NEX-GDDP-CMIP6 sits behind many climate risk platforms, research studies and scenario analyses for climate disclosure. Many people use it through tools that process the files for them, without seeing which scenarios, variables and file versions the tool picked.
What Is NEX-GDDP-CMIP6?
NEX-GDDP-CMIP6 stands for NASA Earth Exchange Global Daily Downscaled Projections, built on the Coupled Model Intercomparison Project Phase 6 (CMIP6). It is a publicly available dataset that provides bias-corrected, statistically downscaled climate projections from global climate models.
Raw CMIP6 climate model output is coarse, with grid cells of roughly 100 to 300 km. NEX-GDDP-CMIP6 refines this to a 0.25 degree grid, about 25 km, which makes it usable for regional and location-level analysis.
The Climate Analytics Group and NASA Ames Research Center produced the dataset on the NASA Earth Exchange, and the NASA Center for Climate Simulation (NCCS) distributes it. It holds daily values from 1950 to 2100: a modeled historical run for 1950 to 2014 and projections for 2015 to 2100 under four emission scenarios. Coverage runs from 60°S to 90°N, so Antarctica is not included. NASA’s technical note for version 2.0 is the reference document.
Sample Climate Risk Assessment
See a full climate risk assessment, end to end
We’ll email you a complete worked example: 12 climate hazards across multiple emission scenarios, out to 2050. It shows what a full climate risk assessment output looks like.
What Are CMIP6 Models Used For?
CMIP6 is the sixth generation of coordinated climate model experiments, organized by the World Climate Research Programme. The CMIP6 registry lists 49 modeling institutions worldwide, and their models simulate how the Earth’s climate system responds to greenhouse gas concentrations.
These models are used for:
- IPCC assessment reports (the AR6 report draws primarily from CMIP6 output)
- National climate adaptation planning (governments use CMIP6 projections to plan infrastructure)
- Corporate climate risk assessment (physical risk analysis for climate disclosure often uses CMIP6 scenarios)
- Insurance and financial sector stress testing
NEX-GDDP-CMIP6 makes these models usable without a supercomputer and without bias-correcting and regridding raw model output yourself. For background on how these scenarios work, see our guides to SSP scenarios and RCP scenarios.
How Many Climate Models Are in CMIP6?
The CMIP6 registry lists more than 130 model versions. NEX-GDDP-CMIP6 downscales 35 of these models, taken from runs that published daily output.
Each model has its own Equilibrium Climate Sensitivity (ECS), which measures how much the planet eventually warms when CO₂ doubles. ECS across the NEX-GDDP-CMIP6 models runs from about 1.8°C to 5.6°C. The IPCC’s assessed very likely range is 2°C to 5°C, so a few models sit outside it. The same emission scenario can produce materially different warming depending on which model you select.
For practical risk analysis, most practitioners do not run every model in the dataset. Common approaches include:
- Single model with fallback: Select one well-validated model (such as ACCESS-CM2, MPI-ESM1-2-HR, or MIROC6) and use alternates if data gaps exist for a given scenario or time period
- Multi-model ensemble: Run 5-10 models and report the median and spread. Provides uncertainty bounds but requires 5-10x more computation
- Hot/cold bracketing: Select one high-ECS and one low-ECS model (for example, CanESM5 and INM-CM4-8) to show the range of plausible futures
The choice of approach depends on computational resources and whether uncertainty quantification is required. For screening, a single well-chosen model is often enough. When a decision depends on how large the change is, report the spread across several models.
In the five-model run below, the models agree on every baseline rating and on cold stress, storms, wildfire and landslides. They split on heat, temperature, rainfall and river flood in the future horizons, where one model alone would show only one side of the range.
