monitoring china's coastal zones and adjacent seas under...
TRANSCRIPT
Monitoring China's Coastal Zones and Adjacent
Seas under Global Change by Satellite Data (10470)
Dr. Tingwei Cui(崔廷伟)
First Institute of Oceanography, SOA, Qingdao, China
Dr. Juergen Fischer
Institut fuer Weltraumwissenschaften, Freie Universitaet Berlin, Germany
Outline
General introduction
Objectives, Research Content, Expected outcome, Team, Data
Recent results
Global Evaluation of Satellite Ocean Color Products from Europe,
America and China: Toward a 15-yr merged global dataset
Monitoring coastline change in Guangxi province since 1973 by Landsat
and HJ-1 CCD images
Coastal wetlands classification by PROBA CHRIS image in Yellow
River Delta
Coastal shallow water depth inversion model of ZY-3 image
Future work
Motivation: China’s coastal zone is experiencing dramatic changes due to significant socio-economical effects and the intense human activities, and the resource and environment condition here is also susceptible to the global change.
Part I: General introduction
To disclose the spatio-temporal variation of land use/cover changes and wetlands along China coast as well as the causes and implications with respect to global change;
------continued work based on Dragon 2 (ID:5292)
To develop new methods to detect submerged macroalgae, river plume by synergistic use of optical, infrared and microwave satellite data;
To validate geographical products of satellite data along China coast to assess the uncertainties of these geophysical products and to offer useful suggestions for higher-order applications.
Objectives
Satellite monitoring and evaluation of land use/cover
changes (LUCC) along China coast
Satellite monitoring and evaluation of wetland along China
coast
Macroalgae bloom monitoring by synergistic use of satellite
data
River plume monitoring by satellite-derived suspended
sediment and sea surface salinity
Validation of satellite data along China coast
Research Content
Chinese Partners- 10
– Dr. Tingwei CUI (PI) ─ First Institute of Oceanography (FIO), State Oceanic Administration (SOA)
– Dr. Yi Ma ─ FIO, SOA, China – Dr. Guangbo Ren ─ FIO, SOA, China – Dr. Ping Qin ─ Ocean University of China (OUC), China – Jianhua Zhu ─ National Ocean Technology Center (NOTC), SOA, China – Huping Ye-NOTC, SOA, China – Bing Mu-OUC, China – Shaoqi Ni ─ FIO, SOA, China – Xiaomin Li ─ FIO, SOA, China – Xiaoqing-Cai─ FIO, SOA, China
European Partners- 3
– Dr. Juergen FISCHER(PI) ─ Institute for Space Sciences, Freie Universitat Berlin, Germany;
– Dr. David DOXARAN ─Laboratoire d'Océanographie de Villefranche, France; – Dr. Lena Kritten ─ Institute for Space Sciences, Freie Universitat Berlin,
Germany
Team
Expected outcome
•New method of monitoring macroalgae bloom by synergistic use of optical, infrared and microwave satellite data
•New method of monitoring river plume by satellite derived suspended sediment as well as synthetic sea surface salinity
•Annual changes of land use/land cover and wetlands along China coast as well as the causes and implications
•Macroalgae bloom dynamics in the Yellow Sea and East China Sea and annual variability
•River plume dynamics in the Yellow River, Yangtze River and Pearl River and the mechanism
•Uncertainties of MERIS, OLCI products along China coast
Satellite data
ESA ◦ MERIS, SAR, AATSR, OLCI (Sentinel-3), SPOT,……
CHINESE ◦ HJ, FY, ZY, HY,……
TPM ◦ ALOS, GOCI, TM,……
a). Analyze spectral characteristics of LUCC types
based on field work; HJ-1 CCD and MERIS images
b). Establish LUCC classification system
consider LUCC type changes in different geographical regions along China coast
c). Extract LUCC types information and generate thematic maps
develop semi-automatic information extraction method
d). Analyze driving factors of LUCC changes
policy, human activity, climate changes, sea level rise….
