After some experimentation we have decided not to use Intake data catalogs, so this repository is not currently being updated or maintained. We may repurpose it for use with another data catalog system, but for the time being, please see the PUDL Data Access Documentation for instructions on how to access the data we publish.
This repository houses a data catalog distributing open energy system data liberated by Catalyst Cooperative as part of our Public Utility Data Liberation Project (PUDL). It uses the Intake library developed by Anaconda to provide a uniform interface to versioned data releases hosted on publicly accessible cloud resources.
- Raw FERC Form 1 DB (SQLite) -- browse DB online
- PUDL DB (SQLite) -- browse DB online
- Census Demographic Profile 1 (SQLite)
- Hourly Emissions from the EPA CEMS (Apache Parquet)
To track ongoing development of the PUDL Catalog you can follow these issues in the main PUDL repository:
See also:
You can install the PUDL Catalog using conda:
conda install -c conda-forge catalystcoop.pudl
or pip:
pip install catalystcoop.pudl-catalog
The pudl_catalog
registers itself as an available data source within Intake when
it's installed, so you can grab it from the top level Intake catalog. To see what data
sources are available within the catalog you turn it into a list (yes this is weird).
import intake
import pandas as pd
from pudl_catalog.helpers import year_state_filter
pudl_cat = intake.cat.pudl_cat
list(pudl_cat)
[
'hourly_emissions_epacems',
'hourly_emissions_epacems_partitioned',
'pudl',
'ferc1',
'censusdp1tract'
]
Printing the data source will show you the YAML that defines the source, but with all the Jinja template fields interpolated and filled in:
pudl_cat.hourly_emissions_epacems
hourly_emissions_epacems:
args:
engine: pyarrow
storage_options:
simplecache:
cache_storage: /home/zane/.cache/intake
urlpath: simplecache::gs://intake.catalyst.coop/dev/hourly_emissions_epacems.parquet
description: Hourly pollution emissions and plant operational data reported via
Continuous Emissions Monitoring Systems (CEMS) as required by 40 CFR Part 75.
Includes CO2, NOx, and SO2, as well as the heat content of fuel consumed and gross
power output. Hourly values reported by US EIA ORISPL code and emissions unit
(smokestack) ID.
driver: intake_parquet.source.ParquetSource
metadata:
catalog_dir: /home/zane/code/catalyst/pudl-catalog/src/pudl_catalog/
license:
name: CC-BY-4.0
path: https://creativecommons.org/licenses/by/4.0
title: Creative Commons Attribution 4.0
path: https://ampd.epa.gov/ampd
provider: US Environmental Protection Agency Air Markets Program
title: Continuous Emissions Monitoring System (CEMS) Hourly Data
type: application/parquet
The source.discover()
method will show you some internal details of
the data source, including what columns are available and their data
types:
pudl_cat.hourly_emissions_epacems.discover()
{'dtype': {'plant_id_eia': 'int32',
'unitid': 'object',
'operating_datetime_utc': 'datetime64[ns, UTC]',
'year': 'int32',
'state': 'int64',
'facility_id': 'int32',
'unit_id_epa': 'object',
'operating_time_hours': 'float32',
'gross_load_mw': 'float32',
'heat_content_mmbtu': 'float32',
'steam_load_1000_lbs': 'float32',
'so2_mass_lbs': 'float32',
'so2_mass_measurement_code': 'int64',
'nox_rate_lbs_mmbtu': 'float32',
'nox_rate_measurement_code': 'int64',
'nox_mass_lbs': 'float32',
'nox_mass_measurement_code': 'int64',
'co2_mass_tons': 'float32',
'co2_mass_measurement_code': 'int64'},
'shape': (None, 19),
'npartitions': 1,
'metadata': {'title': 'Continuous Emissions Monitoring System (CEMS) Hourly Data',
'type': 'application/parquet',
'provider': 'US Environmental Protection Agency Air Markets Program',
'path': 'https://ampd.epa.gov/ampd',
'license': {'name': 'CC-BY-4.0',
'title': 'Creative Commons Attribution 4.0',
'path': 'https://creativecommons.org/licenses/by/4.0'},
'catalog_dir': '/home/zane/code/catalyst/pudl-catalog/src/pudl_catalog/'}}
Note
If the data has not been cached this method might take a while to finish depending on your internet speed. The EPA CEMS parquet data is almost 5 GB.
