OPERATIONAL ENVIRONMENTAL SATELLITE DATA APPLICATIONS:
HOW THE EOC CAN HELP ACHIEVE
THEM IN AUSTRALIA.
David L B Jupp
CSIRO EOC Science Leader
April 1997
1. ENVIRONMENTAL SATELLITE DATA APPLICATIONS
Environmental Satellite Data have provided
the promise of monitoring at regional to global scales for more than 15
years. The data time series has been growing since the late 1970's first
as the NOAA TIROS and later as the AVHRR data series. In recent times, the
NOAA 'Pathfinder' data sets have provided a pointer to maturing Standards
of processing and products.
The data provide regional to global scale information
for environmental monitoring, they have high time resolution and can be
acquired and processed for low cost per sq km. The products have stabilised
and have many applications.
Examples of environmental management &
monitoring which are now heavily dependent on AVHRR data include:
The Environmental Satellite data scale is right
for this wide range of applications. Valuable and immediately useable products
are available commercially and the cost frame is acceptable for a well defined
set of applications.
Despite this "undeniable opportunity",
there is still user resistance to the widespread operational use of AVHRR
based products in Australia. I believe the problems stem from lack of consistent
and accepted user experience with data delivery, quality and interpretability.
This is despite years of extensive international and Australian R&D.
2. KEY ISSUES IN ENVIRONMENTAL SATELLITE
DATA APPLICATIONS
The Key Issues can be separated into questions
of:
What is Necessary for effective operational use?
What is Possible for the best quality operational
use?
Over the years during which high quality R&D
has occurred internationally and in Australia, "what is possible"
has been discussed more than "what is necessary". Yet it is "what
is necessary" that is the basis for the market and operational use
of these data. Briefly, we can state:
Despite considerable work, consistent and standardised
data sets have not been produced. Users and suppliers have been forced to
use empirical standardisation to overcome obvious variations in the data
which are not due to changes in the earth surface properties. Such empirical
and data dependent approaches are not satisfactory for truly operational
applications.
In addition, despite research throughout the
world, it seems difficult to establish an acceptable standard set of advanced
scientific algorithms or an accepted "best practice". However,
for end users and for a market to be established and sustained it is essential
that effective standards for both base products and algorithms become agreed
and standardised.
3. CURRENT BARRIERS TO OPERATIONAL MONITORING
What are the most significant Barriers to successful
operational use of these products in Environmental Monitoring, Assessment
& Measurement? What must we do to achieve "what is necessary"?
The barriers are in three primary areas:
1. Data Access & Communications
The large amounts of data collected over 15
years are dispersed into many different holdings, with varying levels of
processing and with variable levels of accessibility.
2. Data & Product Calibration & Performance
A minimal requirement for monitoring is a level
of consistent and well defined calibration. Ideally, if there are no changes
to the signal reaching the sensor, the base products should not change.
Approaches to data calibration are variable with data suppliers often forced
to accept empirical standardisation and there are few attempts to assess
the accuracy of base products. In addition, to compare information over
time at a given location requires accurate geolocation or navigation of
the data. To separate earth surface changes from those of the atmosphere
minimally requires pixel cloudiness to be flagged. These basic requirements
are not being met with consistent and (in many cases) acceptable accuracy
for user's needs.
3. Data & Product Consistency & Standardisation
Many groups, as well as station operators and
commercial groups provide value added products such as Sea Surface Temperature
(SST) and Normalised Difference Vegetation index (NDVI). However, the algorithms
used vary, are often not documented precisely, are sometimes not able to
be re-processed if changes to (say) calibrations become available and are
rarely validated operationally.
4. CSIRO EOC TASKS ADDRESSING THESE ISSUES
Since CSIRO has collected, researched and used
AVHRR data for much of the past 15 years it is important to ask "What
is CSIRO (EOC) doing about this situation?"
The CSIRO Earth Observation Centre (EOC) (COSSA,
1996) has been set up in response to some of these issues as they occur
in the wider field of earth observation. Its catalyst was the Simpson Report
(Simpson et al., 1995) which dealt specifically with data issues
of the kind outlined above. It pointed to a level of generic fragmentation
in the CSIRO effort as a factor in preventing some of these issues being
addressed. With an EOC in place to overcome this fragmentation, what is
being done?
