Producing Realistic Emissivity maps from SEVIRIMSG

1
Producing Realistic Emissivity maps from
SEVIRI/MSG
L. Peres, C. DaCamara, Isabel F. Trigo, and C.
Madeira
2
Outline

• Land Surface Emissivity
• Relevance for LST
• Methodologies
• Vegetation Cover
• Direct Retrieval of LST LSE
• Broad Band and Spectral maps
• Future Work

3
Why Land Surface Emissivity?
Land Surface Temperature Generalised
Split-Window ? 10.8?m and 12.0?m
4
LST Available for LSA SAF Geographical
Areas MSG Projection 15 min Europe
Since Jan 2005 MSG Disk Since Jul 2005
5
LST Sensitivity to LSE
Error ? 0.01 ?? 0.01
6
LSE Vegetation Cover Method
Pixel LSE for Band (channel) c ?c ?c,d d?c
7
LSE Vegetation Cover Method
Band Emissivity for VEGETATION / SOIL classes
?(Spectral Libraries)

• SEVIRI/Meteosat Channels
• - IR 3.9
• IR 8.7
• IR 10.8
• IR 12.0
• Broad Band
• ?1 3 ?m ? 14 ?m
• (f?1)

8
Channel Broad-Band Emissivity Database for -
IGBP - GLC2000 Land Cover Classes
Peres and DaCamara (2005), IEEE Trans Geosci
Remote Sens.
9
FVC - Fraction of vegetation cover Land SAF
Product Daily / 10-day composites
SEVIRI VIS 0.6 NIR 0.8 SWIR 1.6
http//landsaf.meteo.pt
10
EMISSIVITY
Other SEVIRI channels IR8.7 IR12.0
?d ?c,v FVC ?c,g (1-FVC)
11
Uncertainties in Emissivity 1) Variability of
vegetation state (green /senescent) soil
and vegetation types
Vegetation ? ?LSE 0.001 0.008
IGBP Class 16 Barren/Sparsely vegetated 25
rocks 25 sand quartz 25 desert varnish
25 aridsols
Bare Ground ?LSE 0.35 IR8.7
?LSE lt 0.02 IR10.8 IR12.0
12

• Desert/Semi-desert regions
• VCM provides Constant LSE
• Dry atmosphere ? High LST sensitivity to LSE

13
Direct LSE LST Retrievals (over Arid
Regions)
Recursive methodology Observations SEVIRI
radiances, L, IR10.8 IR12.0,
e.g., at 6UTC 12 UTC of day d Observations
close in time ? LSE remains constant
Unknowns LST_6UTC, LST_12UTC LSE_IR10.8,
LSE_IR12.0
14
Search LST LSE combination which minimizes
f ? Lobs LRTM(LST, LSE)i2 for i
channels IR10.8, IR12.0, 6 and 12 UTC
15
Algorithm I - LSTo LSEo 1st guess ? VCM GSW
algorithms
II.b Recompute LSTn with GSW using LSEn
III Repete II.a II.b until convergence
criteria are met!
16
Direct LSE LST Retrievals - Error Analysis
Method applied to synthetic SEVIRI images
17
September 2005
LSE IR 10.8
LSE VCM 0.950
18
LST Land SAF product (12UTC 12/09/2005)
19
Summary

• LSE maps 1st developed for retrieval of LST
• ? Laboratory measurements (ASTER spectral
  library) used to assign LSE to IGBP surface
  types
• Look-up-Table for SEVIRI/Meteosat channels
  broad-band (3-14 ?m)
• LSE is estimated using the LUT and the Fraction
  of Vegetation Cover (Land SAF)
• Same methodology will be applied for AVHRR/Metop

20
Summary

• LSE uncertainties over semi-arid regions are
  high
• The impact of LSE errors on LST retrievals is
  highest for dry atmospheres
• Method for simultaneous retrieval of LST LSE
• LSTt1, LSTt2, LSEIR108, LSEIR120 which minimize
  the SSD
• ? Lobs LRTM(LSTti,LSEchn) 2
• ti 6, 12UTC chn IR108, IR120
• Application to synthetic images
• bias 0 rmse 0.015

21
Next - Apply simultaneous retrieval over
semi-arid regions - LSE ? Average over 10-20
daily estimations
2007-2012 Land SAF Operational phase - Updated
LSE maps other issues - Angular effects
- LST Validation (in-situ measurements MODIS,
AATSR, Metop, ...)
22
Land SAF - LST ( LSE) - Albedo - Downward
Surface Fluxes (short long-wave) - Snow
Cover - Vegetation (FVC, LAI, fAPAR)
(...) - Evapotranspiration - Wild fire
products (burnt areas, monitoring, RISK)
http//landsaf.meteo.pt