In our series “Maps and mappers of the 2016 calendar” we will present throughout 2016 the mapmakers who submitted their creations for inclusion in the 2016 GeoHipster calendar.
Asger Sigurd Skovbo Petersen
Q: Tell us about yourself
A: I work at a small Danish company called Septima which I also cofounded back in early 2013. I have been in the geo business since 2004 when I received my masters degree (MScE) from the Technical University of Denmark.
I do development, consulting, and data analysis. One of my primary interests is to find new ways of utilizing existing data. This interest really took off when I worked as the sole R&D engineer at a data acquisition company which had a massive collection of data just sitting there and waiting to be upcycled. At this job I got a lot of experience working with quite big LiDAR, raster, and vector datasets, and developing algorithms to process them effectively.
Q: Tell us the story behind your map (what inspired you to make it, what did you learn while making it, or any other aspects of the map or its creation you would like people to know).
A: When processing the second Danish LiDAR-based elevation model, the producing agency released some temporary point cloud data at a very early stage.
My curiousity was too big to leave these data alone, and with a LASmoons license of Martin Isenburg’s LAStools, it was easy to process the 400km^2 las files into 40cm DTM and DSM. And then the usual open source stack helped publishing a hillshaded version as an easy to use web map.
This web map was widely used and cited, as it was the only visible example of the coming national DEM for quite a while. The old model was 1.6m resolution, and with a new resolution of 0.4m a lot of details were revealed, which were not visible in the old model. In the following months we actually received quite a few notes from archaeologists, who had discovered exciting and previously unknown historic stuff just by browsing our map.
Hillshades are the go-to visualisation of DEMs. Probably because they can be easily processed by almost any raster-capable software, and because they are very easily interpreted. However they can also hide even very big structures depending on the general direction of the structure.
This made me want to find a better way to visualise the data so our archaeological friends could get even more information from the new data.
I then read a heap of papers on the subject and decided to try out a visualisation based on Sky View Factor. At the time I didn’t find any implementation that I was able to use, so I ended up implementing my own. (I later discovered that SAGA had a perfectly good implementation, so I could have just used QGIS. But hey, then I wouldn’t have had the fun implementing my own 🙂 )
I did a lot of tests using the Sky View Factor on the new DTM, but I couldn’t make it work as well as I had hoped. By coincidence I ran it on the DSM in an urban area, which gave a very interesting result. This effect is basically what makes the GeoHipster map look different from most other shaded DSMs.
Q: Tell us about the tools, data, etc., you used to make the map.
A: The map consists of several layers: a standard hillshade, a Sky View Factor, building footprints, and water bodies.
The Sky View Factor layer was made using a custom algorithm implemented in Python using rasterio and optimized for speed using Cython. As mentioned this could probably just as well have been processed using SAGA, for instance, through QGIS. The hillshade layer was made using GDAL and the vector layers did not require any special processing.
QGIS was used to symbolize and combine the layers using gradients, transparency and layer blending.
Data used are the national Danish DEM and the national Danish topological map called GeoDanmark. Both datasets are open and can be freely downloaded from Kortforsyningen. Sadly most of these sites are in Danish only – maybe some clever hidden trade barrier.
Here is an online version of my map. For the online version I had to change the symbolization a bit as producing tiles from QGIS Server doesn’t work very well with gradients.
After submitting the map to the GeoHipster 2016 calendar I have been working on coloring the vegetation to get a green component also. There are no datasets for vegetation which include single trees, bushes etc, so I made a python script to extract and filter this information from the classified LiDAR point cloud.
This new map can be seen here in a preliminary version.