develop a model to make the forecast
Land use/cover Changes(LUCC)
a). Analyze HJ-1 CCD and MERIS imaging capabilities of the coastal
wetland
focusing on typical coastal wetland types: reed, spartina alterniflora,
mangrove, tamarix, sea grass,……
b). Extract coastal wetlands information and generate thematic maps
Human-Computer Interaction method, semi-automatic method
c). Analyze wetland landscape structure distribution and dynamics
as well as the driving factors
d). Evaluate wetland ecological quality
health and service function assessment
Wetland
a).Evaluate the effects of environmental conditions and satellite data processing methods on the detection of free floating macroalgae bloom
atmospheric condition, observation geometry, thin cloud, haze,…
atmospheric correction, NDVI method
b).Comparison of detection capabilities of free floating macroalgae bloom by different satellite data
optical, SAR and infrared images
focus on MERIS chlorophyll fluorescence measurements
c).Develop methods to detect submerged macroalgae by concurrent optical, SAR and infrared data
d).Analyze the dynamics from just the beginning to the collapse of the macroalgae bloom in the Yellow Sea
time series of satellite images, detect the bloom as early as possible
Macroalgae bloom
a). Develop regional algorithms to retrieve suspended sediment concentration
in situ observation; calibration and validation of the regional SPM algorithms
for ENVISAT MERIS and HJ-1 CCD
b).Develop regional algorithms to retrieve sea surface salinity(SSS) of the river
plume
in situ observation of SSS and ocean optical properties
calibrate and validate empirical and semi-analytical models to retrieve SSS by
spectral remote sensing reflectance (Rrs)
c).Analyze river plume dynamics by time series of satellite images
in terms of river discharge, hydrodynamic condition and wind.
River plume
a). Inter-comparison and cross-calibration of satellite radiance data
OLCI (Sentinel-3), VIIRS(Suomi NPP), MERSI (FY-3)
Based on L1b data
b). Validation of satellite data product along China coast
OLCI (Sentinel-3), MERIS (ENVISAT), MERSI (FY-3), VIIRS(Suomi NPP)
The Bohai Sea, Yellow Sea, East China Sea, South China Sea
Validation
Global Evaluation of Satellite Ocean Color Products from Europe,
America and China: Toward a 15-yr merged global dataset
Monitoring coastline change in Guangxi province since 1973 by
Landsat and HJ-1 CCD images
Coastal wetlands classification by PROBA CHRIS image in
Yellow River Delta
Coastal shallow water depth inversion model of ZY-3 image
Part II: Recent results
Global Evaluation of Satellite Ocean
Color Products from Europe, America
and China: Toward a 15-yr merged
global dataset
Background
A 15-yr long global ocean color dataset is to be produced
by merging major satellite data from Europe (MERIS),
America (SeaWiFS and MODIS) and China (MERSI,
COCTS).
The first step is to evaluate the accuracy of these satellite
data in the global scale and make a detailed comparison.
Major global ocean color sensors
SeaWiFS/Orbview-2
Bands Signal-to-Noise Purpose
402~422nm 499 CDOM
433~453nm 674 Chlorophyll
480~500nm 667 Pigment, Kd490
500~520nm 640 Chlorophyll
545~565nm 596 Pigment, optical property, SPM
660~680nm 442 Atmosphere correction, chlorophyll
745~785nm 455 Atmosphere correction, chlorophyll
845~885nm 467 Atmosphere correction, chlorophyll
August 1997~December 2010
8 bands from 412 to 865 nm
spatial resolution: 1 km
swath width: 1502km
MODIS/Terra (Aqua)
Bands Signal-to-Noise
405~420nm 880
438~448nm 838
483~493nm 802
526~536nm 754
546~556nm 750
662~672nm 910
673~683nm 1087
743~753nm 586
862 – 877nm 516
1999~ (Terra); 2002~(Aqua)
totally 36 bands, including 9 ocean color bands from
405 to 877 nm
spatial resolution: 1 km
swath width: 2330 km
MERIS/ENVISAT April 2002~ March 2012.
15 bands from 390 to 1040 nm
spatial resolution: 1.2 km/0.3 km (RR/FR)
swath width: 1150km
Central wavelength/nm Band width/nm Design objective 412.5 10 Yellow substance and detrital pigments
442.5 10 Chlorophyll absorption maximum
490 10 Chlorophyll other pigments
510 10 Suspended sediment, red tides
560 10 Chlorophyll absorption minmum
620 10 Suspended sediment
665 10 Chlorophyll absorption & fluo. reference
681.25 7.5 Chlorophyll fluorescence peak
708.75 10 Fluo. Reference, atmosphere corrections
753.75 7.5 Vegetation , cloud
760.625 3.75 O2-branch absorption band
778.75 15 Atmosphere corrections
865 20 Vegetation, water vapor reference
886 10 Atmosphere corrections
900 10 Water vapor, land
MERSI/FY-3 series
The second generation of China’s polar-orbit meteorological satellites.
FY-3A, FY-3B, FY-3C was launched in May 2008, Nov. 2010 and Sep.
2013, respectively.
Totally 20 bands, including 9 ocean color bands
Swath width: 2900 km.
