How to download NEX-GDDP-CMIP6 data as CSV

NASA's NEX-GDDP-CMIP6 dataset is one of the most widely used sources of downscaled climate projections, but it is distributed as many thousands of global NetCDF files. This guide explains what's in the dataset, how to extract a single location yourself, and how to get the same data as ready-to-use CSV and Excel files without writing code.

Updated 16 November 2026

What is NEX-GDDP-CMIP6?

NEX-GDDP-CMIP6 stands for NASA Earth Exchange Global Daily Downscaled Projections for CMIP6. NASA takes the global climate model (GCM) runs from the Coupled Model Intercomparison Project Phase 6 (CMIP6), the generation of models behind the IPCC Sixth Assessment Report (AR6), and bias-corrects and downscales them to a much finer grid. The result lets you look at climate change impacts at local scales, which the coarse original models can't resolve.

NEX-GDDP-CMIP6 at a glance
Spatial resolution0.25° × 0.25° (about 25 km)
Time stepDaily
Coverage180°W–180°E, 90°N–60°S (everywhere except Antarctica)
Historical period1950–2014
Future scenariosSSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5, 2015–2100
VariablesDaily mean, maximum and minimum temperature (tas, tasmax, tasmin), precipitation (pr), relative and specific humidity (hurs, huss), longwave and shortwave radiation (rlds, rsds), wind speed (sfcWind)
ModelsMore than 30 CMIP6 global climate models
SizeAbout 38 TB of NetCDF files
ReferenceThrasher et al. (2022), Scientific Data 9, 262

Why getting a simple CSV is harder than it looks

The data is free to download from NASA, but it's organized for climate scientists rather than spreadsheets:

  • It is stored as NetCDF, a binary format for gridded data that needs specialist software or a script (usually Python or R) to read.
  • There is one file per variable, model, scenario and year, and each file covers the whole globe. A single file looks like tas_day_ACCESS-CM2_ssp245_r1i1p1f1_gn_2050_v2.0.nc.
  • Models differ in the details. They use different ensemble members (for example r1i1p1f1 or r1i1p1f2), grid labels (gn, gr, gr1) and even calendars: HadGEM3-GC31-LL, HadGEM3-GC31-MM, KACE-1-0-G and UKESM1-0-LL use a 360-day year.
  • Values come in scientific units: temperature in kelvin and precipitation as a rate in kg m⁻² s⁻¹.

The file count adds up quickly. Nine models × two variables × 151 years (Historical 1950–2014 plus one SSP from 2015–2100) is 2,718 separate files to subset, convert and stitch into one table.

Option 1: extract the data yourself

If you're comfortable with code, there are three main routes:

  • NASA's THREDDS server. NASA's Center for Climate Simulation (NCCS) serves the files through a THREDDS server whose NetCDF Subset Service can return a single point as CSV, one file (one year of one variable, model and scenario) per request. See the NCCS NEX-GDDP-CMIP6 page for access options.
  • AWS Open Data. The full dataset is mirrored in a public Amazon S3 bucket (see the AWS Open Data registry), which suits cloud workflows with Python's xarray.
  • Google Earth Engine. The dataset is also in the Earth Engine catalog as NASA/GDDP-CMIP6.

For example, this Python snippet asks NASA's subset service for one year of daily mean temperature from one model and scenario at a point in Seattle:

import requests

url = ("https://ds.nccs.nasa.gov/thredds/ncss/grid/AMES/NEX/GDDP-CMIP6/"
       "ACCESS-CM2/ssp245/r1i1p1f1/tas/"
       "tas_day_ACCESS-CM2_ssp245_r1i1p1f1_gn_2050_v2.0.nc")
params = {
    "var": "tas",
    "latitude": 47.65,
    "longitude": -122.34,
    "time_start": "2050-01-01T00:00:00Z",
    "time_end": "2050-12-31T23:59:59Z",
    "accept": "csv",
}
response = requests.get(url, params=params, timeout=60)
response.raise_for_status()
print(response.text[:500])  # one year, one model, one variable, in kelvin

That gets you one year of one variable from one model. To build a usable dataset you would loop over every year, model, variable and scenario, retry failed requests, skip combinations that don't exist, convert units, calculate multi-model means and merge everything into one table. That loop is what Climate Data Download automates.

Option 2: download it as CSV with Climate Data Download

Climate Data Download does the extraction for you and emails you ready-to-use files. It's free while in beta. The form has five steps:

  1. Location. Enter a latitude and longitude (−180…180° and 0…360° longitudes both work) or click the map. You can give the location a name, which appears at the top of your files.
  2. Climate variables. Choose up to four of the nine variables and the units for each, for example °C or °F for temperature and mm/day or in/day for precipitation.
  3. Climate models. Select the Recommended model group or up to nine individual models. Not sure which? See choosing scenarios and climate models.
  4. Scenarios and years. Choose up to two scenarios (Historical counts as one) and set the start and end years: 1950–2014 for Historical and 2015–2100 for the SSPs.
  5. Review and submit. Check your selections and the estimated processing time, enter your email address and submit. When the job is finished you'll get an email with download links, valid for 7 days.

What the CSV looks like

You get three files: a daily time series (CSV), annual averages (CSV), and an Excel workbook with both on separate sheets ("Daily Timeseries" and "Annual Averages"). Each file starts with a short header describing your request. Here is the top of the daily file from our Seattle sample, which has 10 years of Historical and 10 years of SSP1-2.6 data from two models:

Seattle
Date of Analysis: 2025-05-01 08:36 UTC
Latitude: 47.651975
Longitude: -122.336426
"Models: ACCESS-CM2, EC-Earth3"
"Variables: tas, pr"
"Scenarios: Historical (1990-1999), Ssp126 (2045-2054)"
Prepared by: ClimateDataDownload.com
"Source Data Citation: Thrasher, B., Wang, W., Michaelis, A. et al. NASA Global Daily Downscaled Projections, CMIP6. Sci Data 9, 262 (2022). https://doi.org/10.1038/s41597-022-01393-4"

Below the header is the data table. These are the temperature columns for the first three days (the file also has precipitation columns):

dateACCESS-CM2_tas[unit="F"]_historicalEC-Earth3_tas[unit="F"]_historicalMULTI-MODEL-MEAN_tas[unit="F"]_historical
1990-01-0135.9745844.6843140.32945
1990-01-0235.5404941.4577938.49914
1990-01-0334.5606348.2057641.38319
  • Column names follow the pattern MODEL_variable[unit="…"]_scenario.
  • After each group of model columns comes a MULTI-MODEL-MEAN column: the average across the models you selected.
  • Historical and future scenarios share one date column, so Historical cells are empty for future dates and vice versa.
  • The annual file has a year column holding the average of the daily values for each year.
  • If a model doesn't provide a variable or scenario in NEX-GDDP-CMIP6, its column is left out.

Working in Excel? The Excel guide shows how to calculate warming, rainfall change and hot days from these files.

Good to know before you use the data

  • A grid cell, not a weather station. Values describe the 0.25° grid cell (about 25 km across) at your coordinates. Local conditions, such as in mountainous or coastal areas, can differ from the grid-cell average.
  • Use several models. Individual models disagree, and the spread between them is a useful measure of uncertainty. Use the multi-model mean together with the range across models.
  • Data gaps. A few model, variable and scenario combinations aren't in the dataset. See the list of gaps and NASA's technical note.
  • Cite the dataset. Please cite Thrasher et al. (2022) when you publish results. See how to cite.

More guides