Processing scripts for NEX-GDDP and LOCA CMIP5 climate datasets.
This repository contains a script for generating ensemble indicators from the downscaled CMIP5 datasets NEX-GDDP (global) and LOCA (continental US).
These data can be viewed at https://prepdata.org/explore.
Downloads are available at:
- http://wri-public-data.s3.amazonaws.com/resourcewatch/raster/nexgddp.zip
- http://wri-public-data.s3.amazonaws.com/resourcewatch/raster/loca.zip
Learn more at https://www.wri.org/research/making-climate-data-accessible
Citation: Gassert, Francis, Enrique Cornejo, and Emily Nilson. 2021. "Making Climate Data Accessible: Methods for Producing NEX-GDDP and LOCA Downscaled Climate Indicators." Technical Note. World Resources Institute. Washington DC. Available at https://www.wri.org/research/making-climate-data-accessible
This repository contains a script for generating ensemble indicators from the downscaled CMIP5 datasets: NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP) (Thrasher and Nemani 2015) for the globe and Localized Constructed Analogs (LOCA) CMIP5 projections for the continental United States (Pierce, Cayan, and Thrasher 2014; Maurer et al. 2007). NEX-GDDP and LOCA contain daily minimum and maximum temperature and precipitation values from several dozen models for two climate scenarios over the period of 1950-2100. These scripts reduce the large size of source datasets (~12TB and ~9TB, respectively) by several orders of magnitude.
This typically takes place in four steps:
- For each model, scenario, and year: compute annual indicators
- For each indicator, model, and scenario: compute moving averages (typically 30yrs)
- For each indicator, model, and scenario: compute difference and/or change layers
- For each indicator and scenario: compute statistics across the ensemble of models (e.g. median)
These scripts require Docker and an AWS account to run.
This package is designed to compute against the large input data sources using AWS S3 as intermediate storage. This allows it to be run on commodity hardware, with the minimum requirement that some annual indicators may require up to 1.5GB of memory per CPU.
It is possible to compute the listed indicators under the AWS Free Tier.
Computation can generate a large amount of intermediate data. It is recommended that you set a Lifecycle Rule on your S3 Bucket to automatically expire (delete) data older than 30 days to avoid accumulating storage cost, and move or download the resulting data that you want to keep.
Under test conditions, computing the listed indicators generated costs as follows:
- ~$40 - AWS EC2 Spot Instance
- ~$50/month - S3 Storage
- Download the repository
git clone https://github.com/fgassert/process-gddp/
cd process-gddp
- Create a
.envfile in the root folder with the following four configuration variables.
# AWS credentials for accessing S3
AWS_ACCESS_KEY_ID={aws-access-key-id}
AWS_SECRET_ACCESS_KEY={aws-secret-access-key}
# Bucket name for saving data
GDDP_BUCKET={bucket-name}
# Prefix (folder) in which you want to save data i.e. s3://{bucket-name}/{prefix-name}/
GDDP_PREFIX={folder-name}
There is a shell script in the root directory to build and run the main script in a docker container. It takes any number of keynames (output file names) as parameters. For each keyname, it will determine what input data and intermediate steps are needed to generate that output, and produce them in order.
./start.sh- Run without parameters to generate a few test outputs../start.sh [keyname] [keyname]...- Compute specific outputs../start.sh $(cat scripts/outputs.txt)- Compute all outputs listed inscripts/outputs.txt.
Outputs and intermediate results are saved in s3://{GDDP_BUCKET}/{GDDP_PREFIX}/.
Keynames are the output filenames that will be generated, but also define what to compute.
Keynames are composed of the following parts, joined by underscores, with an implied .tif extension:
{formula}_{variable}_{scenario}_{model}_{years}_{dataset}[.tif]
# or expanding {formula} into subcomponents:
{ensemble}-{timefunc}-{indicator}_{variable}_{scenario}_{model}_{years}_{dataset}[.tif]
e.g. q50-abs-annual_pr_rcp85_ens_2035-2065_nexgddp.tif - Median absolute average annual (q50-abs-annual) precipitation (pr) for the high emissions scenario (rcp85) across the ensemble (ens) for the 31yr period centered on 2050 (2035-2065) derived from NEX-GDDP (nexgddp).
Formula
The formula defines what computations to run, the remainder of the keyname defines what data to run the computationts against. The {formula} can be decomposed further as follows:
# {formula} can be decompose into:
{indicator}_...
{timefunc}-{indicator}_...
# or
{ensemble}-{timefunc}-{indicator}_...