NEX-GDDP-CMIP6 Variables and Resolution
The dataset provides nine climate variables at daily resolution:
| Variable | Full Name | Native Unit | Typical Use |
|---|---|---|---|
| tas | Daily Mean Temperature | Kelvin | Temperature change analysis |
| tasmax | Daily Maximum Temperature | Kelvin | Heat wave detection, wildfire risk |
| tasmin | Daily Minimum Temperature | Kelvin | Cold stress, frost days |
| pr | Precipitation | kg/m²/s | Drought, extreme rainfall, flood risk |
| hurs | Relative Humidity (daily mean) | % | Heat stress, fire weather |
| huss | Specific Humidity (daily mean) | kg/kg | Heat stress, humidity trends |
| rlds | Downwelling Longwave Radiation | W/m² | Surface energy balance, heat models |
| rsds | Downwelling Shortwave Radiation | W/m² | Solar resource, evaporation, crop models |
| sfcWind | Daily Mean Wind Speed | m/s | Wind trends, wildfire spread (daily means smooth out gusts) |
Not every model carries all nine, so check the model table in NASA’s technical note before you pick a model.
Unit conversions practitioners need to know:
- Temperature: Celsius = Kelvin minus 273.15
- Precipitation: mm/day = flux value × 86,400 (seconds per day)
- Annual precipitation: mm/year = mm/day × 365
Variables Missing From NEX-GDDP-CMIP6
NEX-GDDP-CMIP6 has no runoff (mrro), soil moisture, snow or sea level variable, and no sub-daily data. Flood analysis that uses it alone has to treat precipitation as a stand-in for river flow, and catchment hydrology needs a dedicated hydrological model. Sea level projections come from separate datasets.
Humidity is included as daily means (hurs and huss). Heat index and fire weather indices can be built from it as daily approximations, since the dataset has no daily minimum or midday humidity. Version 2.0 corrected out-of-range humidity values found in earlier versions.
Spatial Resolution: What 25 km Means in Practice
The 0.25 degree grid gives one value per cell, and each cell is roughly 25 to 28 km across. That resolution is sufficient for:
- Distinguishing coastal vs. inland climate signals
- Capturing regional precipitation patterns
- Identifying broad heat and drought exposure zones
But it cannot resolve:
- Urban heat island effects within a city
- Microclimates from local terrain (valley vs. hilltop within the same grid cell)
- Individual property-level exposure differences
- Small islands, where NASA notes values may be unrealistic because the observation data used for bias correction is not validated over oceans
For site-specific engineering decisions, higher-resolution downscaling or local weather station data should supplement NEX-GDDP-CMIP6 projections. For portfolio screening, the grid is fine enough to compare regional climate signals across many sites.
Scenarios Available in NEX-GDDP-CMIP6
The dataset provides projections under four Shared Socioeconomic Pathways, the four “Tier 1” CMIP6 scenarios, each from 2015 to 2100. The IPCC AR6 best estimates of warming for 2081 to 2100, compared with 1850 to 1900, are:
- SSP1-2.6: Low emissions, about 1.8°C. CO₂ emissions fall to net zero around or after 2050.
- SSP2-4.5: Intermediate emissions, about 2.7°C. A “middle of the road” path where some climate policies are implemented.
- SSP3-7.0: High emissions, about 3.6°C. CO₂ emissions roughly double from current levels by 2100.
- SSP5-8.5: Very high emissions, about 4.4°C. CO₂ emissions roughly double from current levels by 2050.
Each model in the dataset shows more or less warming than these values, depending on its climate sensitivity. Not every model ran every scenario.
The historical run covers 1950 to 2014. It is model output, bias-corrected against an observation-based dataset over 1960 to 2014, so it gives a modeled baseline for measuring future change.
How to Access NEX-GDDP-CMIP6
The data is free to use under a Creative Commons Zero (CC0) license. NASA distributes the full archive through two channels: the NCCS THREDDS server and AWS S3.
AWS S3 (Open Data Registry)
For large-scale or programmatic access, the full archive sits on AWS S3 at s3://nex-gddp-cmip6, and no AWS account is needed. Files are NetCDF4, one variable per year, organized by model, scenario and variable. The bucket keeps older versions next to version 2.0, so check the version in each file name.
