Document not found (404)
+This URL is invalid, sorry. Please use the navigation bar or search to continue.
+ +This URL is invalid, sorry. Please use the navigation bar or search to continue.
+ +This open-source snakemake +workflow can be used to analyse environmental risks to infrastructure +networks using global open data.
+We can use open-gira to analyse open data on roads, railways and
+their exposure to river flooding, or electricity transmission lines and how
+they're affected by tropical cyclones.
open-gira is a work in progress.
This open-source snakemake +workflow can be used to analyse environmental risks to infrastructure +networks using global open data.
+We can use open-gira to analyse open data on roads, railways and
+their exposure to river flooding, or electricity transmission lines and how
+they're affected by tropical cyclones.
open-gira is a work in progress.
This guide will help you to install open-gira and its dependencies on your computer.
Once everything is installed, run the tests to ensure everything is linked up +properly:
+python -m pytest tests
+The major installation steps are to:
+open-giraInstall open-gira by cloning the repository:
+git clone https://github.com/nismod/open-gira.git
+This repository comes with a environment.yml file describing almost all of the
+software dependencies required to run open-gira.
There are several ways to manage Python versions and install libraries.
+conda lets you install different versions of Python
+and Python libraries and other dependencies.micrombamba
+a replacement for conda, and what the open-gira developers use.The recommended approach for open-gira is to install micromamba then use it
+to create and manage environments.
To install the required dependencies on a local machine, create the open-gira
+conda environment:
micromamba create -f environment.yml -y
+Then activate it:
+micromamba activate open-gira
+You're now ready to configure workflows and request outputs.
+If installing on a cluster, you can work as above, or, create a seperate +orchestrating environment containing only snakemake, e.g.
+micromamba create -n snakemake python=3.9 snakemake
+In this context, snakemake itself can manage the other required dependencies,
+creating other environments as necessary. To activate the orchestration
+environment:
micromamba activate snakemake
+To run the workflow on a cluster +you will need to provide a profile, +requesting targets as follows:
+snakemake --profile <path_to_cluster_config> -- <target_file>
+The following tools are not available through conda and must be installed separately.
exactextract is used for zonal statistics. Please see installation instructions here.
+Native installation on Windows is extremely difficult because of the difficulty building the +tools that open-gira relies on (binaries are not available for Windows).
+To avoid this issue, we recommend Windows users install via Windows Subsystems for Linux (WSL). +To install WSL, follow the install +instructions.
+Once WSL is installed, open-gira can be installed inside the Linux subsystem by following the +Linux install instructions.
+It is possible to run open-gira within a container using Docker. You must have Docker
+installed.
git clone https://github.com/nismod/open-gira.git
+docker build -t open-gira .
+We use the -it options to make the container interactive and give it a terminal,
+and the -d option to keep it running in the background (daemon mode).
+We give it a nice name we can refer to later, and make sure it is created using the
+image we just built.
+We tell it to run bash so that it has something to do and doesn't close immediately.
docker run -itd --name og open-gira bash
+We can now lauch a bash shell in our new container:
+docker exec -it og /bin/bash
+And once inside we can navigate to the open-gira folder.
+cd open-gira
+We're now ready to run snakemake, launch a python prompt, or do whatever else we like inside +the container.
+open-gira is comprised of a set of snakemake rules which call scripts and
+library code to request data, process it and produce results.
The key idea of snakemake is similar to make in that the workflow is
+determined from the end (the files users want) to the beginning (the files
+users have, if any) by applying general rules with pattern matching on file and
+folder names.
A example invocation looks like:
+snakemake --cores 2 -- results/wales-latest_filter-road/edges.geoparquet
+Here, we ask snakemake to use up to 2 CPUs to produce a target file, in this
+case, the edges of the Welsh road network. snakemake pattern matches
+wales-latest as the OSM dataset name and filter-road as the network type we
+want to filter for.
To check what work we're going to request before commencing, use the -n flag:
snakemake -n --cores 2 -- results/wales-latest_filter-road/edges.geoparquet
+This will explain which rules will be required to run to produce the target +file. It may be helpful to visualise +which rules are expected to run, too.
+The snakemake configuration details are in config/config.yml. You can edit
+this to set the target OSM infrastructure datasets, number of spatial slices, and
+hazard datasets. See
+config/README.md
+and the documentation for each workflow for more details on the configuration
+variables.
open-gira can currently create connected road, rail and electricity grid
+networks from open data.
