INTRODUCTION

CSIRO Earth Observation Centre (EOC) supported a delegation of four CSIRO scientists to travel to the Second International Workshop on Multiangular Measurements and Models (IWMMM-2). The workshop was held at the European Commission's Joint Research Centre (JRC) in Ispra, Italy, 15th - 17th September 1999.

Following the conference in Ispra, I had further meetings at the JRC in Ispra, DLR in Germany, Carleton University in Ottawa, Pacific Forestry Centre in Victoria (British Columbia) and at Oregon State University in Oregon. Outcomes from these meetings are summarised below.

SUMMARY OF VISITS
Second International Workshop on Multiangular Measurements and Models (IWMMM-2), Italy (15th - 17th September)

The International Workshop on Multiangular Measurements and Models (IWMMM) is considered one of the premier BRDF workshops in the world, bringing together some of the leading scientists in the field. This workshop is a follow-up to the first IWMMM which was held in Beijing, China, September 13-18, 1996.

IWMMM-2 was hosted by the Space Application Institute (SAI), part of the Joint Research Centre (JRC) of the European Commission, located in Ispra, in north-west Italy. Approximately 100 people attended the three-day meeting (see the group photo below).

The CSIRO participants in IWMMM-2 were Norm Campbell (Mathematical and Information Sciences, Perth), Michael Caccetta (Exploration and Mining, Perth), Ian Grant (Atmospheric Research, Melbourne) and Darius Culvenor (Forestry and Forest Products, Melbourne). Posters were presented by Darius Culvenor and Ian Grant.

A detailed report on the IWMMM-2 meeting can be found on the EOC website. The report was written by Ian Grant with contributions from Michael Caccetta, Norm Campbell and Darius Culvenor. However, given my interest and ongoing research into forest canopy bidirectional reflectance modelling, a description of the discussions and presentations relating to forest canopy reflectance is given below.

Kuusk-Nilson directional multispectral forest reflectance model - Andres Kuusk & Tiit Nilson

Andres Kuusk discussed a new directional multispectral forest vegetation reflectance model. Kuusk described how the transfer of radiation in forest canopies is extremely complex, and that we need new models to help determine the driving factors. These new models should incorporate existing forestry data and should produce optical images. The Kuusk-Nilson model represents crowns as combinations of ellipsoids, cylinders and cones. Single-order scattering in the model is computed by numerical integration. Understorey, background, trunks, bark and foliage are all valid reflecting surfaces within the model.

The model uses the biochemical PROSPECT2 model (Jacquemoud and Baret 1990) for computing leaf angle and transmittance. The PROSPECT2 model works well with crops, however unrealistic biochemical parameters must currently be used in association with the Kuusk-Nilson model to give reasonable values of forest reflectance (probably because of the added effects of crown structure and branch reflectance in forest environments). The Kuusk-Nilson model also uses the 6S atmospheric model (Vermote et al. 1997) for calculating incoming fluxes. The use of the PROSPECT2 and 6S models within the Kuusk-Nilson model reflects an apparent trend in combining separate generic models as components of new integrated models.

The Kuusk-Nilson model is used for simulating forest canopy reflectance versus wavelength (400 nm - 2500 nm) as a function of view zenith angle. Results presented by Kuusk showed a very sharp hotspot - a hotspot that Kuusk agreed was probably too sharp for natural scenes. The hotspot should be narrow for leaf-scale scattering elements, and become progressively broader as modelling occurs at the crown scale. Kuusk's conclusions were that the new model works well, but requires a large number of parameters, which may make inversion difficult. The model is available to anyone.

Radiation transfer Model Intercomparison (RAMI) - Benard Pinty

RAMI was an exercise coordinated by Bernard Pinty and Sig Gerstl for the intercomparison of computer codes that model the spectral and angular radiation fields encountered in remote sensing, in particular, remote sensing of vegetation canopies. Heterogeneous and homogenous type models were distinguished for the purposes of comparison. The authors of the models were instructed to provide BRDF results based on a list of specific scene parameters. Because there was no absolute 'truth', all models were compared blindly with respect to relative performance and consistency. For the heterogeneous models, the greatest deviation in results was only 5%. The RAMI comparisons will continue and will include a comparison of inversion capability.

Building a global vegetation ecology from hotspot data - Sig Gerstl

Sig Gerstl discussed the Trianna satellite - a satellite that will be positioned at the L1 point, which is the gravity-neutral point between the earth and the sun (1.6 million kms away). From this point, the entire earth can be viewed over a solid angle of 0.39°. Sig argued that the angular component of remote sensing is really the critical factor that needs further attention. Using the Trianna satellite, the whole earth is viewed at the hotspot at a spatial resolution of 8 kms. This leads to some interesting possibilities, since the width of the hotspot can be used to infer the size of the scattering elements in the scene (a narrower hotspot is indicative of smaller scattering elements, e.g. leaves).

