Guideline to reconstruct with heights (instead of PC) with City4CFD

[gsclara]

Current guidance in City4CFD wiki focuses on high quality point clouds, however most parts of the world do not have that priviledge. The idea would be to extend the wiki to reconstruct with height data, instead of point cloud.

• City to test, could be any city
• We just need a new section in the wiki for reconstructing with heights
• Contact point: Clara Garcia-Sanchez, Akshay Patil

[AkshayPatil1994]

I recently created this using a DEM and DSM to get "coarse point cloud-like" datasets, and I'd be happy to share the scripts I used to generalise the approach if you need!

[cmoon329]

Hi @gsclara , @AkshayPatil1994 ,

Thank you for sharing the new task description.

@gsclara
I would like to double-check if I have the right idea.
For this task, I'll

1. create a new Markdown file for the new 'reconstructing with heights' section
2. add a brief description
3. share it with you so that you can create the section

Could you let me know if I understood the scope correctly?

@AkshayPatil1994
I'd love to take a look at the scripts and datasets you used to get a better understanding of it. Could you please share them with me when you have a chance?

Please let me know if there's anything I missed.

Thank you.

Best regards,
Chaeyeon

[AkshayPatil1994]

Hi Chaeyeon,

I have included a short test case I developed for the city of Exeter, UK, for which I obtained a DTM online and wrote Python scripts to convert it into a point cloud and then reconstruct the domain as needed. The code needs a bit of cleaning up. Perhaps a nice pipeline would be for the user to prompt the name of the city and a radius of interest, or the coordinates of the bounding box, and for the code to automatically download the DTM from available options and directly port everything internally from DTM to LAS, ready for City4CFD. This could be relatively easy to do given high-resolution topographic data from open topography (so some API design for this), but I am sure you have a better way?

Link2Download: https://surfdrive.surf.nl/s/wxzLQdxWS2WSZnH

[cmoon329]

Hi Akshay,

Thank you for sharing the scripts and details.
It sounds like a very interesting task!

I will use your scripts to build a pipeline that automatically downloads the DTM based on user input (city name, spatial extent) and converts it into a point cloud.

I plan to wrap up my current task from Hide this week and will start on this next week. I expect to complete it within 2 weeks, but I will leave a comment if I need more time or any help along the way.

Thank you.

Best regards,
Chaeyeon

[cmoon329]

Hi @AkshayPatil1994 ,

I’d like to ask one question about this task.
I found a way to automate the pipeline to download a DSM and convert it to a PC based on the user input, using either Open Topography or Google Earth Engine as sources.

However, most of the datasets on these platform have a 30m resolution, making it hard to extract exact building heights. If my understanding is correct, the point cloud input for City4CFD should contain points for both the buildings and the terrain.

I will write the code to use local, high-resolution dataset when they are availble, but if a region only has 30m data, would it be okay to use that as a fallback?

[AkshayPatil1994]

Hi @cmoon329,

I think 30m resolution becomes a bit too coarse without a meaningful output, is there perhaps a hard stop /assertion check that the code can make saying "The data source is too coarse for decent reconstruction, so we suggest download higher resolution data from suggested sources "

@gsclara what do you think?

[cmoon329]

Thanks for your input, @AkshayPatil1994 .

From Google Earth Engine (GEE), it seems there are only two datasets available that provide high-resolution DSMs: England (1m res) and the Netherlands (AHN). I will make the code to 1. output DSMs when the input spatial extent falls within these regions.

For other regions, I came up with an alternative idea, which injecting/overlaying building height data (from Overture Maps) on top of a 30m DTM raster (from GEE). It seems technically feasible, but I'd love to hear your thoughts. Do you think this is a valid approach, and would it be useful for the main goal of this task?

[AkshayPatil1994]

That is a fair compromise between data availability (or the lack thereof) and a practical way forward. Overture Maps has an API rate limit on the number of calls you can make, but it's worth checking Open Topography for DSM and DTM data to get ground and building data. Anyways, I trust your judgement on the workflow, and we can always fine-tune once a simple pipeline is in place. Thanks!

[cmoon329]

Hi @AkshayPatil1994 ,

Sharing the code I've made for the point cloud generation.

Link:

• https://github.com/cmoon329/city4cfd_pointcloud_input_generation

Summary:
Since I couldn't find high-resolution DSM datasets that cover the entire world from both Open Topography and Google Earth Engine (GEE), I varied the DSM generation logic based on the target city.

• For cities in England/Netherlands: Downloads regional high-res DSMs directly from GEE.
• For other cities: Generates DSMs by merging DTM (from OpenTopo) with building footprint/height data (from OSM/Overture Maps)

You can check more details in the link above.

If you have any feedback or questions, please let me know!

Thank you.