Valued-Added Albedo and Ecosystem Products Derived From MODIS Data.

1. Valued-Added Albedo and Ecosystem Products Derived From MODIS Data. Eric G. Moody1,2, Michael D. King1, Steven Platnick1,3, Alan H. Strahler4, Crystal Schaaf4 , Feng Gao4 1NASA Goddard Space Flight Center 2L3 Communications EER 3University of Maryland Baltimore County 4Boston University Outline Motivation. Descriptions of Value-Added Products. Methodology for Generating Products. Conclusions.

2. Albedo Characteristics • Albedo is ratio of upwelling to downwelling radiative flux at the surface. • White sky albedo (bihemispherical reflectance) • Absence of direct component when diffuse component is isotropic. • Black sky albedo (directional hemispherical reflectance) • Absence of diffuse component and is a function of solar zenith angle. • Albedo is intrinsically related to ecosystem classification. • Albedo is dependent: • Spatially • Temporally • Spectrally 2

3. Value Added Products - Motivation. • Production and research requirements: • Complete global maps. • High spatial resolution. • Seasonal representation. • Specific wavelengths of interest. • Available in projections standard within community. • Issues with data sets prior to MODIS: • Broadband only. • Lower resolution. • Static maps. • MODIS Land data sets provide needed characteristics 3

4. Terra/Aqua MODIS Data • Relevant global Land Level 3 products: • MOD12Q1 - Ecosystem Classifications Product • 1 km resolution, globally. • Classification Schemes include: • IGBP scheme • 14 class UMD scheme • BGC Biome scheme • LAI/fPAR Biome scheme • Produced using supervised (non-production) approach. • MOD43B3 - Albedo Product (White- and Black-Sky) • 1 km resolution, 16-day global composites. • 7 specific wavelength data sets: • 0.47, 0.555, 0.659, 0.858, 1.24, 1.64, and 2.13 m • 3 broadband data sets: • 0.3-0.7, 0.3-5.0, 0.7-5.0 m 4

5. Terra/Aqua MODIS Data • Stored in equal area Integerized Sinusoidal Projection (ISIN). • Approximately 300 tiles contain some land. • (From the MODLAND website) 5

6. Terra/Aqua MODIS Data • Data Volume: • MOD12Q1 - Ecosystem Classification Product • ~ 300 tiles per global map. • 22 MB/tile, ~6.6 GB per map. • MOD43B3 - Albedo Product (White- and Black-Sky) • ~300 tiles per 16-day composite global map. • ~6,900 tiles per year. • 66MB/tile, ~20 GB per map. • 23 maps per year, ~ 460 GB per year. 6

7. Important Caveats • ISIN Projection. • Large Data Volume. • Large Number of Files. • Incomplete Maps. 7

8. Proposed Value-Added Products. • Complete Albedo maps. • Reformat, apply QA, and fill in missing data. • Albedo/Ecosystem statistics. • For each ecosystem class in NxM degree boxes. • Parameterization of complete Albedo maps. • Store temporal fit coefficients. • Pilot Study -- Y2001 and Y2002 data sets. 8

9. Value-Added Products - Methodology. Series of steps to generate products: • Mosaicing of ISIN tiles to equal angle map. • Application of Quality Assurance. • Generation of Albedo/Ecosystem statistics. • Filling in of missing points. • Parameterization of complete maps. 9

10. Mosaic ISIN Tile Data - Description. • Two Step Process. • “Pure” mosaicing of data onto grid. • Application of QA and exclusion of pixels. • Stored in equal angle grids. • 1 minute resolution (2km at equator, < 1km at poles). • Stored in HDF files: • Data arrays • Geolocation • QA -- original and processing. • Metadata -- original and processing. • Albedo and Ecosystem data are mosaiced. 10

11. Mosaic IGBP Ecosystem Map. • The product contains the following data sets: • IGBP classification map. • Classification assessment. • Secondary ecosystem type and percentage. • QA -- original and processed. • Metadata -- original and processing. • Reduces number of files from ~300 to 1, per map. • Data volume reduced to 600 MB from 6.6 GB • Codes modifiable for: • Other classification schemes. • Different resolution. 11