Among EOC Task activity for 1996/97 (see Jupp,
1997 for more detail) are the following:
1. ILS, IMS, IDN & LAS at Canberra
for the Global 1km data system
As part of the USGS global 1 km data project,
AVHRR data from the stations covering our region (Perth, Darwin, Townsville,
[some Melbourne], Hobart, Casey and some from Manila) have been collected
in a data archive at the EOC in Canberra. The data series is complete for
the Australian land mass and covers from 1992 to present. The management
of such data is being demonstrated by its linkages to the world via
The purpose of these demonstrations is to develop
a capacity for the EOC to guarantee data access to researchers when no other
option exists and to undertake comparative data management studies that
will enable the best common approach to data management to be established
for CSIRO and other Australian data managers.
2. AVHRR Calibration Web Site
Base calibration for AVHRR data is an important
issue. The EOC is tackling the problem by setting up a Web site (CalWatch,
http://www.eoc.csiro.au) that records the current best estimates over time
of calibrations for different sensors. If possible, it will record the actual
or likely calibrations used by some of the more common data suppliers so
that the data users can judge the effects for their applications. This work
is being done collaboratively with the Canadian CCRS. Task leader is Ross
Mitchell of DAR who is working with David Parkin of Canberra EOC to maintain
the Web site facility.
3. Station vs Time Series Strategies
for Calibration
A station operator or near real-time data producer
must choose calibrations without current or future knowledge of sensor operation
and performance. Past data can benefit from information from the invariant
sites of the type described below. Choices of the best strategy for near
real-time calibration are being studied to minimise the problems this can
create. This involves current calibration estimates and the history of calibration
change for a specific satellite and instrument. It is especially important
for near real-time uses.
4. Cal/Val sites for product and methods
testing
A range of sites have been or are being instrumented
for calibration monitoring, atmospheric property measurement and validation
of products (referred to generally as "Cal/Val"). These are the
CIGSN network (Hay and Amburla) managed by Fred Prata of DAR and a new desert
site in northern SA (Tinga Tingana) being instrumented by Ross Mitchell
and Denis O'Brien of DAR and Dean Graetz of Canberra EOC with permanent
sun photometers and other instruments. These sites provide a base for quality
testing as well as calibration monitoring and validation missions. The EOC
is investigating how to maximise the consistency and standardisation of
field and validation mission data by forming a shared instrument resource,
establishing the location and status of instruments at Divisional sites
and the examining current status of measurement protocols in well established
missions. This is being done through a specific instrumentation Task by
David Parkin of Canberra EOC and Bob Cechet of DAR.
6. CAPS for Navigation, Calibration &
Cloud Tagging
The algorithms developed at CSIRO Division
of Atmospheric Research and Marine Labs are of high quality and can provide
a common basis for Australian base and advanced products. The best choices
for Navigation, Geolocation & Remapping, for Calibration and for high
quality Cloud Tagging are being realised as an open and extendable software
suite called CAPS (Common AVHRR Processing System, Prata (1996)). CAPS is
a joint effort between CSIRO Mathematical and Information Sciences, Marine
Division and Atmospheric Research and will provide a level of algorithm
description as well as coded demonstration that will enable it to be used
as it is or provide a blue-print for best practice code development.
5. CURRENT OPTIONS FOR ADVANCED HIGH QUALITY
PRODUCTS
The above discussion was about what is necessary
for the Environmental Satellite Data Series to realise its promise among
users. It is only when these primary issues are addressed that the full
view of what is possible can be had. What is possible in monitoring
is to separate real changes in the earth's surface condition from those
only in the data and to derive geophysical parameters from the measurements
the data provide for those changes. Technically, this involves:
What is the CSIRO EOC doing in this area? Current
EOC Tasks addressing these opportunities are (Jupp, 1997):
1. CAPS
CAPS as outlined above is being defined as
a base system in which to document and implement these advanced processing
methods. It is a software environment but it is not an image processing
system. It is being developed as a tool kit that allows users to work in
most image processing environments.
2. AtCor studies using CalVal sites
The Calibration and Validation site network,
as well as providing a means to quality check base products, is the source
of information for atmospheric correction (AtCor) studies. These are being
related to water vapour and aerosol climatology's (best default choices)
and provide test data for comparing methods. Atmospheric correction is a
major factor in retrieving data series that are primarily responding to
surface effects rather than instrument, sensor view or atmospheric effects.
However, it is still a scientific research area and uninformed use of atmospheric
correction can make data series less standardised and consistent.
