Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Sylvain Capo,Bertrand Lubac and Driss Bru
A semi‐analytical model for SPOT bathymetric inversion:an historical morphodynamical approach of the Arcachon inlet evolution since 20 years
Sylvain Capo,Vulnerability of coastal ecosystems to global change and extreme events
Biarritz 18-21 October 2011
INFOLITTORAL Project
The aim of the project is to develop and promotes operational remote sensing productsfor coastal users and managers through environmental indicators
In coastal regions, bathymetric data are critical to understand the ecosystem
Use of VHR multispectral remote sensing : most effective and low cost solution
Ocean Color approach: Bathymetric inversion algorithm development adaptation of theQAA to low spectral resolution SPOT
Morphodynamics of the Arcachon Lagoon inlet (Kalideos)
Sylvain Capo,
Ocean Color: characterizing ocean color (satellite reflectance) as a function of the water column constituents and depth Objective : linking AOP to IOP (ligth attenuation by absorption a and scattering b)Remote sensing sensors measure the reflected solar radiation
Bathymetric inversion algorithm
Sylvain Capo,
•3 step process : SPOT model derived from the QAA (Lee et al., 2002)•1 : absorption (a) and backscattering (bb) terms are derived from Rrs•2 : The vertical diffuse attenuation (Kd) is derived from a and bb terms•3 : Depth is derived from Kd , bottom substrate albedo and optically deep water reflectance
•Model interest: All informations are derived from SPOT scene…except the water level
Bathymetric Application for pixels where the bottom substrate contribution is significant Bio optical Models to derive water column constituents concentration [C]
Semi-Analytical Model for coastal systems
Sylvain Capo,
Bathymetric inversion algorithm
In energetic areas, Remote sensing = Very High spatial ResolutionSPOT, FORMOSAT, PLEIADES, IKONOS
VHR (2.510m) frequency (~ 1 to 5-6 days) coverage (~ 10 à 60km)
Model ran on the 2007-09-13 SPOT and soundings data
Bathymetric inversion algorithm
Sylvain Capo,
For measured depth +4 to -6m
Relative error ~ 10%Absolute averaged difference ~ 0,9mRoot Mean Square Error ~ 1,2m
(Capo et al., In prep)
Morphological entities
Northern Channel (NC)•~ 0.8 km
Southern Channel (SC)•~0,8 km
Arguin Sand Bank (ASB)~2,15 km
Bernet Sand Bank (BSB)
Tourlinguet Sand Bank
Sylvain Capo,
Morphodynamic evolutionLarge view
Inlet Focus
Sylvain Capo,
Channels mobility
Tourlinguet Sand BankNorthern Channel
Arguin Sand Bank
Southern Channel
Sylvain Capo,
Measurements performed on a NS transect at the Inlet for the zeros m level
Channels mobility
Sylvain Capo,
~130m averaged ~ 6m/y
~170m averaged ~ 8m/y
Widening of the Tourlinguet Sand Bank>1000 m (~ 60m/year)Sand supply by the littoral driftsouthwards migration and anticlockwise rotation of the channel
South-easternMigration ArguinSB~800 mAveraged ~ 40m/year
Conclusions and perspectives
Sylvain Capo,
Conclusions and perspectives
Channel width stability
Sylvain Capo,
Increasing of the meanderingof the Bernet SBSouthwest migration ~410 m~20m/year
Conclusions and perspectives
Sylvain Capo,
PERSPECTIVES
1. Model improvementAtmospheric correction
Bottom substrate discretisation
Implementation of Bio-optical models for water constituants (TSM, Chloro, CDOM)
2. Morphodynamics
Wave forcing to understang the by-pass process
Sylvain Capo,
Thank you
Sylvain Capo,
SPOT SA inversion model
Sylvain Capo,
1st step: IOP
Synthetic data IOCCG Comparison with QAA model
RMSE divided by 2 compared to Lee et al., 2002 model QAA
SPOT CapoR² = 0.98rmse = 0.03
QAAR² = 0.92rmse = 0.06
SPOT CapoR² = 0.99rmse = 0.04
QAAR² = 0.93rmse = 0.07
Bathymetric inversion algorithm
2nd step: Vertical diffuse attenuation derivationSynthetic database IOCCG+ QAA SA model comparison
Better rmse by 25%
especially for high Kd values
rmse divided by 2 for Kd >0.2(more accurate for coastal environments)
SPOT CapoR² = 0.98rmse = 0.04
QAAR² = 0.96rmse = 0.05
Bathymetric inversion algorithm
Sous évaluation des zones profondes turbides (chenaux et panaches) Bonne reproduction par le modèle de la géomorphologie des zones peuprofondes (+2 à -6m)
profondeurs de sonde de 0 à 20m
Modèle semi‐analytique d’inversion
3eme étape: Depth
4eme étape: Nu données synthétiques de référence IOCCG + comparaison avec modèle QAA
Modèle semi‐analytique d’inversion
Erreur moyenne quadratique divisée par deux par rapport au modèle de Lee et al., 2002
SPOT CapoR² = 0.92rmse = 0.17
QAAR² = 0.87rmse = 0.31
4eme étape: dérivation de l’exposant Nu pour le spectre des bbp (λ) à partir de bbp (λ0)données synthétiques de référence IOCCG + comparaison avec modèle QAA
Modèle semi‐analytique d’inversion
4eme étape: dérivation de l’exposant Nu pour le spectre des bbp (λ) à partir de bbp (λ0)données synthétiques de référence IOCCG + comparaison avec modèle QAA
Modèle semi‐analytique d’inversion
Arcachon Inlet Evolution 1986-2011
Arcachon Inlet Evolution 1986-2011
Northern Channel (NC)• Double and large• ~ 1,7 km
1990-03-17
Southern Channel (SC)• Single• ~0,8 km
1990-03-17
Arguin Sand Bank (ASB)• channel• ~2,15 km
1990-03-17
Flood delta
1990-03-17
Water column
Bottom
Albert and Mobley, 2003 Radiative Transfert Model im:plemented in Hydriloght
SPOT Capo (In prep)
QAA Lee et al., 2002
m0 1+0.005θsun 1+0.005θsun
m1 3.84 4.18
m2 0.256 0.52
m3 10.83 10.8
2eme étape: Paramétrisation du Kdet comparaison avec données synthétiques de référence IOCCG + modèle QAA
Modèle semi‐analytique d’inversion
Rrs 4 bands
λ0 =550nm
rrs
λ0 =490nm
rrs+Kd
λ0 =490nm
Rrs 2 bandes
λ0 =545nm
(IOCCG, 2005)
QAA Lee et al., 2002 et 2005
Loisel Jamet2010Kd derived NN
Loisel JametNew+True Kd
SPOT QAACapo (In Prep.)
a(total) Relative_err 2,79% 10,95% 1,74% 1,86%RMSE 0.0603 0.0554 0.0098 0.0305
bb Relative_err 4,48% 10,41% 1,93% 1,22%RMSE 0.066548 0.0534 0.0117 0.0369
Kd Relative_err 6,30% 16,82% 0,49%RMSE 0.0545 0.148 0.0407
Model Performances compared to synthetic IOCCG dataset
En rouge: meilleur résultatEn Vert : meilleur résultat en n’utilisant que les réflectances
QAA Lee et al., 2010 SPOT Capo (In Prep.)rrs(λ)=Rrs(λ)/(0.52+1.7Rrs(λ)) rrs(λ)=Rrs(λ)/(0.52+1.7Rrs(λ))
u(λ)=
g0= 0.089 g1= 0.125
u(λ)=
g0= 0.088 g1= 0.163