DIRECT BROADCAST (DB) ON EARTH OBSERVATION DATA MEETING

7 - 9 DECEMBER 1998

jointly supported by

CSIRO, COSSA, NASA, ACRES and the Bureau of Meteorology

held at

CSIRO Corporate Headquarters, Limestone Avenue, Canberra, AUSTRALIA

Report by David Griersmith (Bureau of Meteorology)

Monday 7 December 1998

Opening

1. Dr David Jupp welcomed participants and drew attention to the Program (details were distributed at the meeting and are attached). It was noted that a Press Release had been organised to coincide with the event. He introduced the first (keynote) speaker Dr Jim Dodge, Program Manager, Global Change Data Analysis, Office of Mission to Planet Earth at NASA Headquarters.

Dr Jim Dodge (NASA):

2. Dr Dodge spoke on "An overview of the NASA Plans for EOS (Earth Observation System) Direct Broadcast". He said that a 5m program track antenna was sufficient for a ground station for direct broadcast (DB). A preliminary version of the EOS-AM1 Direct Broadcast Users Guide is available and has been provided to participants via the Organising Committee. For AM-1 NASA is planning on four ground stations at GSFC, University of Wisconsin, University of South Florida and Hawaii. Dr Dodge described a number of technical documents which he provided for copying.

3. For each overpass of MODIS there will be about 1Gb of data. For MODIS he suggests that initially one could experimentally treat it as a 36 channel AVHRR instrument. NASA is promoting the concept of an international user community which shares information on data reception, processing algorithms, calibration and validation and so on. NASA's policy is to support free exchange for scientific and educational users.

4. The MODIS instrument has 250m resolution for channels 1-2; 500m for 3-7 and 1000m for 8-36. Potential real time products include surface temperature, NDVI, ocean colour and so on. There is a MODIS Airborne Simulator which has 50 bands. Dr Dodge showed examples of images from this instrument which are available on the WWW.

5. A chart of forthcoming missions is available at www.earth.nasa.gov/missions/2002/index.html. EOS AM-1 is expected maybe around August 1999. WWW NASA pages give extensive details such as www.modarch.gsfc.nasa.gov/EOS-AM/BROCHURE.

6. Dr Dodge spoke about SeaWiFS and suggested colour map products might include oceanographic parameters such as SST or ocean colour complemented with vegetation, topography or surface temperature over the land. He suggested this might provide an up to date composite of the environment over a particular region. In order to do any quantitative derivations corrections must be done for smoke or dust (SeaWiFS images were shown which easily delineate these phenomena very well).

7. FY-2 and GMS-5 images were shown and he said that stereographic imagery products could be derived. He was promoting the concept of integration of data from different sensors and satellites.

8. An International Network of EOS DB ground stations might for example comprise: 4 in USA for AM-1; 2-3 Australia; 1 in Japan; 2-3 China; 1 in Brazil; 1 in northern Europe; and there is also interest in Indonesia and Thailand. NASA expects this type of regional international approach to work with local processing otherwise large costs for centralised approaches might be incurred including associated very expensive communications costs.

9. EOS PM-1 will have in addition to MODIS, AIRS, AMSU, AMSR-E, CERES and HSB so meteorological applications will increase. NASA is planning a joint mission to bridge the gap between the EOS series and NPOESS. He said there would be a continuing series of high bandwidth X-band satellites over the next decade or more. It is clear that this will actually become part of a gradual migration toward X-band for all operational satellites.

10. Some of the discussion indicated that for some processing software being developed there would be a copyright on free software and as a result users would have to be prepared to distribute source code if they do add-ons to the original software. The initial focus will be on Level 0 and 1 data (and possibly level 1.5) i.e. calibration and navigation. Dr Dodge and others wish to be fairly conservative about what can be promised early on and initially most products may be fairly simple. Wisconsin will provide some operational software to give at least Level 1. NASA is supporting that effort. Participants noted some of the problems being encountered in Europe concerning restrictions on software access for ATOVS processing.

Richard Smith (Western Australian Government, Department of Land Administration): Operational Applications of near real time NOAA-AVHRR in WA

11. Mr Smith spoke on the range of applications of NOAA AVHRR including NDVI, fire monitoring and sea surface temperatures. He also drew attention to WASTAC's plans to receive X-band data including EOS AM-1 MODIS. WASTAC is the Western Australian Satellite Technology and Applications Consortium which operates a NOAA/SeaWiFS reception and processing facility in Perth, Western Australia, and which comprises DOLA, Curtin University of Technology, CSIRO and the Bureau of Meteorology.

12. Huge areas of tropical savannah are burnt each year in northern Australia in areas that have very low population density especially in comparison with the southeastern coastal regions of Australia or capital cities. Such tropical fires can potentially cause severe environmental degradation. Fuel load, curing (fuel flammability) then fire spotting are major steps involved in the process. Within 2 hours of the NOAA AVHRR overpass processed data are placed on the Internet. Mr Smith showed examples of fire history maps over the Kimberley (northwest Australia). This history mapping has been done since 1993. Curing index maps over the Kimberley were also shown. There is a seasonal bushfire cycle like a normal distribution in which there is a build up, curing, peak, fires, hotspots, then a tapering off with lack of fuel, rain, colder weather and winter.

13. Richard Smith said that his Remote Sensing Services group in WA DOLA could only achieve all these services and products because of its strategic alliances with the Bureau of Meteorology and CSIRO. He specifically cited the very good NOAA reception network operated by the Bureau and the efforts of Bureau staff Andrew Donaldson, Paul Bekker and others in developing an excellent system.

14. Salinity problems on the land were raised by Jim Simpson and Mr Smith indicated this was assessed with Landsat data on the basis of areas that persistently do not respond to rain, are low in the landscape and so on.

