chapter 2 monitoring of coastal environment using satellite images in the united arab emirates

of 12 /12
Monitoring of Coastal Environment Using Satellite Images in the United Arab Emirates Mohammad Yousif Al Jawdar ERWDA, P. O. Box 45553, Abu Dhabi, United Arab Emirates Manabu Shiobara Japan Oil Development Co., Ltd., 1-21-2, Shinkawa, Chuo-ku, Tokyo 104-0033, Japan Takumi Onuma JGI, Inc., 1-5-21, Otsuka, Bunkyo-ku, Tokyo 112-0012, Japan ABSTRACT The diversity of marine wildlife along the coast of the United Arab Emirates is well known as well as the offshore giant oilfields. The protection of sustainable marine environment is the first priority to keep the biodiversity in the region and the Environmental Research and Wildlife Development Agency (ERWDA) has been paying attention on the preservation of coastal environment since it was established in 1996. It is widely accepted that long-term monitoring is mandatory for environmental hazard and satellite images with intermediate spatial resolution covering the broad region are suitable for this purpose. This study is based on such a viewpoint to contribute to the monitoring of environmental change along the coast of Abu Dhabi Emirate. The methodology of producing bottom type maps and vegetation distribution maps from satellite image data are presented. The operational atmospheric correction algorithm to minimize air-water interface was designed and applied to Terra ASTER data to produce Bottom Index images, which were compared with the field survey results to convert the index value to the bottom type. Bottom index value of 2.5 was deemed to discriminate sea grass and seaweed from sand bottom. The distribution of onshore vegetation was extracted by using Soil Adjusted Vegetation Index, one of the indices dealing with vegetative activity, and then was classified into two vegetation types by the supervised classification method using ERWDA s Natural Resources Maps (2000) as the supervisor. 1. Intrduction There are well preserved coral banks, reefs and shallow lagoons along the long coastline of Abu Dhabi Emirate in the southern Arabian Gulf. The coastal area is characterized by the biodiversity of marine wildlife, including endangered species of dugongs and turtles. Mangrove forests and algal mats develop naturally along the tidal limits of the lagoons. The conservation of the biodiversity is major concern on the protection of coastal environment in the Emirate. Oil pollution related with the offshore petroleum development, such as leakage from production facilities and pipelines, oil tanker accidents etc. obviously has severe impact on the coastal environment and is the issue on conservation of the ecosystem in coastal regions. In order to respond to oil spill events the Abu Dhabi Oil Spill Protection Priorities Maps 2000 (ERWDA, 2000) was prepared by using Landsat 5 TM images acquired in 1998 with the spatial resolution of 30m. Recent satellite images, however, show clearly that dredged canals and newly constructed facilities have changed coastal feature since then. Thus updating of the maps using new satellite images with finer spatial resolution is urgently needed for the protection of coastal environment. The study aims to develop the operational algorithms for producing bottom type maps in the shallow waters and vegetation maps of the coastal regions by using Terra ASTER (Advanced Spaceborne Thermal Emission and Reflection radiometer) images. The shallow sea with water depth less than 5m, where Terra ASTER can detect the reflection from the sea floor, covers broad intertidal areas in the Abu Dhabi Emirate. Imaging the detailed distribution of bottom types by using Terra ASTER data with the spatial resolution of 15m is one of the goals of this study. It is well known that the vegetation of the intertidal areas plays a significant role in the ecosystem of coastal regions. ERWDA and oil development companies like Japan Oil Development Co., Ltd. (JODCO) have contributed to mangrove plantation for the conservation of the coastal environment. The study also aims to extract vegetation distribution in the coastal areas from Terra ASTER images, as well as to classify vegetation types. The details of algorithms are described first and followed by the examples which were analyzed from Terra ASTER images and the field verification conducted by ERWDA. 2. Satellite Image Data Twenty scenes of Terra ASTER data acquired from 2000 through 2003 are used in the study. ASTER is a cooperative effort between NASA and Japan's Ministry of Economy, Trade and Industry, and consists of three instrument subsystems referred as VNIR (Visible-Near InfraRed), SWIR (ShortWave InfraRed) and TIR (Thermal InfraRed). Each subsystem operates

