Land Vegetation and Cryosphere: ADEOS Vegetation Biology, ADEOS-Cryosphere & Hydrology
GLI-Land, GLI Vegetation and GLE Cryosphere Group Joint Workshop.
-ADEOS-Cryosphere & Hydrology and GLI-Cryosphere Group Joint Session-

Morning Session Afternoon Session

Morning Session

Room H December 7 1999
  1. Opening Remarks:Dr. F. Nishio
  2. Validation Program and some results for ADEOS/ADEOS -II field campaign (F. Nishio)
  3. Observation of Coastal Zone of Antarctica using ADEOS / AVNIR and microwave sensors (Dr. H. Enomoto and Dr. F. Nishio)
  4. Relationship between spectral reflectance and evaporation from vegetated surface (A. Kondo and A. Higuchi)
  5. Hydrological applications of ADEOS / AVNIR images (K. Takara and Y. Tachikawa)
  6. An anticipation for the ADEOS- II mission over Sea Ice, in view of ADEOS / NSCAT results and the first SeaWind/QuikSCAT data (R. Ezraty and A. Cavani)
  7. Bidirectional reflection properties of flat snow surface at AEOS-II/GLI channels (T. Aoki)
  8. Cloud mask over Snow/Ice surface and Snow Grain Size and Impurity Retrieval using ADEOS-II/GLI measurements (K. Stamnes)
  9. Determination of Snow Grain size and Soot Concentration based on Neural Network (T.Oishi)

Land Vegetation and Cryosphere: ADEOS Vegetation Biology, ADEOS-Cryosphere & Hydrology
GLI-Land, GLI Vegetation and GLE Cryosphere Group Joint Workshop.

-ADEOS-Cryosphere & Hydrology and GLI-Cryosphere Group Joint Session-

Morning Session

Room H December 7 1999

1. Opening Remarks:Dr. F. Nishio

Explained contents of presentation - each 20 minutes - and structure of Q&A session. Introduced each speaker.

Back to index

2. Validation Program and some results for ADEOS/ADEOS-II field campaign (F. Nishio)

First news about plenary session of 6.12.99, and that delay of ADEOS may be up to one year.
Dr. Nishio (Hokkaido University of Education) is working on ADEOS studies, and algorithm for sea ice in both Arctic regions. Professor Nishio's presentation covered the following topics:
  • Sea Ice and Climate in Okhotsk Sea
  • Sea Ice Concentration - key mission of ADEOS-II, using AMSR. Will use bootstrap algorithm.
  • Data flow from brightness temperatures from AMSR, plus temperature of sea ice and depth of snow on ice.
  • Study of distribution of snow on sea ice (AMR and AMSS data) and presentation of results so far.
  • Model for the spectral albedo of pure snow
  • Snow depth distribution algorithm - various data (e.g. snow grain size)
  • AMR, AMSS actual results for snow depth and grain size distribution
  • Wish to continue these studies with ADEOS-II using GLI
Q: (Dr. James Simpson, Scripps Oceanography U.C.S. San Diego) Is retrieval for parameters standard satellite retrieval or model?
A: (Dr. Nishio) Algorithm using GLI data, also including other data, plus models (tables)
Confirmed by BRDF correction for the snow being carried out.

Back to index

3. Observation of Coastal Zone of Antarctica using ADEOS / AVNIR and microwave sensors (Dr. H. Enomoto and Dr. F. Nishio)