Rob Lees (SPOT Services Australia): SPOT Vegetation instrument

15. A new vegetation instrument is being flown on board SPOT-4. The SPOT commercial program began in 1986 with the launch of SPOT-1. Currently the vegetation instrument on SPOT-4 is downlinked via X-band to Kiruna from solid state memory. Also there is S-band transmission to the ground. SPOT Image is responsible for the commercial distribution of the vegetation instrument data, although there is also a research program. For the vegetation instrument data is processed at CTIV in Belgium. Authorised S-band ground stations can do processing in real time (RT). The numbers involved are: 7-8 Gbits per day; 4 to 5 memory dumps per day; 3.4Mbps; daily data processing 10 Gb; delivery of commercial products in 2-4 days which is suitable for most clients.

16. SPOT-4 also has Panchromatic 10x10m 0.61-0.68 microns; multispectral 20x20m with 0.50-0.59, 0.61-0.68. The vegetation instrument has 1.15x1.15km 2200km swath and has four(?) channels including 0.79-0.89, 1.58-1.75 microns. The general registration error is around 0.3km. There are three main goals of surface parameter mapping: terrestrial biosphere mapping; agricultural; and pastoral and forestry production (commercial). The first two of these are more research oriented with application to climate change.

17. There are two main streams of data: VGT-P products in which radiometric corrections, geometric corrections are made with resampling to a 1km grid and simple classification to land, sea ice etc which goes into a catalogue; VGT-S includes daily synthesis computed from all daily passes in the region of interest; decade synthesis is computed from all passes for a ten day period with removal of cloudy pixels, with the highest NDVI being used (i.e. maximum value composite approach). Note that the processing time is 2-4 days.

18. Typical prices are as follows: daily synthesis 1 million sq km - US$150; 1-4 mill sq km - US$200; 10 day synthesis 1 mill sq km US$ 400; 10 day synthesis 1-4 mill sq km - US$500. He argues that many years ago turn-key systems for ground stations were common but now Internet access and product delivery from centralised processing organisations to clients meets many needs. Some clients do not have the interest, resources or people to do it all themselves and they want fast access via ftp.

19. From general discussion and questions it emerged that getting a ground station authorised is a commercial proposition. Some data will be available free for researchers.

Ian Barton (CSIRO Marine Research): Marine applications form Direct Broadcast data from EOS platforms

20. CSIRO Marine Research has centres in Hobart, Perth and Brisbane with around 350 people in total. Hobart uses AVHRR, GMS, SeaWiFS (Hobart and Townsville have the decryption software for use in RT), TOPEX/POSEIDON, ERS-1/2 etc. In the near future there is interest in EOS-AM1, PM-1, Envisat, ADEOS-II, etc.

21. Dr Barton said there were many, many users requiring X-band data or interested in X-band data. He advocated a processing system to get Level 1B products, namely accurately calibrated and navigated data. He also said that a network needs to be established which gives and shares around cal/val data, for example from ground sites. In Australia a lot of validation data will b collected at remote sites called Hay, Tinga Tingana, Lake Argyle and Thangoo.

22. The new Australian small satellite under construction and called Fedsat will involve low data rates and communications experiments plus a minerals imaging satellite called Aries will have an imaging spectrometer. Both of these initiatives signal an important local role in space matters.

23. Dr Barton gave diverse examples of products and scientific studies in areas such as SSTs, chlorophyll, plus ERS-2 scatterometer data around a tropical cyclone. Also shown was TOPEX/POSEIDON data with ocean height overlaid on AVHRR data which shows height correlated with currents. For Antarctica ERS-2 SAR data gives good coverage and scatterometer data can help in monitoring iceberg movement.

24. Lower cost X-band systems are becoming available, for example see www.neonet.nl/rapids. Also SAR processing on a laptop is now possible.

25. Discussion ensued on MODIS and SeaWiFS and oceanographic applications. Dr Dodge explained that NASA had actually been approached with a request for NASA to shut down MODIS DB, however Dr Dodge explained many millions had been invested and if other similar missions had launched in time there would have been no problem via overlap with AM-1, hence DB will go ahead. This policy was applauded by those present.

Graham Harris (Chief, CSIRO Land and Water): Land applications

26. Dr Harris highlighted huge costs for unsustainable farming practices costing Australia hundreds to billions of dollars. He suggested one estimate of Australia's land assets indicated a value of $650 billion. Many basins such as the Murray-Darling in the east of Australia are in trouble. Salinity is a major problem. Adelaide's water supply by 2030 will be disastrous due to annual salinity increases of around 1 to 3 per cent in many parts of the basin in New South Wales. Currently in Australia there are major incentives for plantation forestry which will result in massive changes in agriculture practices. The Federal government is now writing a new national resource management policy. There are major issues on agricultural production and integration of land and water use. Where should new forests be placed?

27. How are alternative land use practices going to be handled e.g. retiring some pastoral areas from use; using public monies on private land? What are the performance measures for improving land use and decreasing salinity?

28. Considerable investment has been done by various Government and CSIRO groups and the involvement of remote sensing in the new initiatives will be important. Examples of the latter include:

SIR DEM for high quality topographic data;

VCL (vegetation canopy lidar, airborne) for tree height and canopy structure;

NDVI, MODIS;

CASI, AVIRIS, Hyperion-Aries, etc

29. Outcomes include strategies to ameliorate water and land degradation plus value-adding technologies for allowing sustainable production and environmental benefit.

