changes in topography. The standard approach is DEM differencing. So we just subtract two digital elevation models. But this is really useful only for comparing the elevation between two elevation models.

But if we are talking about time series, which we now have easily 15-20 even more digital elevation models, then we need to look at some other ways how to summarize and extract the information about the dynamics of this landscape.

So, for example, we have developed some tools to automatically extract features from these digital elevation models and generate a gradient field of the migration of that feature. We also use the time series analysis based on the raster data to extract valuable

information and analysis from this time series. And I will show an example of that. And of course for this we use GRASS GIS temporal framework and you might have been working with it already in some of the workshops.

And some of these tools and analysis are captured in a booklet by Eric Hardin and a team which shows all of these techniques and what commands to use. And some of these scripts, although some updates are of course needed and I will talk about those later.

So let's look at some of these analyses of time series of elevation data and coastal areas are an excellent example, excellent discipline where we can

study these dynamic landscape, because the coastal topography is continuously modified by nature, as you can see here, and the changes are on the order of several meters, even after a single storm. Or significant changes to topography are also made by humans, as they are trying to

do adaptive management in this environment and protect the structures and infrastructure on the islands. So that's why the coastal areas essentially have pioneered LiDAR mapping.

And the LiDAR data in the coastal areas go really back into 1996, where the larger regions were were mapped using the LiDAR technology. But as you can see, at the late 90s, the technology still produced

rather sparse point clouds, and the structures were not really very well represented. But it was much better than anything else that was previously available.

Now, after 2000, as the LiDAR technology evolved, you can see that the point density has increased dramatically, and we can now get structure of the buildings and great detail also in the vegetation and the geomorphology of the coastal features.

Again, all these time series is available as open data through NOAA Digital Coast. It is also cataloged in open topography and in North Carolina spatial data download. And more recently, additional data are being collected using the drones, using the GPS transects, that's the local communities that they do,

and also some agencies collect the aerial imagery and derive topographic data using stereo photogrammetry.

So then, once we take this data and do the processing and interpolation of the digital elevation model into consistent resolutions and consistent levels of detail, we can extract core surface that represents the volume of sand that has not moved, and the operation is extremely simple.

It's just recording the minimum elevation that was measured for each cell over the entire time series. We can also define envelope surface, which is the maximum elevation for each cell,

and then we can define a dynamic layer within which the topography has evolved. And then, if we extract shoreline from core and envelope, we can define the shoreline fluctuation band, and I will show you how that looks like and why it is important.

And we can, of course, run additional statistics on each grid cell time series to compute, for example, the regression and get the rate of change. So here is an illustration or cross section that shows you the relationship between the

core surface, envelope surface, and the actual elevation surface, which is shown in gray color. So you can see that the erosion or retreat of the shoreline is not really a simple process, but it goes back and forth.

Let me check the check. Okay. All right. So here is the shoreline fluctuation band. As I mentioned, it's extracted from the core and envelope shoreline, and you can see that

now we can identify those homes or structures that are within this shoreline fluctuation band, which can be much larger than the long term erosion rate. And I will soon show you how these homes are identified using the core and envelope approach.

And because of these rapidly moving shorelines, there is an adaptive management associated with these regions,

where if there is a really valuable home and it gets into that shoreline fluctuation band, they either need to be moved or they need to be removed. So this was a particular home, which was in a movie, Nights in Rodente, and in 2010, they actually collected enough money to move it.

The relocation of such a home costs about $100,000, so usually it's quite unusual to do it. So this is a more common situation where the homes are let go. This is not

really a good outcome, because they need to close the beach if something like this happens. This was a video that was posted about a month ago, and I think it was shared broadly on the internet. And it is really in this area that we have studied, and this has been happening there for quite a few years.

So how do we map this area or these homes that are being lost and are being either moved or rebuilt? Again, we can do it from a long time series using the core and envelope approach,

where the changes in the elevation or the vertical difference in the core and envelope is beyond certain threshold. We can identify the homes that were lost or rebuilt or new homes that were built.

And then we can use the map of time of minimum and raster map of time of maximum to identify whether that change was lost or whether that change was built.

So this can be a fully automated process without the need to look at the individual orthophotos for each time step. So then we have another type of dynamic features on the coast of North Carolina and many others in the world, and that's the sand dunes.

The challenge in the management of this type of dynamic landscapes in this particular case is that the dune is locked within a boundary of a state park. So as it is moving, it is actually moving beyond the boundaries of this park. So there is a

big question how to preserve this dune, how to preserve this feature, and what would be the best management strategy. And to find out the strategy, we really need to first understand how this dune is moving and quantify the rates of change.

