Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
QA4ECV WP4: an ECV algorithm forfAPAR & LAI
Simon Blessing, Ralf Giering
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
D4.3 ATBD for FAPAR & LAI from albedo
- internally reviewed
- delivered via EUMETSAT
- description of data flow through TIP5D-LUTs
- description of error propagation and budget
- guidance on suitable usage of FAPAR and LAI
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
Progress on ATBD for FAPAR & LAI from albedo
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
The Two-Stream inversion package
Optimisation:- find 2-stream parameters matching BHRVIS and BHRNIR
Error Propagation:- use 2nd derivative at optimum to determine posterior covariances
TIP-Tables:process is tabulated over observation space for performance
From Covariances:uncertainties onLAI and fAPAR
At Optimum:- optimal LAI- simulated VIS- absorption in canopy is fAPAR
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
TIP 5D Table Construction96 TIP look-up-tables for 5D generated for
● NIR, VIS uncertainty levels 0%-5%, 5%-7%, 7%-10%,>10%, respectively,● NIR/VIS correlation levels <-0.5, -0.5% – 0.5%, >0.5%,● snow and no snow prior conditions.
● Tables resolve albedo differences of 10-3.● TIP inversions are trusted for typically observed albedos.● Flag table extended to hold necessary 5D information.● Tables contain posterior uncertainty of each inversion.
Final cost function for no snow (left) and snow (right) using 5 % albedo uncertainty
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
Performance of Re-Processing with GA-BHRs
● In preparation for the “real” data, simple5D was applied to GA-BHRs● FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011,
8-daily, global, 1km resolution● Processing on CEMS' LOTUS cluster took two working days,
including rescheduling of interrupted jobs
histogram of job durations in cpu
seconds
histogram of job durations in cpu
seconds
x: max. memoryy: cpu seconds
x: max. memoryy: cpu seconds
x: max. swapy: cpu seconds
x: max. swapy: cpu seconds
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
global snapshot:
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
fapar/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
From: Pinty B., N. Gobron: Land Surface Albedo Dynamics, in State of the Climate 2014, Special Supplement to BAMS Vol. 96, 2015
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
ΔFAPAR ΔσFAPAR
● comparsion at site-level to 2D-processing re-run for all GA-verification sites
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
How do TIP-5D and TIP-2D results differ for a tile, e.g. h18v03?
=> Small increase in value, reduced uncertainty.
FAPARFAPAR σFAPAR
σFAPAR
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
● comparsion at site-level
ΔFAPAR
ΔσFAPAR
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
FAPAR/LAI from GlobAlbedo BHRs processed for 1998-2011, 8daily, global, 1km resolution
input BHRs
input BHRs
resulting FAPAR
resulting FAPAR
QA4ECV Mid-Term Review Meeting, MSSL, 14-16 June 2016
Summary
● ATBD D4.3 for fapar and LAI delivered
● TIP-5D LUT generated
● GlobalAlbedo 1998-2011 reprocessed with TIP-5D
● processing of daily,1km VIS+NIR BHR (1998-2015)expected to take ~10days on CEMS
● new fapar and LAI have a slightly reduced uncertainty