Jim Simpson (Scripps, UCSD): MODIS product development and validation: an EOS Direct Broadcast Demonstration Project

30. Dr Simpson said that the project principals included himself, Torben Nielsen at Hawaii and others. The project is based on the central tropical Pacific region, centered on Hawaii and will involve a distributed computing and data delivery approach. He is seeking good scientific products and rapid delivery.

31. He described the scientific methodology adopted including cloud clearing. Numerous examples were given illustrating some of the latest techniques available. Various data types have been used to date including AVHRR and ATSR. Dr Simpson spoke about some more advanced products such as velocity structure and combinations with models of flow (published in IEEE) which were also based on AVHRR and ATSR. In one oceanographic example, validation was done using Lagrangian drifters off the coast of California.

32. Dr Simpson described a scheme for determining cloud height which involved angle (geometry) determination on a pixel by pixel basis plus cloud shadows. Validation of cloud detection, shadow detection and height determination was done via lidar. The mean difference between the two is around 400m for low clouds noting the lidar error is around 100m and from AVHRR pixel size the error must be at least 300m. For 10km altitude clouds the error is bigger but still results in very useful data.

33. NDVI with and without cloud and cloud shadow shows that the NDVI distribution is markedly improved after corrections which were actually determined initially for oceanographic applications.

34. Dr Simpson showed examples of volcanic ash detection (T4-T5 negative) using AVHRR which showed that there can be problems in discriminating between ash and normal clouds. He believes that a new technique has been developed by him which may allow use of channels on MODIS to greatly improve the accuracy of volcanic ash detection.

Peter Bayliss (NERC satellite station, University of Dundee): NERC satellite station at the University of Dundee

35. Dr Bayliss explained that the University of Dundee is just north of Edinburgh in Scotland, at about 56N, 3W. They have a 1.8m and a 2.4 m antenna. They had a PDUS in 1977 ready for Meteosat. Recent upgrades have included AVHRR in 1978, CZCS in 1979 and SeaWiFS and very recently a move to a 3.7m dish for X-band. Natural Environment Research Council (NERC) is providing funding. Their archive has about 3 Tb and the early part is unique. Running costs are around US$300-400k pa.

36. A consortium of UK researchers headed by Jim Aitken has enabled access to SeaWiFS in real time: their data is relayed to Goddard (see example on the Web of an image around UK).

37. Products are made in Plymouth for use by users including quicklooks on CD-ROM. The main Web site is at www.sat.dundee.ac.uk. Users span a wide range across the community; about 220 scientific papers have been published during the last 5 years using data from the Dundee satellite data receiving station.

38. NERC is already receiving applications for funding for projects using data from MODIS expected to be received from Dundee. They are using a 3.7m Andrew antenna with zero backlash gear boxes, HPBW 0.7 deg and low noise amplifier. (Dundee also has a 1.7m AVHRR system). 15 deg per sec max slew rate is an issue for overhead passes - the antenna is told to follow an optimised path rather than a predicted trajectory so that errors are minimised. The resultant error of 0.65 deg is too big. Signal loss is -3db for 0.35 deg pointing error and maybe as much as 15db for 0.65 deg pointing error (the latter would result in total loss of signal). They circumvent this problem using a hybrid auto track system.

39. They use a Mitec LNA as part of the X-Y servo controlled tilting feed. The ground receiving station shows an Alcatel (made in Belgium) bit conditioning demodulator for deconvolution. They made their own simulator, frame synch and they use a Sun SPARC workstation. The signal simulator uses Mach231 PALs and 2x64kb e-Proms. 127 frames in e-prom. They also have a satellite look alike output for testing. The frame synchroniser once again has a bit of really good IP in it. The computer includes a 70Gb DLT with heaps of disks: 4Gb, 21Gb and 6x9Gb.

40. Discussion showed that program tracking can result in only a very small pointing error margin. However auto track can work better. A few hundredths of a degree can result in noticeable degradation. Low elevation passes are OK but above 80 deg there is a problem.

Bertil Granberg (University of Karlstad, Sweden): Construction of an X-band receiver station for AM1 DB

41. Dr Granberg described the Swedish system which is around half way to completion. Coverage from Sweden was shown diagrammatically. They used a 3.7 m Cassegrain antenna: the goal is to receive data and make it available at level 0. The total cost is around US$200k and most is a result of software and to a lesser extent engineering. The block diagram for the overall system shows gain 49db; LNA then a comms link and a few components which were a challenge such as the Viterbi decoder, frame sync and Reed-Solomon decoder.

42. The most important challenge has been the antenna electronics design which can go up to 100 deg per sec and 0.05 deg in azimuth and elevation. The front end rectangular waveguide and low noise amplifier were shown. The NEC receiver is retuned from 7.26GHz to 8.212 and bit rate from 26.5 to 34 Mbps. The NEC OQPSK demodulator had to be modified. Equipment includes a Stanford Telecom Viterbi decoder type STEL-2060, plus NASA frame synch type PIFS and NASA Solomon-Reed type RSEC. They use a Pentium II with an AMCC prototyping card for the PCI-bus. A digital transmission analyser is used for testing. Planned applications for MODIS data include vegetation, snow and ice, weather forecasting. See www.kau.se/~gerhabax/modis.html and note Bertil Granberg's email address: bertil.granberg@kau.se

43. Dr Granberg noted a Nokia commercial digital TV receiver with RF-tuner, OQPSK and many other X-band reception components ideal for AM-1, for around US$500!!

