MODIS LAI and FPAR: Project Status and Validation

1
MODIS LAI and FPAR Project Status and Validation
N. Shabanov, W. Yang, B. Tan, H. Dong, R.B.
Myneni, Y. Knyazikhin /Boston University P.
Votava, R. Nemani /NASA Ames Research Center S.W.
Running /University of Montana
LAI Intercomparison Group Meeting, Missoula, MT,
August 16, 2004
2
Contents

• MODIS LAI and FPAR Project Highlights
• Validation Sampling Scheme
• Validation Summary of Results

3
Contents

• MODIS LAI and FPAR Project Highlights
• Validation Sampling Scheme
• Validation Summary of Results

4
Status of MODIS LAI and FPAR Products
5
MOD15_BU LAI and FPAR 1- and 4-km, monthly
6
Prototype of Collection 5 MODIS TERRA, AQUA and
Combined LAI products for North America for July
20-27, 2003
7
Visualization of MODIS LAI and FPAR data over
FLUXNET sites

• Convenient web-based tool developed by ORNL DAAC
  for visualization of ASCII subset of MODIS land
  data over about 300 FLUXNET sites.
• Allows 2 types of data visualization (a) at 7x7
  pixels grid for particular date (b) as a time
  series
• Tool emphasize Quality Control grid tool allows
  selection of data at particular QC range, while
  time series tool separates best quality and all
  the available data
• Available from ORNL DAAC, http//daac.ornl.gov/SMM
  /modis_gr.html

Grids of FPAR at 7x7 km
Grids of LAI at 7x7 km
Time series of LAI for 7x7 km Grid
Time series of FPAR for 7x7 km Grid
8
Contents

• MODIS LAI and FPAR Project Highlights
• Validation Sampling Scheme
• Validation Summary of Results

9
Scaling Procedure
Field Measurements
Fine Resolution Satellite Image
Generate Fine Resolution LAI Map
Aggregate to Coarse Resolution
Compare with MODIS LAI product
10
Data Sampling Scheme Objectives
1000m

• Example field campaign in needle leaf forests,
  Ruokolahti site, Finland. This site is very
  heterogeneous mosaic dense, sparse and
  intermediate needle leaf forests
• The objective of the optimal data sampling scheme
  is to sample dynamic range of natural variability
  in LAI for main vegetation species over the area
  where validation of satellite product will be
  performed (5x5 to 10x10 km)
• Prior to field campaign we analyze fine
  resolution satellite images ETM, IKONOS), aerial
  photo (as in Ruokolahti) and forest surveys,
  available locally to identify areas which will
  serve to represent range of variability in LAI at
  the validation area

1000m
Intermediate Forests
Dense Forests
11
Data Sampling Scheme Patches
ETM Surface Reflectances, 1x1km
Corresponding Patches, 1x1km

• To establish reliable relationship between field
  measured LAI and fine resolution surface
  reflectances, errors of measurements should be
  considered field and satellite data geolocation
  errors, errors of atmospheric correction of
  satellite data, high heterogeneity of forest
  stands at the scale of few meters (mixture of
  dense vegetation elements and gaps).
• To minimize the impact of the above errors,
  Boston University team proposed to split site
  into set of relatively homogeneous patches and
  sample LAI within each patch. Average LAI over
  the patch can be used for correlation with
  surface reflectances.

12
Data Sampling Scheme Patch Surface Reflectance

• The coefficient of variation does not exceed
  10-2 indicating that the segments can be trated
  as homogeneous areas with respect to their RED
  and NIR reflectances.

13
Data Sampling Scheme Patch LAI

• However, the LAI values exhibit higher variation
  within patches. Most of the patches can be
  represented by mean LAI within uncertainty of
  20. However, in some segments (1, 4, 8 and 9)
  the uncertainty can be high.

14
Generation of Fine Resolution LAI Map Correlate
Fine Resolution Surface Reflectances with Field
LAI

• To generate fine resolution map we first
  establish correlation between fine resolution
  surface reflectances and field LAI
• Correlation is established at the patch level
  (not pixel) to reduce errors.
• Various approaches can be used, but empirical
  relationship between reduced simple ratio and LAI
  was found to be most accurate (highest R2)
• RSR includes Shortwave Infrared (SWIR) band and
  can suppress background influence and the
  difference between land cover types.

