Fernando Gilbes1, Vilmaliz Rodriguez, Jose Martinez, and Eidalia ...

Fernando Gilbes 1 , Vilmaliz Rodriguez, Jose Martinez,
and Eidalia Gonzalez
Geological and Environmental Remote Sensing Lab
Department of Geology
University of Puerto Rico at Mayagüez
1
gilbes@cacique.uprm.edu

MAYAGUEZ BAY
Deep and Clear
Waters
Añasco River
Sewage Outfall
Yaguez River
CLIMATOLOGY OF
AÑASCO RIVER DISCHARGE
(20 years)
Shallow and Clear
Waters with
Coral Reefs
Guanajibo River
DRY SEASON
RAINY SEASON

TWO SEASONS IS EQUAL TO
TWO OPTICAL CONDITIONS

MULTI‐SPECTRAL AIRBORNE SENSORS
AOCI
30 m and 10 bands
ATLAS
10 m and 15 bands

MULTI‐SPECTRAL SATELLITE SENSORS
THEMATIC MAPPER
30 m and 7 bands
IKONOS
1 m and 4 bands

MOTIVATION FOR OUR WORK
• Can we use remote sensing to measure suspended
sediments in Mayaguez Bay?
• If so, what sensor and algorithm are more accurate?
• What are the sources of error in measuring suspended
sediments with remote sensors in Mayaguez Bay?
• Can we develop a continuous monitoring system of
suspended sediments in Mayaguez Bay, and therefore
over Puerto Rico?

REMOTE SIGNAL FROM THE OCEAN

SEPARATING THE REMOTE SIGNAL
What we Measure
• Water Optical Properties
• Bottom Reflectance
FromNEMO Overview
Nemo.nrl.navy.gov

REFLECTANCE VS. SEDIMENTS
• Reflectance
• Previous findings
• Increase in SS = Increase in
reflectance
• Red & Near IR wavelengths
show lower slopes in high
SS
• Mayagüez Bay
• Concentration too low to
provide strong response
From Witte et al. (1982)

MAIN OBJECTIVE
To validate the accuracy of several
remote sensors and algorithms for
measuring suspended sediments
(SS) in Mayagüez Bay, PuertoRico.

TECHNICAL APPROACH
Field
Work
Testing
MODIS
Testing
AVIRIS
Seasonal
samplings
Testing the
Miller & Mckee
algorithm using
250 m bands
Develop a local
algorithm with
IR bands
Suspended
sediments by
the filter
method
Develop local
algorithm using
250 m bands
Geobio‐optical
Properties

20 SAMPLING DATES
• April 24-26, 2001
• October 2-4, 2001
• February 26-28, 2002
• August 20-22, 2002
• February 25-27, 2003
• October 7-9, 2003
• January 12-14, 2004
• February 12, 2004
• August 19, 2004
• March 10, 2005
• July 19, 2005
• August 17, 2005
• September 20, 2005
• October 19, 2005
• December 6, 2005
• March 8, 2006
• April 21, 2006
• September 26, 2006
• October 26, 2006
• May 1-2, 2007

GRID OF
SAMPLING
STATIONS

Moderate Resolution Imaging
Spectroradiometer (MODIS)
EOS AM
~10:30 AM
LOCAL TIME
EOS PM
~1:30 PM
LOCAL TIME

MODIS SPATIAL RESOLUTION
Bands 1-2 = 250 m
Bands 3-7 = 500 m
Bands 8-36 = 1000 m

MODIS SPECTRAL RESOLUTION
Primary Use Band Bandwidth
Primary Use Band Bandwidth
Land/Cloud/Aerosols
Boundaries
1 620 - 670
2 841 - 876
SS
Surface/Cloud
Temperature
20 3.660 - 3.840
21 3.929 - 3.989
Land/Cloud/Aerosols
Properties
3 459 - 479
4 545 - 565
Chl-a
22 3.929 - 3.989
23 4.020 - 4.080
5 1230 - 1250
6 1628 - 1652
Atmospheric
Temperature
24 4.433 - 4.498
25 4.482 - 4.549
Ocean Color/
Phytoplankton/
Biogeochemistry
7 2105 - 2155
8 405 - 420
9 438 - 448
10 483 - 493
11 526 - 536
12 546 - 556
13 662 - 672
NASA
Chl-a
Cirrus Clouds 26 1.360 - 1.390
Water Vapor
27 6.535 - 6.895
28 7.175 - 7.475
Cloud Properties 29 8.400 - 8.700
Ozone 30 9.580 - 9.880
Surface/Cloud 31 10.780 - 11.280
Temperature
32 11.770 - 12.270
14 673 - 683
15 743 - 753
Cloud Top
Altitude
33 13.185 - 13.485
34 13.485 - 13.785
16 862 - 877
35 13.785 - 14.085
Atmospheric
Water Vapor
17 890 - 920
18 931 - 941
36 14.085 - 14.385
19 915 - 965

SUSPENDED SEDIMENTS
• Miller and McKee (2004)
• Study of northern Gulf of
Mexico‐Mississippi Delta
• Linear relationship: in situ
Suspended Matter vs.
MODIS Terra Band 1 (620 –
670 nm)
• Provided evidence of
sediment transport
Calibrated images of Mississippi River delta derived
from MODIS Terra Band 1 (Miller & McKee, 2004)

Miller and McKee (2004) Algorithm

RESULTS IN MAYAGUEZ BAY

2 ND APPROACH‐A NEW ALGORITHM
• Image preprocessing
with ENVI (Environment
for Visualization of
Images)
• Geo‐referencing: UTM 19
(Datum NAD 83)
• 17 GPS points
corroborated
• Atmospheric correction:
• Dark Subtract
GPS points used in geo-reference validation