Dr. Enomoto's (Kitami Institute of Technology) presentation covered the following:
  • Antarctic research for Japanese Group - East Antarctica
  • Antarctic research station in Lutzow Holm Bay
  • fast ice area break up studies
  • data from ADEOS / AVNIR and Nimbus-7
  • Speed of flow and fractation of Shirase (over interval of a few years)
  • NDJF Mean Temperature (Showa Station)
  • Surface conditions - breaking up/melting signs
  • distribution of puddles on fast sea ice and open water
  • hydrological process on coastal slope - melting features (using ADEOS / AVNIR image)
  • Meteorological data for 2 observations (31 Jan and 4 Feb) Warmer weather showed rapid changes in
  • retreat/break up of fast ice
  • Data points from Luzthow Holm Bay, and microwave data for brightness temperature gradient ratio and
  • meteorological data
  • 10 year melting index for Showa Station (1987 - 1998)
  • Cracks near the edge of glacier - may indicate end point of break-up
In conclusion, concluded that further observation was necessary, using present and future data for break up of fast ice. See presentation slide.
Q: (Dr. Simpson): Microwave data can give false signature for water on ice, light break up - how is this compensated for?
A: Compared meteorological data and field data (i.e. local observation also used)
Q: (Dr. Nishio) : What is difference between gradient ratio for melting index and gradient ratio of sea ice? How about polarization?
A: 0 (zero) is whole ice cover, and positive gradient measure means water is open. - (minus) 37 GHz was used. Similar signs can be observed when ice is thick - but starting point could not be used in 1998.

Back to index

4. Relationship between spectral reflectance and evaporation from vegetated surface (A. Kondo and A. Higuchi)

Chiba University - group of hydrologists. Presentation covered the above topic.
  • measuring surface state variables and evapotranspiration - empirical relationship between measured flux
  • and satellite data. Intend to use ADEOS / AVNIR.
  • study area of Tsukuba Science City and flux stations (routine data since 1970s)
  • satellite data from LANDSAT/TM and SPOT/XS data (NASA)
  • Seasonal changes in Et and NDVI - relationship to evapotranspiration
  • Red and Infrared Brightness - relation to daily evapotranspiration
  • Upward short wave radiation/PAR and NIR components
  • Seasonal changes in Albedo, IE/Qn and H/Qn
  • Seasonal Variation of Spectral brightness (Red and NIR) and NDVI
  • Volumetric water content of surface soil layer
  • Relationship of NDVI with IE and IE/Qn
  • Relationship of IE with NDVI and red band brightness (in growing season)
In conclusion, good relationship and evapotranspiration rate, spectral brightness in VNIR wavelength and NDVI. More research is needed in relationships for different vegetation.
  • PGLIERC 1999 - preliminary global imager experiment
  • North China Plain Project
Q: (Dr. Andrew Tronin, EORC, JAXA) I believe that NDVI is not the best parameter for forest, only grassland/paddy field. Different parameters are needed (e.g. different short wave bands)
A: (Dr. Kondo) I agree, correlation obtained from forest data is not so good. Also, the forest used is small and mixed, as it is difficult to obtain large forest areas in Japan - this may be the main reason for poor data.

Back to index

5. Hydrological applications of ADEOS / AVNIR images (K. Takara and Y. Tachikawa)

Dr. Takara of Kyoto University, (Disaster Prevention Research Institute)
Started with outline of work carried out by Disaster Prevention Research Institute and explained that he would present hydrological applications of ADEOS / AVNIR images.
Covered:
  • GEWEX - GAME/HUBEX
  • Hydrological model structure (using AVNIR data)
  • image examples from ADEOS / AVNIR
  • Image processing system images
  • Channel network dataset for Shiguang River and procedure for making channel network dataset
  • Model structure of routing model
  • Structure of subsystem models and structure of total system models
  • Results from Shiguang River basin (2 locations)
  • Land cover classification application (with degraded resolution images)
  • Image class cross plot graphs (unsupervised classification -> assigned land cover class)
  • Results of land cover classification (13 classes, best for the region)
  • No of pixels for land cover classification (16-m and 2000-m)
In conclusion - hydrological models are useful for large basins. Land cover classification accuracy for run-off needs to be verified. ADEOS-II /GLI images have more bands, so look forward to data.
Q: (Dr. Wang Zhihua, Aero Geophysical and Remote Sensing Control, Ministry of Land and Resources, China) Do you use topographical maps, is geometrical correction done before making measurements?
A: (Dr. Takara) Use processing system with ground control points (from topographical maps)
Q: (Dr. Wang Zhihua) How many ground control points?
A: (Dr. Y. Tachikawa) - not ground control points in fact. Used four corners of location of image to rectify the image.
A: (Dr. Takara) For this analysis it doesn't matter - compared results of classification on same image, so geometric location is not so important for this analysis. This is an only unsupervised classification result, if confirmed on ground, then geometrical correction will be necessary.
Q: (Dr. Wing Zhihua) I understand, but for land cover classification and formation there may still be problems.

