Minutes from Informal Workshop on BRDF 28-3-96

List of participants

Interest Areas of Participants

Discussion

Modelling Approaches

The Factor of Scale in BRDF

Outcomes

• Gary Bastin, Vanessa Chewings, Nicholas Coops (CSIRO DWE)
• Graham Behn, Jeremy Wallace, Norm Campbell (CSIRO M&S)
• Alan Strahler (Boston University)
• Megan Lewis; B.Sparrow, S.Spackman (Adelaide)
• B.Lowe, David Lamb, Asoka Edirisinghe, John Louis (Charles Sturt Uni)
• Andrew Wilson (Forestry NZ)
• Vittal Shettigara (DSTO)
• Dr Frank Honey,Greg Kirkpatrick (Specterra);
• David Jupp (CSIRO EOC)
• Elizabeth McDonald (ERIN)

Interest Areas of Participants

• BRDF correction from Aerial photography and videography - best methods for solving the problem
• Exploration of operational use of BRDF correction algorithms
• Correction of BRDF for mosaicing purposes
• Use of attitude and GPS from plane to assist in correction algorithms
• Correction of BRDF for monitoring purposes (both environmental and artificial objects) - temporal changes in signatures due to BRDF
• Development of theoretical view angles.
• Reliability indicators of spectral information
• The use of empirical solutions using statistics
• Shapes of curves from different cover types
• Geometrical Optical models
• Advantages and disadvantages of different modelling approach i.e. empirical, semi empirical and physical
• The factor of scale in BRDF

Discussion

At small scan angles < 20 degrees atmospheric effects are much reduced as shown by the CASI.

Scanning between range of +-20 degrees BRDF effect is fairly gentle. BRDF effects are present but much reduced.
At scan angles of +- 40 degrees BRDF is serious
At scan angles of +- 60 degrees BRDF very serious

The output from image scanning devices are better equiped to deal with BRDF effects due to the line scan geometry. In photography or videograpy the central perspective of the viewer becomes a problem.

Three aspects of BRDF:

1. 'hotspot' occlusion effects - hiding of objects
2. volume BRDF- radiative transfer, structural attributes of objects
3. Glint effects from water targets and specular effects over land

Estimating the proportion of each can be difficult.

For monitoring purposes, only interested in the nature of change of the landcover - not the changes in BRDF or in extant atmospheric effects.

State of the Art BRDF Correction for the MODIS satellite

1. Brightness
2. volume scattering - theory on leaves
3. shape from shadowing

Models
Alternatives to correct for different shapes
MODIS uses four combination and finds the best fit
Practical approach to predicting the mean function.

Shapes are variable: tall and thin VS short and fat - where is it positioned in the landscape - a priori knowledge of shape helps

Modelling Approaches

1. Empirical approach

• Use of image variance and extent
• Derivation of specific functions
• Fitting of data to particular functions using a statistical approach
• Can use empirical models based on variance and extent if you have the shape of the function and then can reduce to 1 or 2 parameters. If the functional form of these effects is known then can start to use models. Extraction of functional form, complicated however by heterogeneity of ground cover.
• Implication of just changing the 'colour' of objects ie, physically changing the data to make it look better.

2. Semi-empirical approach

• Mixture modelling
• Combination of empirically derived functions and atmospheric correction
• Provision of opportunities for the inversion of surface reflectance models.

3. Physical Approach

• Atmospheric Correction to obtain reflectance of objects
• Need to have a very good understanding of the nature of BRDF
• Use of physically based models to obtain the absorption spectra of a leaf
• Derivation of composite functions to explain atmosphere and broad land cover effects. However, difficulties due to local mismatch - high frequency effects will remain.

The Factor of Scale in BRDF

• BRDF is a broad scale averaging effect.
• Atmospheric effects for satellite data lead to a low level of smoothing.
• At high resolution (ie sub 1m pixel) fine scale images show broad trends in BRDF
• Investigation of the patterns of light and shade
• Small pixels provide difficult issues
• Sunlit and shaded components at fine scale.
• In big pixels - statistical effect of how component proportions change. For a particular type of landcover how do shadow effects change as a function of the BRDF.

Outcomes

Multidisciplinary approach to solving the problem.
Funding needed for comprehensive study.
Coordination of functions

Use different datasets from different platforms, different scales, sun angles, spectral characteristics.
Correction VS Normalisation for monitoring purposes.

How do we correct BRDF, would this be a good thing?
What does correction really mean?
Comparison of methods.
Surface property effects to look at monitoring issues.
Looking for best approach as a function of imagery.

Earth Observation Centre projects will be structured by recommended projects. Dr Norm Campbell recommended as CSIRO contact to identify a working group for interested parties in the CSIRO

Investigation of the methods used for BRDF correction in the upcoming MODIS launch. Pater outlining methods can be found at (FTP site address).

Copyright CSIRO 1999 ©