Other comments / discussion

Some general comments made during discussion time were:

1. Should we be modelling forest canopies at the leaf level ? For example, the PROSPECT model doesn't appear to work too well in coniferous forest types. Should we discard leaf models, or take the difficulties encountered so far as a challenge to develop new models that can adequately represent forests (both leaf and needle types) at the leaf scale. The consensus seemed to be that we should pursue the development of more models.

2. Should models be developed to specifically investigate more spectral ranges ? One suggestion was that if the reflecting canopy surfaces are considered as having lambertian reflectance characteristics, then there is a generic geometric solution that will be applicable regardless of wavelength. Therefore, the only variable should be the spectral properties of the canopy, not its angular response to incident radiation.

3. The hotpot arising from different scattering elements in a scene are superimposed. An important area of future research should be how to separate the relative contribution of different scattering elements to the hotspot.

4. At the close of the meeting there was a wide discussion regarding the future of BRDF research. It was proposed that researchers in the field of BRDF become more applications-oriented. It was considered by many that the future of BRDF research in fact depends on promoting wider applications, such as in ecosystem characterisation and monitoring. In was conceded however that much of the difficulty lies in convincing non-experts of the need to incorporate BRDF principles in many remote sensing projects. It was suggested that a good start will be to include more BRDF publications in non-BRDF journals.

5. A general comment made during the discussion time was that we should not encourage single-angle measurement remote sensing systems.

Proceedings

The proceedings of IWMMM-2 will be published as a special issue of Remote Sensing Reviews, edited by Michel Verstraete and Bernard Pinty. The target date for this publication is August 2000.

Next meeting

The next meeting will most likely be held in 2002.

Environment and Geo-Information Unit (EGEO) at JRC, Italy (20th September)

Following the IWMMM-2 conference, and while at the JRC, I visited the Environment and Geo-Information (EGEO) Unit of the Space Application Institute with Michael Caccetta from CSIRO Exploration and Mining. The EGEO Unit is primarily responsible for research and development of remote sensing for environmental applications. The focus is on the natural environment of Europe and changes in the urban environment. Particular areas of research include the mapping and monitoring of natural forests and grasslands, and also the monitoring and protection of environmentally sensitive regions, such as the Mediterranean basin. The geographic focus of EGEO is southern Europe and North Africa. Meetings with specific people are discussed below:

Helen Preissler

Helen is responsible for collecting, defining and maintaining a database of spectra measured at selected field sites using handheld spectrometers. Of particular use is the database used to manage the recorded spectra. The database, called MedSpec (for Mediterranean Spectra) was written at EGEO and stores a large range of information for each measurement, such as location, material description (including digital photographs), atmospheric conditions, instrument type and user, an estimate of the material purity (%), and the spectra themselves. All spectra are resampled to 5nm to accommodate the different spectral resolution of the variety of spectrometers used. MedSpec can also be used to query and display the spectra. The database is still undergoing development and there are no plans to release the software commercially.

Thomas Kemper

Thomas is investigating the environmental effects of a recent mine accident in Spain. In April 1998 a dam containing toxic mining effluents burst, releasing five million cubic metres of contaminated sludge into the Guadalquivir River system. The sludge eventually reached the Doñana nature reserve 40 kms downstream causing significant environmental damage. Thomas's research involves hyperspectral remote sensing of the river system (using HyMap) in association with field spectral measurements of soil samples using an ASD spectrometer. The soil samples are later analysed for concentrations of heavy metals. Thomas's work will enable mapping the extent and severity of the environmental damage, as well as provide a benchmark for monitoring the success of cleanup operations.

Pam Kennedy

Pam heads up the forest remote sensing section within EGEO. A major task within the section is the development of a Europe-scale forest inventory system incorporating AVHRR data. The objective is to develop a consistent inventory system for mapping forests and monitoring change across the whole continent. This has involved a lot of 'scaling-up' of data collected from numerous sources, including field data, inventory records and other sources of remotely sensed data.

DLR (German Aerospace Centre), Germany (22nd September)

At the German Aerospace Centre in Munich I visited the Institute of Optoelectronics (IOO), along with Michael Caccetta from CSIRO Exploration and Mining. Meetings with Andreas Müller and Kurt Günther are described below.

