• 10 - 12 June: IWMMM-3
• 14 June: Boston University
• 17 June, NASA Goddard Space Flight Center, Washington, DC
• 18 June, NOAA/NESDIS, Washington, DC

10 - 12 June
Third International Workshop on Multiangular Measurements and Models (IWMMM-3)
Steamboat Springs, Colorado


I attended the Third International Workshop on Multiangular Measurements and
Models (IWMMM-3). When the IWMMM-2 was held in 1999, several new multiangle
satellite sensors were about to become operational; this workshop was the
first time the multiangular community had gathered to review results from
those instruments. There were about ninety participants, including one other
Australian: Michael Hill of the Bureau of Rural Sciences. Two mornings were
devoted to talks summarising results from each of MODIS, MISR, POLDER, and
ATSR-2, and application areas. Work by Fred Prata and Ian Barton with ATSR-2
was noted, as was my hot-spot work with POLDER. The comparison of BRDF
parameters with Australian land cover types by Jenny Lovell and Dean Graetz
was particularly highlighted by Marc Leroy in his presentation on POLDER.

An afternoon was devoted to posters, and I presented a poster on my work in
estimating hot-spot characteristics from POLDER data. Of particular interest
was work by Lewis (University College London) and Alan Strahler's group
towards MODIS products that tracked land cover dynamics (eg burn-scars) better
than the standard analysis in static 16-day windows. Results that could
potentially be applied in Australia include Leblanc's mapping of vegetation
parameters (clumping index, fAPAR) over Canada, and North's mapping aerosol
and vegetation parameters with ATSR.

Bernard Pinty presented the results of the second RAdiation transfer Model
Intercomparison (RAMI-2), which will be placed on the ENAMORS web site. About
a dozen models for radiative transfer in canopies over soil were
compared. Models tended to agree for homogeneous (one-dimensional) cases, but
displayed significant differences on complex (3D) scenes.

Much time was devoted to discussions on where the multiangular community
should focus its efforts. The consensus was that the various instrument teams
had been fully occupied with producing and validating products from the new
sensors, and with distributing products and educating users, and that the
directions mapped out at IWMMM-2 were still appropriate: improve the radiative
transfer models; conduct more field validation campaigns (some done though not
enough, but NASA funding for this is winding down); get the products to the
users and continue to develop ties with potential user communities; and
conduct case studies to demonstrate the value added by multiangular data, even
quantifying it in monetary terms. The best developed class of BRDF
applications is that requiring only empirical modelling of the BRDF: albedo
retrieval and normalisation of wide-swath data. In the extraction of
vegetation structural information from the BRDF, some promising advances have
been made using a purely empirical approach but much more needs to be done;
David Diner and I suggested that the modellers participating in RAMI be
encouraged to systematically use their models to explore how structural
parameters useful in land surface models may be extracted from measured
angular reflectance signatures. Initial efforts to extract aerosol
information, 3-D cloud structure and height-resolved cloud-drift winds from
MISR data have been successful.

The workshop steering committee was keen to hold the IWMMM-4 in Australia in
2004 or 2005, and I, after consulting with David Jupp, tentatively agreed. It
is expected that the IWMMM-4 will place emphasis on the applications of
multiangular remote sensing that are expected to have been demonstrated by
then.


14 June, Boston University

I met with the MODIS BRDF group (Strahler, Schaaf, Gao and Zhang) in the
Department of Geography. I described my tests of linear BRDF models on
Australian land cover types; the BU team would like to compare the albedo
product I am deriving from AVHRR data with their MODIS product.

They have just released a new MODIS global landcover classification; it looked
good in visual comparisons with Australian vegetation maps, and we discussed
performing a detailed digital comparison.

They have also recently developed a vegetation phenology product based on the
MODIS nadir BRDF-adjusted reflectance (NBAR). The 16-day NBAR timeseries has
much lower noise than traditional maximum-value composites. By fitting
analytic temporal profiles, dates of greening and senescence may be derived
that show very consistent spatio-temporal patterns when animated.

The BU team is excited about a potential vegetation structure parameter
derived by Gao from the ratio of the near infrared volume-scattering parameter
and the red-band geometric-shading parameter of the fitted linear BRDF model.

Two workshops will be held at BU in October 2002: the first for users of MODIS
radiative products (albedo, NBAR, fAPAR, LST, etc.); the second on the
validation of MODIS radiative products.


17 June, NASA Goddard Space Flight Center, Washington, DC

I met with Jeff Morisette, who is responsible for validation of MODIS land
products. Uardry is an EOS core validation site. Any data NASA has for the
site is placed on an EOS webpage, and Jeff would like any data we collect at
the site to be forwarded for inclusion. NASA is considering making available
funding to expand the capabilities of existing field sites to serve MODIS
validation objectives, as a more effective use of validation funds than
establishing new sites.

I had an unplanned but intense discussion with David Roy, who is developing
processing to map fire scars with MODIS. The results he showed impressed on me
that a suitable algorithm not only needs to handle BRDF effects and the
transient (~1 week) fire-scar signature, but must be robust to
contamination of the data by undetected cloud and cloud shadow. Fire scar
mapping will find application in carbon accounting, and will benefit from
research in handling BRDF in the face of rapidly changing land cover, and in
using MODIS 500-m resolution data rather than the current 1-km of the MODIS
BRDF products.

I spoke to Eric Vermote about the status of his AVHRR calibrations, and likely
future improvements to the production of MODIS land surface reflectances, most
notably the introduction of a non-Lambertian atmospheric correction. He
expects to release updated calibrations for the AVHRRs on NOAA-14 and NOAA-16
"in a couple of months".

I met with Rachel Pinker at the University of Maryland to discuss the Baseline
Surface Radiation Network meeting that occurred in early June, and our mutual
interests in deriving albedo from AVHRR and geostationary satellites.


18 June, NOAA/NESDIS, Washington, DC

I met with the NOAA team responsible for AVHRR and GOES calibration (Weinreb,
Wu and Cao). They were impressed with how far I had taken the lunar
calibration of the visible-band of GMS. Weinreb agreed with me that the lunar
technique is the most attractive for retrospective intercalibration of all the
geostationary meteorological imagers for long-term measurement of the Earth's
shortwave radiation budget, and was keen to collaborate with me on this. He
could request funding for that activity, but due to the timing of the proposal
cycle, funds could not flow to me until October 2003.

"Fred" Wu has assumed the late Nagaraja Rao's responsibilities for AVHRR
shortwave calibration. The current NOAA-recommended calibration for NOAA-14 is
based on Rao's data on the Libyan desert up to 1998. Wu will use more recent
Libyan desert data (2000, and a few points in 2001) to calibrate NOAA-16, but
updating the NOAA-14 calibration to use these data is a lower priority since
operational users have little interest in NOAA-14 now. A single
cross-calibration of NOAA-16 with MODIS showed channels 1 and 2 had maintained
their prelaunch calibrations, within errors, while channel 3 had changed.

I met with Ken Knapp, who is responsible for the AVHRR Atmosphere Pathfinder
(PATMOS) activity to derive aerosol and surface BRDF and albedo from 20 years
of global AVHRR GAC data. While his surface albedo product has 110-km
resolution and mine for Australia is 5-km, they are based on the same sensor
and many of the processing steps are very similar, so we will compare them.