12. Mosaiced IGBP Ecosystem Map 12

13. Mosaic Albedo Data. • Mosaiced data stored in 10 HDF files per 16-day period. • One file per wavelength (7 specific wavelengths and 3 broadband) • Each file contains: • White- and black-sky albedo data. • QA -- original and processing. • Metadata -- original and processing. • Pixels excluded for: • Simulated seasonal polar darkness. • Seasonal snow/ice coverage. • General and band specific QA. • Reduces number of files per 16-day period from ~300 to 10. • Reduces data volume to 390 GB/year from 460 GB. 13

14. Mosaiced Albedo Map, July 2001, 0.659 mm. Step 1 - “Pure” Mosaicing of MOD43B3 data. 14

15. Mosaiced Albedo Map, July 2001, 0.659 mm. Step 2 - Application of QA and Pixel Exclusion. 15

16. Mosaiced Albedo Map, July 2001, 0.47-2.13 mm. Step 2 - Application of QA and Pixel Exclusion. 16

17. Mosaiced Albedo Map, Year 2001, 1.24 mm. Step 2 - Application of QA and Pixel Exclusion. 17

18. Missing data points. • Remaining issue -- missing data points. • Fill methodologies: • Spatial Fit. • Spectral Fit. • Temporal Fit. 18

19. Temporal data. • Three scenarios: • Nearly complete temporal curve. • One or more missing seasons. • Very few temporal points. 19

20. Fitting Temporal data. • Goal: Fill in with geophysically meaningful data. • How to fit the three scenarios? • Nearly complete temporal curve. • One or more seasons missing. • Very few temporal points. 20

21. Temporal Fitting Using Regional Information. • Ecosystem class behaves according to phenological events. • Region’s ecosystem class has unique temporal behavioral curve. • Image provided by Xiaoyang Zhang, Boston University. 21

22. Temporal Fitting of Eco Class Mean Albedo. • Impose Regional Temporal Behavior curve. 22

23. Preliminary Analysis of Fits. 23

24. Fill Methodology. • Compute Albedo/Ecosystem statistics. • Describe behavioral curve of statistics. • Apply metrics to determine “best” behavioral curve. • Impose shape of behavioral curve to pixel data. 24

25. Albedo/Ecosystem Statistics - Description • Generate albedo statistics for each ecosystem class: • 0.5 x 0.5 degree box • For each specific or broad albedo wavelength • Water albedo set to 5%. • Combine boxes for larger boxes and belts. • Squares of up to 10x10 degree. • Latitude belts of 10 x Lon degree. • Subcontinent belts of 60 x Lon degree. 25

26. Mosaiced Albedo Map with Excluded Pixels. July 2001, 0.858 m. 26

27. Regenerated Albedo Map From Alb/Eco Statistics. July 2001, 0.858 m. 27

28. Box/Belts of Alb/Eco Statistics. July 2001, 0.858 m. 28

29. Albedo/Ecosystem Statistics - Product • Statistics stored in 15 HDF files per 16-day period. • One file per box size. • Each file contains: • White- and black-sky albedo/ecosystem statistical data. • Metadata. • Reduces data volume to 25 GB/year from 460 GB. • Regenerate albedo maps: • Lat/lon and ecosystem classification. 29

30. Parameterization of Albedo Maps. • Albedo maps have large data volume. • Reduce data volume through parameterization. • Data set consists of fit coefficients stored in HDF file. • Reduced data volume ~10-20 GB/year (from 450 GB). • Regenerate maps by inputting lat/lon and julian day. 30

31. Take Home Message • Three proposed value-added products. • Year of complete albedo maps. • Albedo/Ecosystem statistics. • Parameterized complete albedo maps. • Products provide: • Complete global maps. • High spatial resolution. • Seasonal representation. • Specific wavelengths of interest. • Available in projections standard within community. • Reduced data volume and number of files. 31

32. Future Directions. • Complete analysis and development of filling in maps. • Provide Data Sets and FORTRAN Utilities for working with the Data Sets. • MODIS Land team is using this project as pilot for an official product. • Visits and talks with user communities to publicize product. • Several papers in the works. 32