Andreas Müller

Andreas Müller is the head of the IOO, which is well known for research into geophysical remote sensing and sensor development, particularly in the infrared spectral range. The Digital Airborne Imaging Spectrometer (DAIS) was designed and built at the IOO under European Union funding. DAIS has similar specifications to HyMap but also has a thermal infrared channel which has proven useful for acquiring data in areas of geothermal / volcanic activity. There are approximately two significant DAIS campaigns per year, although there are a total of 19 permanent field sites throughout Europe that are systematically monitored using DAIS.

Kurt Günther and Stefan Maier

Kurt Günther is involved in the development of the Medium Resolution Imaging Spectrometer (MERIS) aboard the European Space Agency's ENVISAT platform scheduled for launch in November 2000. MERIS is a pushbroom instrument with a swath width of 1150 kms (acquired from five adjacent, slightly overlapping cameras) and a spatial resolution of 300 m at nadir. Up to fifteen spectral bands can be selected by ground command, each of which has a programmable width and a programmable location in the 400 nm to 1050 nm spectral range. MERIS was originally designed for oceanic imaging, but has now been extended for terrestrial applications. One particular modification is the calculation of an AVHRR-compatible NDVI product. A benefit of this product is that a time-series dataset for the MERIS data effectively exists in the AVHRR data. This is achieved by developing an algorithm that coverts the higher spectral and spatial resolution data from MERIS into the equivalent AVHRR measurement. Correction factors are then applied to give the directly-comparable NDVI.

The correction factors are based largely on a vegetation reflectance model developed by Stefan Maier at IOO. The model, called SLOP (Stochastic Model for Leaf Optical Properties), is a radiative transfer model similar to PROSPECT in that it is a statistical model that takes leaf biophysical parameters as input. Like most radiative transfer models, SLOP takes into account multiple-scattering effects within the canopy, however an interesting feature is the distinction between different leaf sides (i.e. leaves are not necessarily assumed to have the same spectral properties on each side).

Carleton University, Canada (27th September - 1st October)

Doug King is a Professor in the Department of Geography at Carleton University in Ottawa. Doug is a specialist in high spatial resolution remote sensing with emphasis on airborne video and digital camera systems for data acquisition. The main reason for visiting Doug was to discuss topics of current and future research in the field of high spatial resolution data acquisition and analysis. In particular, research being done in Doug's group is looking at correlations between various local image variance statistics and measures of forest canopy health. There are two main field sites where these investigations are occurring. One is the abandoned Kam Kotia mine site northwest of Timmins, Ontario, where acidic mine tailings are causing premature decline of the local mixed boreal forest. Also, following the severe ice storm in Ontario in 1998, research is being performed to map and monitor the extent and severity of forest canopy structural damage. Some trees were covered in layers of ice 100 mm thick, which resulted in severe canopy damage. Initial work is investigating relationships between the number of broken branches in a crown and LAI / clumping index measurements made with the LICOR 2000 and TRAC instruments.

There is strong collaborative interest between Doug King and CSIRO Forestry and Forest Products to apply some of the image analysis techniques developed at Carleton University to imagery acquired as part of a forest health study recently initiated in central NSW.

Canadian Centre for Remote Sensing (30th September)

During the IWMMM-2 meeting in Ispra, I had met Jing Chen and Sylvain Leblanc, and had arranged to meet them at the Canadian Centre for Remote Sensing in Ottawa. Discussions with Jing and Sylvain concentrated on their 4-Scale bidirectional reflectance model (now officially the 5-Scale model), and the Tracing Radiation and Canopy Architecture (TRAC) instrument.

The well known 4-Scale bidirectional reflectance radiative transfer model developed by Chen and Leblanc (1997) has recently been updated to specifically include branching architecture (in addition to the foliage clumping already included), and is now known as the 5-Scale model. This modification was deemed necessary due to the differences in branch and leaf spatial and angular distributions in different species. In particular the new model allows the branch and leaf orientation to be specifically investigated for its effect on forest canopy bidirectional reflectance. The new model is available from Sylvain Leblanc on request and is detailed in (Leblanc et al. 1999).

While at CCRS, Sylvain Leblanc demonstrated the TRAC instrument designed by Jing Chen. The instrument comprises three LICOR sensors (one pointing down and two pointing up) and a data-logger. TRAC records the frequency of intercepted direct light as the operator walks a transect beneath the forest canopy. Post-processing of this data generates estimates of LAI, foliage clumping and FPAR. This instrument is being trailed by Doug King's group at Carleton University in their investigation of canopy structural damage.

The visit to CCRS also provided a useful opportunity to discuss in more detail the forest canopy simulation model (SORTE) I had presented at IWMMM-2. During the discussions, Jing Chen made some valuable comments on the interpretation of BRDF simulations from the SORTE model and suggested a modification to the way in which the brightness of shaded leaves is computed. The changes suggested by Jing Chen have since been incorporated in the SORTE model.