Back to index

6. An anticipation for the ADEOS II mission over Sea Ice, in view of ADEOS / NSCAT results and the first SeaWind/QuikSCAT data (R. Ezraty and A. Cavani)

Presenters were unable to attend, so presentation made by Prof. Nishio on their behalf. Professor Nishio apologized that the topic was reverting to the imaging of sea ice.
  • Important for describing global changes, and this research is very much applied (previously SSMI
  • microwave data), and now using ADEOS, AMSR and SeaWinds data.
  • Covered:
  • NSCAT and SSMI data over 20 years
  • ODDEN NSCAT data
  • Empirical backscatter model over arctic sea ice
  • QuikSCAT simulation from NSCAT data
  • QuikSCAT backscatter maps over arctic ocean, normalized standard deviation maps of arctic region, pixel
  • normalized standard deviation maps
  • QuikSCAT sea ice/open water discrimination images
  • backscatter and normalized standard deviation maps for Antarctic region, pixel values over Antarctic Ocean
In conclusion: SeaWinds very well suited for discriminating isotropic surfaces such as sea ice. (e.g. discrimination of sea ice/water areas). Professor Nishio able to make copies of this presentation as required for further information.

Dr. Nishio announced coffee break until 11 am. (20 minutes)
COFFEE BREAK

Back to index

7. Bidirectional reflection properties of flat snow surface at ADEOS-II/GLI channels (T. Aoki)

Dr. Aoki presented his studies on result of validation, as follows.
  • Objectives
  • Methodology
  • Site of field experiment: Kitami city, Hokkaido
  • Photographs of experiment instruments
  • Results of snow depth work; snow type, snow grain size, density, impurity etc.
  • Snow model
  • Spectral albedo; correlation between snow surface albedo and wavelength
  • Calculated results of BRDF
  • Scattering angle model and phase function
  • Observation results of BRDF; rainbows are not observed on observation results.
  • Conclusion on Snow BRDF Observation
  • Effects of aerosols; rural aerosol model
  • Rayleigh Atmosphere and Atmosphere with rural aerosols
  • BRDF rural - BRDF Rayleigh
  • GLI viewing angle
  • Conclusion for the Effect of Aerosols on Snow BRDF
Q: (Dr. Nishio) What is reason for difference between theoretical and observation results for spectral albedo?
A: (Dr. Aoki) At the present time there is no explanation for the difference in this region (indicating graph). One possible reason could be problems with refractive index data or another possible reason could be a problem in the model. We are not able to explain the difference at this stage. For example, the snow surface albedo shows the same difference in data.
Q:(Dr. Nishio) Sometimes, there is consistent data.
A: (Dr. Aoki) Yes, some areas are consistent. We have checked many snow models, but we can't get agreement in this region (indicating graph)


Back to index

8. Cloud mask over Snow/Ice surface and Snow Grain Size and Impurity Retrieval using ADEOS-II/GLI measurements (K. Stamnes)

Dr. Stamnes presented his studies on cloud mask over high-albedo surfaces.

