Report on the Third

MODIS Workshop

Commonwealth Bureau of Meteorology

1 May 2003

1. Introduction and Welcome

The Third MODIS Workshop was hosted by the Commonwealth Bureau of Meteorology at its Head Office in Melbourne. Ian Shepherd, acting as chairperson, welcomed the group and confirmed the agenda (Attachment A). The list of participants is provided in Attachment B. A list of acronyms is provided in Attachment C.

The group was then addressed by David Jupp, who reviewed the Action Items from the First MODIS Workshop, hosted by COSSA in Canberra in August 2002. The first Action Item reviewed was the formation of a MODIS Coordination Group and the appointment of leaders for individual product groups. This Action was completed, however Ian Grant noted that Fred Prata is not a member of the Land product group. The remaining Action Items would be addressed by group leaders during the meeting.

David Jupp then provided an overview of the MODIS group’s basis and objectives. He highlighted the many years of work on AVHRR products by the various agencies and the positive outcomes resulting from this collaborative effort. Discussions at the First MODIS Workshop highlighted the work done in the USA on the standardisation of products, the benefits of which are well recognised. The excellent progress made at the training course held in Perth in November 2002 was also acknowledged, and the meeting agreed that it was the second major activity of the group. The purpose of this meeting (now recognised as the Third MODIS Workshop) is to determine the level of standardisation required for the Australian region.

2. (and 3.) Agency Reports

Alex Held - CSIRO Land & Water

Alex Held described the main activities at CSIRO Land & Water (CLW), with the focus of the group being on land applications. Current projects comprise cal/val for MODIS and MERIS, and the implementation of existing MODIS products. Other applications of MODIS data at CLW include near-real time land cover dynamics for sediment models, crop forecasting with MODIS and AVHRR time series data, use of a global flux network for hydrological studies, and Sentinel for bushfire applications.

Peter Turner - CSIRO Atmospheric Research

Work underway at CSIRO Atmospheric Research (CAR) includes the reprojection of MODIS data, taking into consideration the bow-tie effect. The group has produced software which reprojects MODIS data into UTM, with commercial interest in the software already established. When asked whether existing reprojection software should be used, Peter responded that it was preferable to have a unified software environment and noted that CAPS has strong remapping capabilities. New developments in CAPS include functionality for computing BRDF and SMAC.

Edward King – CSIRO EOC

CSIRO EOC are interested in archiving and data consistency issues. Plans include extending the long-term archive of data by building on the existing AVHRR archive. CSIRO groups modelling Australian land cover function see MODIS as a prime tool for monitoring vegetation dynamics.

Gary Quinn – ES&S

ES&S are currently planning to build and install X-band ground stations for the Asian region. The company is interested in developing applications packages (Level 2 products), particularly those which are unique to Australian conditions (eg. bushfires), to be made commercially available as a combined groundstation and software system. ES&S are seeking to collaborate with the scientific community on this. On the development and commercial distribution of MODIS products, John Le Marshall (BoM) noted the importance of building on international experience, in particular providing products in the formats that users’ downstream systems can use immediately. Ian Shepherd suggested that international exposure of Australian Level 2 products may be achieved through ES&S.

Ken Suber – CSIRO Marine Research

MODIS-related work at CSIRO Marine Research includes cal/val, ocean colour, atmospheric correction and SSTs. The data is processed to Level 1b using the NASA code. Ocean products will be produced using NASA code.

Leon Majewski – Curtin University

The Remote Sensing and Satellite Research Group (RS&SRG) at Curtin University has secured a PhD student to work on a MODIS atmosphere product. The RS&SRG are also working on cloud detection algorithms, and cal/val exercises with CSIRO Marine Research, DSTO, and the Leeuwin Centre. A current collaborative project with CSIRO is the implementation of US code for Level 2 ocean products, and validation of ocean products using instruments on board a Singapore cruise. The group also has a strong collaboration with Liam Gumley at the University of Wisconsin, working on improving the IMAPP cloud mask and other products for the Australian region.

