GIMMS NDVI¶

Global Inventory Modeling and Mapping Studies (GIMMS) AVHRR NDVI — NDVI3g v1 (1981–2015), its successor GIMMS-3G+ V1.2 (1982–2022) and the Peking University recalibration PKU GIMMS NDVI (1982–2022) (1/12°, half-monthly maximum value composites)

Discontinued

At a glance¶

Note on period: GIMMS-3G+ V1.2 covers 1982-01-01 to 2022-12-31 according to NASA CMR (the ORNL dataset title says '1981-2022', but files start in 1982). PKU GIMMS NDVI has an AVHRR-only version for 1982-01 to 2015-12 and a MODIS-consolidated version for 1982-01 to 2022-12. The discontinued NDVI3g v1 covered 1981-07 to 2015-12 (per the NCAR Climate Data Guide), and the earlier NDVI3g (v0) ran to 2013-12.

Overview¶

GIMMS NDVI is a global 1/12°, half-monthly NDVI record built since 1981 by joining AVHRR observations from NOAA and MetOp satellites and correcting for sensor differences, orbital drift and volcanic aerosols; it has been a reference dataset for more than four decades of vegetation greenness trends and phenology. The third-generation NDVI3g stopped at v1 (through 2015), and in 2023 NASA's GIMMS team released GIMMS-3G+ V1.2 at the ORNL DAAC, extended to 2022 with corrections. Separately, a Peking University group recalibrated NDVI3g against Landsat and continued it with MODIS from 2003 as PKU GIMMS NDVI (1982–2022). None of these has been extended beyond 2022.

What changes with the successor¶

GIMMS NDVI3g v1 → GIMMS-3G+ V1.2¶

Transition: 2023-08

The successor released at the ORNL DAAC by the same NASA GIMMS team, extending the NDVI3g record to 2022 with corrections. The grid (1/12°) and half-monthly maximum-composite scheme are unchanged, but distribution moved to the ORNL DAAC with a DOI, MetOp AVHRR data were added, and some corrections were revised. This version also stops in December 2022.

Sources: ORNL DAAC user guide — Global Vegetation Greenness (NDVI) from AVHRR GIMMS-3G+, 1981-2022 · Pinzon & Tucker (2014), NDVI3g · NCAR Climate Data Guide — NDVI3g

GIMMS NDVI3g v1 → PKU GIMMS NDVI V1.2 (Peking University)¶

Transition: 2023-08

Not an official NASA release but a recalibration by a Peking University group. NDVI3g v1 was recalibrated with biome-specific neural networks (BPNN) trained on about 3.6 million Landsat NDVI samples, reducing orbital drift and sensor degradation artefacts, and extended to 2022 with MODIS NDVI from 2003. It appeared as an alternative after NDVI3g updates stopped in 2015, and it too stops in 2022.

Sources: Li et al. (2023), ESSD — PKU GIMMS NDVI · Zenodo — PKU GIMMS NDVI V1.2

Available variables¶

GIMMS-3G+ follows the netCDF4 variable table in the ORNL DAAC user guide; the integer variables use −32768 as fill. PKU GIMMS NDVI is distributed as unnamed GeoTIFF (uint16) bands, so only their content is listed (from the Zenodo description). The variable layout of NDVI3g v1 could not be confirmed in official documentation and is not listed.

instantaneous mean accumulated maximum minimum static — instantaneous: value at the time stamp · mean/accumulated/maximum/minimum: over the indicated interval · static: does not change with time

GIMMS-3G+ V1.2 (NASA, 1982–2022)¶

PKU GIMMS NDVI V1.2 (Peking University; AVHRR-only 1982–2015 / MODIS-consolidated 1982–2022)¶

The full variable list is in the official documentation.