  • Outline
  • Objectives
  • Why difficult; insufficient albedo and thermal contrast
  • To arrive at a better confidence level, combine a set of tests
  • Surface classification
  • Snow/sea-ice discrimination
  • Flow chart for processing one pixel
  • Application cloud mask algorithm to satellite data
  • Cloudy / clear discriminator image
  • Summary of cloud mask algorithm
  • Snow grain size and impurity retrieval
  • Penetration depth of radiation
  • Application of snow grain size and impurity retrieval algorithm to satellite and simulated data
  • Selection of aerosol model
  • Flow chart for aerosol model discrimination
  • Figure 3.4: the structure of the simulated GLI image
  • Summary of snow grain size and impurity retrieval
Q: (Dr. Aoki) You say that snow size may change according to wavelength. If you think inversely, can you estimate profile of snow grain size by using different channels?
A: (Dr. Stamnes) In principle it can be done, and a 2/3 layer model is possible, but the question is how easy that may be. We have tried the same concept for chlorophyll retrieval from the ocean, but an operational algorithm would take some work.
Q: (Dr. Tronin) Is retrieval valid for snow on land or on sea ice?
A: (Dr.Stamnes) Both. We only wish to distinguish between snow over sea ice and bare sea ice.
Q: (Dr. Simpson) Do you eliminate weeds in sea ice?
A: (Dr. Stamnes) We need to distinguish between open water and snow covered sea ice, so we check that first. The pixel can be a mixture of everything.
Q: (Dr. Simpson) I agree that a mixed signal is possible. I just wanted to clarify. Thank you.
Q: (Dr. Tronin) When snow is wet during melting, is snow retrieval possible?
A: (Dr. Stamnes) Firstly, there is a great increase in grain size. There will be a point where it is so wet that it doesn't look like snow any more to the wavelengths. Not sure how that would influence reflectance. There is no water index in the model, so that needs to be investigated.

Back to index

9. Determination of Snow Grain size and Soot Concentration based on Neural Network (T.Oishi)

Contributors: Tomohiko Oishi (Tokai Univ.)
Teruo Aoki (MRI)
Akihiko Tanaka (Tokai Uni)
  • Main purpose
  • Approach
  • BRDF wavelength patterns
  • General Procedure of Neural Network; Radiative Transfer Computation Doubling and Adding Method
  • (Aoki's model)
  • Geometry
  • Conditions of Computation
  • Two Step Neural Network Mode (First and Second Step) and Equations
  • Single Step Neural Network Mode
  • Results of Two Step Mode; snow grain size and soot concentration: estimate and true values
  • Results of Single Step Mode (as above)
  • Conclusions
  • Problems
Q: (Dr. Simpson) Topographical structure can be extremely sensitive; sensitivity tests are necessary. Have you carried out sensitivity tests?
A: Yes, it is necessary. From results, we can see that the Single Mode seems to be better. We have tried it many times and results seem to be OK but sensitivity test must be carried out.
Q: (Dr. Simpson) Did you look at details of output probability density function?
A: (Dr. Aoki) That is difficult to answer.
Q: (Dr. Simpson) OK we can talk about it later.
Q: (Dr. Nishio) Do you have validation criteria for this algorithm?
A: We have carried out experiments since 1995/6 and have experienced how to collect data. We need to continue to get more reliable data and methods.
Dr. Nishio concluded the morning session with a reminder to participants that JAXA is calling for manuscripts and images to help with ADEOS and rocket launch problems. Please submit any papers by March 31st 2000.

Afternoon Session

Back to index

GLI-Land, GLI Vegetation and GLI Cryosphere Group Joint Workshop.
-ADEOS-Cryosphere & Hydrology and GLI-Cryosphere Group Joint Session-

Room H, December 7 1999

13:00 - 17:20

  1. Validation Site for ADEOS and ADEOS-II (Y. Honda)
  2. Semi-automatic land cover classification using ADEOS / AVNIR multi spectral data(N.D. Duong) (Changed Title)
  3. ADEOS / OCTS Mosaic data for vegetation monitoring (K. Kajiwara, Chiba University)
  4. Progress and Problems in NPP Estimation (changed title from Forest monitoring using ADEOS and ADEOS-II) (Y. Awaya)
  5. Agricultural Monitoring with Oblique Viewing Remote Sensing (G Saito and T. Murakami)
  6. Innovative land surface parameters from ADEOS-II (Changed title, presented by Marc Leroy)
  7. Precise Geographical Position (Dr. T. Hashimoto)
  8. Global Land Surface Radiation Budget Products from the ADEOS-II/GLI (Dr. Fred Prata)
  9. Development of Topographic Correction Algorithm for ADEOS-II GLI Data, for Use in Forest Carbon Cycling and Primary Production Studies (C. Trotter)
  10. Development of Spectral Indices Optimized for the GLI Instrument (M. M. Verstraete)
  11. Development of New Vegetation Indices and Algorithms for Detecting Vegetation Change (N Fujiwara)
  12. Land Cover Classification by GLI 250m - Data (R. Tateishi)