David Griersmith, Anthony Rea, John Le Marshall – Commonwealth Bureau of Meteorology

David Griersmith noted the BoM’s involvement in WASTAC, and BoM’s plans to expand the X-band network to Crib Point, Darwin and Casey in the next few years. Regarding MODIS, the BoM is primarily interested in atmosphere products, however there is also much interest in oceanographic and land applications. MODIS applications of interest for both research and operations include volcanic ash, ozone, fog/low cloud, fire hotspots, flood monitoring, pollution/smoke, surface albedo, SSTs, cloud properties, sea ice, high latitude winds, and aerosols. David also encouraged the group to adopt an inclusive, integrated approach which included all disciplines such as atmosphere, oceans and land.

Anthony Rea noted that the Space Based Observations Section has tested IMAPP, and that Ian Grant (previously CAR) had joined the section.

John Le Marshall expressed interest in low cloud and SST products due to their operational status in the BoM. Regarding MODIS Atmospheric Motion Vectors (AMVs), studies from Wisconsin have shown positive impact, and high level AMVs from MODIS are being used operationally in the USA. Also of interest are AIRS products for the BoM’s Numerical Weather Prediction (NWP) models. An impact study using AIRS data will begin when the AIRS processing package becomes available.

Craig Smith – Geoscience Australia (ACRES)

ACRES is acquiring all MODIS day and night passes and processing them to Level 1b within 1-2 hours, with the Lambert Conformal Conic projected images available on the ACRES website 7-8 hours later. On-line products to be made available in the near future include full swath Level 1b HDFLook Quicklook images, Segmented 10º Level 1b gridded product with quicklooks, and archived full swath Level 1b with quicklooks. Proposed Level 2 products include vegetation indices, land cover products, chlorophyll and SSTs. ACRES also plans to increase the bandwidth from Alice Springs to Canberra to 4x the current capability. This issue raised comments from the group, mainly that communications are a concern for many organisations considering the large data volumes of new sensors. Processing the data in situ, before dissemination (the Sentinel model), is a viable option.

Richard Smith, Stefan Maier – SRSS DOLA (also on behalf of WASTAC)

Richard Smith noted that half of his group’s 19 people are involved with MODIS. He remarked that involvement of state agencies in DB is important, since the States have the prime responsibility for land management, but that the State agencies could not be involved without Commonwealth Government support. Richard noted that the WASTAC deed is being negotiated to cover not only reception and archive of MODIS data (as for AVHRR) but also processing. He stated that work is underway on combined Landsat/MODIS products for land applications.

Stefan Maier summarised MODIS activities at DOLA:

Preliminary work with the IMAPP Level 2 cloud mask algorithm has revealed that the algorithm may be unsuitable for Australian conditions.

Gridded Level 1b data and a gridded cloud mask covering Western Australia, and NDVI covering Australia, is delivered daily to the WA Government Department of Conservation and Land Management (CALM). Planned ingest of ACRES data will expand the coverage to the eastern states.

Surface reflectances are produced operationally using only 7 MODIS bands. SMAC code is used to perform the atmospheric correction; initial findings reveal some problems with aerosols. Computed surface reflectances are fed into a BRDF model which is still under development.

Discussions at the end of the session centred around the development of MODIS products. Alex Held (CSIRO EOC) recommended that a generic surface reflectance product for land and ocean be developed. Alex suggested the desirability of a spectrally consistent atmospheric correction that treats MODIS as a hyperspectral instrument, although David Jupp noted that MODIS is essentially 3 instruments and so the bands needed to be treated in groups. Richard Smith suggested that responsibilities for individual products be distributed amongst the agencies; for example the BoM could provide the aerosol depth product.

4. (and 5.) Geophysical (Level 2) products

Table 1 provides a summary of the baseline geophysical (Level 2) products identified by the group, and the organisations committed to undertake development of these products. Issues raised by the group during this discussion are summarised below.

Application	Lead Group	Contributors
Atmosphere
Aerosol (MOD04)	BoM	DOLA/CRC-SI GA CAR
Water Vapour (column & profile)	DOLA BoM	GA
IMAPP Cloud Mask (fine tuning required)	Curtin DOLA GA BoM	GA
Atmospheric Profile	not to be implemented, waiting for AIRS product
Total Column Ozone	DOLA	BoM
Ocean
Ocean Colour	CMR	Curtin
SST	CMR	Curtin
Ocean Primary Production	CMR	Curtin
Land
Surface Reflectance (MOD09) BRDF (MOD43)	BoM (ACRES DOLA CAR)
Land Surface Temp & Emissivity (MOD11)	CLW (Tim McVicar)
Land Cover & Land Cover Change (MOD12)	GA CLW
NDVI (MOD13)	CLW
Thermal Anomalies, Fires & Biomass Burning (MOD14)	Completed (Sentinel)	Validation - DOLA

