Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
vito MUMM
B r o c k m a n n C o n s u l tTel: + 49 4152 889300 • E-Mail: [email protected]
t<
Validation SIMEC adjacency correction forCoastal and Inland Waters
Sindy Sterckx1, Els Knaeps1, Susanne Kratzer2 4, Kevin Ruddick3, Ana Ruescas41 Flemish Institute for Technological Research (VITO), Belgium
2 Department of Systems Ecology, Stockholm University, Sweden 3 Royal Belgian Institute for Natural Sciences (RBINS), Dept VI Management of the North Sea Mathematical Models (MUMM), Belgium
4 Brockmann Consult, GermanyIntroductionSeveral new satellites such as Sentinel-2, Sentinel-3 and the hyperspectral satellites EnMAP and PRISMA will be launched in the near future. These EO sensors will provide a wealth of new data at increased spatial, spectral and temporal resolutions. Although not all conceived as being ocean colour missions, the inland and coastal water community could benefit considerably from these new data sources. A higher spatial resolution extends the existing monitoring efforts to cover even the first nautical mile from coast where the Water Framework Directive (WFD) is still in force or to lakes which are small and have irregular shapes and can not be monitored with the existing missions. However for these nearshore coastal and inland waters adjacency effects complicate the atmospheric correction process. Light reflected from the nearby land can be forward scattered by the atmosphere into the sensor’s field of view. This causes a “blurring” of the signal and the effect, generally known as adjacency, background or environment effect, modifies the spectral signature of the observed pixel. Here we present SIMEC (SIMilarity Environment Correction), a new approach for the correction of adjacency effects.
SIMECThe SIMEC correction method was first proposed by Sterckx et al. (2010) for the correction of airborne hyperspectral imagery. The correction algorithm estimates the contribution of the background radiance based on the correspondence with the NIR similarity spectrum (Ruddick et al., 2006). A key aspect of the method is that no assumptions have to be made on the NIR albedo.
Validation approachSIMEC is applied to correct the TOA radiance signal for adjacency effects for a series of MERIS FR data covering the Belgian North Sea coastal waters, the turbid Scheldt estuary and lake Vänern in Sweden, the third largest lake in Europe. Next, the standard MERIS processor (MEGS) is applied using the ODESA (Optical Data processor of ESA) software, in order to retrieve the water reflectance and to compare with in-situ measured water reflectance.
76
5©oe= 4
âJ 0,6
Water pixel influenced by adjacency effects
0,5
3 NIR similarity
2NIR similarity +/- 1 stdev
10
650 700 750 800 850 900
4 >O
Cs
Bc/>
C(Su
% *©D*1/1©
B<
Wavelength (nm)
-
i __________________
^ ^ 1 iRHO_WN
SIMEC
r
i TOA LIB r ODESA-MEGS 8.1 atm. cor.
Results and discussion i
RHO_WN_noSIMEC
r
The normalized water reflectance retrieved from ODESA-MEGS processing on MERIS FR images with and without first applying SIMEC are compared to the in-situ measured normalized water reflectance. For several sampling stations the MERIS retrieved normalized water reflectance is strongly underestimated without SIMEC preprocessing with often negative water reflectance values for the first two MERIS bands. The correspondence between MERIS retrieved normalized water reflectance and in-situ measured normalized water are quantified on the basis of the Root Mean Square Error (RMSE) and the correlation coefficient (R2). A significant decrease in RMSE and increase in R2 is observed for several stations after SIMEC pre-processing.
2 0 0 8 0 7 2 9 V I ODESA_RHO_WN_noSIMEC 2 0 0 8 0 7 2 9 * V I ODESA_RHO_WN_noSIMEC
ODESA RHO WN SIMECODESA_RHO_WN_SIMEC
Wavelength (nm) Wavelength (nm)
2 0 0 8 0 9 0 3 V I2 0 0 8 0 8 2 2 V I ODESA_RHO_WN_noSIMEC ODESA_RHO_WN_noSIMEC
ODE5A_RHO_WN_SIMEC ODESA_RHO_WN_SIMEC
Wavelength (nm) Wavelength (nm)
2 0 0 8 0 9 2 9 V I ODESA_RHO_WN_noSlMEC
ODESA_RHO_WN_SlMEC
ODESA_RHO_WN_noSIMEC2 0 0 9 0 6 2 9 V I
ODESA RHO WN SIMEC
Wavelength (nm) Wavelength (nm)
ODESA RHO WN noSIMEC2 0 0 8 1 0 0 4 V I 2 0 1 1 0 8 0 2 V I ODESA_RHO_WN_noSIMEC
ODESA_RHO_WN_SI M ECODESA RHO WN SIMEC
Wavelength (nm) Wavelength (nm)
ODESA_RHO_WN_noSIMEC2 0 1 1 0 8 0 3 V IODESA_RHO_WN_SlMEC0.008
^ -in -s itu0.006
0.004
400 450 550 6ÖÏ 700 300 850 900
5 -0.002
-0.004
-0.006
-0.008
-0,01Wavelength (nm)
0.012 0 1 1 0 8 0 3 V 3 -»-ODESA RHO WN noSIMEC
ODESA_RHO_WN_SIMEC
-♦-in-situ0.005
400 550 600 800 850
-0.01
-0.015Wavelength (nm)
8°0'0"E 12°0'0"E 16°0'0"E
V3V2V1
54°0'0"N- -54°0'0"N
Image date Sampling Station
R : (NoSIMEC)R : (SIMEC)RMSE (NoSIMEC) RMSE (SIMEC)
% increase R:% decrease RMSE
20030423N I
0.88070.88200.00700.0069
0.150.63
20030423N 2
0.91610.91570.00380.0039
-0.04-0.61
20030616N 3
0.84810.84810.01350.0135
0.000.00
20030806N4
0.57520.91650.01680.0060
37.24181.22
10030806N5
0.27330.62110.01220.0062
55.9997.96
20060713N6
0.91550.93550.01040.0088
2.140.03
20090616N7
0.92490.99260.00750.0029
6.82153.88
20090616SI
0.76680.96310.02260.0072
Quality indicators for spectral correspondence between in-situ normalized water reflectance and MERIS MEGS retrieved normalized water reflectance with and without SIMEC pre-processing.