We can create a topologically connected road network for a given area from +OpenStreetMap (OSM) data. The resulting +network can be annotated with data retrieved from OSM (e.g. highway +classification, surface type), along with data looked up from user-supplied +sources (e.g. rehabilitation costs).
+.osm.pbf file to input data directory.osmium tags-filter for elements matching the filter file (see configuration, below).keep_tags as configured.To specify a desired network:
+bundled_data/transport, these supply
+information that is used to gap-fill or extend what can be determined from OSM alone.config/config.yaml:
+infrastructure_datasets map should contain a key pointing to an .osm.pbf
+file URL for desired area. There are currently entries for the planet,
+for (some definition of) continents and several countries. We use
+the geofabrik service for continent and
+country-level OSM extracts.network_filters.road.
+This file specifies which elements
+(nodes, ways or relations) to keep (or reject) from the multitude of data
+in an OSM file. See the filter expressions section
+here
+for more information on the syntax of these files.keep_tags.road. This list of strings specifies which
+tags (attributes) to retain on the filtered elements we extract from
+the .osm.pbf file.slice_count. This controls the degree of parallelism possible.
+With it set to 1, there is no spatial slicing (we create the network in
+a single chunk). To speed network creation for large domains, it can be
+set to a larger square number. The first square number greater than your
+number of available CPUs is a good heuristic.And to create the network, by way of example:
+snakemake --cores all -- results/egypt-latest_filter-road/edges.geoparquet
+The process for creating a rail network is essentially the same as for road.
+We can create a topologically connected rail network for a given area from +OpenStreetMap (OSM) data. The resulting +network can be annotated with data retrieved from OSM, along with data looked up +from user-supplied sources (e.g. rehabilitation costs).
+.osm.pbf file to input data directory.osmium tags-filter for elements matching the filter file (see configuration, below).keep_tags as configured.To specify a desired network:
+bundled_data/transport, these supply
+information that is used to gap-fill or extend what can be determined from OSM alone.config/config.yaml:
+infrastructure_datasets map should contain a key pointing to an .osm.pbf
+file URL for desired area. There are currently entries for the planet,
+for (some definition of) continents and several countries. We use
+the geofabrik service for continent and
+country-level OSM extracts.network_filters.rail.
+This file specifies which elements
+(nodes, ways or relations) to keep (or reject) from the multitude of data
+in an OSM file. See the filter expressions section
+here
+for more information on the syntax of these files.keep_tags.rail. This list of strings specifies which
+tags (attributes) to retain on the filtered elements we extract from
+the .osm.pbf file.slice_count. This controls the degree of parallelism possible.
+With it set to 1, there is no spatial slicing (we create the network in
+a single chunk). To speed network creation for large domains, it can be
+set to a larger square number. The first square number greater than your
+number of available CPUs is a good heuristic.And to create the network, by way of example:
+snakemake --cores all -- results/egypt-latest_filter-rail/edges.geoparquet
+Note that the nodes file, results/egypt-latest_filter-rail/nodes.geoparquet
+will by default contain the stations and their names as recorded in OSM.
To create an electricity grid we rely heavily on gridfinder +data. This dataset provides transmission and distribution edges with +substantial global coverage. It also contains a set of 'targets' or electricity +consuming areas, derived from Night Time Lights +(NTL) satellite imagery. Our other major data source for electricity grid creation is the World Resources Institute's (WRI) +powerplants database.
+The workflow currently divides network creation by country. One may request one +country, or several. Note that neighbouring countries' networks are not +connected with one another.
+There aren't currently any options to configure when creating an electricity +grid.
+Here's an example grid creation command for Puerto Rico:
+snakemake --cores all -- results/power/by_country/PRI/network/edges.geoparquet
+Note that the folder name under by_county is an ISO 3166 Alpha-3 country
+code,
+specifying the country in question.
Having created networks, we can analyse the risks that certain hazards +(currently tropical cyclones and flooding) may present to these networks.
+The flooding risk analysis pipeline starts by creating an infrastructure +network (road or rail) as described previously. Please refer to this section to +configure the network creation.
+The pipeline has the following core steps:
+The hazard component of the analysis is configurable in config/config.yaml:
hazard_datasets contains hazard names pointing to files of hazard layers.
+These layers are currently flood rasters (inundation depths for a given return
+period). Add or amend an entry pointing to file containing the rasters you
+wish to consider.hazard_types contains an identical key referencing the hazard types.