Table 1. MODIS Level 2 and higher products

(ACRES: Australian Centre for Remote Sensing, BoM: Commonwealth Bureau of Meteorology, CLW: CSIRO Land & Water, CMR: CSIRO Marine Research, CAR: CSIRO Atmospheric Research, DOLA: Department of Land Administration, GA: Geoscience Australia, CRC-SI: Co-operative Research Centre – Spatial Information)

Atmospheric Products

The group acknowledged that there were selected IMAPP algorithms with the capability of processing Level 1b data to Level 2, however the success of these algorithms in Australian conditions was limited. David Jupp pointed out that the group should determine whether the IMAPP structure for Level 2 products will be used to achieve the goals of this group. The other option is to implement IMAPP Level 1 processing then develop Australian Level 2 algorithms. After some discussion, the group agreed that initially, IMAPP Level 2 algorithms would be tested and validated, and if these algorithms fail to produce acceptable results for the Australian region then they will be replaced with algorithms suitable for local conditions.

On specific atmospheric products, Ian Barton recommended that the aerosol product be handled together with the atmospheric correction product. He suggested that one organisation take on this task. John Le Marshall offered to be a contact person for the validation of atmospheric products.

Peter Turner initiated discussion about the cloud mask product. He explained that no validation of this product had been carried out yet, and that “fine tuning” would be required to make the algorithm suitable for use in Australia. Leon Majewski informed the group that Curtin University had secured a PhD student to work on cloud top properties and cloud mask algorithms. John Le Marshall pointed out that the BoM has a scheme to validate satellite cloud clearing.

Ian Grant noted that the BoM was interested in atmospheric profiles of temperature and humidity, but John Le Marshall commented that there are better instruments (for example, AIRS) for these parameters. It was agreed that MODIS atmospheric profiles were a low priority.

John Le Marshall noted that the BoM produces column ozone amounts daily from TOVS and validates it with Dobson measurements.

Richard Smith noted DOLA’s interest in aerosols due to the impact of smoke plumes on air quality.

Ocean Products

Responsibility for ocean products was assigned to Ian Barton. Initially, code from the University of Miami Remote Sensing Group will be used for processing the ocean products. Ian requested close coordination of activity for atmospheric composition and atmospheric correction. In the experience of those developing ocean products, the MODIS aerosol product has problems over southern hemisphere oceans. Ian suggested that investigation of problems with ocean aerosol retrievals would be helpful for land aerosol retrievals in the Australian region.

Richard Smith observed that while land issues hold DOLA’s attention now, there will be a strong demand for ocean colour products in the future: land applications have a twenty-year lead on ocean applications.

Land Products

Alex Held explained that initial testing of the Land Cover and Land Cover Change Product (MOD12) demonstrated incompatibility of the algorithm with Australian conditions: much of Australia was incorrectly classified.

On the NDVI product, the group agreed that a daily NDVI product could be Level 2, and an 8-day NDVI product such as MOD13 would be Level 3 since it depended on compositing choices.

6. Technical Issues

Ian Grant raised some of the technical issues that are common to all MODIS products. He acknowledged that it was wise to build on the existing IMAPP system, and follow IMAPP standards if they exist, so that any code produced by Australia could be instantly adopted by other IMAPP users. He suggested that algorithms should be implemented as they are in the USA DAACs, and that local tuning of algorithms could be performed at a later date.

Ian noted that NASA has mandated the MODIS science teams to release the code used in the DAACs (the “institutional” code) by the end of 2003. He suggested two approaches to developing new code modules: either base them on the institutional code, or start from scratch from the ATBD with the assistance of the appropriate MODIS science team. The group discussed this issue, and agreed that the group responsible for each module would separately decide on whether to build on institutional code or start from scratch.

Ian noted that whereas DAAC products are in HDF format, the IMAPP Level 2 outputs are not, and asked whether a decision needed to be made that outputs of Australian modules would be in HDF format or not. Leon Majewski noted that converters from the IMAPP binary formats to HDF exist, and the meeting felt the question did not need a rigid decision.

