1. Uncertainties of global moderate resolution Leaf Area Index (LAI) products derived from satellite data Hongliang Fang a , Shanshan Wei a,b , Shunlin Liang c a LREIS, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, 100101, China. b Department of Geography, School of Urban and Environmental Sciences, Northeast Normal University, Changchun, Jilin Province, 130024, China. c Department of Geography, University of Maryland, College Park, Maryland, 20742, USA. IGARSSā01, Vancouver, Canada, Jul 24-27, 2011
2.
3.
4. Lacaze, 2005. User Manual Masson, 2003. JOC Deng, 2006. TGRS Baret, 2007. RSE Knyazikhin, 1998. JGR Ref. NN (MCRM: PROSPECT+SAIL+PRICE/WALTHALL) Linear regression of literature LAI and NDVI SR method for forest and non-forest based on model NN (PROSPECT+SAIL+5 TYPICAL SOIL) LUT (3D model)+VI Alg. 11 directional reflectance in G, R, NIR bands and angular config. AVHRR NDVI VGT, ATSR, (MERIS) ref. SZA, TOC ref in RED, NIR, & SWIR RED, NIR Input 1996.11.5-1997.6.25; 2003.4.5-2003.10.25 1998-2007 [Sep 2006] 1999-2007 1999-present Time 10-day (L3) and monthly (HDF) Monthly Monthly 10-day 8-day Temporal 1/9Ā°(monthly) 30ā to 1D 1km, 10km, 0.25D,0.5D 1/112D 1km, 4km, 0.25Ā° Spatial POLDER ECOCLIMAP GLOBCARBON CYCLOPES EOS Project
5. Four stages of validation defined by the Committee on Earth Observation Satellites (CEOS) Adapted From LPV/WGCV/CEOS (http://lpvs.gsfc.nasa.gov/) Systematic and global Validation results systematically updated with new releases and new data. 4 Weiss et al. (2007) Garrigues et al. (2008) Aim of our study Global Product accuracy assessed systematically and globally. Product uncertainties well established. 3 Verger et al. (2011; 2009), Luo et al. (2004) Regional-continental Validation over a widely distributed set of locations, and validation efforts; 2 Fang and Liang (2005), Cohen et al. (2006), Hill et al. (2006), Pisek and Chen (2007), Sprintsin et al. (2009) Local-regional Validation in a small number of selected locations, time periods and validation efforts 1 Support studies Scale Explanation Stage
6.
7.
8.
9.
10.
11.
12. Global field LAI measurement sites from campaigns and literature 219 observations over 129 sites Fang et al., to be submitted.
13. MODIS and CYCLOPES quality indicators B3 (NIR) saturation: ok 0 / no 1 Bit 9 B2 (red) saturation: ok 0 / no 1 Bit 8 B0 (blue) saturation: ok 0 / no 1 Bit 7 Parameter validity: ok 0 / no 1 Bit 6 Aerosol source: MODIS 0 / climatology 1 Bit 5 Aerosol status: pure 0 / mixed 1 Bit 4 Cloud/shadow: no 0 / suspected 1 Bit 3 Snow status: no 0 / snow 1 Bit 2 Land 0 / sea 1 Bit 1 SM<16 CYCLOPES SM Couldn't retrieve pixel 100=4 128 ļ£ QC<255 Empirical method used (Main method failed due to problems other than geometry) 011=3 96 ļ£ QC<128 Empirical method used (Main method failed due to geometry problems) 010=2 64 ļ£ QC<96 Main (RT) method with saturation 001=1 32 ļ£ QC<64 Main (RT) method with the best possible results 000=0 QC<32 MODIS QC (DN range) and SCF_QC (binary, decimal values) Quality description Binary, DN range
14.
15. Statistics of field measured LAI Fang et al., to be submitted. 1.98 (1.61) 217 2.30 (1.57) 77 1.81 (1.61) 140 Overall 2.19 (1.43) 56 1.87 (1.10) 46 3.65 (1.88) 10 6. Needleleaf forest 3.44 (1.65) 51 3.64 (1.74) 20 3.31 (1.61) 31 5. Broadleaf forest 0.99 (1.14) 42 3.08 (2.53) 3 0.83 (0.84) 39 4. Savanna 2.36 (1.11) 4 2.36 (1.11) 4 3. Broadleaf crops 0.68 (0.70) 25 1.08 (0.79) 8 0.50 (0.58) 17 2. Shrubs 1.63 (1.09) 39 1.63 (1.09) 39 1. Grasses and cereal crops Mean (SD) n Mean (SD) n Mean (SD) n Overall Effective True Biome type
16. MODIS/Terra C4 (QC<128) Main: 85.8% R 2 =0.435 RMSE=1.42 MODIS/Terra C5 (QC<128) Main: 92.5% R 2 =0.307 RMSE=1.53 MODIS/Terra+Aqua C5 (QC<128) Main: 97.6% R 2 =0.526 RMSE=1.09 VGT/CYCLYPES V3.1 (LAI<6.0) R 2 =0.557 RMSE=0.97 Field true LAI
17. MODIS/Terra C4 (QC<128) R 2 =0.234 RMSE=2.08 MODIS/Terra C5 (QC<128) R 2 =0.290 RMSE=1.74 MODIS/Terra+Aqua C5 (QC<128) R 2 =0.186 RMSE=1.63 VGT/CYCLYPES V3.1 (LAI<6.0) R 2 =0.399 RMSE=1.34 Field effective LAI
18. Comparison of MODIS and CYCLOPES LAI with field LAI 1.34 0.399 63 1.39 0.348 56 0.82 0.005 7 CYCLOPES (effective) 0.97 0.557 111 1.05 0.629 58 0.87 0.449 53 CYCLOPES (true) 1.09 0.528 81 (97.6%) 1.05 0.599 57 (96.6%) 1.16 0.042 24 QC<64 1.09 0.526 83 1.06 0.593 59 1.16 0.042 24 MCD15 C5 1.17 0.465 98 (92.5%) 1.27 0.382 46 (88.5%) 1.09 0.221 52 (96.3%) QC<64 1.53 0.307 106 1.82 0.140 52 1.18 0.171 54 MOD15 C5 1.19 0.559 115 (85.8%) 1.13 0.718 63 (79.7%) 1.25 0.061 52 (94.5%) QC<64 1.42 0.436 134 1.50 0.481 79 1.29 0.137 55 MOD15 C4 RMSE R 2 n RMSE R 2 n RMSE R 2 n All biomes Woody Herbaceous
19. Comparison of best MODIS (QC=0) and CYCLOPES (SM=0) with field LAI 1.74 0.043 20 1.774 0.004 19 1.00 ā 1 CYCLOPES ( Effective ļ¼ 0.99 0.557 76 1.12 0.655 37 0.84 0.508 39 CYCLOPES ( True) 0.898 0.542 33 0.797 0.674 21 1.053 0.083 12 MCD15 C5 1.001 0.478 78 1.012 0.509 32 0.994 0.270 46 MOD15 C5 1.101 0.534 48 1.249 0.513 25 0.914 0.087 23 MOD15 C4 RMSE R 2 n RMSE R 2 n RMSE R 2 n All biomes Woody Herbaceous
25. Thank you! Questions, comments? Hongliang Fang ( ę¹ēŗ¢äŗ®ļ¼ Institute of Geographical Sciences and Natural Resources Research (IGSNRR), Chinese Academy of Sciences (CAS) Email: fanghl@lreis.ac.cn