Morning Session

Land Vegetation and Cryosphere: ADEOS Vegetation Biology, ADEOS-
Cryosphere & Hydrology
GLI-Land, GLI Vegetation and GLI Cryosphere Group Joint Workshop.
-ADEOS-Cryosphere & Hydrology and GLI-Cryosphere Group Joint Session-

Room H, December 7 1999

13:00 - 17:20

Dr. Honda opened the afternoon session with an announcement that more ADEOS vegetation and GLI vegetation joint sessions will be held on morning of 8.12.99, a free discussion session to talk about harmonisation with GCOM. Also, there will be an evening PI meeting on the same day.

1. Validation Site for ADEOS and ADEOS-II (Y. Honda)

Dr. Honda presented his studies on the topic above.
  • City Light
  • Vegetation map made by IGBP
  • Validation site
  • Explanation about sites
  • Mobile system
  • Distribution of Measurement points
  • BRDF Measurement system
  • Grass picking
  • Measurement by 3D Scanning system; 3D laser scanner
  • Monitoring using RCH(Radio-controlled helicopter); a helicopter which has various measurement
  • instruments.
  • Programming flight; flight course
  • Reflectance change by sensor zenith angle
  • Digital Camera Images
  • DEM (Digital Elevation Model)
  • Other application; can be applied for forest and wetland
  • Conclusion
Q: (Dr. Michel Verstraete, Space Applications Institute, Italy) How long does it take to make the measurement?
A: (Dr. Honda) It takes less than one second to make the measurement, but the helicopter has to be stable for 8 seconds.
Q: (Dr. Verstraete) When taking measurements, is the helicopter moving or the sensor?
A: (Dr. Honda) The sensor is moving

Back to index

2. Semi-automatic land cover classification using ADEOS / AVNIR multi spectral data(N.D. Duong) (Changed Title)

Dr.Duong's presentation covered the following:
  • Introduction
  • Algorithm Development; improvement of GASC algorithm and 8 image invariance for classification
  • Program GASC_99.F90
  • Computation flow chart
  • Structure of CTRFILE
  • Land Cover Definition by Image Invariant
  • New Model for Color Composite of GLI Data
  • Simulation of new color composite model compared with true color composite and standard false color composite
  • Classification Results
  • Monthly Land Cover Monitoring
  • Rice cultivation monitoring
  • Secondary crop monitoring
  • Conclusion and discussion
Q: (Dr. Honda) For your time series data, how long an interval do you use for your dataset?
A: (Dr. Duong) A suitable interval is difficult to achieve in tropical areas, with data being hard to obtain during the rainy season. At the moment data interval is dependent on this problem, and timing is not always applicable to the agricultural situation. Ideally interval should be once a month, but this depends on GLI status, and we hope to review and make suggestions after the results of this year’s research.

Back to index

3. ADEOS / OCTS Mosaic data for vegetation monitoring (Dr. K. Kajiwara ,Chiba University

Q: (Dr. Simpson) I have a couple of comments to make. Using maximum composite MDI for cloud detection can be problematical, especially in tropical region, where there is no temperature contrast. Could you also please comment on the following: using single ground control point file doesn't produce coregistration of images, with different passes (and therefore different tracks) of the satellite - is there any adjustment carried out to ensure the same geographical location?
A: (Dr. Kajiwara) Do you mean an overall large image?
Q: (Dr. Simpson) We can talk about it later.
Q: (Dr. Wang Zhihua) Which software are you using for mosaic images?
A: (Dr. Kajiwara) They are handmade and they are not commercial. It has mostly been developed by Dr. Hashimoto.