Ian Barton questioned whether IMAPP products include Level 2. Peter Turner noted that IMAPP is tidy to Level 1b, and less tidy for Level 2 products. The group agreed that delineation was required between IMAPP Level 1b products and Level 2 products (which are produced by SSEC, NASA/Goddard, University of Miami, and later, Australian organisations).

Other technical issues raised by Ian Grant include:

Many DAAC products are in L2G format. L2G format stores uninterpolated values from a Level 2 (swath grid) product on a remapped (for example longitude-latitude) grid, together with the geometric information needed to remap. This allows Level 3 processes complete flexibility in interpolation of Level 2 data. The meeting felt this was unlikely to be needed in DB processing.

The internal MODAGG product is used in the DAAC to accumulate surface reflectances in spatiotemporal bins for input to the BRDF processing.

The need to check local products against their DAAC counterparts

A software repository for version control, traceability and quality control is required.

Consideration to determining a target time for near real time (NRT) production, eg.1-2 hours?

On the issue of NRT products, Ian Barton explained that specific to ocean colour products, a better navigated product is available after 3 days. He questioned whether the NRT product should be made available in addition to the improved 3-day product. Alex Held claimed that this would depend on communications.

The group then examined the navigation differences between Terra and Aqua. Due to Terra’s on-board GPS, navigation for this instrument was satisfactory, with an accuracy of 50m up to 40 degrees from nadir. Aqua has no on-board GPS, therefore further work is required to accurately navigate the data. DOLA is collaborating with Liam Gumley on this.

Stefan Maier stated that navigation of Aqua data revealed a consistent offset amount. Craig Smith explained that navigation of Aqua data NRT products may always be a problem as predicted ephemeris for AQUA was not available until 1-2 days after image acquisition. If the offset amount is consistent then it should be possible to remove it, however quality control will be required. Craig Smith noted that users need confidence of pixel-to-pixel registration for accurate multi-temporal analysis of MODIS data, and that while ACRES could develop an automatic process for locational accuracy assessment based on chips or coastlines they would need guidance on correcting IMAPP navigation when anomalies are detected. Ian Shepherd offered to collaborate with DOLA and Wisconsin on this issue. Further examination will be carried out at DOLA to ascertain the scale of this problem.

Richard Smith recommended that a NRT system of detecting errors is required, which includes a facility for detecting data that has not met quality criteria. David Jupp agreed, adding that areas with room for improvement include metadata, land/water categories, and DEMs. The group agreed that these issues fall under “fine tuning” of IMAPP.

Leon Majewski raised the issue of ancillary data. Some Level 2 products require ancillary data, which is different for each product. He suggested that a central repository for ancillary data be established, rather than individuals downloading data from US sites. David Jupp recommended that a list of ancillary datasets be compiled, including relevant providers of the data, for either common download or replacement by local datasets. Ian Shepherd stated that ACRES can provide a website listing the location of ancillary data. Tim McVicar pointed out the need for historical ancillary data for reprocessing archive MODIS data.

Ken Suber suggested that a central repository for information on the progress of the various groups was needed.

Action Item 1: Ian Grant and John Le Marshall to compile list of ancillary data available from the BoM.

Action Item 2: Edward King to set up list server for the MODIS Workshop group.

7. Local algorithm for atmospheric correction and BRDF normalisation

David Jupp prefaced this session by proposing that the products which were the highest priorities for implementation were cloud mask, surface reflectance, aerosol, column water vapour, and column ozone.

Ian Grant briefed the group on the current status of the MODIS atmospheric correction and BRDF processing. The MOD43 BRDF/Albedo Product is produced every 16 days at 1-km spatial resolution. Input requirements for the product include:

Cloud mask (MOD35)

Column ozone (MOD07) and water vapour(MOD05) amounts

Aerosol (MOD04)

Atmospheric correction (MOD09)

BRDF (MOD43)

Ian noted that a proposal to NASA to develop a daily BRDF product for DB is being led by the Boston University BRDF group which is responsible for MOD43, with Ian Grant and David Jupp involved as co-investigators, and Shanti Reddy (ACRES) and Stefan Maier (DOLA) also involved. Hence a number of Australian agencies are supporting the proposal.