44. A full copy of overheads presented during the presentation was provided.

Fritz Hasler, (NASA GSFC): NASA/NOAA/AMS Earth Science Electronic Theatre

45. Dr Hasler gave a colour presentation emphasising visualisation tools for EOS data. He showed a 25Gb movie of hurricane George. He noted that the Science and Public Outreach program has an impressive client-server IT system for this type of visualisation. He showed an example of VIS-5D indicating animated cloud motion winds around a hurricane. A JPL ocean model movie was shown plus a colour topographic map illustrating El Nino ocean heights. Many colour animations were shown which have direct application to monitoring environmental phenomena such a hurricanes, fires, El Nino, drought, biomass burning, severe thunderstorms, tornadoes. This spectacular visual feast illustrated how major print and electronic media are very eager to use publicity material to communicate results from satellite missions and visualisation tools.

46. Discussion reinforced the value of visualisation, Internet access and animations.

Tour: ACRES (Australian Centre for Remote Sensing)

47. A tour of the Australian Centre for Remote Sensing (ACRES) took place in the afternoon of the first day (Monday). ACRES operates two X-band facilities (one at Alice Springs near the centre of Australia, and the other in Hobart, Tasmania (southeast region of Australia - ACRES operates the latter on behalf of a Consortium). It also has an end to end operation involving reception, processing, archival, product preparation, services delivery, marketing and so on. Mr Peter Holland, the General Manager of the Australian Surveying and Land Information Group (of which ACRES is a part) and the Manager of ACRES, Mr Paul Trezise, provided welcoming comments and Mr Trezise gave an overview of ACRES operations and activities before the tour of facilities. Participants greatly appreciated the hospitality of ACRES and were shown numerous examples of products from a range of satellites such as Landsat, Adeos, Spot and ERS. Details of ACRES activities may be found via its WWW site at http://www.auslig.gov.au/acres/index.htm

Tuesday 8 December 1998

John Woolner and Andrew Longhorn (ACRES): Description of TERSS (Tasmanian Satellite Earth Resources Station) and ACRES activities

48. John Woolner provided an overview of the technical and design characteristics of the ACRES and TERSS facilities. ACRES operates TERSS for the TERSS Consortium comprising ACRES, CSIRO, Bureau of Meteorology, Antarctic Division and University of Tasmania, which owns the station.

49. TERSS is a fairly unique remote autonomous facility for X-band reception with a 9m antenna located in Hobart, Tasmania. If wind reaches a dangerous level the antenna is shut down; also from Canberra the schedule and programming can be changed. Data is saved direct to 20Gb RAID. The 9m antenna can do program track, auto track and conical scanning. There is a microwave link from the antenna at Droughty Hill over the water some km to CSIRO Marine Research at Battery Point - the latter contains the main computing equipment. Full details are provided in the copy of overheads provided to accompany the presentation.

50. In general the entire TERSS facility is non-satellite specific. The new ERSDEM demodulator will be installed in 1999 and allows capability for SPOT, ERS, RADARSAT, LANDSAT, EOS AM-1 and ENVISAT (see table in overheads).

51. There was some discussion of possible scheduling clashes for example in relation to MODIS. At TERSS one can drop a satellite and go to another within 60 seconds. Currently Alice Springs (the other Australian X-band station operated by ACRES) is slower for a changeover time between satellites but is being improved. For PM-1 John Woolner suggested that probably the only possible scheduling conflict might be NOAA-14 (ACRES is considering putting in a small NOAA AVHRR system at Alice Springs). Landsat acquisition is toward evening or dusk.

52. For the second half of the presentation Andrew Longhorn gave an overview of recent ACRES development from the perspective of a software engineer. The upgrade project is likely to end around February 1999. The traditional model involves manual tape loading, manual handling for shipping and at Alice Springs manual acquisition - there can be a delay of a week before metadata is available. Under the new model metadata (e.g. quicklooks without cloud assessment - the latter will be completed within hours) will be sent electronically and available within about one hour. Two DLTs will be written, one for backup at the site and the other will be shipped.

53. The new acquisition architecture uses a PCI card which takes the serial stream from the demodulator. There is a small robotic DLT stacker which holds around 30 tapes. The total budget for upgrading the Alice Springs facility is around $400k. SGI computers are used in the processing. New Product Generation Systems are available. If rapid access is required to data then a request can be sent to the acquisition station and a part of a pass can be sent by Internet. Currently the link to Alice Springs is 64kbps and 2Mbps to TERSS. Rather than necessarily upgrading all comms, requests can be made to a reception station and some local processing at the antenna might be possible in the future. Dr Simpson suggested this was a viable desirable strategy.

Dr John Le Marshall (Bureau of Meteorology): EOS plans and activities in the Bureau and the International TOVS Working Group (ITWG) and the NASA EOS

54. In 1987 the Bureau made two proposals in response to a NASA Announcement of Opportunity on the Earth Observation System (EOS), one on AIRS and one an interdisciplinary proposal. The rationale behind these proposals involved the value to Australia of accessing high resolution meteorological and related data which can lead to significant improvements in the accuracy and timespan of numerical weather prediction models.

55. Activities to date for interdisciplinary work include algorithm development using an interactive technique (AIRS has over 2000 channels and delta lamda/lambda of 0.001). Also investigations have been done using the radiative transfer equation in perturbation form plus use of AI techniques to parameterise errors in the retrieval system. Direct assimilation of radiances into the analysis scheme is also being studied. Hence for the interdisciplinary work, activities to date may be summarised as: (a) the algorithm development including:

and data assimilation technique development:

It is noted that in the Bureau the cutoff time for an operational regional scale model is about one hour.

56. Dr Le Marshall gave examples of tropical cyclone forecasting using 4D VAR and demonstrated marked improvements in track forecasting as a result of increased data from satellites temporally and spatially (the conventional wind data is often negligible in the vicinity of some cyclones). Positioning a cyclone can be done very accurately using scatterometer data. A 48 hour forecast of tropical cyclones shows errors down to 200km in position error which is excellent. Also predictions of pressure and wind speed with radial distance from a cyclone centre are very significantly improved, using high resolution satellite data winds.