To read data from the source you call it with some arguments. Here we’re
supplying filters (in “disjunctive normal form”) that select only a subset of
the available years and states. This limits the set of Parquet files that need
to be scanned to find the requested data (since the files are partitioned by
year
and state
) and also ensures that you don’t get back a 100GB
dataframe that crashes your laptop. These arguments are passed through to
dask.dataframe.read_parquet()
since Dask dataframes are the default container for Parquet data. Given those
arguments, you convert the source to a Dask dataframe and the use .compute()
on that dataframe to actually read the data and return a pandas dataframe:
filters = year_state_filter(
years=[2019, 2020],
states=["ID", "CO", "TX"],
)
epacems_df = (
pudl_cat.hourly_emissions_epacems(filters=filters)
.to_dask()
.compute()
)
cols = [
plant_id_eia,
operating_datetime_utc,
year,
state,
operating_time_hours,
gross_load_mw,
heat_content_mmbtu,
co2_mass_tons,
]
epacems_df[cols].head()
plant_id_eia | operating_datetime_utc | year | state | operating_time_hours | gross_load_mw | heat_content_mmbtu | co2_mass_tons |
---|---|---|---|---|---|---|---|
469 | 2019-01-01 07:00:00 00:00 | 2019 | CO | 1.0 | 203.0 | 2146.2 | 127.2 |
469 | 2019-01-01 08:00:00 00:00 | 2019 | CO | 1.0 | 203.0 | 2152.7 | 127.6 |
469 | 2019-01-01 09:00:00 00:00 | 2019 | CO | 1.0 | 204.0 | 2142.2 | 127.0 |
469 | 2019-01-01 10:00:00 00:00 | 2019 | CO | 1.0 | 204.0 | 2129.2 | 126.2 |
469 | 2019-01-01 11:00:00 00:00 | 2019 | CO | 1.0 | 204.0 | 2160.6 | 128.1 |
For more usage examples see the Jupyter notebook at notebooks/pudl-catalog.ipynb
We’re in the process of implementing automated nightly builds of all of our data products for each development branch with new commits in the main PUDL repository. This will allow us to do exhaustive integration testing and data validation on a daily basis. If all of the tests and data validation pass, then a new version of the data products (SQLite databases and Parquet files) will be produced, and placed into cloud storage.
These outputs will be made available via a data catalog on a corresponding
branch in this pudl-catalog
repository. Ingeneral only the catalogs and data
resources corresponding to the HEAD
of development and feature branches will
be available. Releases that are tagged on the main
branch will be retained
long term.
The idea is that for any released version of PUDL, you should also be able to install a corresponding data catalog, and know that the software and the data are compatible. You can also install just the data catalog with minimal dependencies, and not need to worry about the PUDL software that produced it at all, if you simply want to access the DBs or Parquet files directly.
In development, this arrangement will mean that every morning you should have access to a fully processed set of data products that reflect the branch of code that you’re working on, rather than the data and code getting progressively further out of sync as you do development, until you take the time to re-run the full ETL locally yourself.
The Intake docs list a bunch of potential use cases. Here are some features that we’re excited to take advantage of:
The Intake catalog provides a human and machine readable container for metadata describing the underlying data, so that you can understand what the data contains before downloading all of it. We intend to automate the production of the catalog using PUDL’s metadata models so it’s always up to date.
Rather than downloading the same data repeatedly, in many cases it’s possible to transparently cache the data locally for faster access later. This is especially useful when you’ve got plenty of disk space and a slower network connection, or typically only work with a small subset of a much larger dataset.
Intake data catalogs can be packaged and versioned just like Python software packages, allowing us to manage depedencies between different versions of software and the data it operates on to ensure they are compatible. It also allows you to have multiple versions of the same data installed locally, and to switch between them seamlessly when you change software environments. This is especially useful when doing a mix of development and analysis, where we need to work with the newest data (which may not yet be fully integrated) as well as previously released data and software that’s more stable.
All the data sources of a given type (parquet, SQL) would have the same interface, reducing the number of things a user needs to remember to access the data.
Having users access the data through the catalog rather than directly means that the underlying storage location and file formats can change over time as needed without requiring the user to change how they are accessing the data.
- Intake Repo
- Intake Docs
- Intake Examples
- Intake talk from AnacondaCon 2019
- Intake Parquet Repo
- Intake Parquet Docs
- Intake SQL Repo
- Intake SQL Docs
- PUDL intake issues
CarbonPlan is a non-profit research organization focused on climate and energy system data analysis. They manage their data inputs and products using Intake, and the catalogs are public.
Pangeo Forge is a collaborate project providing analysis read cloud optimzed (ARCO) scientific datasets, primarily related to the earth sciences, including climate data. The motiviation and benefits of this approach are described in this paper: Pangeo Forge: Crowdsourcing Analysis-Ready, Cloud Optimized Data Production
Our code, data, and other work are permissively licensed for use by anybody, for any purpose, so long as you give us credit for the work we've done.
- For software we use the MIT License.
- For data, documentation, and other non-software works we use the CC-BY-4.0 license.
- For general support, questions, or other conversations around the project that might be of interest to others, check out the GitHub Discussions
- If you'd like to get occasional updates about our projects sign up for our email list.
- Want to schedule a time to chat with us one-on-one? Join us for Office Hours
- Follow us on Twitter: @CatalystCoop
- More info on our website: https://catalyst.coop
- For private communication about the project or to hire us to provide customized data extraction and analysis, you can email the maintainers: [email protected]
Catalyst Cooperative is a small group of data wranglers and policy wonks organized as a worker-owned cooperative consultancy. Our goal is a more just, livable, and sustainable world. We integrate public data and perform custom analyses to inform public policy (Hire us!). Our focus is primarily on mitigating climate change and improving electric utility regulation in the United States.
This work is supported by a generous grant from the Alfred P. Sloan Foundation and their Energy & Environment Program
Storage and egress fees for this data are covered by Amazon Web Services's Open Data Sponsorship Program.