Upload: independent

Post on 28-Feb-2023

1 views

Category:

Documents


0 download

Embed Size (px)

TRANSCRIPT

Monitoring of Coastal Environment Using Satellite Images in the United Arab Emirates

Mohammad Yousif Al Jawdar

ERWDA, P. O. Box 45553, Abu Dhabi, United Arab Emirates

Manabu Shiobara

Japan Oil Development Co., Ltd., 1-21-2, Shinkawa, Chuo-ku, Tokyo 104-0033, Japan

Takumi Onuma

JGI, Inc., 1-5-21, Otsuka, Bunkyo-ku, Tokyo 112-0012, Japan

ABSTRACT

The diversity of marine wildlife along the coast of the United Arab Emirates is well known as well as the offshore giant

oilfields. The protection of sustainable marine environment is the first priority to keep the biodiversity in the region and the

Environmental Research and Wildlife Development Agency (ERWDA) has been paying attention on the preservation of

coastal environment since it was established in 1996. It is widely accepted that long-term monitoring is mandatory for

environmental hazard and satellite images with intermediate spatial resolution covering the broad region are suitable for this

purpose. This study is based on such a viewpoint to contribute to the monitoring of environmental change along the coast of

Abu Dhabi Emirate. The methodology of producing bottom type maps and vegetation distribution maps from satellite image

data are presented. The operational atmospheric correction algorithm to minimize air-water interface was designed and applied

to Terra ASTER data to produce Bottom Index images, which were compared with the field survey results to convert the index

value to the bottom type. Bottom index value of 2.5 was deemed to discriminate sea grass and seaweed from sand bottom. The

distribution of onshore vegetation was extracted by using Soil Adjusted Vegetation Index, one of the indices dealing with

vegetative activity, and then was classified into two vegetation types by the supervised classification method using ERWDA’s

Natural Resources Maps (2000) as the supervisor.

1. Intrduction

There are well preserved coral banks, reefs and

shallow lagoons along the long coastline of Abu Dhabi

Emirate in the southern Arabian Gulf. The coastal area

is characterized by the biodiversity of marine wildlife,

including endangered species of dugongs and turtles.

Mangrove forests and algal mats develop naturally

along the tidal limits of the lagoons. The conservation

of the biodiversity is major concern on the protection

of coastal environment in the Emirate.

Oil pollution related with the offshore petroleum

development, such as leakage from production

facilities and pipelines, oil tanker accidents etc.

obviously has severe impact on the coastal

environment and is the issue on conservation of the

ecosystem in coastal regions. In order to respond to oil

spill events the Abu Dhabi Oil Spill Protection

Priorities Maps 2000 (ERWDA, 2000) was prepared

by using Landsat 5 TM images acquired in 1998 with

the spatial resolution of 30m. Recent satellite images,

however, show clearly that dredged canals and newly

constructed facilities have changed coastal feature

since then. Thus updating of the maps using new

satellite images with finer spatial resolution is urgently

needed for the protection of coastal environment.

The study aims to develop the operational

algorithms for producing bottom type maps in the

shallow waters and vegetation maps of the coastal

regions by using Terra ASTER (Advanced Spaceborne

Thermal Emission and Reflection radiometer) images.

The shallow sea with water depth less than 5m, where

Terra ASTER can detect the reflection from the sea

floor, covers broad intertidal areas in the Abu Dhabi

Emirate. Imaging the detailed distribution of bottom

types by using Terra ASTER data with the spatial

resolution of 15m is one of the goals of this study. It is

well known that the vegetation of the intertidal areas

plays a significant role in the ecosystem of coastal

regions. ERWDA and oil development companies like

Japan Oil Development Co., Ltd. (JODCO) have

contributed to mangrove plantation for the

conservation of the coastal environment. The study

also aims to extract vegetation distribution in the

coastal areas from Terra ASTER images, as well as to

classify vegetation types.

The details of algorithms are described first and

followed by the examples which were analyzed from

Terra ASTER images and the field verification

conducted by ERWDA.