MODIS MERIS
SeaWiFS MERSI
8.43% 9.68%
6.25% 11.75%
Evaluation of Rrs products Satellite v.s. In-situ (SeaBASS)
-150° E -100° E -50° E 0° E 50° E 100° E 150° E
-50° N
0° N
50° N
MODIS
MERIS
SeaWiFS
Match-ups: ±3 hours, 3×3pixels
SeaWiFS
412nm 443nm 490nm
530nm 555nm 670nm
MODIS
412nm 443nm 488nm
531nm 547nm 667nm
MERIS
413nm 443nm 490nm
510nm 560nm 665nm
N R2 RMSE(sr-1) APD( %)
MODIS
Rrs412 1691 0.76 0.0015 30.05
Rrs443 2109 0.85 0.0011 16.73
Rrs488 1809 0.91 0.0011 14.97
Rrs531 713 0.92 0.0012 13.87
Rrs547 890 0.93 0.0012 13.42
Rrs667 345 0.92 0.0002 35.25
MERIS
Rrs413 974 0.67 0.0020 41.97
Rrs443 882 0.79 0.0014 23.49
Rrs490 1090 0.86 0.0014 18.17
Rrs510 306 0.85 0.0011 14.41
Rrs560 205 0.91 0.0011 16.22
Rrs665 234 0.92 0.0004 44.40
SeaWiFS
Rrs412 1204 0.77 0.0018 28.31
Rrs443 1273 0.78 0.0014 18.76
Rrs490 1456 0.81 0.0013 15.53
Rrs510 769 0.85 0.0010 14.71
Rrs555 1157 0.85 0.0013 15.19
Rrs670 525 0.84 0.0006 32.15
Satellite VS. Satellite MERSI/FY-3 v.s. MODIS/AQUA
MERSI daily products in 2011 from FY-3A and FY-3B were
compared with concurrent MODIS.
Close correlation and significant bias (>52%) were found.
443nm 490nm 565nm
FY-3B
FY-3A R=0.487 R=0.370 R=0.557
R=0.806 R=0.580 R=0.547
Satellite Rrs from SeaWiFS, MODIS and MERIS show comparable accuracy.
◦ Rrs uncertainties have the same spectral feature for, e.g. higher (28%~45%) in blue
and red bands, and lower in green bands (13%~23%).
◦ Performance of MODIS and SeaWiFS is slightly better than MERIS, especially for the
blue (412nm) and red band (670nm).
MERSI/FY-3 Rrs data needs to be improved before being merged.
Summary
Evaluation of CHL products
In situ data:
from Ocean Biology Processing Group (OBPG)
period: 2000~2013
47944 observation data (8:00-16:00) were selected
Satellite v. s. In-situ
Mach-ups R2* APD-median(%) RMSE*
MERIS 655 0.81 38.80 0.33
MODIS 688 0.82 38.25 0.32
SeaWiFS 1654 0.85 36.59 0.29
FY-3A CHL 172 0.27 129.49 0.70
FY-3B CHL 111 0.43 69.15 0.38
* in log10 scale
MERIS
MODIS
SeaWiFS MERSI FY-3A MERSI FY-3B
Satellite v. s. Satellite
t
m ji
mji
jimean
STDCV
1 ,
,,
,
Based on the daily satellite chlorophyll in 2010, the consistency of
MERIS, MODIS and SeaWiFS products were analyzed with the
variation of coefficient CV.
MERIS VS. SeaWiFS
MERIS VS. MODIS
MODIS VS. SeaWiFS
For most regions, CV of MERIS, MODIS and
SeaWiFS chlorophyll a content products is less
than 0.2.
For the clear water of sub-tropic where the
chlorophyll content is low all the year round, CV
is higher.
MERSI/FY-3 VS. MODIS/AQUA
R* RMSE* Bias* APD-median FY-3A vs. MODIS 0.84 0.48 0.33 43.5% FY-3B vs. MODIS 0.84 0.49 0.31 42.0% * in log10 scale
FY-3A FY-3B
Satellite Chl from SeaWiFS, MODIS and MERIS show comparable accuracy
(35%~40%).
MERSI/FY-3 Chl data needs to be improved before being merged.
Summary
Improvement of MERSI/FY-3 CHL
products
Step 1: Evaluate the accuracy of MERSI-derived ocean
color indices (e.g. Rrs(443)/Rrs(565), Rrs(490)/Rrs(565), CI)
based on the MODIS counterparts.
Step 2: Modify the MERSI-derived indices based on their
regression relationships with MODIS counterparts.
Step 3: Retrieve chlorophyll concentration using the
modified MERSI-derived.