57. Dr Le Marshall suggested that these results highlighted the importance of accessing AIRS and MODIS data to enable NWP improvements.

58. For AIRS work activities to date include:

(a) algorithm development:

(b) data assimilation technique development

(c) data acquisition

establishment of AIRS preprocessing capability assistance from JPL NASA to allow data for AIRS to be processed to Level 1B;

(d) calibration/validation activities

(e) assessment of impact on NWP models.

59. Dr Le Marshall described the International TOVS Working Group (ITWG) which began in 1983 and meets every 18 months. It deals with sounder data and processing and has many subgroups and working groups to address various sensors and NWP assimilation issues. Software is freely distributed to scientists. Dr Le Marshall and Dr Guy Rochard are Co-Chairs. The next meeting is in Boulder Colorado in 1999.

60. During discussion in response to a question from Jim Dodge, it emerged from general discussion that NASA MODIS processing software is expected to result in calibration and navigation and basically Level 1B data. All channels will be done. Dr Dodge asked about rapid access to data products and Dr Le Marshall said that in Australia a UNIX computer at each site (say Perth and Hobart) would do some processing - one and a half hours is probably sufficient but for a research study a little longer would be sufficient. According to NASA the standard AIRS processing system is self contained and uses microwave channels for cloud handling. Dr Barton asked about soundings from MODIS and according to Dr Le Marshall MODIS has 3km vertical resolution and computers would probably be needed at local reception facilities. It is also noted that AIRS is colocated with MODIS on EOS PM-1.

Dr Torben Nielsen (School of Ocean and Earth Sciences, University of Hawaii): University of Hawaii DB Activities

61. Dr Nielsen said the University has a 5m antenna system designed and built internally. He passed around several colour hard copy examples of ERS-2 and Landsat-5 images. The ground station is designed for 8 to 8.5 GHz with a G/T > 32.5 at 8.25 GHz. They frame synchronise in software. The ingest board runs up to 800 Mbps but frame synchronisation cannot be done at that rate yet! During development, particular attention was paid to pointing and tracking with an outcome achieved of 0.01 degrees pointing accuracy relative to a geocentric coordinate system (they periodically scan the sun and moon - that was last done around a year ago which is believed to be sufficient). The tracking accuracy is about 0.02 deg and tracking is pure program track.

62. Pointing and tracking is crucial in X-band and errors as small as a few hundredths of a degree can be undesirable or even devastating. The SGP4 model is not accurate enough - in other words there must be considerable care involved in choice/calculation of orbital elements and predicted path of the satellite. NORAD Two Line Elements (TLE) are not really as good as would be liked - the University of Hawaii goes to spacecraft owners who do ranging and their velocity and position data is fed into an accurate orbital determination model. A high order geopotential model is needed that takes account of at least the sun, moon, and solid earth tides. Satellite owners regularly do manoeuvres and do not schedule passes then.

63. The University of Hawaii is currently ready for Landsat-7, EOS AM-1 and EOS PM-1. Software has been developed to do real time decoding and processing of Landsat-7, AM-1 and PM-1. They can do Reed-Solomon decoding at 40 Mbps AM-1 and PM-1 and BCH decoding for Landsat -7 at 400 Mbps. Generally they are open to sharing software with others. They have done a fair amount of the low level work. For Landsat-7 they want a lag time of 5 secs or so after each line.

64. The total cost to date for their system is around US$550k which provides a turnkey system. Dr Nielsen said it was possible to build cheaper systems for AM-1 and PM-1. The technical architecture for Hawaii is similar to that for Dundee and also TERSS but it is noted that they try to do nearly everything in software and avoid hardware. BER seems to be about 1.0E-08.

Liam Gumley (SSEC, University of Wisconsin): EOS Direct Broadcast at the University of Wisconsin-Madison

65. Dr Gumley works with Dr Paul Menzel who is a PI for MODIS. SSEC aims include:

They will also develop an EOS direct broadcast processing package to:

66. About 6 full time people work on MODIS and as a result of Paul Menzel's role in the MODIS Science Team Atmosphere Group they plan to deliver:

67. SSEC is really an applications site and will purchase a DB reception system so that its plans include : 3m antenna and radome, 40ft tower, telemetry processing hardware, UNIX system to track satellite and process data to Level 0. The expected cost is around US$350k and expected installation around 2Q99 (NASA is providing funding assistance). The tower would be located on the roof of the SSEC building beside and above other reception equipment e.g. GOES.

68. For MODIS processing from Level 0 to 1A to 1B SSEC will be using software supplied by Daesoo Han at NASA GSFC and SSEC will simplify/modify the code as necessary. Operational calibration requires 5 mins of data on either side of the scene to be processed which is difficult for DB: possibly 60-70 sec of data at the start of a pass may be sufficient. The DB algorithm from Daesoo does not do terrain correction or scene matching. For implementation the software requires the ECS toolkit and will be AM/PM compliant.

69. SSEC will use existing in-house algorithms from MODIS, POES and GOES to generate regional products from MODIS Level 1B. Products that are provisionally planned may include a cloud mask, clear sky products such as total column water vapour, atmospheric temperature and moisture layer averages, total column ozone, vegetation index, aerosol optical thickness, snow/ice index, SST, LST and emissivity, and so on.

70. SSEC calls the software the International MODIS Processing Package (IMPP) which goes to Level 1B and Level 2 and its heritage is from ITPP (AVHRR/HIRS/MSU) which has been available from SSEC for over 10 years. The target platform is a low cost UNIX workstation (SGI, IBM, Sun, HP) and the timescale for release is about September 1999 around the time of launch of AM-1. Processing might take 30 mins with quicklooks in 5 minutes to a WWW page: data will be in HDF format.