2. Satellite Image Data

Twenty scenes of Terra ASTER data acquired from

2000 through 2003 are used in the study. ASTER is a

cooperative effort between NASA and Japan's

Ministry of Economy, Trade and Industry, and consists

of three instrument subsystems referred as VNIR

(Visible-Near InfraRed), SWIR (ShortWave InfraRed)

and TIR (Thermal InfraRed). Each subsystem operates

in a different spectral region, and has its own

telescope. In this study VNIR data is mainly used

because green (Band 1) and red (Band 2) can reach the

sea bottom down to the depths of 15m and 5m

respectively, while the energy at Bands 3 through 14 is

completely absorbed. As chlorophyll absorption is an

important input for vegetation models and a key factor

in many vegetation indices, infrared (Band 3) and red

bands are frequently used for vegetation analysis.

SWIR and TIR data are also used to classify

vegetation and to detect clouds. ASTER also has a

pointing capability enabling the revisit of 5 days

despite of swath width of 60 km that is narrower than

the orbital spacing of 172 km. Because of this

capability, the geometry of the sun–sea surface–sensor

must be considered for precise measurement of the

reflectance of surface object and so-called

water-leaving radiance.

Northeast to the Abu Dhabi Island was selected as a

“pilot” area (Fig. 1) for verifying algorithms to be

applied. Several scenes were used to cover the area

based on time span for surface monitoring and data

quality including acquisition condition.

3. Methodology

3-1. Water-leaving radiance

The signal received over the ocean surface consists

of atmospheric backscattering components and

water-leaving radiance. The radiance backscattered

from the atmosphere and/or the sea surface has

typically at least an order of magnitude larger than the

desired radiance scattered out of the water. Therefore,

the process of retrieving water-leaving radiance from

the total radiance is needed. According to Gordon et al

(1983), the sensor radiance Lt() can be divided into

its components:

Lt() = Lr()+ La() + t()Lw() (1)

where Lr() is the contribution arising from Rayleigh

scattering, La() is the contribution arising from

aerosol scattering, Lw() is the water-leaving radiance,

and t() is the diffuse transmittance of the atmosphere

between sea surface and the sensor. This

approximation is assumed for application to CZCS

data that has the tilting capability to avoid sun glint. In

order to apply the algorithm to Terra ASTER data, a

component of sun glint Lg() must be accounted in Eq.

(1):

Lt() = Lr()+ La()+ Lg() + t()Lw() (2)

Lr() is a function of wavelength, Rayleigh optical

thickness and ozone optical thickness. Rayleigh

optical thickness is also a function of wavelength.

Ozone optical thickness is calculated by using the

amount of ozone. In the study, monthly average of

TOMS data were used for the amount of ozone. t() is

the function of Rayleigh optical thickness, ozone

optical thickness and zenith angle of a vector from a

pixel to the sensor. Details of the derivation of terms

in Eq. (1) are introduced by Gordon et al (1983).

As for the La(), assuming the Lw(IR) (infrared) is

equal to zero due to the absorption by the ocean, then

from Eq. (2);

La(IR)=Lt(IR) – Lr(IR) – Lg(IR)

Lt(IR) is satellite data and Lr(IR) is known. Lg(IR) is

a function of Rayleigh optical thickness, ozone optical

thickness, aerosol optical thickness a(IR) and wind

velocity. That is, Lg(IR) is needed to calculate La(IR),

and a(IR) is needed to calculate Lg(IR). But La(IR)

must be known in order to calculate a(IR), because

a(IR) is inversely derived from La(IR). To solve this

problem, the optical thickness data of typical aerosol

(Elterman, 1968) was referred in our algorithm as the

start value of iterative search. For IR band, the

algorithm first calculates Lg(IR) using the start values

of aerosol optical thickness for infrared band and

derives Lt(IR) – Lr(IR) – Lg(IR). Then it calculates new

a(IR) and recalculates Lg(IR) using thisa(IR). Repeat

the calculation until the difference of first a and new a

becomes less than 0.05. When the condition met, take it

as the aerosol optical thickness of the pixel and

calculate Lg(IR) as the final.

Scattering phase function by Haltrin (1998) for

aerosol of maritime atmosphere was used in the study

to calculate a(IR). The single scattering albedo of

aerosol for the seawater is stable for one Terra ASTER

scene and can be considered as a constant. The result of

simulation shows about 0.9 around the Arabian

Peninsula (Takemura and Nakajima, 2002).