And to do this, we were able to build a time series of elevation models that we keep adding to as the new data are becoming available. So the first data sets are from digitized contours and photogrammetry. So you can see that 20 years time interval, then that shortens to

just a couple of years, and then we have these really dense time series of LiDAR data that is augmented by the drone imagery. And we were also able to find some spot elevation for the 1953 and for the 30s.

So using these data and using the processing tools in GRASS GIS, we were able to create a dynamic model of this dune that we keep updating every year when we get the new data sets.

So this just shows the snapshot between 1974 and 2017. So you can see the dune moving southward, we can quantify its rate of movement, also the transformation from crescentic

dune to parabolic dunes, and we can also compute the sand volumes and other characteristics. So what we were able to also do, look at the loss of elevation. What was quite

interesting that the loss of elevation has been very steady, really linear, just about 30 centimeters per year. It is slowing down a little bit because it's already quite low. And looking at the historical data, what we were able to identify was that this feature was actually a short term feature. It was a transient, very large

dune that is now coming back to its original state that was a much lower dune. And eventually, there were periods of time when this dune was forested. So what are the implications now for management?

So the entire idea of the state parks is to preserve the landscapes. But as I mentioned here, because the dune is migrating, this is very challenging. So the dune is managed as a living landscape. And to preserve the dune and

manage it as a living landscape, when the dune moves towards the boundaries of the park, then they do a sand relocation project. That means that the sand that reaches the park boundaries is

actually moved and relocated to the area from where the wind can bring the sand back onto the dunes. And that way, the volume of sand within the park has been preserved over time. So there were two such projects in 2003 and more recently in 2019.

And as I mentioned, we keep updating the workflows and a subset of the tools and the processing of this temporal data is now available through a Jupyter notebook.

And you can find it under this link. And the Jupyter notebook will cover just a small part of the dune and some of that analysis of core and envelope and also the space time cube.

So let's now move on to a different environment. As you might have noticed, not everything is changing everywhere. So doing the LIDAR mapping, which is quite expensive, is really done only every five years or even less than that.

But to lower the cost and keep the DEM up to date, we can identify those areas that have significant, that probably have significant elevation change. And we can update only those using much more lower cost technologies, such as drone mapping with structure from motion.

So for coastal regions, we can do sea level rise prediction updates or storm impacts mitigation by updating only those areas that have changed.

And it is even more important in urban regions where we have human induced topographic change, which can be quite significant and can also influence, for example, the stormwater runoff. So here is an area in the, like a typical suburban development around the

triangle in North Carolina where people are moving in, the school is being built. So you can see that this area has the trees removed, but also the topography was graded

and changed, and essentially the flow of the streams has changed quite a bit as well as the digital surface model. So how can we do this? Corey White has came up with an idea of generating a priority queue for DEM mapping and updates.

And what he has proposed is to use the satellite imagery to identify a land cover change and classify this land cover change into categories that most likely lead to also a three dimensional change.

So for example, if you are seeing the forest clearing, there may be a high probability that also that there will be a development if it is an urban area, and that also the land will be graded, leading to a change in topography.

And given the density, the temporal, like short time step that we have for satellite data, this can be done pretty frequently, and we can find and locate and identify those areas that need to be

flown, let's say by drone at very low cost, and the digital elevation model then can be updated very efficiently. So here you can see one watershed in our area, and those pink areas show the location where forest has been removed, so quite a few areas.

And one of the biggest areas which was identified as a high priority is this location, and you can see that change within one year.

And we have assumed that it will be associated also with a three dimensional change. So then we can define actually an over area like a polygon for drone mapping, including an overlapping area, so that we can generate the digital surface model point cloud from the structure from motion, and then use this to update the digital elevation model.

So here you can see the, the location that is being flown in the middle is a five centimeter resolution digital surface model that shows in great detail, the sedimentation pond on the construction site.

And this is a classified point cloud that allows us to extract the structures from the point cloud and derive the background elevation model that would be needed for updating the existing DEM.

So here you can see the, the extracted and mapped the digital elevation model like a background digital, let's call it digital terrain model.

This is the existing three meter resolution model with a very natural topography. And here you can see an updated digital elevation model where we are using a smooth fusion to ensure the smooth connection between along the boundaries of these two digital elevation models.

And this has been developed using the GRASS GIS modules. It is available in as a Jupyter notebook, and it is, it is designed as a web ODM plugin.

What is really nice about this, that the cost of such mapping is such so low and the drones are so widely available that these updates can be done either through the construction monitoring

because many construction sites these days are monitored using drones, or it can be even done using citizen science. Here you can see the change in the slope, so quite a dramatic change in the topography.

And also that leads to quite a different distribution of surface water flow. So instead of these naturally flowing streams, we now have a construction site, which before, before the drainage is installed would have some standing water.