+This is currently limited to flood only.direct_damages.asset_types contains any assets you wish
+to calcuate direct damage costs for. Currently implemented assets are
+available in src/open_gira/assets.py as the classes inheriting from
+Assets.direct_damages.curves_dir is set to a path containing damages
+functions per asset type, organised by hazard type. See
+bundled_data/damage_curves for an example.<asset_type>.csv, e.g.
+road_unpaved.csv. Their format is exemplified by
+bundled_data/damage_curves/flood/road_unpaved.csvTo generate a network, split by the hazard raster grid we might issue something like:
+snakemake --cores all -- results/splits/<dataset_name>_filter-<network_type>/hazard-<hazard_name>/slice-<slice_number>.geoparquet
+For the egypt-latest OSM dataset, road network, aqueduct-river hazard set and slice 0, that would be:
snakemake --cores all -- results/splits/egypt-latest_filter-road/hazard-aqueduct-river/slice-0.geoparquet
+To request an evaluation of Expected Annual Damages (EAD) as a function of
+hazard Return Period (RP) for a given slice, we can request something like:
For example (with a config.slice_count of 9):
snakemake --cores all -- results/direct_damages/egypt-latest_filter-road/hazard-aqueduct-river/EAD_and_cost_per_RP/slice-5.geoparquet
+And to compute all the slices for a given domain and then aggregate to country level (admin level 0):
+snakemake --cores all -- results/egypt-latest_filter-road/hazard-aqueduct-river/EAD_and_cost_per_RP/agg-sum/admin-level-0.geoparquet
+For more possible outputs please refer to the detailed documentation and the
+rules defined in workflow/rules/.
This analysis intersects electricity grids with tropical cyclone wind speed +risk. Network creation is described previously. The hazards are event-based (as +opposed to return period map based).
+config/config.yaml:
+storm_sets should contain a storm set name, pointing to a JSON file.
+This JSON file should contain an empty list to process all storms for
+this storm set, or a list of string storm IDs if only a subset is
+required.transmission_windspeed_failure as a list of wind speeds in
+meters per second at which to fail edges.See comments in the config/config.yaml file for other less crucial
+configuration options.
Below you will find example commands necessary to create outputs for the case of +PRI (Puerto Rico) and the storm 2017242N16333 (Irma, 2017) from the +IBTrACS historic cyclone dataset. Later steps will trigger earlier, +dependent steps if necessary.
+snakemake --cores 1 -- results/power/by_storm_set/IBTrACS/storms_by_country_impacted.jsonsnakemake --cores 1 -- results/storm_tracks/IBTrACS/tracks.geoparquetsnakemake --cores 1 -- results/power/by_country/PRI/storms/surface_roughness.tiffsnakemake --cores 1 -- results/power/by_country/PRI/storms/IBTrACS/max_wind_field.ncsnakemake --cores 1 -- results/power/by_country/PRI/exposure/edges_split.geoparquetsnakemake --cores 1 -- results/power/by_country/PRI/exposure/IBTrACS/2017242N16333.ncsnakemake --cores 1 -- results/power/by_storm_set/IBTrACS/by_storm/2017242N16333/exposure_by_target.ncsnakemake --cores 1 -- results/power/by_storm_set/IBTrACS/exposure.txtTODO: Add additional rules to facilitate comparison across countries and cyclone datasets.
+TODO: Document these additional rules.
+snakemake --cores 1 -- results/power/by_storm_set/IBTrACS/by_storm/2017242N16333/outage_map/outage_map_by_threshold.gifsnakemake --cores 1 -- results/power/by_storm_set/IBTrACS/by_storm/2017242N16333/wind_field.pngThere also exist other plotting and mapping steps to visualise intermediate and final outputs. Refer to workflow/rules/analysis for a description of these.
open-gira comes with a few small utilities for data processing and file management.
You can remove intermediate files by running the clean rule.
snakemake --cores 1 -R clean
+As standard we use the .geoparquet format to store vector data on disk.
+Unfortunately common GIS software such as QGIS may not yet support this file
+format. To convert file(s) to geopackage, use:
python workflow/scripts/pq_to_gpkg.py <path_to_geoparquet_1> <path_to_geoparquet_2> <...>
+This will write .gpkg files beside their source .geoparquet.
matplotlib plots can be interactive (zoom, pan, etc.), but not as static
+images. Some rules produce pickled plot files. To view these, use:
python workflow/scripts/unpickle_plot.py <path_to_pickled_plot>
+The bash script archive_results.sh can be used to back up analysis results.