Each of these components is a reducing function that depends on the previous. For example, computing ensemble median average annual (q50-abs-annual) precipitation from individual models' average annual (abs-annual) precipitation:
| key | is derived from |
|---|---|
q50-abs-annual_..._ens_... | abs-annual_..._ACCESS1-0_...abs-annual_..._BNU-ESM_...abs-annual_..._CCSM4_...abs-annual_..._CESM1-BGC_...abs-annual_..._CNRM-CM5_... ... |
Or computing the 31yr average (abs-annual) precipitation from annual (annual) precipitation:
| key | is derived from |
|---|---|
abs-annual_..._2035-2065_... | annual_..._2035_...annual_..._2036_...annual_..._2037_... ... annual_..._2065_... |
- ensemble
| value | description |
|---|---|
q25 | 25th percentile |
q50 | median |
q75 | 75th percentile |
iqr | interquartile range |
mean | mean |
- timefunc
| value | description |
|---|---|
abs | absolute i.e. the average value over the range of years |
diff | linear change from baseline (1960-1990) (i.e. abs_..._{startyear}-{endyear}_... minus abs_..._1960-1990_...) |
ch | multiplicative change from baseline (1960-1990) (i.e. abs_..._{startyear}-{endyear}_... divided by abs_..._1960-1990_...) |
- indicator
| value | description |
|---|---|
annual | annual average |
q98 | 98th percentile for the year |
q99 | 99th percentile for the year |
gt-q98 | number of days/yr exceeding the 98th percentile for the baseline period (1960-1990) |
gt-q99 | number of days/yr exceeding the 99th percentile for the baseline period (1960-1990) |
gt50mm | number of days/yr exceeding 0.0005787037037 (kg/m2/s) (50mm/day) |
gt95f | number of days/yr exceeding 308.15 (Kelvin) |
gt90f | number of days/yr exceeding 305.37 (Kelvin) |
gt85f | number of days/yr exceeding 302.59 (Kelvin) |
gt32f | number of days/yr exceeding 273.15 (Kelvin) |
frostfree | length of longest run of days in a year exceeding 273.15 (Kelvin) |
drydays | length of longest run of days in a year less than 0.000011574 (kg/m2/s) (1mm/day) |
dryspells | number of runs of at least 5 days in a year with less than 0.000011574 (kg/m2/s) (1mm/day). Each day over 5 days counts as 0.2 of a run. |
tavg-tasmin | Use with tasmax. Daily average temperature. (Specifically, daily value plus tasmin divided by 2.) |
hdd65f-tasmin | Use with tasmax. Heating degree days. The sum of 291.48 - value (Kelvin) (65ºF), where positive, for each day in a year. |
cdd65f-tasmin | Use with tasmax. Cooling degree days. The sum of value - 291.48 (Kelvin) (65ºF), where positive, for each day in a year. |
- variables
| value | description |
|---|---|
pr | precipitation (kg/ms/s) |
tasmin | daily minimum temperature |
tasmax | daily maximum temperature |
- scenario
| value | description |
|---|---|
rcp45 | Low emissions scenario |
rcp85 | High emissions scenario |
historical | Retrospective data (1990-2005 only) |
- model
| value | description |
|---|---|
ACCESS1-0BNU-ESMCCSM4CESM1-BGCCNRM-CM5CSIRO-Mk3-6-0CanESM2GFDL-CM3GFDL-ESM2GGFDL-ESM2MIPSL-CM5A-LRIPSL-CM5A-MRMIROC-ESM-CHEMMIROC-ESMMIROC5MPI-ESM-LRMPI-ESM-MRMRI-CGCM3NorESM1-Mbcc-csm1-1inmcm4 | Models included in NEX-GDDP |
ACCESS1-0ACCESS1-3CCSM4CESM1-BGCCESM1-CAM5CMCC-CMCMCC-CMSCNRM-CM5CSIRO-Mk3-6-0CanESM2EC-EARTHFGOALS-g2GFDL-CM3GFDL-ESM2GGFDL-ESM2MGISS-E2-HGISS-E2-RHadGEM2-AOHadGEM2-CCHadGEM2-ESIPSL-CM5A-LRIPSL-CM5A-MRMIROC-ESMMIROC-ESM-CHEMMIROC5MPI-ESM-LRMPI-ESM-MRMRI-CGCM3NorESM1-Mbcc-csm1-1bcc-csm1-1-minmcm4 | Models included in LOCA |
ens | For ensemble functions |
- years
| value | description |
|---|---|
A single {year} between 1950 and 2100 | Year of data (functions without a {timefunc}) |
A range of years {startyear}-{endyear} | For multi-year averages (functions with a {timefunc}) |
- dataset
| value | description |
|---|---|
nexgddp | Derive from NEX GDDP |
loca | Derive from LOCA |
Formula and indicators are defined in src/processgddp/formulae.py. The baseline period for indicators that use a baseline is defined here.
Thrasher, Bridget, and Rama Nemani. 2015. “NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP).” https://nex.nasa.gov/nex/projects/1356/.
Pierce, David W., Daniel R. Cayan, and Bridget L. Thrasher. 2014. “Statistical Downscaling Using Localized Constructed Analogs (LOCA).” Journal of Hydrometeorology 15 (6): 2558–2585. doi:10.1175/JHM-D-14-0082.1.
Maurer, Edwin P., Levi Brekke, Tom Pruitt, and Philip B. Duffy. 2007. “Fine-Resolution Climate Projections Enhance Regional Climate Change Impact Studies.” Eos, Transactions American Geophysical Union 88 (47). John Wiley & Sons, Ltd: 504–504. doi:10.1029/2007EO470006.
We acknowledge the World Climate Research Programme's Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups for producing and making available their model output. For CMIP the U.S. Department of Energy's Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals.