[Live-demo] LANDSAT-8 sample data of osgeo tutorials and workshops

Hamish hamish_b at yahoo.com
Sun Sep 8 01:49:53 PDT 2013

Hi all,

With the availability of LANDSAT-8 imagery now coming on line[1] it seems a good time to make an accessible common sample dataset available for everyone to play with and adapt+race their various softwares against. More acutely for the OSGeo Live DVD we're looking for a modern multi- or hyper-spectral imagery standard dataset that we can share between quickstarts. The newly online MERIS data[2] might be nice for that but the EU is not as gifting as the US gov't wrt copyright redistribution issues (although this new release is really encouraging!) and the MERIS ground resolution is 10x coarser than that of LANDSAT.

The obvious choice for a ground target is to continue with the Raleigh, NC osgeo-edu landsat time-series[3], and on 2013-05-14 there was a nice image taken:

In the above link the red box is the bounds of the earlier imagery crop. This time I propose that the earlier scenes were a bit small, and to increase the image size to 1000x1000 cells. Raleigh-Durham Airport is present in the NW corner of the proposed new bounds for visual scale and human interest.

projection issues: I'd note that these come from USGS's GloVis as "level 1" data, but actually they are already reprojected into local UTM zone and gridded at 30m resolution. I'm not sure I'd call already georectified level 1, and might be interested in working with what could be called "level 0" in that scheme, but whatever. To match the earlier datasets the above would have to be reprojected from UTM 17N to LCC (NC State Plane) with associated resampling fidelity loss.

I'd like to throw in another nearby imagery target to the proposal though, i.e., Cape Charles at the mouth of the Chesapeake Bay. The reason for choosing the site is that the Chesapeake has been the focus of many previous studies & is well known in the literature; human interest (gee it's gorgeous); a chance to explore the new LANDSAT-8 coastal band 1 [4] (personal interest/bias :-); and finally good overlap with one of the HICO hyperspectral imager's[5] calibration target locations for sensor to sensor comparisons.

here's the 1000x1000 crop of that: (UTM 18N, 2013-07-19)

If that were only a bit taller to the north we'd get to see some cumulus clouds & be able to pick off some top-of-cloud reflectances, FWTW. Both the Raleigh and Cape Charles images are nice clear days so little chance to play with the new Band 9 cirrus filter or see much in the accompanying QA product (* which seems rather nicely done btw, suggesting de-clouding haze and cloud edges will be much better than in earlier LANDSATs). I shy away from providing students with already-perfect images as then they've got very little processing to do/chances to learn but I guess that's a matter of what your lesson's needs/goals are.

other LANDSAT-8 points of interest:
* new 15m panchromatic band 8 and many more IR bands to work with
* 11 spectral bands and 1 QA product to play with
* data is now 16-bit
* 1000x1000 cells at 2 bytes per cell and shipping 12 bands will make the download about 23 MB uncompressed; to save a little the associated QA product might be reclassed to UInt8 instead of UInt16 since only a few values seem to be used in practice.

in-line links:
[1] http://landsat.usgs.gov/about_LU_Vol_7_Issue_4.php
[2] https://earth.esa.int/web/guest/missions/esa-operational-eo-missions/envisat/news/-/article/envisat-entire-meris-frs-l1-archive-processed-for-online-distribution
[3] http://www.grassbook.org/data_menu3rd.php
[4] http://landsat.usgs.gov/band_designations_landsat_satellites.php
[5] http://hico.coas.oregonstate.edu


hoping to start some discussion (pls reply to the geodata list),

More information about the Osgeolive mailing list