I visited the US between April 13 and 29. My shoes were very closely inspected many times as was my laptop which is just holding together but still working after all the dismantling and re-constitution. I am sure they were very concerned due to my foreign habits of walking instead of driving leading to uncommon chemistry in the shoes...

But once through the security it is still just as good and exciting to be in the US and be in touch with the energy and intellectual drive that we could well have a bit more of here - but with the great Australian style they so admire.

1 Andover, Ma

The first week was spent at a workshop in Andover, Ma. Most people I talked to responded to being told this by "what on earth took you to Andover?". The reason is that it is the ideal place for a workshop - nothing else to do. The topic was an Ontar Workshop on Modtran/Fascode/EOSael/LidarPC etc etc. The week was very good and the others there were all heavy users of the various codes so the questions and depth of discussion were just what the PhD ordered.

2 Greenbelt, Md

Then the second week was a talk at the ASPRS as a user who thinks EO-1 is great and has outcomes (yours) to prove it followed by an EO-1 CalVal workshop and then 3 days of SVT meeting.

For a more pictorial response have a look on the EOC website (http://www.eoc.csiro.au) under "What's New?". There are two presentations to look at. One was my general talk on your behalf and the other a talk given by Jay Pearlman but actually the work of the Coly-Crew and also some very late nights just before the talk. There are some very interesting results there and in the later activities. Have a look and see what was said. The general talk was sort of the ASPRS talk with the real stuff (CSIRO science side) added in. Both talks went well, our work is excellent and I was proud to be delivering them on behalf of the EO-1 work you have done. Australia really is at the forefront of activity in Hyperion and the US side of the world knows it even if CSIRO does not always appreciate what has happened this last couple of years.

3 SVT summary

This was not, as feared, the last SVT meeting. There will be another - truly the last - in Hawaii in the week of November 18-21. I only learned this after I returned.

3.1 The Cal_Val workshop

This was well attended by non-SVT people. Both ALI and Hyperion seem to have dark current, instrument effects (such as echo and smear for Hyperion) and contamination under control and the monitoring is also effective and in-place. Hyperion is very stable in its calibration. Later in the meeting a question was raised about smile stability and extent but more of that later.

Both instruments have excellent PSF characteristics. ALI Pan is of very high quality both radiometrically and geometrically. There is some orbit motion "blur" but generally it is at a level most normal users will not complain. The 10 metre pan enhanced ALI data is in quite high demand apparently.

Landsat, ALI and Hyperion have been brought into consistency if not correctness by a "committee" decision to change Hyperion by 8% in the VNIR and 18% in the SWIR. Actually, Coleambally and Frome were used to help support and make the decisions. People with Level 1B1 have the modified calibrations. TRW put a "ramp" between the calibrations across the overlap area. It is unclear if the ramp is good or not. Not everyone knows it is there and any attempt to convert Level 1An data to Level 1B1 needs to get access to undocumented information and files.

However, if you look at the presentation by Jay on our web site you will see an example that suggests Landsat and Hyperion are now very well aligned. What is the "truth" is yet to be decided but at least they all seem to be consistent. If you look closely at that graph and the table associated with it you will see that the biggest difference is in Band 4. Brian Markham gave a talk on the Landsat calibration and monitoring. He mentioned that Bands 4 and 8 (8 is the Pan) have drifted the most. So the difference may be due to the Landsat calibration not being recent enough.

3.2 Coleambally

This was the work Jay presented. The Coleambally site has become a key international site now. It has been the focus of effort by a lot of people to provide it with an EO-1 time series which would now cost an un-affordable (for us ozzies) amount of money. It has a near-permanent Aeronet/CIMEL installation (which will continue until at least September this year and possibly on into the future after that) and it has been used as one of the data sets used in NASA/TRW Hyperion courses as well as in the Australian workshops EOC is running. It is certainly the most complete "purposeful" (in this case agricultural) time-series of EO-1 data that has been collected.

The local efforts have been aimed mostly at agricultural applications obviously but also it has been used for Hyperion/ALI/Landsat cross calibration monitoring and it will be used in the spectral "smile" study I will discuss below. As well, using a technique for spatially well defined ASD measurement developed at Lake Frome by EOC, Coleambally has been used as a reference site to compare atmospheric correction codes such as ACORN, FLAASH and HATCH. The outcomes of these tests were discussed at the Greenbelt meeting and some of the results are in the presentation on the web.

It has become clear that although Hyperion does not have as good SNR as (say) Aviris, it can be made into an excellent data set with care taken to ensure bad pixels and the push-broom streaks are taken care of. At this time there is a lot of attention to the spectral smile as this seems to be limiting the atmospheric correction efforts and adding unnecessary noise in the process. But with care to the more systematic "noise" quite stunning data sets can be obtained. Fred Kruze reported that this was only so when the environmental conditions were good (eg high sun angle and clear atmosphere) but that is the case with all remote sensing.

3.3 The SVT meeting - general report

It was a great meeting. Almost every group is processing data, discovering the same issues and generally taking much the same approach to overcoming them. This includes the bad pixels, streaking and other noise reduction strategies. There are various groups looking at binning strategies and already very interesting applications going on. For example, "DAR" Roberts can use Hyperion much as Aviris has been used to map liquid water in vegetation. This has immediate applications to fire potential. Hyperion is more regional and less precise than Aviris of course but this application at least usually is associated with clear skies and high sun angles! If EO-1 is around next oz-summer it may be an application.