R=0.869
RMSE=2.54
Bias=-2.03
APD=45%
R=0.809
RMSE=1.16
Bias=-0.9
APD=28%
R=0.933
RMSE=4.33*10-3
Bias=1.71*10-3
APD=144%
R=0.628
RMSE=4.78
Bias=-4.43
APD=63%
R=0.411
RMSE=1.94
Bias=-1.79
APD=42%
R=0.795
RMSE=5.62*10-3
Bias=3.44*0-3
APD=112%
Y=2.6311*X-1.2292 Y=1.7953*X-0.6672 Y=0.3109*X-0.0011
Y=2.4814*X+0.8442 Y=1.5584*X+0.4615 Y=0.2101X-0.0027
FY-3B
FY-3A
Evaluation and modification of MERSI-derived ocean color indices
Chlorophyll retrieval based on MERSI-derived indices
and the new algorithm (Hu et al., 2012)
R* RMSE* Bias* APD-median(%) FY-3A v. s. MODIS (CHL) 0.90 0.22 -0.085 10.21 FY-3B v. s. MODIS (CHL) 0.74 0.17 -0.0061 8.16
* In log10 scale.
MERSI
standard
After
Modification
FY-3A
FY-3A
FY-3B
FY-3B
MODIS
Comparison before and after modification
before
after
before
after
Merged by MODIS and
MERIS (2011/01/01), with
13.6% valid coverage
Merged by MODIS, MERIS
and FY-3 (2011/01/01), with
24.6% valid coverage
Benefit from merging FY-3 MERSI
Study area
- Important coastal zone for mangrove
protection in China
- Neighboring Vietnam
- Guangxi Beibu gulf economic zone
Satellite images
Acquisition date Imge type Number Spatial resolution (m)
1973-12 Landsat-1 MSS 2 60
1977-2/1978-12 Landsat-3 MSS 2 60
1990-12/1991-11 Landsat-5 TM 2 30
2000-9 Landsat-7 ETM+ 2 30
2013-9 HJ-1A 1 30
Coastline interpretation
•Bedrock coastline and its location
- The shape of coastline is
irregular
- The color of the rock tidal flat
is brighter and usually
without vegetation
•Sandy coastline and its location
- The tidal flat is narrow and
long
- The shape of coastline is
smooth
- The color of the tidal flat is
brighter and usually without
vegetation
•Muddy coastline and its location
- The tidal flat is wide
- The shape of coastline is
smooth and irregular
- The color of the tidal flat is
darker
- The coastline is located at the
seaside of the vegetation
•Artificial coastline and its location
- The shape of coastline is
straight
- The color of the land-use types
landside is brighter
- The coastline is located at the
seaside of the artificial
constructions
Coastline monitoring results since 1973
1973 1978 1990
2000 2013
The length and ratio of different coastline types in different times
since 1973 (km)
Bedrock Artificial Sandy Muddy Overall
1973 136.52 224.69 167.73 757.49
1376.61 9.92% 16.32% 12.18% 55.03%
1978 106.28 221.97 168.70 779.05
1360.60 7.81% 16.31% 12.40% 57.26%
1990 92.45 575.98 144.75 474.88
1357.26 6.81% 42.44% 10.66% 34.99%
2000 68.28 616.08 137.07 388.90
1367.38 4.99% 45.06% 10.02% 28.44%
2013 32.81 919.88 97.02 85.52
1378.78 2.38% 66.72% 7.04% 6.20%
The length and ratio of Artificial coastline increased from 1997 to 2013.
The other three kind of coastline decreased.
1973-1978 1978-1990
1990-2000 2000-2013
The changes in the Fangchenggang city
0
500
1000
1500
2000
2500
3000
3500
1973to1978 1978to1990 1990to2000 2000to2013
Are
a(h
a)
In each period, the coastline of Fangchenggang city moved seaward, generating a
changing area from 198.07hm2 to 2950.64 hm2.
The changes in the Qinzhou city
0
200
400
600
800
1000
1200
1400
1600
1800
1973to1978 1978to1990 1990to2000 2000to2013
Are
a(h
a)
In each period, the coastline of
Qinzhou city moved seaward,
generating a changing area from
167.5hm2 to 1569.65 hm2.
Summary
The coastline changes in Guangxi province since
1973 was extracted by Landsat and HJ-1 CCD
images.
The length and ratio of Artificial coastline increased
while the other three kind of coastline decreased.
The coastline of Fangchenggang and Qinzhou city
increased regularly.