71. The cloud mask product has categories of "confident clear", "probably clear", "uncertain" and "cloudy" (see Ackerman et al in JGR 1998 and also the SSEC WWW site) and does not use a BRDF correction. For visualisation a range of options is available. For outreach there will be some WWW information such as images of Wisconsin, FAQs etc.

72. Dr Simpson noted that in low light conditions in polar regions scattering should be factored in especially for derivation of snow/ice parameters. Mr Beard asked about a specification for the X-band reception system and Dr Gumley said that specifications would end at the point at which the station provided Level 0 data.

Double presentation by Eugene Schaffer (NASA/Goddard Space Flight Centre): EOS AM-1 and its Direct Broadcast; and Paul Hwang (NASA/Goddard Space Flight Centre): EOS PM-1 and its Direct Broadcast

73. Dr Schaffer described the current system at GSFC for DB reception and processing, noting that his work focus is on applications. Smartech is providing a 3m antenna including antenna control s/w. Monarch E s/w includes STAMP s/w which may replace the Smartech antenna control s/w because the latter may not provide sufficient accuracy. Current status of the system is that most hardware has arrived and front end subsystem tests are being undertaken. EOS AM-1 software is being tested on a PC NT platform. The antenna is on the roof of a building at GSFC. Dr Simpson asked about success of using 3m antennas and it was agreed that it is unclear worldwide although in this case the manufacturer claims it will work (certainly for AM-1 and PM-1).

74. Source code for MODIS processing will be provided to the international community free and release 1.0 is now available. The software is currently running on an SGI Origin 200. A MODIS granule is 5 minutes of data. The final version will be available around August 1999. The software developer is Daewoo Han at daesoo.han.1@gsfc.nasa.gov The total cost of the station is around US $250k.

75. Dr Hwang described the expected launch of PM-1 in December 2000. This is a more meteorological satellite with AIRS, AMSU-A and B, HSB, CERES and MODIS (the latter are copies of the AM-1 instruments). The NASA document on PM-1 direct broadcast describes all technical details provided in the talk. The data rate is 15Mbps and includes all instruments. For USA users there may be blackout periods (EPGS) because of interference with the USA Defense communications system. Data rates for PM-1 are around 6200 kbps for MODIS, 1440 kbps for AIRS and lower for the other instruments. All data is Reed-Solomon encoded. PM-1 has the capability to use a pseudo randomiser which would allow relay via TDRSS hence design of ground stations should cater for this although there is no plan to activate this. As an aside it was noted that AM-1 could have interference with deep space network stations and DB might be turned off (there are three such stations at Goldstone, California; near Madrid, Spain; and Tidbinbilla near Canberra, Australia).

76. The AIRS Science Team is responsible for algorithms for AIRS.

77. In USA spectrum management issues are important and certification is being sought. There is a concern that the PM-1 frequency band is very close to DoD communications systems which are mobile. The National Telecommunications and Information Administration cannot at the moment guarantee all users will be free of interference. This is very important so that USA ground station users cannot be guaranteed by US Government that they will get reception free of interference. This type of problem has highlighted the importance of NASA strategy for data sharing and regional reception stations for DB. Mr Beard noted that WRC-97 made the relevant 8.025 - 8.4 GHz EESS band primary worldwide. For DoD the actual ground based mobile communications system is actually a backup system and is used during exercises. It can happen in most countries according to Dr Nielsen who has observed interference in Hawaii. It is possible that there may be an interference problem only with AM-1 when it goes near to Tidbinbilla Deep Space Tracking station in Australia. There may be a dropout for around 5 minutes, hence Jim Dodge was suggesting in Australia that Perth or Alice Springs might be desirable as DB ground stations.

Dr Bjorn Lambrigtsen (JPL, Pasadena): The AIRS Perspective

78. Dr Lambrigtsen mentioned that JPL has a large number of sponsors. At JPL they are not building their own DB ground station but as the instrument provider they will get data from a relevant DAAC within 48 hours plus near RT access to housekeeping data. They are interested in collaboration with DB people for validation and by way of example are working with Dr Le Marshall in the Bureau of Meteorology and Dr Calheiros in Brazil. They wish to talk with others. The AIRS processing software is still evolving. JPL can assist DB users to get copies of source code.

79. AIRS measurement objectives include soundings up to 80 per cent cloud cover up to 1mb in 1800 km swathes. They seek radiosonde accuracy, namely temperature soundings to 1K per 1km and humidity about 10% accuracy in 2 km layers.

80. The AIRS Sounding System actually comprises four instrument components namely AIRS with 2378 channels in IR and four visible plus AMSU-A1 and AMSR-E with 13 and 2 channels respectively. HSB has 4 channels. The data rates for the four instruments are respectively: 1.44 Mbps, 3.2 kbps, 20 kbps and 4.2 kpbs. Quality information is included in the products which range from Level 0 to 1A, 1B, 2 and 3.

81. The algorithm development is continuing via science team members who deliver code to JPL; software engineering team members do subsequent integration. Version 1 will be released around mid 1999 and initial flight software around mid 2000. About 160,000 lines of code are involved.

82. JPL has a recommended approach for AIRS DB users (for Levels 0, 1A and 1 B) which includes: duplicating the EDOS system; duplicating the DAAC system; obtaining s/w from AIRS team; make modifications to allow short orbit segments; or obtain DB version of s/w from others. For Level 2 one should obtain AIRS retrieval algorithms and develop one's own optimised version of that software. Caveats include: must have a copy of static ancillary data (e.g. calibration parameters); must have source code for dynamic ancillary data such as surface pressure; must access navigation data; and must monitor instrument and operational status; etc. In summary he suggests that there would have to be substantial resources put into AIRS processing.