NASA NCCS THREDDS Server
NASA’s own data server at NCCS offers OPeNDAP access and a subset service that cuts out a point or area and a date range on the server before you download. The server now lists only version 2.0 files.
Bulk Download From NCCS
For bulk downloads, NCCS publishes a checksum index of every version 2.0 file, and NASA’s version 1 technical note includes a short Python script that downloads each file and checks it. Download only the models, scenarios and variables you need.
NASA asks users to cite Thrasher et al. (2022) in Scientific Data and the dataset DOI 10.7917/OFSG3345.
Free Climate Risk Report
Run a free climate risk assessment report
12 hazards under three emission scenarios, for a baseline and 2030, 2040 and 2050, at any location. Start free, no call required.
Strengths and Limitations of NEX-GDDP-CMIP6
| Strengths | Limitations |
|---|---|
| Bias-corrected against observation-based data | ~25 km grid misses local microclimate |
| Up to 35 models per scenario for uncertainty ranges | Not every model has every scenario or variable |
| Daily values, 1950 to 2100 | Daily values only: no sub-daily rain or wind gusts |
| Nine variables, including humidity and radiation | No runoff, soil moisture or sea level variables |
| Free under a CC0 license | Full archive is too large to download whole |
| Two NASA access routes (NCCS and AWS) | Assumes 1960 to 2014 spatial patterns hold |
| Versioned fixes, latest version 2.0 (2025) | Values for small islands may be unreliable |
For screening physical climate risk across many sites, NEX-GDDP-CMIP6 offers a practical balance of resolution, access and model coverage. Continuuiti’s climate risk API rates 12 physical hazards for any site under three emission scenarios, from a baseline out to 2050.
Our climate data API comparison covers how different APIs expose this data, and our climate models explainer covers how the GCMs behind NEX-GDDP-CMIP6 work.
Frequently Asked Questions
What variables are in NEX-GDDP-CMIP6?
Nine daily variables: tas, tasmax and tasmin (temperature), pr (precipitation), hurs and huss (relative and specific humidity), rlds and rsds (downwelling longwave and shortwave radiation) and sfcWind (daily mean wind speed). Not every model carries all nine. Runoff, soil moisture and sea level are not included.
What is the difference between CMIP5 and CMIP6?
CMIP6 is the newer generation of coordinated climate model experiments. It replaced CMIP5’s Representative Concentration Pathways (RCPs) with Shared Socioeconomic Pathways (SSPs), and its models show a wider range of climate sensitivity with a higher average. NASA’s datasets follow the same split: NEX-GDDP, built on CMIP5, has 21 models, two RCPs and three variables, while NEX-GDDP-CMIP6 has 35 models, four SSPs and nine variables.
What is the resolution of NEX-GDDP-CMIP6?
About 25 km (0.25 degrees) spatial resolution, achieved through statistical downscaling of coarser CMIP6 output. Temporal resolution is daily. It suits regional risk screening, and site-specific engineering analysis needs finer local data.
How do I access NEX-GDDP-CMIP6 data?
NASA distributes it through two channels. The NCCS THREDDS server supports server-side subsetting and HTTPS download, which suits a point or small area. The AWS S3 bucket (s3://nex-gddp-cmip6) needs no account and suits large programmatic downloads.
Which scenarios does NEX-GDDP-CMIP6 include?
Four: SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5, each from 2015 to 2100, plus a modeled historical run from 1950 to 2014. Not every model ran every scenario. Both NASA channels, NCCS and AWS S3, carry all four.
How does CMIP6 change the picture for climate projections?
CMIP6 models show higher climate sensitivity than CMIP5 on average, so they project more warming for the same emissions. Their ECS runs from 1.8 to 5.6°C, against an IPCC AR6 assessed very likely range of 2°C to 5°C. CMIP6 also introduced SSPs, which combine socioeconomic narratives with radiative forcing and replaced the simpler RCPs from CMIP5. For more on the differences, see our climate models guide.