Study area
National Nature Reserve of the Yellow River Delta
Data sets
• Remote sensing data - PROBA CHRIS
Acquisition date 2012-06-01
Spatial resolution 17m
Spectral bands 400~1050nm
Bands
Central
wavelength
(nm)
Band width
(nm)
Band 1 411.3 10.9
Band 2 443.6 10.6
Band 3 491.8 11.7
Band 4 511.5 12.9
Band 5 532.0 11.6
Band 6 563.6 13.8
Band 7 576.1 11.0
Band 8 593.2 16.0
Band 9 625.3 13.7
Band 10 654.3 15.4
Band 11 672.9 10.9
Band 12 684.1 11.4
Band 13 692.7 5.8
Band 14 710.7 18.5
Band 15 760.4 14.2
Band 16 786.1 22.8
Band 17 878.6 27.6
Band 18 1026.7 44.1
• Field spectral data
ASD handheld 2:
- Spectral wavelength range:300-1100nm
- Spectral resolution:1nm
- Viewing angle:25°
Method
Hyperspectral feature
extraction based on
association mining
Hyperspectral remote
sensing image
Hyperspectral
feature sets of the
typical vegetation
Classification method
(decision tree)
field spectrum
feature analysis
field spectrum
data sets
Typical wetlands
classification on hyperspectral
remote sensing image
Classification results
Overall accuracy: 82.26% Kappa coefficient: 0.79
Comparison with SVM
Class
Decision tree SVM
Producer accuracy
(Percent)
User accuracy
(Percent)
Producer accuracy
(Percent)
User accuracy
(Percent)
Reeds 78.97 94.72 62.21 99.43
Nude beach 89.9 90.8 78.08 95.88
Clear water 83.74 93.5 85.84 94.31
Muddy water 79.92 87.01 87.23 74.11
Suaeda salsa 83.89 74.2 59.66 73.47
Tmarix chinensis 75.82 70.45 58.21 49.1
Spartina anglica 79.46 28.45 97.62 17.85
Bare land 46.01 26.15 74.3 9.39
Overall acuuracy 82.26% 70.90%
Kappa coefficient 0.79 0.66
Summary
Wetlands in Yellow River Delta were classified by
PROBA CHRIS image based on decision tree
classification method.
The classification results were validated by in-situ
data, which showed that Overall accuracy was
82.26% and the kappa coefficient was 0.79.
The decision tree method was better than SVM.
Yongxing island
Study area
The water around Yongxing island is very clean, which is applicable to water
depth inversion research.
Data sets
•Satellite remote sensing image-ZY-3(资源三号)
•Water depth data-Marine chart
Acquisition date 2012-12-19
Spatial resolution 2.1m
Spectral bands 4 bands:
485nm(Blue), 555nm(Green),
660nm(Red), 830nm(NIR)
Measure date 1971
Scale 10000
Method
Linear regression model (LRM)
-Single band -Dual bands
Water depth for regression (N=122) Water depth for validation (N=60)
Water depth inversion results
•Single band inversion result-validation table
Water
depth
region
Blue band Green band Red band
Absolute(m) relative Absolute(m) relative Absolute(m) relative
0 ~ 2 4.60 1162.6% 2.54 494.7% 3.14 816.1%
2 ~ 5 3.03 101.7% 2.98 98.5% 3.06 96.5%
5 ~ 10 2.40 29.6% 2.44 32.5% 1.79 23.9%
10 ~ 20 7.34 54.3% 4.84 35.2% 7.13 52.7%
> 20 4.37 60.0% 3.47 53.2% 4.10 56.0%
All 4.43 365.3% 3.22 175.4% 3.84 266.5%
Overall, the inversion results of green band was the best.
•Dual bands inversion result-validation table
Water
depth
region
Blue-Green bands Blue-Red band Green-Red band
Absolute
(m) relative
Absolute
(m) relative
Absolute
(m) relative
0 ~ 2 1.02 210.0% 3.31 808.3% 3.07 611.4%
2 ~ 5 2.31 70.1% 3.14 102.7% 2.94 97.2%
5 ~ 10 2.32 31.2% 2.02 26.6% 2.15 28.0%
10 ~ 20 3.55 25.1% 7.08 52.5% 4.16 30.0%
> 20 2.77 40.6% 4.19 58.6% 3.11 49.3%
All 2.28 87.5% 3.94 266.2% 3.10 205.0%
Overall, the inversion results of green band was the best.
Different coastal shallow water depth inversion
model was established and validated.
The inversion results of green band was the best.
Summary
Future work
Analyze driving factors of LUCC and wetlands changes.
Develop methods to detect submerged macroalgae by multi-
source satellite data.
Analyze river plume dynamics by time series of satellite
images.
Thanks for your attention!