Back to index

4. Progress and Problems in NPP Estimation (changed title from Forest monitoring using ADEOS and ADEOS-II) (Dr. Y. Awaya)
  • Test mapping of NPP using satellite data
  • Relationship between NPP and sum of NDVI
  • NDVI of AVHRR data
  • Seasonal changes of NDVI
  • Methodology for NPP estimation
  • flow of NPP estimation in forest
  • Location map of the meteorological station
  • NPP distribution estimation
  • Sum of NDVI by solar radiation
  • NPP estimation in forest
  • Seasonal changes of SPAD values
  • Relationship of DN of chlorophyll meter vs. NDVI
  • Land surface resolution
  • Comparison of spatial resolution
  • Seasonal vegetation changes
  • Relationship between reflectance data and stand parameters
  • Correlation coefficients between DN and stem volume
  • Seasonal spectral changes
  • Results of biomass estimation
  • Problems in NPP estimation (conclusion/validation data)
Q: (Dr. Verstraete) I would like to ask how you do your sums, as NDVI is a different sum each day?
A: (Dr. Awaya) We use the data taken monthly as the basis of our calculation.

Back to index

5. Agricultural Monitoring with Oblique Viewing Remote Sensing (G Saito and T. Murakami, National Institute of Agro-Environmental Sciences)
  • Backgrounds
  • Objectives; using ADEOS/AVNIR and SPOT/HRV
  • Data and Study area: Saga Plains (in Kyushu Island): the largest granary in Kyushu Island
  • Crop Calendar; 6 cropping systems: Rice+, Rice, Soybean+, Soybean, Lotus and Rush+
  • NDVI profiles for six cropping systems
  • Separability Analysis
  • Optimal Scene Selection; NDVI color composite image
  • NDVI Color Composite Images; Rice and Soybean
  • Area Estimation of Barley and Wheat planting fields; Area estimation with VSW index
  • Flow of Area Estimation
  • VSW index; originally developed by Yamagata et al. in 1997
  • VSW index image; comparing ADEOS/AVNIR image and VSW index image
  • Relationships between V index and W index for barley and wheat planting fields
  • Pixel sampling method for barley and wheat planting fields
  • From VSW Image to sampled image
  • Overall Accuracy of Area Estimation
  • Conclusions (1) and (2)
Q: (Dr. Ranga Myneni, Boston University) What is the significance of the relationship between the v and w indices?
A: (Dr. Murakami) As mentioned, the shadow/plant canopy mixture is related to (affects) the v and w index relationship.

Back to index

6. Innovative land surface parameters from ADEOS-II (Marc Leroy)

  • Objectives: Biophysical parameter retrieval and snow water equivalence
  • Approach
  • Inversion protocol
  • Algorithm applied to the data (leaf area index)
  • Temporal analysis of LAI (leaf area index)
  • Surface BRDF
  • Methodology for POLDER, GLI, SeaWinds and AMSR (snow water equivalence)
  • Conclusion - biophysical parameter retrieval (methodology)
Q: (Dr. Simpson) I think that the POLDER pixel is large compared to GLI pixel - could you comment on the issue of cloud contamination?
A: (Dr. Leroy) Yes, cloud contamination is a source of error in POLDER data, but it is possible to make a selection of data by identifying areas and time periods where cloud contamination is minimal.
Q: (Dr. Simpson) But this requires careful selection of data?
A: (Dr. Leroy) Yes, that is true. It requires close data monitoring.

Back to index

7. Precise Geographical Position (Dr. T. Hashimoto)

Dr. Hashimoto presented his studies on the topic above starting off talking about Geometric Accuracy of OCTS.
  • Geometric Accuracy of OCTS
  • Precise Geometric Correction
  • Flow for collecting GCP
  • Flow for exterior orientation
  • Flow for mapping; mapping projection types
Q: None

Dr. Tateishi of Chiba University took over the chair at this stage and apologized for arriving late.
He announced a 'change' in the program, to please note that the presentation from Dr. Trotter is only 20 minutes, not the scheduled 40 minutes which is a mistake in the agenda