It was unanimously agreed that the atmospherically-corrected and BRDF-normalised surface reflectance described by Ian Grant would be the baseline geophysical product for the land surface. Furthermore, agencies would work towards generating this product in a collaborative, pre-commercial way. The prerequisite products for this process would similarly be baseline geophysical products for common development.

It was apparent that implementation of the MODIS aerosol module (MOD04) was an early task. However, it was felt some validation of aerosol products from the DAAC for the Australian region should be done first, since those who had worked with ocean products had found there were problems with MODIS aerosols over southern oceans.

Ian Barton suggested that work on atmospheric corrections should cover ocean regions as well as land regions. In this manner, users of both land and water products will benefit.

8. Development of an operational system for DB MODIS data using the above algorithms, to provide consistent national data in the public interest

Ian Shepherd asked the group to determine when the process of nationally consistent data will begin. Richard Smith indicated that DOLA would wait until products are at a final stage before they consider dissemination or sharing of products. Ian Shepherd acknowledged that all organisations will have these issues, and that there will need to be definition between business products, collaborative products, and products which are a service or will benefit the general community. Richard Smith suggested that standalone products (such as FPAR) which require no ongoing calibration be made available to the general public. Ian Shepherd agreed on behalf of ACRES, that public interest value-added products should be shared. Commercial value-added products will need to be handled at a later date. David Jupp acknowledged that the resources required to undertake the work described today may not be instantly or currently available in each organisation.

The group then discussed the release of interim products which may not be perfectly calibrated or validated. Some organisations are willing to release such products as they are of great value to many users (i.e. an imperfect product is better than no product at all), and they expose users to the benefits of NRT products. ACRES however was not willing to release interim products until they are validated and calibrated. Due to the value of interim products for testing purposes, the group agreed to release interim products.

Ian Barton noted that CSIRO Marine Research were archiving MODIS Level 1b granules (ACRES archive the raw data only). He sought opinions from the group on whether this is a worthwhile practice, noting that the data can be made available (on-line) if users request it. Edward King confirmed that CSIRO EOC were willing to take this on.

Ian Shepherd agreed that ACRES should be responsible for the software repository and management of versions. Further discussion with Richard Smith will be required to resolve quality control and web site issues.

9. Preparation of an informal document

Ian Shepherd recommended that in order to bind the group and provide a national DB service, an informal letter describing the group’s purpose be signed by each organisation. David Jupp agreed, suggesting that the letter should include objectives, opportunities, and the agencies supporting the MODIS coordination activity and associated service. Ian Barton suggested that some agencies may have difficulty in having any formal document signed off. John le Marshall proposed that the outcomes of the Workshop be documented as a publicly available report, showing agencies' support for the outcomes. This approach was accepted by all. Ian Shepherd and Richard Smith commented that they were pleased with progress in the nine months since the First MODIS Workshop.

The Third MODIS Workshop Report to be drafted by the BoM is to be circulated, including a copy to Wisconsin.

The Fourth MODIS Workshop is to be held in November 2003 at a venue to be confirmed.

The meeting was closed at 4:00pm.

MODIS MEETING - 1 MAY 2003
List of Attendees

NAME		ORGANISATION

Dr Ian Grant		BoM
Dr Ian Barton		CSIRO Marine Research
Ken Suber		CSIRO Marine Research
Ian Shepherd		Geoscience Australia (ACRES)
Craig Smith		Geoscience Australia (ACRES)
Phillip Tickle		Raytheon
Dr Edward King		CSIRO EOC
Glen Gould		BoM
Alan Marks		CSIRO Land & Water
Dr Tim McVicar		CSIRO Land & Water
Michael Schmidt		CSIRO EOC
Dr Peter Turner		CSIRO Atmospheric Research
Leon Majewski		RS&SRG, Curtin Uni.
Dr Denis O'Brien		CSIRO Atmospheric Research
Gary Quinn		ES&S Chief Engineer
Chris Skelsey		ES&S Senior Programmer
Ian Campbell		ES&S Applications Programmer
Dr David Jupp		CSIRO EOC
Dr Alex Held		CSIRO Land & Water
Richard Smith		DOLA SRSS
Dr Stefan Maier		DOLA SRSS
Dr David Griersmith		BoM
Anthony Rea		BoM
Prof John Le Marshall		BoM