Assuming that the aerosol optical thickness follows

i = i–

where i and i are the aerosol optical thickness and the

central wavelength, respectively for band i; i = 1-3 for

ASTER case. is so called the angstrom index. By

taking ratio of the aerosol optical thickness between IR

band and another band,

a() = a(IR)*(IR/)

In most studies -1

function is assumed for the

spectral variation of the aerosol optical thickness

(Ouaidrari and Vermote, 1999; Gordon, et al., 1983;

Zhang, et al., 1999), in which is 1. In the ASTER case,

therefore,

a() = a(VNIR3)*0.807/

where is the center wavelengths (in m) of ASTER

VNIR 1 and 2 (=0.556 and 0.659m). Then, a(VNIR1)

and a(VNIR2) are derived from a(VNIR3), and La(VNIR1),

Lg(VNIR1), La(VNIR2) and Lg(VNIR2) can be calculated

from a(VNIR1) and a(VNIR2), finally the water-leaving

radiance is led from the equation (2) as

Lw(i) = {Lt(i) – Lr(i) – Lg(i) – La(i)}/t(i) (24) The accuracy of Lw depends on that of a(IR). Lw sometimes becomes negative due to overestimation of a(IR), which is resulted from specification of inappropriate wind velocity that is unknown parameter in the algorithm. The algorithm takes an offset value to adjust the negative

value. Under the non-turbid clear water condition, the

lights with wavelengths of ASTER VNIR 1 and 2 can

penetrate down to the depth of 15m and 5m,

respectively. The offset value should be chosen for the

open sea that is adequately deep.

3-2. Bottom index

As for the shallow water of the same bottom type,

the relationship between water-leaving radiance of

two bands in visible-near infrared region can be

expressed as

ln(Lwi) = (ki/kj)ln(Lwj) + a (3)

where Lw is water-leaving radiance, k is an

attenuation coefficient, “a” is a constant, i and j are

band numbers (Lyzenga, 1981). The attenuation

coefficient is independent on bottom types and the

ratio (ki/kj) is constant. Then, the magnitude of the

constant “a” depends on bottom types, and it is called

Bottom Index (BI). From Eq. (3) BI is expressed as

BI = ln(Lwi) - (ki/kj)ln(Lwj)

In the case of Terra ASTER, i and j are 1 and 2,

respectively. The ratio of attenuation coefficients was

estimated as follows. For many small water areas

with the same substratum (sand) at different depths by

visual inspection (as shown in Fig. 2), water-leaving

radiances for bands 1 and 2 are calculated as described

in the previous section. Average values of

(log-transformed) water-leaving radiance for bands 1

and 2 are cross-plotted, then the ratio of attenuation

coefficients is derived as the slope of the regressed

line.

Terra ASTER data of the same area but with

different acquisition dates were tested in the study, and

it was found that the slopes were almost identical

(0.44) and could be treated as a constant (Fig. 3).

Therefore, BI is expressed using Terra ASTER for the

shallow waters of Abu Dhabi as

BI = ln(Lw1) – 0.44* ln(Lw2)

Fig. 4 shows an example of bottom index produced

from Terra ASTER data acquired on November 18,

2000. It can be said that the light of ASTER VNIR

band 2 penetrates down to the water depth of 5m, and

the processing result is quite concordant with the

iso-depth line of 5m (blue line in Fig. 4).

3-3. Vegetation Index

Vegetation index is a parameter widely adopted in

vegetation researches on remotely sensed data, and

many kinds of indices have been proposed for

different purposes. SAVI (Soil Adjusted Vegetation

Index) is thought to be suitable for the areas where

vegetation distribution is sparse. As all the indices

including SAVI are calculated from reflectance values

of infrared and red bands, the atmospheric correction

and the conversion from DN (Digital Number) to

reflectance of ASTER data must be done first. In this

study, the apparent reflectance at the sensor was used

instead of one at the ground, because full atmospheric

correction of Terra ASTER data using radiation

transfer codes is quite time consuming and hence is

not operational. The Improved Dark Object Method

(Chavez, 1988) was adopted for the atmospheric

correction. The method removes the atmospheric

scattering based on the image data itself, and the result

of the process is given in the unit of DN. In order to

calculate SAVI, therefore, the result of atmospheric

correction must be converted into the apparent

reflectance as follows;