3. Scene Brightness
The surface effects such as BRDF (Bi-directional
Reflectance Distribution Function) and BEDF (Bi-directional Emission Distribution
Function) can all be looked at as varying scene brightness that does not
indicate changes in the surface type and condition but rather the varying
sun and view geometry of the data. AVHRR and airborne data both share the
problems of widely varying scene brightness. AVHRR in particular must account
for these effects. Canadian researchers have indicated about 30% of the
variation in AVHRR based NDVI can be due to sun and view angle effects.
No wonder deserts can 'green' in the winter! The EOC has a specific Task
which is tackling BRDF and varying atmospheric effects at a range of scales
and which aims to define a BRDF 'typology' for Australian landscapes that
can be used to correct Environmental Satellite Data for these effects. The
potential to effectively combine calibration, atmospheric correction and
scene brightness corrections to create a standard base data set is one of
the major goals of the EOC.
4. Best Practice Demonstrations
The end-to-end combination of data management,
base processing, advanced information extraction and data series production
is being demonstrated in time series of AVHRR, GMS and Landsat data. In
the case of AVHRR data, a collation of the current CSIRO best practice is
being supported in the Data Consistency & Standardisation Thread. This,
and similar demonstrations, aim to provide demonstrations of what is possible
based on best practice in what is necessary and best current choices
of CSIRO advanced algorithms.
6. BRINGING IT TOGETHER
The EOC and its "Earth_Observing_Crew"
in CSIRO can take actions but they will not develop and expand the market
for operational use of Environmental Satellite Data unless the actions are
coordinated with user needs and supplier needs. We need to consider the
options for a nationally consistent effort on product consistency and standardisation
in (minimally) AVHRR data. This will be a framework that can be applied
to other satellite data but in the immediate future it has a special need
for action.
The main groups that need to discuss this issue
are CSIRO (through the EOC), the Bureau of Meteorology, QDPI, WASTAC, AIMS,
ACRES and representatives of the small but significant user community. There
have already been important meetings and discussions concerning these issues
(McVicar, 1996; Jupp, 1996).
Principles we must keep in the forefront are
that by coming to agreed approaches and standards at the base product and
supplier level we are investing in product quality to build the whole market.
It is against everybody's interest for the current situation of product
inconsistency and lack of standardisation to continue. It is in everyone's
interest for the base product issues to become non-issues. The real markets
and benefits are in the value added products. At present, the impact of
these is being compromised by the varying quality and standards among the
data archives and supplies across Australia and the differences between
these and the data available via the internet from the US.
The CSIRO EOC is investing in Data Consistency
& Standardisation because of its underpinning effects on all products.
However, the EOC is science focussed and its primary aims are to develop
the improvements in the "What is possible" area. If a concerted
effort can be brought about, it would be possible for some of the current
EOC Tasks to move from a research time frame to an operational time frame.
These include the Calibration studies, CAPS and the [data series]. This
move will need support of various kinds including collaboration of staff
from the different organisations and some specific funds to focus Tasks
and bring their outcomes into the required time line of the Australian product
suppliers and users.
Some steps to starting these discussions have
occurred but they are not enough. It is time for the parties, with the pressure
and support of the user community to nail the issues and see environmental
Satellite Data realise its "Undeniable Opportunity".
REFERENCES
[EOC Web Site: http://www.eoc.csiro.au/eoc/eoc.htm]
COSSA (1996). CSIRO Earth Observation Centre
Strategic Science Plan. Vision, Mission and Strategies 1996-2000. (EOC Web
Site)
Jupp, D.L.B. (1996). Land AVHRR Data Users
Group Workshop - Towards Product Standards, ARSC8, Canberra. (EOC Web Site)
Jupp, D.L.B. (1997). EOC primary Tasks and
forward looking Working Groups for 1996/97. (EOC Web Site).
McVicar, T. (1996). Land AVHRR Workshop, ARSC8,
Canberra, Monday 25th March 1996. Meeting Report. (EOC Web Site).
Prata, A.J. (1996). Working Group on Australian
AVHRR Science. Final Report: Recommendations and Proposal for a Common AVHRR-data
Processing Scheme - CAPS. Revised April 1996.
Simpson, J.J., Barton, I.J., Kingwell, J., Neal, A.B. and Wallace, J. (1995). Review of CSIRO Multi-Divisional Program on satellite data acquisition and utilisation (MDP-19)/ Analysis of CSIRO Earth Observation needs., COSSA Publication 34, May 1995.