FIELD SS AND MODIS‐B1 %REFLECTANCE
30
25
Suspended Sediment Concentration (mg/l)
20
15
10
5
y = 105.19x + 2.6373
R 2 = 0.1443
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
MODIS Terra Band 1 Reflectance (%)
All Stations: (2001 – 2006)

FIELD SS AND MODIS‐B1 %REFLECTANCE
30
30
25
25
Suspended Sediment Concentration (mg/l)
20
15
10
5
y = 80.703x + 3.614
R 2 = 0.0695
Suspended Sediment Concentration (mg/l)
20
15
10
5
y = 136.58x + 0.8668
R2 = 0.2788
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
MODIS Band 1 Reflectance (%)
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
MODIS Terra Band 1 Reflectance (%)
Dry Season (2001 – 2006) Rainy Season (2001 – 2006)

FIELD SS AND MODIS‐B1 %REFLECTANCE
30
16
14
25
Suspended Sediment Concentration (mg/l)
20
15
10
5
y = 63.781x + 5.962
R 2 = 0.0473
Suspended Sediment Concentration (mg/l)
12
10
8
6
4
2
y = 55.796x + 3.2736
R 2 = 0.0468
0
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
MODIS Terra Band 1 Reflectance (%)
MODIS Terra Band 1 Reflectance (%)
In-shore Stations (2001 – 2006) Off-shore Stations (2001 – 2006)

A NOVEL APPROACH
SS = 337.26 *(Band 1) + 854.12 *(Band 2)
40.00
35.00
30.00
25.00
20.00
15.00
10.00
5.00
0.00
Suspended Sediments conc (mg/l)
A1
A2
G2
A1
A2
AAA1
Y1
G1
G2
A1
A2
AAA1
G2
A1
A2
AAA1
Y1
A1
A2
Y1
G1
G2
A1
A2
AAA1
Y1
G2
A1
A2
AAA1
G2
A1
A2
AAA1
Y1
G1
July 17
05
August 17 05 September
20 05
October 19
05
December 6 05 April 21 06 September
26 06
TSS (Observed)
TSS (Estimated)
October 26 06
30.00
25.00
20.00
15.00
10.00
5.00
0.00
Observed
Estimated
A1
A2
AAA
Y1
G1
G2
S05
S09
S11
S13
S15
S17
S19
S21
S23
S01
S02
S03
S13
S14
S15
S21
S22
S23
S01
S04
S05
S07
S17
S19
S21
S23
Suspended Sediments conc (mg/l)
y = 0.4033 * band 1 – 0.0006
SS (mg/l) = 452.41 * y + 2.9603
Mar-06 Jan-13-04 Jan-14-04 Feb-12-04 Oct-07-03 Feb-27-03

SS WITH NEW ALGORITHM
October 7, 2003 January 14, 2004

AVIRIS
AIRBORNE VISIBLE/INFRARED
IMAGING SPECTROMETER
‣ 224 Spectral Bands
‣ Range from 400 to 2500 nm
‣ Strip lines of 11 km wide
‣ Spatial resolution from 4 m to 20 m
(Puerto Rico mission was ~17 m)

AVIRIS MISSION OVER PUERTO RICO
AUGUST 19, 2004

AVIRIS IMAGES USED IN THIS STUDY
TO ESTIMATE SUSPENDED SEDIMENTS

Suspended Sediments measured
during AVIRIS overflight
18
16.7
Suspended Sediments (mg/l)
16
14
12
10
8
6
15.5
7.2
?
5.1
5.5
4
2
1.4
2.4
3.0
1.7
1.1
0
S01 S02 S04 S05 S07 S13 S15 S21 S23 S24
Stations

Anasco River discharge during August 2004

IMAGE PROCESSING
Raw Image
from NASA
Atmospheric
Correction
using ACORN
ENVI
Extraction of
Reflectance
from stations
20 λ’s
Application of
site-specific
algorithm to
the image
Development of
the algorithm
Correlation of image
data with field data
Masking
of
bad pixels
Mosaicking
Final Image
of
Sediments

Relative Reflectance as measured
with AVIRIS
2000
1500
Añasco
1000
Guanajibo
500
Yaguez
0
404.0
432.0
463.0
493.0
524.0
554.0
585.0
616.0
646.0
677.0
707.0
738.0
769.0
797.0
828.0
858.0
889.0
919.0
950.0
981.0

Testing different bands

Suggested Algorithm to estimate suspended
sediments in Mayaguez Bay using AVIRIS
Concentration of Suspended Sediments (mg/l) = a (x) + b
Where, a = 0.0829
x = AVIRIS Reflectance at 777 nm
b = 0.0325

SUSPENDED SEDIMENTS AS DETERMINED WITH AVIRIS

CONCLUSIONS
• Image processing and analyses clearly demonstrated that
MODIS is not the most appropriate ocean color sensor for
Mayagüez Bay.
• Another sensor with better temporal, spatial, and spectral
resolutions is still needed for the estimation of SS in
tropical coastal waters.
• However, MODIS Band 1 gives good qualitative results of SS
when a site‐specific algorithm is applied in Mayagüez Bay.
• Future work with MODIS in this region must include (but
not limited):
• Improve the atmospheric correction
• Consider other sources of error like bottom and land signal
• Mineral composition and grain characteristics of SS

CONCLUSIONS
• AVIRIS proved to be a good sensor for the estimation of
suspended sediments in Mayaguez Bay.
• The applied techniques produced an algorithm that uses
the 777 nm band with a R 2 of 0.82.
• Some fine tuning is necessary along the image processing
in order to improve the estimations. We are waiting for reprocessed
images (by NASA) in order to repeat the work.
• In order to monitor SS in Mayaguez Bay using remote
sensing we still need to understand their relationship with
the reflectance signal.