Pacific Forestry Centre, Canada (4th - 6th October)

At the Pacific Forestry Centre (PFC) in Victoria, British Columbia, I visited Françios Gougeon, Don Leckie and David Goodenough. Outcomes from these meetings are summarised below.

Françios Gougeon

Françios Gougeon developed the first automated tree delineation algorithm in the mid-1980's. Since then, there has been considerable research and development in this field in many countries around the world. The meeting with Françios provided a valuable opportunity to compare results from his valley-following algorithm, and the spatial-clustering Tree Identification and Delineation Algorithm (TIDA) developed in Australia. Of particular interest was the common assumption that northern-hemisphere species, especially conifers, were more suited to automated delineation than Australian forest types. Françios and I compared tree delineation results from a diverse area of conifer forest acquired at 0.32 m spatial resolution from the MEIS sensor. Initial results suggest that there is not a significant difference in the number of canopy objects identified, but there may be a significant difference in shape. There are no plans at this stage to make a quantitative comparison, however a comparison of results from Australian imagery is planned for the near future.

Don Leckie

Don Leckie is overseeing an operational trial of tree delineation technology in B.C. The trial is primarily designed to determine the benefits of using tree delineation for identifying species on an individual tree scale. Statistics generated from tree delineation algorithms typically include both spatial and spectral statistics, e.g. brightness, variance and shape. The emphasis at this stage of the trial is to establish if there is a spectral-only relationship between delineated tree crowns and tree species. Tree delineation is also being trailed for detecting individual trees affected by root disease. These studies are continuing and are expected to be published shortly.

David Goodenough

David Goodenough is currently leading a major undertaking to map Canadian forest cover and forest change using remote sensing techniques. The 'Earth Observation for Sustainable Development of Forests' (EOSD) project is sponsored by the Canadian Space Agency, and is expected to last approximately 10 years. There are four main components to the EOSD project:

1. Forest cover (Mike Wulder);
2. Biomass (Richard Fournier);
3. Change detection (Don Leckie);
4. Automation (Dave Goodenough);

The project is still undergoing design and development to ascertain the best remote sensing practices, and to ensure the project is flexible enough to adapt as remote sensing technology changes over the next decade. As indicated by component (4), EOSD will ultimately involve automated processing using quantitative mapping and change detection algorithms.

At this stage, the primary source or remotely sensed data will be Landsat 7. As David Goodenough explained, one of the beneficial aspects of using Landsat 7 data is the removal of copyright restrictions. This means that all Landsat data purchased as part of EOSD will become publicly available, and can be disseminated via the internet. David sees copyright restrictions as being a major limitation to the use of some new generation satellite data, e.g. the 1 m spatial resolution IKONOS data. This imagery is well suited to automated analysis using Françios Gougeon's tree delineation algorithm, but it is unlikely to be embraced by PFC due to the copyright limitations applied.

Oregon State University, U.S.A. (7th - 12th October)

Nicholas Coops from CSIRO Forestry and Forest Products is temporarily working in the Forest Department at Oregon State University (OSU) in Corvallis. Nicholas is working with Professor Richard Waring on applying 3PG-S, a spatial version of the physiological process model originally developed by Richard Waring and Joe Landsberg. My visit provided a good opportunity for discussing Nicholas's current work in Oregon, and also strengthened the collaborative links between CSIRO and OSU as new research proposals are raised. Nicholas is still involved in numerous projects in CSIRO, and my time at OSU was particularly useful for reviewing the status of these ongoing projects.

REFERENCE LIST

Chen, J. M. and S. G. Leblanc. 1997. A Four-Scale Bidirectional Reflectance Model Based on Canopy Architecture. IEEE Transactions on Geoscience and Remote Sensing 35(5):1316-37.

Jacquemoud, S. and F. Baret. 1990. PROSPECT: a Model of Leaf Optical Properties Spectra. Remote Sensing of Environment 34:75-914 .

Leblanc, S. G., P. Bicheron, J. M. Chen, M. Leroy, and J. Cihlar. 1999. Investigations of Directional Reflectance in Boreal Forests With an Improved Four-Scale Model and Airborne POLDER Data. IEEE Transactions on Geoscience and Remote Sensing 37(3):1396-414.

Vermote, E., D. Tanré, J. L. Deuzé, M. Herman, and J. J. Morcette. 1997. Second Simulation of the Satellite Signal in the Solar Spectrum: An Overview. IEEE Transactions on Geoscience and Remote Sensing 35:675-86.