Image date 200S0729 200S0729* 200SOS22 200S0903 2 00 SO 9 29 20110S03 20090629 20110S02 2 0110 SO 3 20110S03Sampling Station V I V I V I V I V I V I V I V I V2 V3
R1 (No SIMEC) 0.4176 0.5905 0.74S9 0.9459 0.7565 0.3901 0.S656 0.5036 0.3 9 04 0.9267R1 (SIMEC) 0.62S3 0.S559 0.S547 0.9S2S 0.S492 0.74 IS 0.6SS1 0.94S1 0.7260 0.94S4RMSE (NoSIMEC) 0.0123 0.00S2 0.0040 0.0032 0.0079 0.0054 0.00 2 6 0.0055 0.0052 0.00 6 9RMSE (SIMEC) 0.0076 0.0039 0.0039 0.0017 0.0054 0.0043 0.0054 0.0010 0.003S 0.003S
% increase R- 33.55 31.01 12.37 3.76 10.91 47.41 -25.79 46.SS 46.23 229% decrease RMSE 62.23 111.90 3.01 SS.76 45.79 26.65 -52.23 436.40 3S.12 7S.09
*after removal o f 2 outlier spectra
ConclusionsA new approach for the correction of adjacency effects which was originally developed for high resolution hyperspectral airborne data was presented in this poster. SIMEC is a sensor-generic approach and can therefore be applied to future sensors such as Sentinel-2 MSI, Sentinel-3 OLCI and EnMAP/PRISMA data. The performance of SIMEC was tested on MERIS FR images acquired over coastal areas, estuaries and lakes. SIMEC had a positive or neutral effect on the retrieved water reflectance calculated with the MERIS MEGS processor. A decrease in the RMSE up to 400 % was observed after SIMEC pre-processing.
VITO NV Acknowledgement :Boeretang 200 — 2400 MOL — BELGIUM — Tel. + 32 1433 55 11 —Fax+ 32 1433 55 99 — [email protected] — www.vlto.be This work was funded in the framework of the INTERREG 2Seas ISECA project
2°50'0"E y 0 '0 "E 3°10'0"E 3o20TrE 3°30'0"E 3°40'0ME 3°50'0"E
B e l g i u m
1-----------------------1-----------------------1-----------------------1-----------------------1-----------------------1-----------------------r 12°50’0”E 3°0'0"E 3°10’0"E 3o30'0"E 3°40'0"E 3°50'0"E
51°20'0"N-
51®10'0"N-
°30'0"N
°20'0"N
°10'0"N
ODESA_RHO_WN_noSIMEC2 0 0 3 0 4 2 3 N l O DESA_R HO_WN_noS IM E C 2 0 0 3 0 4 2 3 N 2ODESA RHO WN SIMECODESA_RHO_WN_SIMEC
Wavelength (nm) Wavelength (nm)
2 0 0 3 0 8 0 6 N 42 0 0 3 0 6 1 6 N 3 ODESA_RHO_WN_noS!MEC ODESA RHO W N noSIMECODESA RHO WN SIMEC ODESA_RHO_WN_SIMEC
Wavelength (nm) Wavelength (nm)
ODESA_RHO_WN_noSIMEC2 0 0 3 0 8 0 6 N 5 -•-ODESA RHO WN noSIMEC2 0 0 6 0 7 1 3 N 6ODESA RHO W N SIMEC ODESA RHO_WN SIMEC
Wavelength (nm) Wavelength (nm)
2 0 0 9 0 6 1 6 S I ODESA_RHO_WN_noSlMEC
O DESA_RHO_WN_SI M EC
2 0 0 9 0 6 1 6 N 7ODESA_RHO_WN_noSIMEC
* ODESA RHO W N SIMEC
-0.01
Wavelength (nm)
-0,03
-0.04Wavelength (nm)
0.008
0.006
0.004
0.002
0
-0.002
Wavelength (nm)
2 0 1 1 0 8 0 3 V 2 O DE5A_RHO_WN_noSI M EC
O DESA_RHO_WN_Sl MEC
in-situ