Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
1
GOES-R AWG Land Team: NDVI
Algorithm
June 8, 2010
Presented By: Peter Romanov1
1 CICS/UMD
2
NDVI Product Team
Peter Romanov (Product Lead)Hui Xu, IMSGFelix Kogan, NESDIS/STARDan Tarpley, Short and Assoc.Bob Yu (Land Team Lead)Kevin Gallo, NESDIS/STARWei Guo, IMSG
3
Outline
• Executive Summary
• Algorithm Description
• ADEB and IV&V Response Summary
• Requirements Specification
• Validation Approach
• Validation Results
• Summary
4
Executive Summary
•
This NDVI Algorithm generates Normalized Difference Vegetation Index
•
The NDVI product has been moved from Option 1 to Option 2 list.
•
Algorithm development and validation is performed with MSG SEVIRI.
•
Versions 4 and 5 were delivered. CUTR has been conducted.
•
Validation tool has been developed and applied to common datasets.
•
NDVI Algorithm Package at 100% maturity will be delivered later this year
5
Algorithm Description
6
Algorithm Summary
•
This ABI NDVI Algorithm generates the Option 2 product of Normalized Difference Vegetation Index.
•
NDVI is derived at TOA
•
Standard two-channel algorithm, uses ABI bands 2 (0.64µm) and 3 (0.86µm)
•
NDVI is generated hourly
• Chlorophyll absorbs solar radiation in the red region of the spectrum
• Mesophyll structure of green leaves increases reflectance in the near infrared
• Difference or normalized difference of VIS and NIR reflectance is indicative of the vegetation condition
NDVI: Physical Basis
GOES-R ABI channel 20.59µm - 0.69µm
GOES-R ABI channel 30.85µm - 0.88µm
GOES-R ABI channel 20.59µm - 0.69µm
GOES-R ABI channel 30.85µm - 0.88µm
NDVI Algorithm
•
Top of the atmosphere NDVI is defined asNDVI=(Rnir -Rvis )/(Rnir +Rvis )
•
ABI ch.2 and 3 data will be used•
Retrievals are performed over land surface
•
Retrievals require » Clear-sky conditions» Daylight» No snow cover
8
NDVI Algorithm Output
False color composite Derived NDVI map
MSG SEVIRI April 12, 2007, 12:15 UTC 9
10
Algorithm Changes from 80% to 100%
•
Metadata output will be added
•
Quality flags will be added
•
Scheduled delivery is in September 2010
11
ADEB and IV&V Status
•
No reviews have been received for NDVI Product
•
Therefore not actions have been taken so far
12
Requirements
Product Observational Requirement
Accuracy Precision Latency Horizontal Resolution
NDVI Full Disk 0.04 0.04 60 min 2 km
In 2009 the observational requirement was changed from CONUS to Full Disk
13
Qualifiers
NDVI product has the following qualifiers.•
Solar zenith angle < 70 degrees
14
Validation Approach
NDVI Validation: Accuracy
•
NDVI is a radiance-based parameter•
NDVI accuracy is defined by radiance measurement accuracy
•
For instrument noise equivalent of 0.1% in ch.2 and 3, NDVI accuracy is ~ 0.004
15
NDVI Validation: Precision
•
Spatial and temporal change in NDVI should be consistent with variation of the state of the vegetation cover
•
Precision criterion: Low (lack of) high-frequency (day-to-day) changes in NDVI
16
NDVI Validation: Details (1)
•
Method: Evaluate daily change of NDVI» NDVI precision specification: 0.04
•
Approach: Compare NDVI estimates made in cloud- clear conditions at the same time of the day (same geometry) on two consecutive days
17
NDVI Validation: Details (2)
•
Quantitative measures- Daily RMS change of NDVI
•
Validity criteria:- Daily RMS change of NDVI below 0.04
18
19
Validation Results
20
NDVI Maps from Common Dataset
2006213 2006218 2006223 2006218
2006233 2006238 2006243
MSG SEVIRI 11.45 UTC
August 2006
21
NDVI Time Series from Common Dataset
210 215 220 225 230 235 240 245Day of the year 2006
0
20
40
60
ND
VI *
100
Lat: 39.8 N Lon: 4.2 W
210 215 220 225 230 235 240 245Day of the year 2006
0
20
40
60
ND
VI *
100
Lat: 44 N Lon: 21 E
210 215 220 225 230 235 240 245Day of the year 2006
0
20
40
60
ND
VI *
100
Lat: 17.7 N Lon: 13.2 E
210 215 220 225 230 235 240 245Day of the year 2006
0
20
40
60
ND
VI *
100
Lat: 6.0 N Lon: 29.2 E
MSG SEVIRI 11.45 UTC
August 2006
Large day-to-day variation in derived NDVI values may be due to Missed cloudsCloud shadowsOther factors (coregistration, variable aerosol, etc.)
22
NDVI Temporal Variation
Mixed Veg.Desert
Trop. Forest
Savannah
Mean daily variation of derived NDVI
Mean daily change of NDVI over full disk is below 0.04 •
NDVI estimates over deserts are more consistent than over densely vegetated areas
MSG SEVIRI
11.45 UTC210 215 220 225 230 235 240 245
Day of the Year 2006
0
0.04
0.08
0.12
0.16
0.2
ND
VI P
reci
sion
Full diskMixed VegetationDesertTropical ForestSavannah
23
Validation Results Summary
Surface type (location) Precision (requirement)
Desert 0.012 (0.04)
Mixed vegetation 0.039 (0.04)
Savannah 0.048 (0.04)
Tropical forest 0.087 (0.04)
Overall 0.032 (0.04)
MSG SEVIRI Common Dataset, August 2006
24
Summary
•
The ABI NDVI Algorithm has been developed to provide monitoring of vegetation cover properties with GOES-R ABI
•
Version 5 of the algorithm has been delivered. 100% Algorithm Package is scheduled for delivery in September 2010
•
Overall NDVI product is close to meeting the precision specification
•
More conservative cloud mask over densely vegetated areas may improve NDVI precision. Cloud shadow flags may also help.