[gdal-dev] Total Viewshed and GDAL

Luis Felipe Romero felipe at uma.es
Mon Apr 29 02:15:53 PDT 2024


Hi,

After several months working on other things, I have returned to an application of the total viewshed.

What I have done is use Tamas Szekeres' C++ code as a template and, at least on Windows, I have managed to compile and run. With just a GPU it is quite fast: on a 6000x6000 raster it does the calculations in less than a minute, although it can take two or three hours without a GPU on a PC with 8 
cores. In multiple viewshed mode, the time is reduced proportionally.

On this page I have uploaded more info and some images of a national parks project I am working on. https://github.com/luisfromero/ppnn

It's still a little premature to share it, because I need to do many more tests, compile in Linux, etc., but my main doubt right now is that I need to shape the part of the tool necessary for calculating solar radiation:

In the so-called "mode 3" the visible horizon from each point is calculated, which together with other parameters such as tilt and head allow the calculation of solar radiation in the presence of shadows.

My question is the file format for the horizon data. For now, it's just an array with dimx·dimy·360 byte (using a one degree precision), and I'm considering different formats. For example, save it simply as an output file (compressed, preferably), but it could be compacted into a geopackage (as a 
metadata file or in a database, I don't know). They could even be saved in a multiband raster along with the elevation, tilt and head data from the  DEM. In short, everything necessary to then easily calculate the radiation in a fixed period of time.

Thanks for any suggestions!

Luis F Romero


More information about the gdal-dev mailing list