There are some pretty big stormwater and sediment control ponds, so we can check whether the water from this construction site actually gets into these stormwater ponds. So, so you can see that it is quite important to update this information, and the changes are on the order of several meters.

Now to actually study these changes and provide simple intuitive tools to, let's say, work with the stakeholders to understand how the landscape and water flow will change when we change topography.

We have developed over the years, an environment called tangible landscape, and Anna Petrasova is really took over the original idea and is now the lead developer of this system.

So here you can see how it works, we have the, it combines a physical model of the landscape with Kinect scanner, where the model is scanned, then it is imported geo reference and combined with existing geospatial data, if necessary, or we just ran the models on the, on this data,

and then the results are reprojected on the surface of the model. So in this case, it essentially we have, we are computing digital elevation models deriving contours computing surface water flow

and filling in the depressions all in real time, as the, as the topography is being changed. And again, you can find more about the system and how to build it and how it all works in the, in the book by Anna and team.

So this just shows the structure of the, of the tangible landscape software. So, so it is again like the physical model and Kinect scanning is coupled with GRASS GIS, and that makes all of the modules that are in GRASS available for coupling with the system.

Then we also couple the system with Blender for realistic three dimensional rendering, and I will show you some examples how that works.

So the first question that I always get about the tangible landscape is how to create the three dimensional models. There are many ways. So, most of the models that we do, especially those that are large are CNC routed. And we are using equipment that is available at our design school, and it can be a very precise digital physical model of the landscape.

And if we need a malleable digital elevation model so we want to actually change the topography. We use CNC routing to create the mold, then we fill the mold with polymeric sand, which holds its shape very nicely.

And then we can create a quite precise digital physical model of the landscape, which can be changed by moving around the sand. Then, of course 3D printing is another option. It's suitable more for smaller digital elevation models because the material is quite costly.

And the simplest way that we actually use quite often is just to take a pile of sand, provide the contours and provide the feedback from the existing digital elevation model on where to add the sand and

where to take it off and create the digital elevation model, create the physical model just by hand using that feedback.

And it works pretty well. So another thing that Anna has done over the past couple years is to develop many different interactions for the model, well beyond modifying the surface. So you can use the markers to identify points, for example, from which to compute a view shed or to identify trails and optimize the trail location.

This is another way how to define lines is using the laser pointer. We can also define areas either using the colored clay or using colored patches of felt.

So here is one application that has been developed for studying changing topography and interaction between changing topography and landscape processes.

And in this case, this was a simple game to provide the people and stakeholders on the coast better understanding with storm surge flooding impacts. So we give people a little bit of sand and they are supposed to protect their homes, and then

we randomly breach the fortune. And then they can see how even a very small breach can cause widespread flooding. And what was really interesting is that in this aerial photo, we show this flooded area and several breaches over these fortunes that happened in 2018 during Hurricane Florence.

And we have been providing this game and they have been using it for two years already. So they had a really good understanding about what would happen and how that would work.

We also have used this and developed applications for education. And here we developed structured activities. For example, here the students are supposed to find a highest point from where water would flow into a given area.

Then they were supposed to change the topography in such a way so that it holds the most water. And in the last one, in the previous one, they were identifying the morphometric features. And this was the last one, which was actually the most successful, where they had to use the

information and feedback from tangible landscape to do grading, to change the topography according to the prescribed contours. So this was done for landscape architecture students. And there is, again, like a paper about it with this small video.

As I mentioned, we have coupled tangible landscape with 3D realistic rendering using Blender. So what you can do, you can

put various colored patches on the landscape and each color is associated with different types of vegetation or structures. And then this information is scanned and classified. And then the associated features are rendered in three dimensions on the screen. And you can do either a view

from the air or you can define a viewing point and direction and you can view the landscape from the ground. And as we have it coupled with GRASS GIS, we can compute all kinds of quantitative

metrics that would define this landscape. For example, the area of the lakes, the volume that is in the volume of water in the lakes, but also various structural landscape properties.

And then finally, you can explore these different models, tangible landscape setups by going through the interactive 3D scan of our laboratory, which has not been yet officially released, but I'm providing here the link

so that you can maybe even give us some feedback if you have and what your experience was. So when you click on some of these points of interest, then a Google

Slides will pop up and will give you more information about that particular setup, about that particular application. And some of these slide decks have also linked to Jupyter Notebooks, where you can actually do that particular modeling. So here are some links to the Geo Visualization Lab virtual tour, and to

our laboratory and the Center for Geospatial Analytics, where all this is being done. Okay, so that's all my presentation, and I will welcome your questions. And this is just the

three-dimensional model of the, like the virtual model of the lab that I wanted to show you. As I mentioned, it's not yet fully finished, but you can go really from one setup to another.

This one is with the Blender, and really explore how these different tangible landscape applications are set up. Okay, thank you, thank you for your attention.