VENS: a new EO mission at

1

• VENµS a new EO mission at
• High spatial resolution (5-10m)
• High revisit frequency (2 days)
• Constant viewing angle
• G.Dedieu1 (CNES PI), O.Hagolle1, S. Garrigues1,
  et al (VENµS project team)
• 1 CNES, Toulouse, France

2
Outline

• The mission
• Products and pre-processing
• Applications

3
I The Mission
4
Venµs Mission

• Mission in cooperation between France and Israel
  
• Scientific demonstrator
• Super-spectral
• High spatial resolution
• Multi-temporal observations (every 2 day)
• Constant viewing angles
• Technological mission Test of an electrical
  propulsion system

250 kg
5
Mission specifications

• Venµs image characteristics
• Resolution 5m-10m
• Field of View 27 km
• 12 spectral bands from 412 to 910 nm
• Geometric revisit frequency 2 days
• Systematic acquisition 50 sites
• 2 stereoscopic bands with a low angle difference
• Constant viewing angle gt Directional effects
  are minimised

• Current similar commercial satellite Formosat-2
  (NSPO, Taiwan)
• Launched in 2004
• Resolution 8m, Field of View 24 km
• 1 day repeat cycle
• 4 Spectral bands 488, 555, 650, 830 nm
• Constant viewing angle

6
Formosat-2 time seriesYaqui, Mexico
7
Formosat-2 time seriesYaqui, Mexico
8
Formosat-2 time seriesYaqui, Mexico
9
Formosat-2 time seriesYaqui, Mexico
10
Formosat-2 time seriesYaqui, Mexico
11
Formosat-2 time seriesYaqui, Mexico
12
Formosat-2 time seriesYaqui, Mexico
13
Formosat-2 time seriesYaqui, Mexico
14
Formosat-2 time seriesYaqui, Mexico
15
Formosat-2 time seriesYaqui, Mexico
16
Formosat-2 time seriesYaqui, Mexico
17
Impact of constant view angle
Wheat field Yaqui
VENµS constant view angle gt Smooth time series
Wheat field Romania
SPOT non constant view angle gt Noisy time series
18
II VENµS Products
19
VENµS Products
Cloud detection
Atmospheric correct.
Temporal compositing

• Level 1 (L1)
• single acquisition
• georeferenced
• calibrated TOA reflectance

• Level 2 (L2)
• single acquisition
• georeferenced
• calibrated TOC reflectance

Level 3 (L3) temporal synthesis
20
Multi-temporal processing algorithms

• Multi-temporal (recurrent) algorithms for
• Water detection
• Cloud/shadow detection
• Aerosol estimation
• L2 composite product of date D-1 used as input in
  the algorithm

21
Cloud detection

• Venµs algorithm characteristics
• Use of stereoscopy (620 nm spectral bands)
• Detection of surface reflectance variations in
  the blue
• Detection of multi-temporal decorrelation

22
Atmospheric correction Multi-temporal AOT
retrieval algorithm

• Assumptions
• No directional effects
• Surface reflectances vary
• Quickly with distance
• Slowly with time
• Aerosol optical properties vary
• Quickly with time
• Slowly with distance (few km)?

Search AOT(D) and AOT(D2) minimizing differences
between the 3 surface reflectances (a priori, D,
D2)?
23
FORMOSAT-2 time series
Retrieved AOT
Initialisation image
24
FORMOSAT-2 time series
Retrieved AOT
25
FORMOSAT-2 time series
Retrieved AOT
26
FORMOSAT-2 time series
Retrieved AOT
27
FORMOSAT-2 time series
Retrieved AOT
28
FORMOSAT-2 time series
Retrieved AOT
29
FORMOSAT-2 time series
Retrieved AOT
30
FORMOSAT-2 time series
Retrieved AOT
31
FORMOSAT-2 time series
Retrieved AOT
32
FORMOSAT-2 time series
Retrieved AOT
33
Validation
Surf. Ref. Validation
AOT Validation
NIR
Estimated AOT
Surface reflectance
RED
BLUE
Measured (AERONET) AOT
Hagolle, Dedieu et al., Correction of aerosol
effects on multi-temporal images acquired with
constant viewing angles Application to
Formosat-2 images, Remote Sensing of
Environment, vol. 112, Apr. 2008, pp. 1689-1701.
34
III Applications
35
Dynamic land cover monitoring
HR (10m) Multi-temporal imagery No
directional effect
Classif. (SVM)
FORMOSAT time series
MAIZE
90 of maize pixel detected gtpredicting crop
water demand
Classification accuracy
Ducrot et col. 2008
36
Change detection(FORMOSAT examples)
Image Before
Image After
Burned area
Klaus storm effect
37
Crop water demand monitoring1 Method
Multitemp. Classif
Land cover
FORMOSAT time series
Biomass
RT inversion
Crop Model (SAFYE)
Leaf Area Index
Evapotrans.
Change detection
Sowing dates
B. Duchemin et al., "A simple algorithm for yield
estimates Evaluation for semi-arid irrigated
winter wheat monitored with green leaf area
index," Environmental Modelling and Software,
vol. 23, 2008, pp. 876-892.
38
Crop water demand monitoring2 Results
Over-estimation due to not declared groundwater
pumping
Simulated irrigation (model driven by FORMOSAT-2
data)
Sink
Overestimation
Declared irrigation
39
Conclusions

