Mapping Land Ice
From Byrd Polar Research Center - Research Wiki
The Global Land Ice Measurements from Space (GLIMS) has made important progress on a very detailed yet incomplete global land ice inventory [see e.g. Kargel et al. 2005, Raup et al. 2007, and GLIMS Web Site. A new NASA-funded project (1 September 2011 - 31 June 2014) entitled Mass budget closure on the global inventory of mountain glaciers and ice caps: past and future sea-level rise and streamflow variability has a sub-component led by Jason Box and Christine Chen at Byrd Polar Research Center.
We are to finalize and publish a globally-consistent inventory for total glacierized area and individual contiguous ice masses outside of the conterminous Greenland and the Antarctic ice sheets. We have a prototype methodology already in use by the scientific community via a data set of a Greenland Land Surface Classification Mask
Our most recent land ice classification is made at 1.25 km ground resolution using thresholds applied to a.) MODIS band combinations such as NDSI and b.) MODIS individual bands. Using 12 annual land surface classifications (2000, 2001 .. 2011), we determine a land ice area for Greenland to be 1.76428±0.016 x 10E6 sq km. The number to the right of the ± symbol corresponds with the 8.3% difference in probability of land ice between 100% certainty (in our classification) and the next lowest class 91.7%. This increment is 1/12, as we develop probability from 12 annual land surface classifications. If the classification determines land ice in all 12 years, the probability is 100% or 12/12. If land ice in 11 of 12 years, the classification probability for this 12 years, according to our simple method is 91.7%. This probability image is downloadable from http://bprc.osu.edu/~jbox/LSC. The data file is binary float with dimensions 1860 x 1740. This page contains other relevant information for reading and re-projecting this grid. This 'mask' is on the same grid as the Greenland Accumulation Grids.
The NASA project is co-led by Regine Hock of the University of Alaska, Fairbanks. Postdoc Andrew Bliss joined the team and has begun analyzing our ice classification for the Antarctic Peninsula.
We have input from and collaboration with Graham Cogley of Trent, University., who is a leader in global land ice inventory work (see Cogley, 2009).
Our inventory is based on a 0.25 km land surface classification that uses NASA Moderate Resolution Imaging Spectroradiomenter (MODIS) data. Using a segmentation algorithm, we have automatically given an ID to 6494 isolated ice caps in Greenland that are at least 0.625 sq. km in area.
As many as 12 annual (2000-2011) land ice classifications each incorporate a number of daily cloud-free images from the end of summer each year, late enough in the year that winter snow is melted away revealing the land ice. The images are classified as land, sea, or ice using information from MODIS band reflectance, multi-spectial ratios, Normalized Difference Snow Index (NDSI), and Normalized Difference Vegetation Index (NDVI) [e.g. Solomonson and Appel, 2004].
Our work neither replaces nor duplicates GLIMS efforts. Despite admirable accomplishments, the progress towards a complete inventory of the world’s glaciers has been slow due to emphasis on high precision. In contrast, our proposed work alleviates the urgent need of inventory completeness though at the expense of coarser resolution than GLIMS. For global sea-level closure it is important to have a complete inventory now whose accuracy can then be progressively improved through, e.g. GLIMS. While considered “moderate” resolution, our 0.625 sq km land ice grid has a much higher precision than used by any past global land ice mass budget closure assessments. Hock et al.  used a 110 km global grid.
Having multiple years of MODIS-based classifications, we may determine ice area masks accompanied by uncertainty statistics, for we find an inter-annual variance in ice cap area. To be more sure of the calculated permanent ice area, we are pursuing multiple approaches for validation and classification refinement. These include:
1) comparing MODIS land surface classifications with those applied to ASTER imagery using published approaches by GLIMS investigators [e.g. Paul and Kääb 2005];
2) using published GLIMS outlines.
3) Debris covered ice may be identified by combining DEMs into multispectral classifications [Paul et al. 2004].
See also the older Greenland Land Surface Classification Mask.
- Cogley, J.G., 2009, A more complete version of the World Glacier Inventory, Annals of Glaciology, 50(53), 32-38.
- Hock, R., M. de Woul, V. Radic and M. Dyurgerov, 2009. Mountain glaciers and ice caps around Antarctica make a large sea-level rise contribution. Geophys. Res. Lett. 36, L07501, doi:10.1029/2008GL037020.
- Kargel, Jeffrey S.; 16 others (2005). Multispectral Imaging Contributions to Global Land Ice Measurements from Space. Remote Sensing of Environment 99, 187--219. doi:10.1016/j.rse.2005.07.004.
- Paul, F., C. Huggel and A. Kääb. 2004. Combining satellite multispectral image data and a digital elevation model for mapping debris-covered glaciers. Remote Sens. Environ., 89(4), 510–518.
- Paul, F. and A. Kääb. 2005. Perspectives on the production of a glacier inventory from multispectral satellite data in Arctic Canada: Cumberland Peninsula, Baffin Island. Ann. Glaciol. 42, 59–66.
- Raup, B.H.; A. Racoviteanu; S.J.S. Khalsa; C. Helm; R. Armstrong; Y. Arnaud (2007). The GLIMS Geospatial Glacier Database: a New Tool for Studying Glacier Change. Global and Planetary Change 56:101--110. (doi:10.1016/j.gloplacha.2006.07.018).
- Solomonson V.V, and I. Appel, 2004. Estimating fractional snow cover from MODIS using the normalized difference snow index. Rem. Sens. Envt. 89, 351–360.