Continuous MODIS Leaf Area Index (LAI) Science Data Set for Temporal and Spatial Analysis

1. Toward temporal and spatial continuous MODIS Leaf Area Index (LAI) science data set Feng Gao, Jeff Morisette, Robert Wolfe ACCESS Project: Improving access to Land and Atmosphere science products from Earth Observing Satellites: helping NACP investigators better utilize MODIS data products PIs: Jeff Morisette and Robert Wolfe

2. Providing Smoothed and Gap-filled MODIS LAI from MOD15A2 • Use TIMESAT software to describe LAI temporal variations and produce smoothed LAI values • Develop gap-filling algorithm to fill data gaps which are missing from MODIS or TIMESAT • Compose LAI using high quality (main RT) LAI from MOD15A2 but replacing other quality with smoothed and gap-filled LAI • Produce phenological parameters from MOD15A2

3. TIMESAT & MODIFICATION • A program to analyze time-series satellite data developed by Per Jönsson and Lars Eklundh. It provides three functions for data fitting: Asymmetric Gauss (AG) Double Logistic (DL) Savitzky Golay (SG) • TIMESAT provides 11 phenological parameters based on fitting function • We modified TIMESAT to accept MOD15A2 product as inputs • Weight LAI according to SCF quality flags in MOD15A2 1.0 = Main RT 0.5 = Main RT with saturation 0.1 = Backup algorithm 0.0 = Fill or Cloud

4. A TIMESAT example based on five years (2001-2005) MODIS LAI

5. TIMESAT fitting of one pixel using five years MODIS LAI as inputs Main RT = 1.0; Main RT with saturation = 0.5; Backup Alg. = 0.10; from AG

6. GAP Filling Algorithm • Compute typical TIMESAT curve L(ci, t) for each IGBP within this tile • if there exists many high quality curves, use average • else use the one with best quality within the tile • 2. Find a best quality TIMESAT curve Lb(t)for LAI missing pixel (x0, y0) • obtain IGBP classes for this pixel (x0, y0, c0) • search candidate pixels (xi, yi, c0) within searching window wj • if not found, then increase wj+1 = wj * sqrt(2) • if wj > max_win_size, then use: Lb(t) = L(c0, t) • select best quality curve Lb(t) from (xi, yi, c0) • Adjust TIMESAT curve Lb(t) using high quality LAI L0(t) from (x0, y0) • (Note: num_of_L0(t) < num_of_Lb(t) ) • build linear relation Lo(t) = a + b * Lb(t) using least square approach • based on local fitting (one year data) or global fitting (all available • HQ data) if local fitting approach fails

8. Outputs • Six output layers • Original MODIS LAI • Smoothed and Gap-filled LAI • Composed LAI • Aggregated Quality for Original MODIS LAI • Quality for Smoothed and Gap-filled LAI • Quality of Composed LAI • In both HDF-EOS and binary format

9. TIMESAT LAI (AG) h11v04 A2001001 black = no values Main RT = 1.0; Main RT SAT = 0.5; Backup Alg. = 0.1 0.0 1.5 3.0

10. Gap-filled LAI h11v04 A2001001 black = no values 0.0 1.5 3.0

11. Original MODIS LAI MOD15A2 h11v04 A2001001 black = no values 0.0 1.5 3.0

12. Composed LAI Using high quality LAI from MODIS & Replacing low quality LAI from gap-filled and smooth LAI h11v04 A2001001 black = no values 0.0 1.5 3.0

13. Composed LAI QA Red (1): Original MODIS HQ LAI Green (2): Timesat smooth and gap-filled LAI Blue (3): Fill value use MODIS value h11v04 A2001001

15. Smoothed LAI on 2004001 (1/1/2004) 0.0 3.5 7.0

16. Smoothed LAI on 2004097 (4/6/2004) 0.0 3.5 7.0

17. Smoothed LAI on 2004193 (7/11/2004) 0.0 3.5 7.0

18. Smoothed LAI on 2004289 (10/15/2004) 0.0 3.5 7.0

19. Phenological Parameters • 1. time for the start of the season; time for which the left edge has increased to a user defined level (often 20% of the seasonal amplitude) measured from the left minimum level. • 2. time for the end of the season; time for which the right edge has decreased to a user defined level measured from the right minimum level. • 3. length of the season; time from the start to the end of the season. • 4. base level; given as the average of the left and right minimum values. • 5. time for the mid of the season; computed as the mean value of the times for which, respectively, the left edge has increased to the 80 % level and the right edge has decreased to the 80 % level. • 6. largest data value for the fitted function during the season; • 7. seasonal amplitude; difference between the maximal value and the base level. • 8. rate of increase at the beginning of the season; calculated as the ratio between the values evaluated at the season start and at the left 80 % level divided by the corresponding time difference. • 9. rate of decrease at the end of the season; calculated as the ratio between the values evaluated at the season end and at the right 80 % level divided by the corresponding time difference. • 10. large seasonal integral; integral of the function describing the season from the season start to the season end. • 11. small seasonal integral; integral of the difference between the function describing the season and the base level from season start to season end.

20. Start of Season from 2001 to 2005

21. - 180 - 120 P1 2002 MODIS - 60

22. - 180 - 120 P1 2003 MODIS - 60

23. - 180 - 120 P1 2004 MODIS - 60

24. All Phenological Parameters for 2004

25. - 180 - 120 P1 2004 MODIS - 60 1. time for the start of the season

26. - 360 - 300 P2 2004 MODIS - 240 2. time for the end of the season

27. - 270 - 195 P3 2004 MODIS - 120 3. length of the season

28. - 15 - 7.5 P4 2004 MODIS - 0 4. base level

29. - 240 - 210 P5 2004 MODIS - 180 5. time for the mid of the season

30. - 100 - 50 P6 2004 MODIS - 0 6. largest data value

31. - 70 - 35 P7 2004 MODIS - 0 7. seasonal amplitude

32. - 15 - 7.5 P8 2004 MODIS - 0 8. rate of increase at the beginning of the season

33. - 15 - 7.5 P9 2004 MODIS - 0 9. rate of decrease at the end of the season

34. - 1400 - 700 P10 2004 MODIS - 0 10. large seasonal integral

35. - 700 P11 2004 MODIS - 0 11. small seasonal integral