15
Generation of Fine Resolution LAI Map
ETM LAI, 1x1 km
LAI

• Using relationship Reduced Simple Ratio - LAI and
  fine resolution satellite data we can generate
  fine resolution LAI map at 1km, and then
  extrapolate this relationship at a larger area of
  10x10 km
• 1x1 km area serve as a good representative of
  variability of LAI at 10x10 km area
• 10x10 km fine resolution LAI map need to be
  degraded to MODIS resolution and can be directly
  compared top MODIS LAI product

16
Contents

• MODIS LAI and FPAR Project Highlights
• Validation Sampling Scheme
• Validation Summary of Results

17
Summary of Publications on LAI and FPAR Validation
18
Example 1 Coniferous Forests

• Site Ruokolahti, Finland
• Measurements Dates June 14-21, 2000
• Land Cover Type Coniferous forests
• Results Comparison of the aggregated
  high-resolution LAI map and corresponding MODIS
  LAI retrievals suggests satisfactory behavior of
  the MODIS LAI algorithm although variation in
  MODIS LAI product is higher than expected.
• Publication Wang et al. (2004). Evaluation of
  the MODIS LAI algorithm at a coniferous forest
  site in Finland. Remote Sensing of Environment,
  91, 114-127.

LAI
19
Example 2 Croplands

• Site Alpilles, France
• Measurements Dates February 26 March 15, 2001
• Land Cover Type Croplands
• Results Collection 4 MODIS LAI is accurate to
  within 0.3LAI precision is 20, uncertainty is
  25. Biome misidentification deteriorates the
  accuracy by factor of 2.
• Publication Tan et al. (2004). Validation of
  MODIS LAI product in croplands of Alpilles,
  France. Journal of Geophysical Research, (
  Submitted).

Surf. Reflectances, 3x3 km
20
Example 3 Grass Savanna

• Site Senegal, Western Sudano-Sahelian zone
• Measurements Dates June-November in years 2001
  and 2002
• Land Cover Type grass savanna
• Results Seasonal dynamics of both in situ LAI
  and FPAR were captured well by MODIS LAI and
  FPAR. MODIS LAI is overestimated by approximately
  2-15 and the overall level of FPAR is
  overestimated by 8-20
• Publication Fensholt, R., Sandlholt, I.,
  Rasmussen, M.S. (2004). Evaluation of MODIS LAI,
  fAPAR and the relation between fAPAR and NDVI in
  a semi-arid environment using in situ
  measurements. Remote Sens. Environ. vol. 91, pp.
  490-507

21
Data Sharing

• Safari 2000 Botswana, Africa. Savanna, open
  shurbland and grassland. June 25-July 4, 2000.
  LAI, canopy transmittance, PAR
• Ruokolahti 2000 Ruokolahti, Finland. Needleleaf
  forest. June 14-21, 2000. LAI, canopy
  transmittances and reflectances (helicopter data)
• Harvard Forest 2000 and 2001 Harvard Forest,
  Massachusetts, USA. Broadleaf forest. July 21-25,
  2000 and August 7-9, 2001. LAI, canopy
  transmittances, PAR
• Alpilles 2001 Alpilles, France. Croplands.
  February 26 March 15, 2001. LAI
• Flakaliden 2002 Flakaliden, Sweeden. Needleleaf
  forest. June 25-July 4, 2002. LAI, canopy
  transmittance and reflectance (helicopter data)

Data are posted at MERCURY system
http//mercury.ornl.gov/ or can be downloaded
directly from Boston University FTP
ftp//crsa.bu.edu/pub/rmyneni/mynenimodisvalidatio
n/
22
MODIS LAI and FPAR Future Directions

• Further improve consistency of simulated by
  LAI/FPAR algorithm surface reflectances with
  MODIS observations to enhance quality of product
  retrievals and agreement with field measurements.
  Substantial changes are expected especially for
  woody vegetation.
• At least three versions of LAI and FPAR products
  will be generated TERRA, AQUA and Combined
  TERRA/AQUA.
• Compile all available sources on field
  measurements of LAI and FPAR into one reference
  data base to use in validation of the LAI/FPAR
  algorithm. This data base will help to sample LAI
  and FPAR over a variety of geographical locations
  and climatic conditions. Possible source will
  include FLUXNET, Mercury system, LAI
  intercomparison project and other ongoing
  projects.