Other applications were very nice results from Antarctica on the Ross Ice Shelf, applications to volcanoes from Luke Flynn, some interesting Hyperion/Aster interactions and comparisons by Jim Crowley, and good work mapping invasive species such as Leafy Spurge and other less nasty sounding vegetation types. The modellers are also becoming obvious as the dominance of CalVal and instrument checkout decreases. Jim Smith showed how 3D canopy modelling at Hyperion resolution was very different from SAIL in the red-edge and NIR regions.

Onto atmospheric correction, Alex Goetz described the progress with Hatch-2D. There are some issues and it is not yet available. He said it will be soon but the fact is that the benefits of the "calibration" phase (where central wavelengths are changed to minimise variance in the major atmospheric absorption regions) and of the 2D approach (where this is done separately for each column) are not clear. Nor are the new codes really fully validated. Nevertheless, it seems from the Coleambally tests that Acorn, Hatch and Flaash are very similar to the level of limit due to Hyperion SNR and the current major issue of what the spectral calibration really is? In theory, the Hatch calibration can decide this but it has not yet doe so. There are some possible reasons for this and Alex Goetz is currently addressing them.

Joe Boardman is working on a "super-effort" that will remove the spikes left after atmospheric correction and also the streaking all in one fell swoop of an ENVI plugin. He also reported on SNR effects and concluded (like Fred Kruze surprisingly enough) that a clear day with high sun is essential for good geological mapping with Hyperion. However, given those conditions the results are far better than was originally expected from Hyperion. He claimed that the things of concern to practical users and uses in order of importance were:

SNR, Uniformity, Stability & Calibration

Sounds like "consistent and standardised data sets" maybe?

There was discussion of spatial effects and geometry. Dave Goodenough reported excellent registration using PCI software and sinc function interpolation to sub-pixel accuracy. Most SVT members are happy to use the Level 1B1 corrected data and ENVI with GCPs. However, there was some discussion of the possible mis-registration among the VIS bands. John Mustard suggested that the visible "breathing" most people have seem as they use ENVI "animate" on Hyperion is evidence of this. I suspect it is also partly spectral and involves the change in "smile" but band mis-registration is also possible.

There is reason to believe that band 35 will be displaced in the "y" direction relative to bands 1 and 70. But since we normally only use bands 8-57 it may not be much to worry about. But it is worth testing bands 10, 35 and (say) 56. These tests are being done but I am not sure now by whom.

In the wet areas, John Mustard reported work they were doing which provoked the usual discussion of whether the strong feature at about 710 nm is in any way due to fluorescence. But it was nice to see such water quality signatures being so well captured by Hyperion. The Venice Lagoon study had bad luck with images but the Ikonos images were really something!

3.4 The smile issue

I will discuss this separately in another report. Rob Green presented some work he did with data from the Arizaro dry lake in Argentina. This is very high up and the air is very dry. The surface is "visible" in ALL bands. It is also bright and has a flat spectrum at all wavelengths. Its main problem is access and that the current characterisation is a bit limited in extent. However, there are Aviris and Hyperion data to use.

Rob tracked a number of bands, the O2-a band at about 763 nm, the smaller O2 band at about 1270 nm and the CO2 band(s) at about 2000 nm and found that the apparent "smile" was similar in shape to the TRW published form but offset by about 1.5 nm in the VNIR and 3 nm in the SWIR. This, if true, is very serious.

Barbara Carlson of GIS also tracked the O2-a band and came up with a similar result in shape but not as shifted as Rob found (more like 1 nm at the worst case but "crossing" the TRW curve). I also did some work on the Frome data (where a nice band of salt goes right across the image) and got different results in offset but similar in shape. The issue is that the method used may lead to an offset due to continuum effects and the difficult nature of the correlation at 10 nm resolution.

There is an international exercise going on now to look at the smile at various atmospheric features across relatively uniform or well-characterised areas. One issue is whether there is an offset? (If there is it will make all our atmospheric corrections spiky); Another is whether it has been changing with time? (this is a bit scary); Another is can such vicarious studies really establish what the situation is over all the bands? (this is unlikely!).

If any of you wish to be involved in the ongoing discussions and activities please let me know. Frome is being used but it occurs to me that Uardry may also be very useful for this task.

4 Other SVT bits arising during the meeting

It seems that the software being developed at NASA to convert from Level zero to Level 1B1 will become available but not immediately and possibly only if we can get an extension to our current agreement with NASA. I am continuing to pursue this and will keep you informed.

Both Hatch and Flaash will do wavelength calibration now (but not commercial Hatch) and both groups are involved in the smile study. Obviously, Acorn is involved. I think it is clear that atmospheric correction has matured very suddenly and hence can be included easily in our Hyperion workshops.

Places where results can be communicated are IGARSS and the next SVT meeting. I think we should see how to get a good representation at the final Hawaii meeting. I suspect your Divisions will have to value it so make a case or shore up/budget the funds for it. We are right up at the front and we should ensure it stays that way.

5 Acknowledgement

The EOC Travel Committee supported the visit to 50%. I hope the outcomes will justify their investments.

David L B Jupp
CSIRO EOC