+Here's an example usage:
./archive_results.sh results/ /mnt/backup/open-gira
+This will create a timestamped folder in path of the second argument, e.g.
+/mnt/backup/open-gira/2023-07-24T101756+0100, containing an archive of results
+(excluding input files downloaded from the internet) and a README describing the
+state of the repository at the time of archive creation.
This guide will help you to install open-gira and its dependencies on your computer.
Once everything is installed, run the tests to ensure everything is linked up +properly:
+python -m pytest tests
+It is possible to run open-gira within a container using Docker. You must have Docker
+installed.
git clone https://github.com/nismod/open-gira.git
+docker build -t open-gira .
+We use the -it options to make the container interactive and give it a terminal,
+and the -d option to keep it running in the background (daemon mode).
+We give it a nice name we can refer to later, and make sure it is created using the
+image we just built.
+We tell it to run bash so that it has something to do and doesn't close immediately.
docker run -itd --name og open-gira bash
+We can now lauch a bash shell in our new container:
+docker exec -it og /bin/bash
+And once inside we can navigate to the open-gira folder.
+cd open-gira
+We're now ready to run snakemake, launch a python prompt, or do whatever else we like inside +the container.
+ +The major installation steps are to:
+open-giraInstall open-gira by cloning the repository:
+git clone https://github.com/nismod/open-gira.git
+This repository comes with a environment.yml file describing almost all of the
+software dependencies required to run open-gira.
There are several ways to manage Python versions and install libraries.
+conda lets you install different versions of Python
+and Python libraries and other dependencies.micrombamba
+a replacement for conda, and what the open-gira developers use.The recommended approach for open-gira is to install micromamba then use it
+to create and manage environments.
To install the required dependencies on a local machine, create the open-gira
+conda environment:
micromamba create -f environment.yml -y
+Then activate it:
+micromamba activate open-gira
+You're now ready to configure workflows and request outputs.
+If installing on a cluster, you can work as above, or, create a seperate +orchestrating environment containing only snakemake, e.g.
+micromamba create -n snakemake python=3.9 snakemake
+In this context, snakemake itself can manage the other required dependencies,
+creating other environments as necessary. To activate the orchestration
+environment:
micromamba activate snakemake
+To run the workflow on a cluster +you will need to provide a profile, +requesting targets as follows:
+snakemake --profile <path_to_cluster_config> -- <target_file>
+The following tools are not available through conda and must be installed separately.
exactextract is used for zonal statistics. Please see installation instructions here.
+ +Native installation on Windows is extremely difficult because of the difficulty building the +tools that open-gira relies on (binaries are not available for Windows).
+To avoid this issue, we recommend Windows users install via Windows Subsystems for Linux (WSL). +To install WSL, follow the install +instructions.
+Once WSL is installed, open-gira can be installed inside the Linux subsystem by following the +Linux install instructions.
+ +open-gira is comprised of a set of snakemake rules which call scripts and
+library code to request data, process it and produce results.
The key idea of snakemake is similar to make in that the workflow is
+determined from the end (the files users want) to the beginning (the files
+users have, if any) by applying general rules with pattern matching on file and
+folder names.
A example invocation looks like:
+snakemake --cores 2 -- results/wales-latest_filter-road/edges.geoparquet
+Here, we ask snakemake to use up to 2 CPUs to produce a target file, in this
+case, the edges of the Welsh road network. snakemake pattern matches
+wales-latest as the OSM dataset name and filter-road as the network type we
+want to filter for.
To check what work we're going to request before commencing, use the -n flag:
snakemake -n --cores 2 -- results/wales-latest_filter-road/edges.geoparquet
+This will explain which rules will be required to run to produce the target +file. It may be helpful to visualise +which rules are expected to run, too.
+The snakemake configuration details are in config/config.yml. You can edit
+this to set the target OSM infrastructure datasets, number of spatial slices, and
+hazard datasets. See
+config/README.md
+and the documentation for each workflow for more details on the configuration
+variables.
open-gira can currently create connected road, rail and electricity grid
+networks from open data.
To create an electricity grid we rely heavily on gridfinder +data. This dataset provides transmission and distribution edges with +substantial global coverage. It also contains a set of 'targets' or electricity +consuming areas, derived from Night Time Lights +(NTL) satellite imagery. Our other major data source for electricity grid creation is the World Resources Institute's (WRI) +powerplants database.
