Sup3rCC Examples

Sup3rCC Examples#

Super-Resolution for Renewable Energy Resource Data with Climate Change Impacts (Sup3rCC) is one application of the sup3r software. In this work, we train generative models to create high-resolution (4km hourly) wind, solar, and temperature data based on coarse (100km daily) global climate model data (GCM). The generative models and high-resolution output data are publicly available via the Open Energy Data Initiative (OEDI) and via HSDS at the bucket nrel-pds-hsds and HSDS path /nrel/sup3rcc/. This set of examples lays out basic ways to use the Sup3rCC models and data.

Sup3rCC Data Access#

For high level details on accessing the NREL renewable energy resource datasets including Sup3rCC, see the rex docs pages

The Sup3rCC data and models are publicly available in a public AWS S3 bucket. The data files and models can be downloaded directly from there to your local machine or an EC2 instance using the OEDI data explorer or the AWS CLI. A word of caution: there’s a lot of data here. The smallest Sup3rCC file for just a single variable is 18 GB, and a full year of data is 216 GB.

The Sup3rCC data is also loaded into HSDS so that you may stream the data via the NREL developer API or your own HSDS server. This is the best option if you’re not going to want the full annual dataset over the whole United States. See these rex instructions for more details on how to access this data with HSDS and rex.

Directory Structure#

The Sup3rCC directory contains downscaled data for multiple projections of future climate change. For example, a file from the initial data release sup3rcc_conus_ecearth3_ssp585_r1i1p1f1_wind_2015.h5 is downscaled from the climate model MRI ESM 2.0 for climate change scenario SSP5 8.5 and variant label r1i1p1f1. The file contains wind variables for the year 2015. Note that this will represent the climate from 2015, but not the actual weather we experienced.

Within the S3 bucket there is also a folder models providing pre-trained Sup3rCC generative machine learning models.

Example Sup3rCC Data Usage#

The jupyter notebook in this example shows some basic code to access and explore the data. You can walk through the example notebook. You can also clone this repo, setup a basic python environment with rex, and run the notebook on your own.

Running Sup3rCC Models#

In a first-of-a-kind data product, we have released the pre-trained Sup3rCC generative machine learning models along with the sup3r software so that anyone working at the intersection of energy and climate may create their own high-resolution renewable energy resource data from GCM input. You might want to do this if you have your own GCMs or climate scenarios that you’re interested in studying.

To run the Sup3rCC models, follow these instructions:

  1. Decide what kind of hardware you’re going to use. You could technically run Sup3rCC on a desktop computer, but you will need lots of RAM (we use compute nodes with 170 GB of RAM). We recommend a high-performance-computing cluster if you have access to one, or an AWS Parallel Cluster if you do not.

  2. Download the Sup3rCC models to your hardware using the AWS CLI: $ aws s3 cp s3://nrel-pds-sup3rcc/models/

  3. Download the GCM data that you want to downscale from CMIP6

  4. Setup the Sup3rCC software. We recommend using miniconda to manage your python environments. You can create a sup3r environment with the conda file in this example directory: $ conda env create -n sup3rcc --file env.yml

  5. Copy this examples directory to your hardware. You’re going to be using the folder structure in /sup3r/examples/sup3rcc/run_configs as your project directories (/sup3r/ is a git clone of the sup3r software repo).

  6. Navigate to /sup3r/examples/sup3rcc/run_configs/trh/ and update all of the filepaths in the config files for the source GCM data, Sup3rCC models, and exogenous data sources (e.g. the nsrdb_clearsky.h5 file).

  7. Update the execution control parameters in the config_fwp.json file based on the hardware you’re running on.

  8. You can either run sup3r-batch to setup multiple run years, or sup3r-pipeline to run just one job. We recommend starting with sup3r-pipeline (more on the sup3r CLI).

  9. To run sup3r-pipeline, make sure you are in the directory with the config_pipeline.json and config_fwp.json files, and then run this command: python -m sup3r.cli -c config_pipeline.json pipeline

  10. If you’re running on a slurm cluster, this will kick off a number of jobs that you can see with the squeue command. If you’re running locally, your terminal should now be running the Sup3rCC models. The software will create a ./logs/ directory in which you can monitor the progress of your jobs.

  11. The sup3r-pipeline is designed to run several modules in serial, with each module running multiple chunks in parallel. Once the first module (forward-pass) finishes, you’ll want to run python -m sup3r.cli -c config_pipeline.json pipeline again. This will clean up status files and kick off the next step in the pipeline (if the current step was successful).

Nuances of Sup3rCC#

The Sup3rCC dataset is quite unlike the legacy NREL historical wind and solar datasets. As such, we expect there will be some confusion about how to use the data. There are some nuances of the data enumerated below. If you have any questions about how to apply the Sup3rCC data to your work, please reach out to Grant Buster (Grant.Buster@nrel.gov).

  1. Sup3rCC data is based on global climate model (GCM) data, which does not represent historical weather, only historical climate. So for example, Sup3rCC 2015 does not represent the actual historical weather in 2015, just the historical climate in 2015.

  2. The GCM data was bias-corrected using the NSRDB and WTK data. GCM irradiance, temperature, and humidity are bias corrected using the NSRDB for the years 2015-2021. GCM windspeeds from 2015-2021 are bias corrected using the WTK from 2007-2013 (we don’t currently have modern years of high-resolution wind data). Note that temperature and humidity from the NSRDB are actually originally sourced from MERRA2, a reanalysis product. Additional bias may still exist in the high-resolution outputs and a secondary bias correction step may be valuable in downstream applications

  3. Sup3rCC data represents just one possible future climate subject to deep uncertainties. Do not use the Sup3rCC data as an accurate prediction of future weather. Some uncertanties about our future climate can be quantified by exploring a large ensemble of GCM data across multiple climate scenarios and multiple models.

  4. Sup3rCC cannot represent many meteorological events that are not skillfully represented in GCM data (e.g., hurricanes, tornadoes, mesoscale convective storms, wildfires, etc…).

  5. Sup3rCC does not currently use land use data and only understands phenomena like urban heat islands via the bias correction of the GCM data with historical reanalysis data. Application of Sup3rCC to individual cities may benefit from statistical validation using historical ground measurement data.

Sup3rCC Versions#

The Sup3rCC data has versions that coincide with the sup3r software versions. Note that not every sup3r software version will have a corresponding Sup3rCC data release, but every Sup3rCC data release will have a corresponding sup3r software version. This table records versions of Sup3rCC data releases. Sup3rCC generative models may have slightly different versions than the data. The version in the Sup3rCC .h5 file attribute can be inspected to verify the actual version of the data you are using.

Version

Effective Date

Notes

0.1.0

6/27/2023

Initial Sup3rCC release with two GCMs and one climate scenario. Known issues: few years used for bias correction, simplistic GCM bias correction method, mean bias in high-res output especially in wind and solar data, imperfect wind diurnal cycles when compared to WTK and timing of diurnal peak temperature when compared to observation.

Acknowledgements#

This work was authored by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding provided by the DOE Grid Deployment Office (GDO), the DOE Advanced Scientific Computing Research (ASCR) program, the DOE Solar Energy Technologies Office (SETO), and the Laboratory Directed Research and Development (LDRD) program at the National Renewable Energy Laboratory. The research was performed using computational resources sponsored by the DOE Office of Energy Efficiency and Renewable Energy and located at the National Renewable Energy Laboratory. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes.