GLASS¶

Global LAnd Surface Satellite (GLASS) product suite (AVHRR long-term records from 1981 / MODIS records from 2000; global land surface products from Shunlin Liang's group at Beijing Normal University and the University of Hong Kong)

Available

At a glance¶

Note on period: According to the HKU download table, AVHRR products run from 1981 (or 1982) to 2018–2022 (LAI, FAPAR and net radiation to 2018, ET to 2022), and MODIS products from 2000 to 2024–2025 (LAI and FAPAR to 2025; albedo, net radiation and ET to 2024). The combined albedo product (MODIS & AVHRR, 0.05° and 0.5°) covers 1981–2024. No fixed update schedule is published; new years usually appear in a batch the following year.

Overview¶

GLASS is a suite of global land surface satellite products developed mainly at Beijing Normal University and now distributed and updated by Shunlin Liang's group at the University of Hong Kong. Using the AVHRR long-term data record (LTDR), MODIS reflectances and other satellite and ground data, it provides more than 20 spatially continuous variables — leaf area index (LAI), FAPAR, fractional vegetation cover (FVC), broadband albedo, net and downward shortwave radiation, evapotranspiration (ET), gross primary production (GPP) and more — from 1981 (AVHRR) and 2000 (MODIS). Because many variables come from related algorithms on consistent grids, GLASS is widely used in energy, water and carbon cycle studies.

Available variables¶

Each product comes in separate files (HDF-EOS4), one variable (product) per file. For LAI, FAPAR, FVC, net radiation, shortwave radiation, ET and GPP, neither the BNU product pages, the HKU file naming document nor the archive metadata (hdf.xml) give scientific dataset (SDS) names, so the variable-name column is left empty (the pages describe only the content, e.g. 'leaf area index dataset' and 'quality flag'). Check them in your files with gdalinfo or similar. Albedo layer names follow the GLASS02B05 table on the BNU albedo page (which states that GLASS02B06 and A06 are the same); whether the current combined MODIS–AVHRR version (GLASS02B03 V50 etc.) uses the same names could not be confirmed in official documentation. Scale factors and valid ranges follow the BNU product pages and may differ between versions (V4, V5, V6), so always check the HDF attributes (scale_factor, _FillValue, valid_range).

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

Leaf area index LAI (MODIS V6 250 m, 500 m, 0.05° from 2000 / AVHRR V4 0.05° 1981–2018, 8-day)¶

Fraction of absorbed PAR, FAPAR (MODIS V6 250 m to 0.5° from 2000 / AVHRR V4 0.05° 1982–2018, 8-day)¶

Fractional vegetation cover FVC (MODIS 500 m to 0.5° 2000–2024 / AVHRR 0.05° and 0.5° 1981–2020, 8-day)¶

Broadband albedo (combined MODIS–AVHRR 0.05° and 0.5° 1981–2024, MODIS 250 m and 500 m, 8-day)¶

Radiation (daily)¶

Evapotranspiration and carbon (8-day)¶

The full variable list is in the official documentation.

How to get the data¶

• HKU GLASS download (archive) — HTTPS directory download (no login; bulk download with wget etc.)
  • The layout is https://glass.hku.hk/archive///// (e.g. archive/LAI/MODIS/0.05D/2024/). Each file has an HDF, a metadata file (.hdf.xml) and a preview JPG; the 500 m and 250 m tiled products have an additional day-of-year folder under each year. The site asks users to read the 'HDF File Naming Convention' document before downloading.
• BNU GLASS product pages (Chinese) — Algorithm and data format descriptions per product / links to geodata.cn and HKU
  • The per-product pages (e.g. introduction/LAI.html) give scale factors, valid ranges and quality bit definitions.
• National Earth System Science Data Center (geodata.cn) — Web search and download (mostly in Chinese)
  • Named as a distribution site in the GLASS overview paper (Liang et al., 2021). Some earlier versions may only be found here.

Example code¶

Downloads the MODIS LAI 0.05° file for the first 8 days of January 2024 from the HKU archive and opens it as HDF4. Layer names are not documented officially, so list them first. Reading HDF4 needs GDAL built with the HDF4 driver, or pyhdf.

import requests
from osgeo import gdal   # conda install -c conda-forge gdal (build with HDF4 support)

fname = "GLASS01B01.V61.A2024001.2026063.hdf"   # LAI, MODIS, 0.05°, 1-8 January 2024
url = f"https://glass.hku.hk/archive/LAI/MODIS/0.05D/2024/{fname}"
with open(fname, "wb") as f:
    f.write(requests.get(url, timeout=120).content)

ds = gdal.Open(fname)
for name, desc in ds.GetSubDatasets() or [(fname, "single layer")]:
    print(name, "|", desc)                      # check the layer names

sub = gdal.Open((ds.GetSubDatasets() or [(fname, "")])[0][0])
md = sub.GetMetadata()
print({k: v for k, v in md.items() if k.lower() in ("scale_factor", "_fillvalue", "valid_range", "units")})
arr = sub.ReadAsArray().astype("float32")
# apply the scale factor and fill value from the attributes before use

