2) Creating a systemd service (automatic startup)

The goal: run OpenTopoData as a service that restarts automatically and is managed with systemctl.

Create the service:

sudo nano /etc/systemd/system/elevation.service

Content:

[Unit]
Description=OpenTopoData elevation API
After=network.target

[Service]
Type=simple

User=deploy
Group=deploy

WorkingDirectory=/home/www/opentopodata
ExecStart=docker run --rm --volume '/home/www/opentopodata/data:/app/data:ro' --network host opentopodata:1.10.0

# if we crash, restart
TimeoutSec=30
Restart=on-failure

TimeoutStartSec=120
TimeoutStopSec=30

# output goes to the journal
StandardOutput=journal
StandardError=journal

SyslogIdentifier=opentopodata-elevation

[Install]
WantedBy=multi-user.target

Key points for the service to work:

• pin the version (1.10.0)

• do not use the -t and -i options

• use the host network (--network host) so the API can be called locally on port 5000; the -p option is then useless

• the service user must belong to the docker group

Enable and start it:

sudo systemctl daemon-reload
sudo systemctl enable elevation.service
sudo systemctl start elevation.service
sudo systemctl status elevation.service

Alternative: installing without Docker

It can be installed without Docker:

https://www.opentopodata.org/notes/running-without-docker/

Performance is the same with or without Docker.

3) Choosing and configuring the topographic data

OpenTopoData supports several datasets. Two useful pages:

• dataset list (public API):
  https://www.opentopodata.org/ (§Public API)

• dataset sizes:
  https://www.opentopodata.org/notes/dataset-sizes/

After adding data, the Docker image has to be rebuilt.

Datasets in use: history and current setup

In the current version, only SRTM30 data is used, for the whole world.

Previous setup:

• Europe: EU-DEM

• North America: SRTM30

• Elsewhere: SRTM90

Comparing the data

• SRTM30: the most precise data (2x slower than SRTM90)

• SRTM90: approximate data (the fastest)

• EU-DEM: 25 m, but less precise than SRTM30 (3x slower than SRTM90)

4) World: SRTM (SRTM30m / SRTM GL1)

OpenTopoData dataset (SRTM):

• https://www.opentopodata.org/datasets/srtm/

Create an account on OpenTopography:

• https://portal.opentopography.org/dataCatalog?group=global

Total size of the tif files: 126 GB.

The data sits on S3-compatible storage; to download it, first install aws-cli:

curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
sudo ./aws/install

Then, in the target directory:

aws s3 cp s3://raster/SRTM_GL1/ . --recursive --endpoint-url https://opentopography.s3.sdsc.edu --no-sign-request

The global VRT file is downloaded too (-32768 for areas without data).

Optimizing: converting to compressed GeoTIFF

If the files are not tif already (or to optimize I/O), convert them to compressed GeoTIFF:

Install GDAL:

sudo apt install gdal-bin

Conversion:

for f in *.hgt.zip; do gdal_translate -co COMPRESS=DEFLATE -co PREDICTOR=2 "$f" "${f%.*}.tif"; done

Cleanup:

rm *.zip

Time to generate the TIF files for North America: 30 min.

Deploying the data to the server and rebuilding

Copy the generated tif files to the server, update the config.yaml file and rebuild the Docker image:

make build

Restart the service:

sudo systemctl restart elevation.service

Test (if the server accepts local requests):

curl "http://localhost:5000/v1/srtm30m?locations=57.688709,11.97"

5) Europe: EU-DEM

OpenTopoData dataset (EU-DEM):

• https://www.opentopodata.org/datasets/eudem/

Download: since January 2024, EU-DEM is no longer distributed by Copernicus. The OpenTopoData page above says where to get a copy of the dataset.

Steps:

1. Download all the data at once

2. Extract all the files

3. Keep only the TIF files

Total size of the tif files: 38 GB.

Adding EU-DEM to OpenTopoData

In the opentopodata directory:

mkdir ./data/eudem

Copy the TIF files into it (rename them as described here: https://www.opentopodata.org/datasets/eudem/)

Create the config.yaml file declaring the dataset:

datasets:
- name: eudem25m
  path: data/eudem/
  filename_epsg: 3035
  filename_tile_size: 1000000

Rebuild and restart the container:

make build && make run

Test (if the server accepts local requests):

curl "http://localhost:5000/v1/eudem25m?locations=57.688709,11.976404"

6) Getting a specific continent for SRTM

Reference:

https://github.com/graphhopper/graphhopper/blob/master/core/src/main/resources/com/graphhopper/reader/dem/

Rename the file extension and prepend the URL to each line.

7) Other elevation services (alternatives)

• https://pypi.org/project/elevation/

• https://open-elevation.com/