A collection of tools for making 3D models and similar things out of GIS data. Designed to be run as a pipeline, though the individual tools can be run too.

Define a pipeline in a text file, with one command per line (there are examples in the examples/ directory). The commands should not have the landcarve prefix on them; for example:

# Decimate down to a small size
decifit --xy-steps=50

# Render to an STL
realise --xy-scale=0.01 --z-scale=0.01 --base=0.2 --solid

Then, run the pipeline with an input and output path, like so:

landcarve pipeline national-parks.txt input.geotiff output.stl

You can read more about the individual commands below. If you want to get going directly with an example, download the .asc file linked from the examples/london-tiles.txt pipeline, and run:

landcarve pipeline examples/london-tiles.txt tq3780_DSM_1m.asc output.stl

This isn't on PyPI yet, so clone the repository and run pip install -e ..

Installing GDAL can be a particular pain; if your OS offers it, I highly recommend installing a Python 3 GDAL package from there. On Ubuntu/Debian, this is python3-gdal.

There is no true "right scale" for geographic data due to map projections, so you need to take care if you want to get things coming out at an exact scale, or consistent with other prints.

Generally, the horizontal (X & Y) scale is determined by the number of rows/columns in your data, and all the tools will try and maintain its aspect ratio. decifit allows you to shrink it down while maintaining the ratio, while realise will simply map one column/row to one unit in the final model unless you use --xy-scale.

The vertical (Z) scale is more flexible, as generally printing with the same scale on Z as you have on X and Y will result in things that look too flat -the sense of perception of height we have for models is weird.

There's two ways of dealing with Z - always expanding it to a certain height in the resulting model (zfit), or scaling it consistently with a factor (--z-scale on realise).

zfit is for small, one-off items that aren't going to be directly compared to each other, and are large-scale geography -this is what I use for National Park miniatures, for example, as you always want to see the geographic detail, and I don't want the same scale across all the parks (otherwise the mountainous ones will cause the rest to look basically flat).

The --z-scale option, on the other hand, is for when you want to print a set of tiles that will all sit next to each other. Keep it the same, and your heights will all line up.

If you're printing an area that is all at elevation, you will need to either use zfit (which will auto-trim to the lowest point you have), or pass --minimum to realise to set the "base level" of your model. Anything below the minimum will be rendered as flat; this is also important if you have holes in your model that go below sea level (e.g. excavations).

Finally, realise that the runtime (and memory usage) of this code goes up rather quickly as you increase the size of the grid being used. Always use decifit as the first element in a pipeline, and try to keep under 1000 in each dimension; 500 tends to be a good tradeoff.

Takes a text file containing a list of URLs and downloads only the "interesting" ones to a local folder. Designed for use with the USGS National Map download feature.

Options:

• --xy-steps: Size to fit the raster within. Default: 1000

Takes an input raster and fits it within certain limits of X and Y, so there's at most --xy-steps rows and columns. Does not touch Z/values.

It will preserve aspect ratios - non-square inputs are scaled to the size of their longest edge.

Options:

• --nodata: NODATA boundary for input data. Default: 0

All the rest of the tools in the suite assume a NODATA value of -1000. If you have source data that is not aligned, use this pipeline step to set anything equal or lower to the value of --nodata you pass to the internal NODATA value.

Options:

• --snap: Quantization factor in projection units (XY symmetrical). Default: 1
• --void-distance: How far to search for void-filling neighbours. Default: 10

Takes one or more LAS (or LAZ) files, thins them (by highest return elevation), and turns them into a GeoTIFF DEM.

Options:

• --xy-scale: Scale factor for STL model in x/y axes. Default: 1
• --z-scale: Scale factor for STL model in z axis. Default: 1
• --minimum: Level to cut off detail and assume as base of model. Default: 0
• --base: Thickness of the base below the bottom of the model. Default: 1
• --simplify/--no-simplify: If simplification should be run. Default: --simplify
• --solid/--not-solid: If the model should be forced to a square tile with no holes. Default: --not-solid

The realise step takes a heightmap and renders it out as an STL file.

By default, the STL will have a dimension matching that of the input grid in all dimensions - so if your input is a 500x500 heightmap with values from 0 - 50, the resulting STL will have a dimension of 500x500x50.

To scale this linearly, use --xy-scale and --z-scale.

If all points on your model are above a certain elevation, use --minimum set at that elevation to shift the whole model downwards. The value of minimum will be what ends up at zero height on the model, on top of the base thickness. Any features that are below the minimum (but that have data) will be rendered flat.

--base sets the thickness of the base of the model in output units. It's recommended you have a base as most forms of manufacturing will need one.

Simplification is run on the STL model to try and merge flat areas together; if you don't want this, pass --no-simplify. The resulting model will have a lot more polygons, but you'll save the slow simplification step. The built-in simplification is quite basic; you may want to run it through another program and do a shape-preserving simplification if your model is too detailed to load into a slicer/pathing tool.

By default, areas that are set as NODATA in your heightmap will not be rendered with a base; this is to allow non-rectangular outputs from the model. If your goal is a set of tiles, though, set --solid to ensure you get a base; this will help make sure your output is perfectly square.

Options:

• --factor: Smoothing factor. Default: 1

Smooths heightmap data to remove jagged heights caused by reflections or laser errors. Only use if your data is not already cleaned up.

The higher the factor, the more the model is smoothed.

Options:

• --fit: New target height. Default: 1

Re-scales the Z axis (value) data so that it ranges between 0 and the value passed for --fit. As well as scaling the Z axis, this also includes shifting the whole model down so the lowest value is the new 0 (for data which is entirely at elevation).

Models printed using this will not have the same Z scale as each other. Only use this for models that are not meant to be joined together.