Practical tips¶

• The same product name can hide different versions (V40, V50, V60, V61) and sensors (AVHRR, MODIS) with different algorithms and scale factors. Record the V number in the file name and do not mix versions in one time series. The version in the HKU download table (e.g. LAI V60) can differ from the actual file names (V61).
• AVHRR and MODIS products are different datasets. To run a record from 1981, first check their differences over the post-2000 overlap. Where an official combined product exists (albedo MIX), it is the safer choice.
• LAI and FAPAR fill cloud gaps with a time-series model. A value being present does not mean there was an observation, so use the quality layer (cloud, snow, fill bits).
• File numbers in the HKU archive can differ from the documentation (FAPAR files start with GLASS09, ET with GLASS11, net radiation with GLASS07). Go by the actual folder contents rather than the numbers in the 'HDF File Naming Convention' table.
• The MODIS products are affected by Terra and Aqua orbital drift and their end of operations. Interpret trends from the 2020s together with these input changes.
• A single global 0.05° file ranges from a few MB to over 100 MB. For the tiled products (500 m, 250 m), select regions by MODIS Sinusoidal tile number (hXXvYY); the Korean Peninsula is around h28v05.

Notices and change history¶

No notices yet.

References for your paper¶

Reference papers¶

GLASS product suite overview (BAMS) — DOI 10.1175/BAMS-D-18-0341.1

@article{liang2021glass,
  author  = {Liang, Shunlin and Cheng, Jie and Jia, Kun and Jiang, Bo and Liu, Qiang and Xiao, Zhiqiang and Yao, Yunjun and Yuan, Wenping and Zhang, Xiaotong and Zhao, Xiang and Zhou, Ji},
  title   = {The Global Land Surface Satellite ({GLASS}) Product Suite},
  journal = {Bulletin of the American Meteorological Society},
  volume  = {102},
  number  = {2},
  pages   = {E323--E337},
  year    = {2021},
  doi     = {10.1175/BAMS-D-18-0341.1}
}

GLASS long-term dataset (IJDE) — DOI 10.1080/17538947.2013.805262

@article{liang2013glass,
  author  = {Liang, Shunlin and Zhao, Xiang and Liu, Suhong and Yuan, Wenping and Cheng, Xiao and Xiao, Zhiqiang and Zhang, Xiaotong and Liu, Qiang and Cheng, Jie and Tang, Hairong and Qu, Yonghua and Bo, Yanchen and Qu, Ying and Ren, Huazhong and Yu, Kai and Townshend, John},
  title   = {A long-term Global {LAnd} Surface Satellite ({GLASS}) data-set for environmental studies},
  journal = {International Journal of Digital Earth},
  volume  = {6},
  number  = {sup1},
  pages   = {5--33},
  year    = {2013},
  doi     = {10.1080/17538947.2013.805262}
}

GLASS LAI V6 (250 m, Bi-LSTM) — DOI 10.1016/j.rse.2022.112985

@article{ma2022glasslai,
  author  = {Ma, Han and Liang, Shunlin},
  title   = {Development of the {GLASS} 250-m leaf area index product (version 6) from {MODIS} data using the bidirectional {LSTM} deep learning model},
  journal = {Remote Sensing of Environment},
  volume  = {273},
  pages   = {112985},
  year    = {2022},
  doi     = {10.1016/j.rse.2022.112985}
}

GLASS FAPAR V6 (250 m, 8-day) — DOI 10.5194/essd-14-5333-2022

@article{ma2022glassfapar,
  author  = {Ma, Han and Liang, Shunlin and Xiong, Changhao and Wang, Qian and Jia, Aolin and Li, Bing},
  title   = {Global land surface 250 m 8 d fraction of absorbed photosynthetically active radiation ({FAPAR}) product from 2000 to 2021},
  journal = {Earth System Science Data},
  volume  = {14},
  number  = {12},
  pages   = {5333--5347},
  year    = {2022},
  doi     = {10.5194/essd-14-5333-2022}
}

Acknowledgements¶

Most GLASS products have no dataset DOI, so cite the suite overview (Liang et al., 2021) and the algorithm paper for each variable you use (see the HKU 'Algorithms' page), and state the product name, version (V number), resolution, download source (e.g. the HKU archive) and access date in the text.

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¶

• HKU GLASS download page (products, versions, periods)
• HKU GLASS overview
• HKU GLASS algorithms and references
• GLASS HDF file naming convention (PDF)
• BNU GLASS LAI page (scale factor, quality bits)
• BNU GLASS albedo page (layer names, scale factor)