How to get the data¶

• NASA ORNL DAAC — GIMMS-3G+ (DOI 10.3334/ORNLDAAC/2187) — HTTPS download / earthaccess / Earthdata Search (free Earthdata Login)
  • The CMR short name is Global_Veg_Greenness_GIMMS_3G_2187, version 1. Files sit under https://data.ornldaac.earthdata.nasa.gov/protected/global_vegetation/Global_Veg_Greenness_GIMMS_3G/data/ as ndvi3g_geo_v1_2_YYYY_0106.nc4 (v1_1 for 1982–2014).
• Earthdata Cloud OPeNDAP — OPeNDAP — variable and spatial subsetting
  • Each granule has its own address; replace the file name part of the URL above. Requires an Earthdata Login.
• Zenodo — PKU GIMMS NDVI V1.2 — HTTPS download (no login; zip files per decade)
  • Includes the MODIS-consolidated version (PKU_GIMMS_NDVI_AVHRR_MODIS_consolidated_.zip, 1982–2022) and the AVHRR-only version (PKU_GIMMS_NDVI_AVHRR_solely_.zip, 1982–2015). About 5 GB in total.
• Google Earth Engine — old NDVI3g (v0) — Earth Engine catalog (NASA/GIMMS/3GV0)
  • The older NDVI3g (v0) for 1981-07 to 2013-12. It is neither v1 nor GIMMS-3G+ and should not be used for new analyses.

Example code¶

Downloads the GIMMS-3G+ file for the second half of 2021 with earthaccess and decodes NDVI and the quality flag over Korea. On first run you will be asked for your Earthdata username and password. The file is opened without automatic scaling so the storage rules (÷10000, −32768 fill, −5000 not computable) can be applied explicitly.

import earthaccess
import xarray as xr

earthaccess.login()  # free NASA Earthdata account
results = earthaccess.search_data(
    short_name="Global_Veg_Greenness_GIMMS_3G_2187", version="1",
    temporal=("2021-07-01", "2021-12-31"),
)
files = earthaccess.download(results, "gimms3g_plus")

ds = xr.open_dataset(files[0], mask_and_scale=False)
sub = ds.sel(lon=slice(124, 132))
sub = sub.sel(lat=slice(33, 39)) if sub.lat[0] < sub.lat[-1] else sub.sel(lat=slice(39, 33))

raw = sub["ndvi"]
ndvi = (raw.where((raw != -32768) & (raw != -5000)) / 10000.0)
pct = sub["percentile"]
flag = pct // 2000                      # 0 good, 1 spline, 2 seasonal profile, 3 missing
ndvi_good = ndvi.where(flag == 0)
print(ndvi_good.mean(dim=("lat", "lon")).values)   # area mean of the 12 half-month composites

Practical tips¶

• Half-monthly maximum value composites are higher than the mean NDVI for the same period. When building monthly values, state in the paper whether you took the maximum or the mean of the two composites.
• NDVI3g v1 (to 2015), GIMMS-3G+ (to 2022) and PKU GIMMS NDVI come from the same AVHRR raw data but are calibrated differently, so their values differ. Do not mix versions, and cite the exact version name and DOI.
• The official user guide states that atmospheric effects such as Mt. Pinatubo (1991) and sensor calibration, orbital drift and solar zenith biases were not completely removed. Check long-term trends against satellite changeover dates.
• The PKU authors report that orbital drift and sensor degradation artefacts in NDVI3g were largest over tropical evergreen broadleaf forest. For tropical forest trends, compare the PKU version with GIMMS-3G+.
• The PKU consolidated version uses MODIS values from 2003, so its source differs before and after. If AVHRR-only consistency matters for 1982–2015, use the AVHRR-only (solely) version. The Zenodo description recommends removing sparse vegetation (NDVI below 0.1) in trend analyses.
• There are no GIMMS-family data from 2023 onwards. If you continue with VIIRS or MODIS NDVI, first check their differences over an overlap period (e.g. 2012–2022).