83. Dr Ian Barton said there was one group of NMHS operational users and another group with less real time or operational emphasis or requirements and the two should work closer together. It was noted that at the ITOVS meeting in 1999 two JPL people will be present.

84. JPL reiterated that it would assist DB users to access software and documentation. He invited visits and collaboration with DB users.

Tour: Canberra Deep Space Communication Complex, Tidbinbilla (near Canberra, Australia)

85. A tour of the Canberra Deep Space Communication Complex (DSCC) at Tidbinbilla took place in the afternoon of the second day (Tuesday). At Tidbinbilla, an initial presentation session was headed by Peter Churchill, the Director, and questions followed including discussion on the important issue of radio frequency allocations and interference. It was noted that the certification/licencing process for EOS AM-1 was not yet finished and that there were wide sidelobes for its X-band transmission. It was agreed that the Canberra (Tidbinbilla) DSCC would liaise with JPL on the issue, noting that further detailed technical considerations needed to be undertaken before the issue of interference could be progressed (noting that as a last resort the DB from EOS platforms might have to be turned off during passes which go over Tidbinbilla and the other two DSCCs). It was also noted that the Canberra DSCC has a plan over about a 10 year period to migrate to the Ka band for future missions. General information may be obtained from the NASA JPL Deep Space Home pages at for example http://deepspace.jpl.nasa.gov/dsn/

86. After talks participants were given a tour of facilities including the operations centre and an inspection or first hand view of the large number of antenna systems now on site. The efforts of Dr Miriam Baltuck, the NASA Senior Representative in Australia and Peter Churchill and Richard Jacobson at the Canberra DSCC in making the tour possible were greatly appreciated by participants.

Wednesday, 9 December 1998

Dr Malcolm McIntosh (Chief Executive, CSIRO)

87. Dr McIntosh welcomed participants and noted that Australia has been an enthusiastic user of remote sensing for a long time. CSIRO and Government would just about cease function without access to remotely sensed data. A problem such as salinity highlights the extreme importance of RS data. The notion of using data in real time is also important.

88. His most important point was on free access to data and he cited the cooperation around the world with meteorological data but mentioned that with RS data in general there are threats to free exchange in the global community. A Treaty in the international arena on IP protects genuinely commercial databases but there has been a risk that more general scientific data might also be protected which might limit access greatly. The Europeans suggest scientific data (and its free exchange) is safeguarded but the fear is that it will be copyrighted and therefore have restricted access unless one can prove an exemption, whereas CSIRO believes this is the wrong way around. If the wrong direction is pursued then data sharing will be destroyed. The hope is that in the USA the legislation will include free access to scientific data.

89. Australia contributes only a relatively small fraction of the internationally exchanged free data (around 5%) and gets access to a lot but especially in areas like geology or meteorology our data and contributions are very important. In closing he looked forward to the on-line new data availability from EOS and wished participants well in the conference.

Fritz Hasler (NASA, GSFC): NASA/NOAA/AMS Earth Science Electronic Theater and Dr Palaniappan (University of Missouri): Distributed Spreadsheet Paradigm and Multispectral Land Cover Classifictaion and Visualisation

90. Dr Hasler and Marit Jentoft-Nilson gave a 3D presentation of earth observation data. Dr Hasler noted that the presentation would also be given at the Bureau of Meteorology in Melbourne and then in other centres in Australia. He suggested that in the near future hyper data would be available with 10-200 channels, resolution 2-500 m, and up to 1 min frequency. A GOES animation which he showed had 300 Mb per image. Similarly VIS 5D was used to show a 3D animation closeup of a hurricane with 1min images plus winds - rather spectacular. For the demonstration they used an SGI with 1Gb RAM but in the USA they recently used 6 Gb for more rapid animations and flexibility (cost for a computer for typical displays is around $15-20k).

91. Dr K Palaniappan (University of Missouri, Columbia) has been examining how IT and Remote Sensing can be used to enhance commercial opportunities with near real time applications. He focussed on a spreadsheet approach (the "distributed spreadsheet paradigm") which involves a number of cells in each of which there can be various datasets with animations or whatever. That paradigm is about organisation and integration of data such as those from MODIS. Bill Hibbard at SSEC has been involved. The work began 5 years ago.

92. The aim is to allow interactive manipulations of the data, access to data via distributed networks, plus analysis and intercomparisons. The spreadsheet allows access to 3D datasets. Up to around 10,000 datasets in a 3D cell matrix may be accessed with all interactive tools available. One purpose is to enable access to terabyte MTPE archives. [Editorial comment: This sounds like a next generation hybrid of DDE and DCE with a highly sophisticated GUI that allows comprehensive interactive manipulation of datasets].

93. Research is being undertaken on new products using MODIS DB data. One example involves (see BAMS) multi-satellite winds and multi-satellite stereo imagery. At Missouri they do a land cover automatic classification procedure based potentially on Landsat-7 or MODIS data. Example datasets in the form of 4 panels of Landsat data include a 12 level land classification, colour enhanced over the Missouri area with full zoom down to the 30m resolution of the raw data. Land use management policy may be influenced via access to near real time land classification remotely sensed datasets.

94. According to Dr K Palaniappan the Vslcca algorithm out-performs mpeg for compression - an example was given of a panel of hurricane images compressed in different ways. Dr Simpson said that for Scripps work on multispectral classification they found difficulties. Dr Palaniappan said ground truthing was now taking place for examination of various sites across a range of ecological environments. In response to a question by Richard Smith it was noted that crop forecasting and stress could be monitored better.