+The workflow currently divides network creation by country. One may request one +country, or several. Note that neighbouring countries' networks are not +connected with one another.
+There aren't currently any options to configure when creating an electricity +grid.
+Here's an example grid creation command for Puerto Rico:
+snakemake --cores all -- results/power/by_country/PRI/network/edges.geoparquet
+Note that the folder name under by_county is an ISO 3166 Alpha-3 country
+code,
+specifying the country in question.
The process for creating a rail network is essentially the same as for road.
+We can create a topologically connected rail network for a given area from +OpenStreetMap (OSM) data. The resulting +network can be annotated with data retrieved from OSM, along with data looked up +from user-supplied sources (e.g. rehabilitation costs).
+.osm.pbf file to input data directory.osmium tags-filter for elements matching the filter file (see configuration, below).keep_tags as configured.To specify a desired network:
+bundled_data/transport, these supply
+information that is used to gap-fill or extend what can be determined from OSM alone.config/config.yaml:
+infrastructure_datasets map should contain a key pointing to an .osm.pbf
+file URL for desired area. There are currently entries for the planet,
+for (some definition of) continents and several countries. We use
+the geofabrik service for continent and
+country-level OSM extracts.network_filters.rail.
+This file specifies which elements
+(nodes, ways or relations) to keep (or reject) from the multitude of data
+in an OSM file. See the filter expressions section
+here
+for more information on the syntax of these files.keep_tags.rail. This list of strings specifies which
+tags (attributes) to retain on the filtered elements we extract from
+the .osm.pbf file.slice_count. This controls the degree of parallelism possible.
+With it set to 1, there is no spatial slicing (we create the network in
+a single chunk). To speed network creation for large domains, it can be
+set to a larger square number. The first square number greater than your
+number of available CPUs is a good heuristic.And to create the network, by way of example:
+snakemake --cores all -- results/egypt-latest_filter-rail/edges.geoparquet
+Note that the nodes file, results/egypt-latest_filter-rail/nodes.geoparquet
+will by default contain the stations and their names as recorded in OSM.
We can create a topologically connected road network for a given area from +OpenStreetMap (OSM) data. The resulting +network can be annotated with data retrieved from OSM (e.g. highway +classification, surface type), along with data looked up from user-supplied +sources (e.g. rehabilitation costs).
+.osm.pbf file to input data directory.osmium tags-filter for elements matching the filter file (see configuration, below).keep_tags as configured.To specify a desired network:
+bundled_data/transport, these supply
+information that is used to gap-fill or extend what can be determined from OSM alone.config/config.yaml:
+infrastructure_datasets map should contain a key pointing to an .osm.pbf
+file URL for desired area. There are currently entries for the planet,
+for (some definition of) continents and several countries. We use
+the geofabrik service for continent and
+country-level OSM extracts.network_filters.road.
+This file specifies which elements
+(nodes, ways or relations) to keep (or reject) from the multitude of data
+in an OSM file. See the filter expressions section
+here
+for more information on the syntax of these files.keep_tags.road. This list of strings specifies which
+tags (attributes) to retain on the filtered elements we extract from
+the .osm.pbf file.slice_count. This controls the degree of parallelism possible.
+With it set to 1, there is no spatial slicing (we create the network in
+a single chunk). To speed network creation for large domains, it can be
+set to a larger square number. The first square number greater than your
+number of available CPUs is a good heuristic.And to create the network, by way of example:
+snakemake --cores all -- results/egypt-latest_filter-road/edges.geoparquet
+Having created networks, we can analyse the risks that certain hazards +(currently tropical cyclones and flooding) may present to these networks.
+ +This analysis intersects electricity grids with tropical cyclone wind speed +risk. Network creation is described previously. The hazards are event-based (as +opposed to return period map based).
+config/config.yaml:
+storm_sets should contain a storm set name, pointing to a JSON file.
+This JSON file should contain an empty list to process all storms for
+this storm set, or a list of string storm IDs if only a subset is
+required.transmission_windspeed_failure as a list of wind speeds in
+meters per second at which to fail edges.See comments in the config/config.yaml file for other less crucial
+configuration options.
Below you will find example commands necessary to create outputs for the case of +PRI (Puerto Rico) and the storm 2017242N16333 (Irma, 2017) from the +IBTrACS historic cyclone dataset. Later steps will trigger earlier, +dependent steps if necessary.