iiii G*)HDN(L

)2

cos(*E

d*L*

i

2i

i

where i is the band number, DNi is the digital count of

raw data, Hi is the offset calculated by the Improved

Dark Object Method, Gi is the conversion coefficient

for ASTER band i, Ei is exoatmospheric solar

irradiance, d is the astronomical distance between the

sun and the earth, and is the sun elevation angle.

Parameters for Improved Dark Object Method are

shown in Table 1. The starting haze value (SHV) is

calculated by subtracting the offset value of Table 1

(always 1 in the case of ASTER) from the minimum

value of VNIR band 1. Then Hi is given by

)i(Offset)i(NORM*)i,case(MLF*SHVHi

And SAVI is given by

)L(

)(*)L1(SAVI

32

23

where2 and 3 are the reflectance of ASTER VNIR

bands 2 and 3, respectively and L is soil adjustment

parameter and 0.5 in this study.

4. Bottom Type Map and Vegetation Map

A goal of the study is to develop the operational

methodology for producing the maps of bottom types

and vegetation, aiming at the application in

environmental monitoring. Water-leaving radiance,

bottom index, apparent reflectance and vegetation

index are the parameters needed for the production of

these maps, of which derivation algorithms are

described in the above. From the viewpoint of

verification, the results of data processing should be

crosschecked by the ground truth. ERWDA carried out

the field survey to collect data of bottom types and

vegetation distribution in the pilot area in February

2003. The processing results of satellite image data

and the maps has been compared by using Natural

Resources Maps (ERWDA, 2000) before the survey to

investigate the applicability of the algorithms

developed in the study.

4-1. Bottom Index to Bottom Type Map

Threshold values of bottom index were estimated

through the comparison of the Natural Resources Map

and the bottom index calculated from Terra ASTER

data. The area to the northeast of Abu Dhabi city was

chosen for the pilot area, and regions of sandy bottom,

sea grass-shallow, and mangrove were digitized

separately (Fig. 5). Then, the histograms of the bottom

index image (Fig. 6) for sandy bottom and sea

grass-shallow areas were compared. By the try and

error manner two thresholds of 2.49 and 2.72 were set,

and provisional bottom type map was produced (Fig.

7). When compared with the Natural Resources Map,

the distribution of the sand in Fig. 7 looks quite

concordant with that of sandy bottom and sheltered

tidal flat in Fig. 5. However, because the area does not

contain a coral patch and the accuracy of the Natural

Resources Map is unknown, the field survey data is

needed to clarify the relationship between the ASTER

bottom index and the real bottom types.

Fig. 8 shows the locations of the survey points and

Table 2 shows the summary of the survey. Assuming

the accuracy of handy GPS positioning is 50 m, that is

to say equivalent to 3 pixels of ASTER VNIR spatial

resolution, a bottom index value of the target point

was calculated by mean value of 7 by 7 pixels

surrounding the center pixel of the given coordinate as

the target point. Inside of brackets of Bottom Index

column in Table 2 represents number of non-null

pixels in 7 by 7 window. Calculated bottom index has

been evaluated with the field data and summarized as

follows:

・ Bottom Index greater than or equal to 2.5 seems

to correspond to sandy bottom, based on the

comparison between substrata and Bottom Index.

・ However, bottom type with Bottom Index greater

than 2.4 may also be the sandy bottom (No. 1 &

No. 6 in Fig. 9), the detailed study may be needed

to reveal the relationship between the coverage of

the vegetation and the Bottom Index.

・ The evaluation suggests that time difference of

approximately three years between the field

observation and satellite data acquisition made the

accurate correlation between the field data and

image analysis difficult.