Back to index

8. Global Land Surface Radiation Budget Products from the ADEOS-II GLI (F. Prata)

Dr. Prata explained that this presentation was just an update from the last meeting to cover the following:
  • Land surface temperature (global GLI/LST algorithm)
  • LST Regression Coefficients
  • Global land cover glasses and fractional vegetation cover
  • GLI land surface temperature algorithm coefficients
  • terrain height - digital elevation model and DEM variance flag
  • precipitable water - water vapor path, radiosonde precipitable water
  • Emissivity - under investigation, using interferometer, measurements
  • CSIRO scanning radiometer, measurements and results
  • Field measurements, location, data, and topography
  • Lake Tahoe (USA) radiometer, ATSR data
  • Validation, algorithm against GMS data
  • composite land surface temperature
  • Surface albedo, validation of against surface albedo measurements
  • POLDER BRDF parameters (work still in progress)
  • Albedo sampling
  • Composite of short wave data from ATSR
  • Summary/future plans
Q: (Dr. Mervyn T. Lynch, Curtin University of Technology, Perth, Australia) What are the major changes which brought about the albedo sequence?
A: (Dr. Prata) Mostly related to rainfall.
Q: (Dr. Rachel T. Pinker, University of Maryland) With your bi-directional models, and satellite radiance, do you carry out atmospheric correction first?
A: (Dr. Prata) We have mostly clear skies and few aerosols, so simple parametric models seem to work. This may not be the case in more polluted areas, we hope that others will be able to make atmospheric corrections.

Back to index

9. Development of Topographic Correction Algorithm for ADEOS-II/GLI Data, for Use in Forest Carbon Cycling and Primary Production Studies (C. Trotter)

Dr. Trotter explained that he is mainly concerned with experimental measurements to be presented here:
  • Objectives of the research
  • Topographic correction - current status
  • Conifer Canopy, LANDSAT/TM
  • Topographic Effect studies on model canopies
  • Conifer and Broad leaved canopy - nadir view
  • Observed and calculated radiance (off-nadir view also)
  • Topographic correction, LANDSAT/TM
  • Radiance Variation, off-nadir view, and with topography
  • Validation of topographic correction using imagery
  • Conclusions
Q: (Dr. Verstraete) I didn't understand the dual flight procedure.
A: (Dr. Trotter) You don't need to have two planes. If there is a short flight (around 5 km), two flights could be done at midday within 15 minutes of each other with the same transect on the ground, taking data from the right and left sides. Can't think of another possible definitive test for this algorithm.
Q: (Dr. Tateishi) What GLI data will be suitable?
A: (Dr. Trotter) We are not yet sure. We have 2 models, so need to think about scaling before requirements for ancillary data.
Q: (Dr. Tateishi) What amplification is required for 1km resolution data?
A: (Dr. Trotter) Can't give definite answer. Standard is 20m contour data, but not sufficient, maybe 10m would be OK. GLI is 250m, so at least 100m contour data (or data from other sensors)