+snakemake --cores 1 -- results/power/by_storm_set/IBTrACS/storms_by_country_impacted.jsonsnakemake --cores 1 -- results/storm_tracks/IBTrACS/tracks.geoparquetsnakemake --cores 1 -- results/power/by_country/PRI/storms/surface_roughness.tiffsnakemake --cores 1 -- results/power/by_country/PRI/storms/IBTrACS/max_wind_field.ncsnakemake --cores 1 -- results/power/by_country/PRI/exposure/edges_split.geoparquetsnakemake --cores 1 -- results/power/by_country/PRI/exposure/IBTrACS/2017242N16333.ncsnakemake --cores 1 -- results/power/by_storm_set/IBTrACS/by_storm/2017242N16333/exposure_by_target.ncsnakemake --cores 1 -- results/power/by_storm_set/IBTrACS/exposure.txtTODO: Add additional rules to facilitate comparison across countries and cyclone datasets.
+TODO: Document these additional rules.
+snakemake --cores 1 -- results/power/by_storm_set/IBTrACS/by_storm/2017242N16333/outage_map/outage_map_by_threshold.gifsnakemake --cores 1 -- results/power/by_storm_set/IBTrACS/by_storm/2017242N16333/wind_field.pngThere also exist other plotting and mapping steps to visualise intermediate and final outputs. Refer to workflow/rules/analysis for a description of these.
The flooding risk analysis pipeline starts by creating an infrastructure +network (road or rail) as described previously. Please refer to this section to +configure the network creation.
+The pipeline has the following core steps:
+The hazard component of the analysis is configurable in config/config.yaml:
hazard_datasets contains hazard names pointing to files of hazard layers.
+These layers are currently flood rasters (inundation depths for a given return
+period). Add or amend an entry pointing to file containing the rasters you
+wish to consider.hazard_types contains an identical key referencing the hazard types.
+This is currently limited to flood only.direct_damages.asset_types contains any assets you wish
+to calcuate direct damage costs for. Currently implemented assets are
+available in src/open_gira/assets.py as the classes inheriting from
+Assets.direct_damages.curves_dir is set to a path containing damages
+functions per asset type, organised by hazard type. See
+bundled_data/damage_curves for an example.<asset_type>.csv, e.g.
+road_unpaved.csv. Their format is exemplified by
+bundled_data/damage_curves/flood/road_unpaved.csvTo generate a network, split by the hazard raster grid we might issue something like:
+snakemake --cores all -- results/splits/<dataset_name>_filter-<network_type>/hazard-<hazard_name>/slice-<slice_number>.geoparquet
+For the egypt-latest OSM dataset, road network, aqueduct-river hazard set and slice 0, that would be:
snakemake --cores all -- results/splits/egypt-latest_filter-road/hazard-aqueduct-river/slice-0.geoparquet
+To request an evaluation of Expected Annual Damages (EAD) as a function of
+hazard Return Period (RP) for a given slice, we can request something like:
For example (with a config.slice_count of 9):
snakemake --cores all -- results/direct_damages/egypt-latest_filter-road/hazard-aqueduct-river/EAD_and_cost_per_RP/slice-5.geoparquet
+And to compute all the slices for a given domain and then aggregate to country level (admin level 0):
+snakemake --cores all -- results/egypt-latest_filter-road/hazard-aqueduct-river/EAD_and_cost_per_RP/agg-sum/admin-level-0.geoparquet
+For more possible outputs please refer to the detailed documentation and the
+rules defined in workflow/rules/.
open-gira comes with a few small utilities for data processing and file management.
You can remove intermediate files by running the clean rule.
snakemake --cores 1 -R clean
+As standard we use the .geoparquet format to store vector data on disk.
+Unfortunately common GIS software such as QGIS may not yet support this file
+format. To convert file(s) to geopackage, use:
python workflow/scripts/pq_to_gpkg.py <path_to_geoparquet_1> <path_to_geoparquet_2> <...>
+This will write .gpkg files beside their source .geoparquet.
matplotlib plots can be interactive (zoom, pan, etc.), but not as static
+images. Some rules produce pickled plot files. To view these, use:
python workflow/scripts/unpickle_plot.py <path_to_pickled_plot>
+The bash script archive_results.sh can be used to back up analysis results.
+Here's an example usage:
./archive_results.sh results/ /mnt/backup/open-gira
+This will create a timestamped folder in path of the second argument, e.g.
+/mnt/backup/open-gira/2023-07-24T101756+0100, containing an archive of results
+(excluding input files downloaded from the internet) and a README describing the
+state of the repository at the time of archive creation.