4-2. Vegetation Index to Vegetation Map

Discrimination of the vegetation areas from

non-vegetation areas such as desert, salt pan, sabkha,

bare soil, water, etc., is obviously important for

environmental monitoring. Multi-temporal change of

the vegetation distribution cannot be monitored

without identification and extraction of vegetation

areas. Vegetation indices are widely used for this

purpose, as well as for rough estimation of vegetation

activity and/or healthiness. In this study threshold

value discriminating vegetation from non-vegetation

was sought by comparing the vegetation distribution

map produced from Terra ASTER data with the

Natural Resources Map (ERWDA, 2000). To find an

appropriate threshold value, the mangrove areas were

digitized from the Natural Resources Map first, and

then SAVI values of those areas were examined. Salt

marshes in the Natural Resources Map were not

digitized because they likely include non-vegetated

areas based on different criteria. By taking minimum

value of SAVI for digitized areas as a tentative

threshold value, pixels having values larger than or

equal to this threshold were extracted as vegetation

pixels. Then the threshold was adjusted so that

vegetation areas showing reddish color on the VNIR

false color composite were completely extracted. Fig.

10 shows an example of extraction of vegetation areas

in the same region as Fig. 5. It reveals that the areas of

mangroves and salt marshes of the Natural Resources

Map are satisfactory extracted. Mangrove areas seem

to correspond to the areas with SAVI over 0.2 (Fig.

11). From theoretical point of view, however, it is

impossible to discriminate the mangroves from salt

marshes merely by the magnitude of SAVI. SAVI

values bigger than 0.2 are likely resulted from high

density of distribution of trees.

Although the extraction of vegetation areas is a

mandatory process for monitoring of coastal

environment, vegetation types must be classified from

operational viewpoint because multi-temporal change

in the distribution of certain vegetation type is focused

in the course of actual monitoring. Vegetation types

are relatively simple such as mangroves, halophytes

other than mangroves, and algal mats in the coastal

region of Abu Dhabi. Mangroves consist of single

species Avicennia marina in the study area.

Arthrocnemum macrostachyum is representative

species of the halophytes except mangroves. Because

each category has specific signature on the ASTER

VNIR false color composite, discrimination among

categories will be successfully made by the

classification with appropriate supervisors. Areas of

mangroves and salt marshes (halophytes) on the

Natural Resources Map (ERWDA, 2000) were used as

the supervisors (Fig. 11). Supervised classification

using 9 bands of ASTER VNIR through SWIR was

performed for the vegetation areas that were extracted

by using certain threshold value of SAVI. Distribution

of classified mangroves and halophytes seems to be

fairly concordant with the Natural Resources Map (Fig.

12). However, field verification shows that, when trees

are relatively low and sparsely distributed, they are not

recognized as vegetation (Fig. 12, 13), suggesting one

of limitations on the spatial resolution of 15m. In turn,

this may mean that dense forest of relatively taller

mangroves is classified as mangroves by supervised

classification.

5. Summary

Operational algorithms for producing bottom type

and vegetation type maps have been developed using

Terra ASTER image data in Abu Dhabi coastal region.

Bottom type can be classified by using bottom index,

which is calculated mainly from water-leaving

radiance of Terra ASTER VNIR bands 1 and 2. An

algorithm for extracting water-leaving radiance

accounting water surface reflection and sun glitter

effect, together with corrections of atmospheric

scattering, from signal received by ASTER has also

been developed. Bottom index is calculated by

BI = ln(Lw1) – 0.44*ln(Lw2)

where Lw1 and Lw2 are water-leaving radiance for

VNIR bands 1 and 2 respectively. By the comparison

of bottom index image and results of sea truth

performed by ERWDA, BI=2.5 was provisionally

deemed to be a threshold between sand and sea

grass/seaweed. Sand has bottom indices more than or

equal to 2.5 while sea grass / seaweed are less than

2.5.

The evaluation suggests that time difference of

approximately 3 years between the field observation

and satellite data acquisition made the accurate

correlation between the field data and image analysis

difficult. The result also suggests a field observation

synchronized with satellite observation is necessary to

obtain more accurate results. In addition, it became

clear that a large number of field samples would be

required to analyze accurately. In this pilot study some

uncertainty is still remained because of limited

number of the field samples.