95. Liam Gumley reminded participants that VIS5D was developed at SSEC and is available on the Web and free to all users.

Ian Barton (CSIRO): Summary and Plenary discussion

96. Dr Barton provided a summary of talks, discussions and outcomes so far.

97. Dr Jim Dodge outlined AM-1 and PM-1 plans and NASA's plan for four DB stations, software under development and blending MODIS data with other instruments. As an aside it is noted that NASA's policy on data and software access is very clear in that everything is free for research and educational use (non-commercial) but for commercial users charges would apply (discussions would be entered into). Dr Dodge's talk was followed by five talks on applications from Richard Smith, Rob Lees, Ian Barton, Graham Harris and Jim Simpson (who described a MODIS algorithm to detect volcanic ash which was an improvement over current AVHRR techniques).

98. There were five presentations on X-band stations from Peter Bayliss (3.7 m antenna, novel feed), Bertil Gransberg (3.7 m cassegrain, from defence), John Woolner (ACRES - 19 yrs operation and TERSS), Andrew Longhorn (upgrades of ACRES and TERSS and Internet usage) and Torben Nielsen (Hawaii 5m turnkey system).

99. Next there were 3 talks on PM-1. John Le Marshall described the AIRS as a next generation instrument and gave results of impact studies. He described the International TOVS Working Group and supported the desirability of enhanced coordination with the AIRS Team. Paul Hwang and Gene Schaffer described the EOS PM-1 and its 15 Mbps data stream and software to be available. Bjorn Lambrigtsen from JPL described the AIRS sounding system (four instruments) in detail.

100. Liam Gumley gave one of the most important talks especially in relation to goals and access to software for processing which required some software from NASA which would pre-process to Level 1. Dr Barton suggested software from the SSEC was an example of a most likely source for the scientific community.

101. Discussion points/key issues proposed by Dr Barton included:

Dr Barton noted that it was unfortunate that representatives from Eumetsat, NASDA, China, Thailand and a few other areas were unable to come (hence the discussion point on involvement of other agencies).

102. David Jupp chaired ensuing discussions for two hours or so and some of the points raised are described below.

103. Jim Simpson suggested that SSEC could have open architecture which has calibration for MODIS in one module. Ancillary files for MODIS which cover temporal variation should be put on a Web site for DB users. The updating would be similar to the way AVHRR calibration information is placed on the WWW.

104. In response to questions from David Jupp, Jim Dodge said NASA is encouraging and will be as helpful as possible for various regions around the world to establish DB stations.

105. Discussion followed on a broad range of topics. For example an extensive interchange took place on exchange of data and software. Dr Le Marshall noted an example of IP issues in relation to processing new sounding data (e.g. AMSU) which seem to have been ameliorated via having many countries as beta sites. It was agreed that software for exchange should be timely and freely available. Torben Nielsen suggested that a GNU (General) Public Licence (GPL) could be applied at the outset which could enable distribution as freeware which would prevent a commercial entity taking free software, modifying it and selling it back to the originators. David Jupp noted two key issues namely that IP was real and present plus liability had to be recognised. A GPL covers the liability issue since it goes with each piece of s/w. The WWW site is http://gnu.org

106. NASA sees a smaller number of serious users (perhaps larger organisations) maybe 35 around the world with DB sites. Ideally such sites should have products on Internet in one hour and more detailed data by Internet or off-line. NASA and others are encouraged to provide Level 1B data for free: Liam Gumley said SSEC will do that.

107. Liam Gumley asked what the consensus was on what the requirements (e.g hardware and software systems) to receive DB data. Ian Barton suggested that maybe a paper could be published. It was also noted that all technical details to assist in the process for reception were now available for each spacecraft (e.g. AM-1 or PM-1). Paul Hwang will put documents for AM-1 and PM-1 on his Web site. Jim Dodge offered to email anyone who did not yet have the technical documents.

108. Discussion followed on reception equipment. It was acknowledged that probably 3m antennas would be sufficient for MODIS and AIRS but for high data rates such as for Lansat-7 and ERS SAR then a 5m antenna minimum is needed. Gary Quinn questioned the viability of commercial opportunities in reception station manufacture however Jim Dodge said that if a station could be sold for $100k it would be popular. Torben Nielsen said receiving MODIS from AM-1 and PM-1 are two different propositions and the latter might be easier. Also X-band radomes are expensive: one might need honeycomb with few tenths of a db attenuation rather than around 1db from cheap radomes.

109. Discussion took place on the critical adverse impact of radio interference on reception. Peter Bayliss said mobile phones were making S-band reception for them in Scotland difficult in spite of having purchased protection from their local authority.

110. Discussion confirmed the need for availability of software for MODIS for Level 1B processing (Level 1B is sensor sampled (not re-projected) data which is calibrated and navigated). David Griersmith cited the Australian example of the Common AVHRR Processing System (CAPS) which provided free access to software to give calibrated, navigated AVHRR imagery to users.

111. The timing and venue for the next meeting (noting the three events so far in Hawaii, Pasadena and Australia) were discussed and University of Dundee offered to host the next event around mid 2000. Ian Barton agreed to provide a summary of the present meeting to Jim Dodge for publication in EOS Observer as a record of the current event. David Griersmith volunteered an electronic summary of the meeting which he had prepared.

112. David Jupp closed the meeting and thanked participants. Jim Dodge thanked the organising committee and administrative people in particular who made the event such a success (the main people involved were: Ian Barton (CSIRO), David Jupp (CSIRO), Cheryl Gay (CSIRO), Paul Trezise (ACRES), Miriam Baltuck (NASA) and David Griersmith (Bureau of Meteorology)).