Back to index

10. Development of Spectral Indices Optimized for the GLI Instrument (Dr. M. M. Verstraete)

Dr. Verstraete presented his studies on the topic above especially focusing on the evaluation of performance of the indices
  • Objectives
  • Approach
  • Construction of the LUT
  • Optimization of the GLI VI (1)
  • Implementation issues
Q: (Dr. Prata)Are you assuming sky is clear?
A: Yes, in principle, it is clear, but it is a different issue.
Q: (Dr. Simpson) How is cloud cover contamination handled?
A: (Dr. Verstraete) Scheduled to be done by GLI
  • Performance and evaluation of the GLI VI
  • Comparisons between the FAPAR and the GLI VI (TOA), NDVI(TOA) and EVI
  • Intercomparisons between the FAPAR derived from GLI VI (GLI), OVNI (VEGETATION) and MGVI (MERIS)
  • Sensitivity of vegetation indices with respect to extreme atmosphere/aerosol types and amounts
  • Visualization of heterogeneous 3-D images
  • FAPAR derived from SeaWiFS data acquired under two extreme viewing conditions
  • NDVI distribution etc.
  • Land Cover map from FMERS-I Projects
  • Spectral Analysis
I'd like to go into discussion and talk about details of my studies which I presented just now.
Q: (Dr. Prata) Has the algorithm been tested in the presence of noise (white noise and structured noise)?
A: (Dr. Verstraete) The last example that I gave is actual data - the simulation uses mode as-is and this leads to the data shown, and this is just one example.
Q: (Dr. Rachel T. Pinker) What were the vegetation indices used in the comparisons for FAPAR? (How was FAPAR estimated?)
A: (Dr. Verstraete) We use the same models for both FAPAR and top atmosphere reflectance. The diagram shows FAPAR as estimated by direct model (horizontal axis) against the index value (vertical axis), so showing the FAPAR retrieved after applying index on simulated data.
Q: (Dr. Prata) Why have you not tested on POLDER data?
A: (Dr. Verstraete) We are working with 9 different instruments, and each instrument has its own characteristics, so cannot apply across board. Your question implies that index should be optimised for each instrument, and we are working on that, but not for POLDER yet.

Also, the scale of pixel is so large, there may be difficulties.

Back to index

11. Development of New Vegetation Indices and Algorithms for Detecting Vegetation Change (Dr. Fujiwara , Nara Women's University)
  • Pattern Decomposition Method for GLI Data Analysis
  • Normalized standard patterns
  • Data used in this analysis
  • Results of approximately 600 samples shown on the graph
  • Radiation balance on the terrestrial area; mainly developed by Dr. Muramatsu
Dr. Furumi took over the presentation
  • Motivation and purpose
  • Vegetation Index (VIPD); characteristics of VIPD
  • Model of NPP estimation using VIPD; GPP Estimation, Respiratory losses and Net Primary Production
  • Simulation results; photosynthetic rate at strong and weak PAR
  • Validation of NPP values for paddy fields
  • Annual NPP map around tropical rain forest
  • Conclusion and Next Plan
Q: (Dr. Craig Trotter) Was your NPP field data above ground data only?
A: (Dr. Furumi) In this relationship (indicating graph), NPP values mean whole - above and above ground.
Q: (Dr. Trotter) But you seem to be comparing with above ground data only
A: (Dr. Furumi) For the 2 forest areas, there was whole NPP. For the paddy fields, above ground biomass data only.
Q: (Dr. Trotter) How are you planning to capture various changes in your model (e.g. as the forest ages, hydraulic conductivity reductions, temperature deviations etc)
A: It was agreed to discuss this issue later.
Q: (Dr. Duong) What about universal spectral pattern of water/saturated matter?
A: (Dr. Furumi) 3 universal patterns were taken from LANDSAT data - water, vegetation and soil patterns. They are fixed, so don’t move for satellite data.
A: (Dr. Tateishi) Universal pattern is not needed - using 3 basic patterns, other patterns can be applied.
Q: (Dr. Duong) Did you try an urban area?
A:(Dr. Furumi) Urban area is similar to soil patterns.
Q: (Dr. Duong) I suggest that you don't use soil component, but include bare surface (asphalt, concrete etc)

Back to index

12. Land Cover Classification by GLI 250m-Data Dr. Tateishi explained how he made land cover classification by using every HR data.

  • AARS Asia 30-second land cover data set
  • Methodology
  • GLI channels for LAND; more channels than AVHRR
  • Ground truth data
  • Background for GLCGT-DB
  • Proposal of GLCGT-DB
  • original and generalized ground truth data
  • Conclusion, wishes to discuss GLI-PI in group meetings tomorrow.
Q: (Dr. Prata) I could be a user of your product, but I can perceive 3 possible problems. Firstly, how soon can this be produced after launch? Secondly, how can we manage the large dataset produced from 1km database.
A: (Dr. Tateishi) My method uses 1 year of data, so production will be at least 1 year after launch, plus around 6 months processing time.
Q: (Dr. Prata) The third issue is that if it is based on the satellite data that I use, then there is a risk that I could use the same information twice.
A: (Dr. Tateishi) This needs to be clarified.