Notices and change history¶

References for your paper¶

Dataset citation¶

GIMMS-3G+ (ORNL DAAC) — DOI 10.3334/ORNLDAAC/2187

@misc{pinzon2023gimms3gplus,
  author    = {Pinzon, J. E. and Pak, E. W. and Tucker, C. J. and Bhatt, U. S. and Frost, G. V. and Macander, M. J.},
  title     = {Global Vegetation Greenness ({NDVI}) from {AVHRR} {GIMMS-3G+}, 1981-2022},
  year      = {2023},
  publisher = {ORNL Distributed Active Archive Center},
  address   = {Oak Ridge, Tennessee, USA},
  doi       = {10.3334/ORNLDAAC/2187}
}

PKU GIMMS NDVI V1.2 (Zenodo) — DOI 10.5281/zenodo.8253971

@misc{li2023pkugimms_data,
  author    = {Li, Muyi and Cao, Sen and Zhu, Zaichun and Wang, Zhe and Myneni, Ranga B. and Piao, Shilong},
  title     = {Spatiotemporally consistent global dataset of the {GIMMS} Normalized Difference Vegetation Index ({PKU GIMMS NDVI}) from 1982 to 2022 ({V1.2})},
  year      = {2023},
  publisher = {Zenodo},
  version   = {V1.2},
  doi       = {10.5281/zenodo.8253971}
}

Reference papers¶

NDVI3g data paper — DOI 10.3390/rs6086929

@article{pinzon2014ndvi3g,
  author  = {Pinzon, Jorge E. and Tucker, Compton J.},
  title   = {A Non-Stationary 1981--2012 {AVHRR} {NDVI3g} Time Series},
  journal = {Remote Sensing},
  volume  = {6},
  number  = {8},
  pages   = {6929--6960},
  year    = {2014},
  doi     = {10.3390/rs6086929}
}

GIMMS 8 km NDVI (previous generation) — DOI 10.1080/01431160500168686

@article{tucker2005gimms,
  author  = {Tucker, Compton J. and Pinzon, Jorge E. and Brown, Molly E. and Slayback, Daniel A. and Pak, Edwin W. and Mahoney, Robert and Vermote, Eric F. and El Saleous, Nazmi},
  title   = {An extended {AVHRR} 8-km {NDVI} dataset compatible with {MODIS} and {SPOT} vegetation {NDVI} data},
  journal = {International Journal of Remote Sensing},
  volume  = {26},
  number  = {20},
  pages   = {4485--4498},
  year    = {2005},
  doi     = {10.1080/01431160500168686}
}

PKU GIMMS NDVI data paper — DOI 10.5194/essd-15-4181-2023

@article{li2023pkugimms,
  author  = {Li, Muyi and Cao, Sen and Zhu, Zaichun and Wang, Zhe and Myneni, Ranga B. and Piao, Shilong},
  title   = {Spatiotemporally consistent global dataset of the {GIMMS} Normalized Difference Vegetation Index ({PKU GIMMS NDVI}) from 1982 to 2022},
  journal = {Earth System Science Data},
  volume  = {15},
  number  = {9},
  pages   = {4181--4203},
  year    = {2023},
  doi     = {10.5194/essd-15-4181-2023}
}

Acknowledgements¶

For GIMMS-3G+, cite the ORNL DAAC dataset DOI (10.3334/ORNLDAAC/2187) following the EOSDIS data citation guidance, and add Pinzon & Tucker (2014) for the method. For PKU GIMMS NDVI, cite the Zenodo DOI and Li et al. (2023) in ESSD. State in the text which version you used (NDVI3g v1, GIMMS-3G+ V1.2 or PKU V1.2).

Information on the datasets used in this study was compiled with the help of the Korean Climate Data Navigator (KCDN) (Sim, 2026; https://doi.org/10.5281/zenodo.23300923).

Links¶

• ORNL DAAC — GIMMS-3G+ dataset page
• ORNL DAAC — GIMMS-3G+ user guide (variables, flags)
• NASA Earthdata — GIMMS-3G+ catalog entry
• Li et al. (2023) ESSD — PKU GIMMS NDVI
• NCAR Climate Data Guide — NDVI3g