• Venµs, new EO mission concept
• Benefit for accurate surface reflectance time
  series
• Cloud discrimination
• Aerosol optical thickness estimation
• Benefit for EO applications
• Vegetation/crop monitoring
• Water resources monitoring
• Surface change detection
• Potential for the validation of global veg.
  product
• 12 FORMOSAT-2 time series available for
  scientific use
• Preparing future operational EO mission
  (SENTINEL-2/ESA)

high resolution (10m) high revisit frequency
(2 d) constant viewing angle
40
VENµS simulated LAI (FORMOSAT images 2006)
DEMAREZ et al., 2009.
41
ADDITIONAL SLIDES
42
Evapotranspiration et rendementMaroc CESBIO

• 50 images Formosat-2 (11-2005 gt 11-2006)?
• Méthode
• 1. Carte d'occupation des sols (Blé, jachères)?
• 2. Estimation du LAI (Leaf Area Index) sur les
  données Formosat-2
• Formule empirique à partir fu NDVI, calée sur
  données in-situ
• 3. Détection des interventions agricoles
• Labour gt date de semis
• 4. Modélisation avec SAFY (modèle simplifié
  développé au CESBIO)?
• Calage du modèle par le LAI et la date de semis
• Calcul de rendement du blé en t/ha
• Calcul de l'évapotranspiration et de la demande
  en eau
• Calcul des apports d'eau
• 1 B. Duchemin et al., "A simple algorithm for
  yield estimates Evaluation for semi-arid
  irrigated winter wheat monitored with green leaf
  area index," Environmental Modelling and
  Software, vol. 23, 2008, pp. 876-892.

43
VENµS Products

• Niveau 1
• Mono acquisition, pas d'hypothèse sur le paysage
  observé
• Données étalonnées (Réflectances TOA) et en
  projection géographique. Les données
• Niveau 2
• Mono acquisition, hypothèses sur le paysage
  observées possibles
• 2a réflectances de surface (après détection des
  nuages et corrections atmosphériques)
• Ce produit devrait être le produit de base pour
  les utilisateurs de Venµs
• 2b variables biophysiques, peu d'efforts dans
  le cadre de Venµs
• Niveau 3
• Synthèse temporelle pour une courte période

44
Composite images

• Composite image contains the latest available
  reflectance
• At 100 m resolution
• TOA reflectance
• Surface reflectance

45
Clouds reflectance variation

• Surface reflectances in the blue are usually
  stable
• Cloud presence increases the reflectance
• Criterion
• Detection of reflectance increase in the blue
• greater than in Near Infrared
• But some false detections must be avoided
• Water (sunglint, foam, turbid waters)gt water
  mask
• Effect of rain, when the soil dries gt rain flag
• Modifications of land cover gt coarse resolution
  (100m)?
• Ploughing, crops
• Undetected cloud shadows
• Cirrus flag based on the cloud altitude
• Variations of aerosol content, or vegetation
  growth
• gt use of a correlation mask
• If good correlation with composite gt not a
  cloud

46
Atmospheric correction
47
Atmospheric correction

• Accurate models for radiative transfer exist
• Reference codes
• 6S for gaseous transmission gt SMAC
• SOS (LOA-CNES) for scattering gt Look-up tables
  
• Main difficulties
• Water vapour
• Low impact
• Presence of 910 nm spectral band
• Use of POLDER algorithm, LOA agrees to compute
  coefficients
• Aerosols
• Main source of errors
• Adjacency effect
• Necessary at Venµs resolution
• A simple correction was tested (works well)?
• Slope illumination correction
• A simple correction works well

48
Atmospheric correction algorithm

• Reference codes
• 6S for gaseous transmission gt SMAC
• SOS (LOA-CNES) for scattering gt Look-up tables
• Multi-temporal algorithm for Aerosol (AOT)
  estimation
• Accounts for Adjacency effect (atmospheric PSF)
• Necessary at Venµs resolution
• A simple correction was tested (works well)?
• Slope illumination correction

49
AOT retrieval

• Simulations of aerosol inversion
• Good results even with stable aerosol conditions
• Convergence requires varying aerosol conditions
• A small negative bias is observed

50
Atmospheric correction

• Algorithmic details
• Inversion is done at 100m resolution
• Inversions performed for a neighbourhood of 50
  pixels
• The aerosol model is constant for one site
• Only the AOT is inverted
• If reflectances in the NIR change too much,
  pixels are discarded
• A minimum number of 25 pixels is necessary
• If the standard deviation of reflectances is too
  small
• Not enough information
• The neighbourhood is extended
• More details provided in Mireille's Presentation

51
Aerosol estimates with FORMOSAT Muret(New
version)?
52
Aerosol estimates with FORMOSAT La crau(New
version)?
53
Validation of atmospheric correction

• Reflectance comparison for two dates with
  different AOT

54
Atmospheric correction

• Validation at La Crau Calibration site (ROSAS)?

Dots La crau station Triangles Formosat
2 Formosat 2 was very useful to correct bugs
on ROSAS software
55
Effects of varying surface reflectance

• Causes of variation
• quick variations of reflectances
• Ploughing, harvest...
• Detection of water bodies and of rain events
• Sometimes, reflectances in the whole image evolve
  quickly
• Semi arid zones, mono cultures
• Solutions
• Detection of varying surface when possible
• Would work better with a SWIR band
• constrain AOT to stay within plausible values

56
Constraining AOT

• 2 constraints for AOT estimate
• Floor minimum value of AOT
• AOT gt0.03
• Ceiling maximum value of AOT
• Use of  Dark Pixel  method
• Search for the darkest pixel within image (blue
  spectral band)?
• Suppose its surface reflectance is 0.01
• Compute the necessary AOT to obtain the TOA
  reflectance
• gt maximum AOT
• The ceiling constraint has a low weight in the
  least squares
•   soft ceiling 
• Enables variation of AOT within the image

57
Without Ceiling constraint on AOT
Yaqui, Mexique
58
With ceiling constraint on AOT
Yaqui, Mexique
59
F2 time series
60
Venµs Formosat 2

• Venµs image characteristics
• Resolution 5m-10m, Field of View 27 km
• 2 day repeat cycle
• 12 Spectral bands from 412 to 910 nm
• Systematic acquisition of 50 sites every second
  day
• Constant viewing angle
• Formosat-2 images NSPO (Taiwan) satellite
• Resolution 8m, Field of View 24 km
• 1 day repeat cycle
• 4 Spectral bands 488, 555, 650, 830 nm
• Constant viewing angle
• 1000 / image
• Launched in 2004

61
Data availability

• More than 10 Formosat2 time series are available
• Data are available for free for future Venµs
  Users
• LPV is already member the Venµs User list (S.
  Garrigues proposal)?
• Interested ?
• Please contact olivier.hagolle_at_cnes.fr
• Or gerard.dedieu_at_cesbio.cnes.fr

62
Accessibility
Altitude 720 km Inclination 98.27
Sun-synchronous Local Time 10h30
63
(No Transcript)
64
(No Transcript)
65
Spectral bands
Simulated vegetation reflectances for 3 different
Chlorophyl content
66
Quelques exemples d'imageursHaute répétitivité,
Haute résolution, Angle de visée constant
67
Available FORMOSAT-2 time series
68
Morocco

• CESBIO/IRD site

• 11/2005 à 11/2006
• Sunphotometer
• One gap in spring

69
SudOuest

• CESBIO site Near Toulouse

• 03/2005 à 12/2007
• Sunphotometer
• Many data gaps

70
La Crau

• 02/2006 à 10/2006
• Sunphotometer
• Many data gaps
• VZA 41

• Calibration site

71
Cestas, France

• From 24/05/2005 to 21/07/2005
• 21 images
• No sunphotometer

72
Montréal, Canada

• Period
• From 05/06/2005 to 03/07/2005
• 9 dates
• No sunphotometer
• Sunglint conditions

73
Bardonecchia, Italie

• Period
• From 04/09/2005 to 03/11/2005
• 11 dates peu nuageuses
• Snow
• No Sunphotometer

74
Agoufou, Mali

• Period
• From 06/2007 to 10/2007
• One image /4 days
• Sunphotometer
• Half of the time
• VZA 51
• AOT up to 1.

75
Drôme, France

• Period
• From 06/2007 to 08/2007
• One image /week
• Sunphotometer

76
Yaqui , Mexico

• Period
• From 11/2007 to 06/2008
• One image /5 days
• Sunphotometer
• AOT up to 0.1 !

77
Svalbard , Norway

• Period
• From 03/2007 to 09/2007
• Few images
• Some snow
• No Sunphotometer

78
Multi-temporal processing algorithms

• Multi-temporal (recurrent) algorithms for
• Water detection
• Cloud/shadow detection
• Aerosol estimate
• L2 composite product of date D-1 used as input in
  the algorithm

79
Cloud detection

• Venµs algorithm characteristics
• Use of stereoscopy (620 nm spectral bands)
• Detection of surface reflectance variations in
  the blue
• Detection of multi-temporal decorrelation