Onshore vegetation is easily discriminated from

bare soil, water, artificial construction, etc., based on

SAVI. Simple atmospheric correction specially

modified for Terra ASTER data was applied to convert

DN values into the reflectance. After screening out of

non-vegetation by SAVI, supervised classification

using ASTER VNIR and SWIR bands has been

performed for onshore vegetation to classify

vegetation types such as mangroves and halophytes

(except mangroves). As mentioned above, a future

field observation is needed to be synchronized with

satellite data acquisition for verifying the algorithm

adopted in this research.

Acknowledgement

This study is a part of the collaboration research

between Japan Oil Development Co., Ltd. (JODCO)

and the Environmental Research and Wildlife

Development Agency (ERWDA), called ”ASTER

Remote Sensing Alliance (ARSA)”. Authors would

like to express their thanks to ERWDA and JODCO

for granting permission to submit this paper, and to

U.A.E. University for providing the opportunity to

present the paper at OPEIC 2003.

References

Chavez, P.S., Jr., 1988, An Improved Dark-Object

Subtraction Technique for Atmospheric

Scattering Correction of Multispectral Data.

Remote Sensing of Environment, vol. 24, p.

459-479.

Elterman, L., 1968, UV, Visible, and IR Attenuation

for Altitudes to 50 km, 1968. Environmental

Research Papers, no. 285, AFCRL-68-0153, Air

Force Cambridge Research Laboratories, 56 pp.

Environmental Research and Wildlife Development

Agency (ERWDA), 2000, Abu Dhabi Coastline

Oil Spill Protection Priorities 2000, Natural

Environmental Resources and their Oil Spill

Protection Priorities for the Coastline of Abu

Dhabi Emirate, U.A.E. published by ERWDA,

U.A.E., 120 pp.

Gordon, H.R., Clark, D.K., Brown, J.W., Brown, O.B.,

Evans, R.H., and Broenkow, W.W., 1983,

Phytoplankton pigment concentrations in the

Middle Atlantic Bight: comparison of ship

determinations and CZCS estimates. Applied

Optics, vol. 22, no. 1, p. 20-36.

Haltrin, V.I., 1998, Empirical relationship between

aerosol scattering phase function and optical

thickness of atmosphere above the ocean.

Propagation and Imaging through the

Atmosphere II, Proceedings of SPIE, vol. 3433, p.

334-342.

Lyzenga, D.R., 1981, Remote sensing of bottom

reflectance and water attenuation parameters in

shallow water using aircraft and Landsat data.

International Journal of Remote Sensing, vol. 17,

p. 155-166.

Ouaidrari, H., and Vermote, E.F., 1999, Operational

Atmospheric Correction of Landsat TM Data.

Remote Sensing Environment, vol. 70, p. 4-15.

Takemura, T., and Nakajima, T, 2002,

Single-Scattering Albedo and Radiative Forcing

of Various Aerosol Spacies with a Global

Three-Dimensional Model, Journal of Climate,

15, 4, 333-352.

Zhang, M, Carder, K., Muller-Karger, F.E., Lee, Z.,

and Goldgof, D.B., 1999, Noise Reduction and

Atmospheric Correction for Coastal Applications

of Landsat Thematic Mapper Imagery. Remote

Sensing Environment, vol. 70, p. 167-180.

Fig. 1 Terra ASTER VNIR mosaic image covering coastal region of Abu Dhabi Emirate.

Fig. 2 Example of selection of sand bottom at different water depths.

Abu Dhabi

Fig. 3 Ratios of attenuation coefficients between ASTER VNIR bands 1 and 2.

Fig. 4 Bottom index image superimposed on Terra ASTER VNIR false color composite.

Table 1. Parameters for Improved Dark Object Method.

very clear clear moderate hazy very hazy

1 1.00 1.00 1843.00 1.000 1.000 1.000 1.000 1.000

2 0.95 1.00 1555.00 0.532 0.729 0.854 0.895 0.924

3 0.78 1.00 1111.70 0.286 0.534 0.731 0.803 0.855

*1: unit = W/m*m*sr*m

Normalized Multiplication Factors (MLF)ASTER

Band

Normalized

Gain *1

(NORM)

Offset Irradiance

Fig. 5 Natural Resources Map (after ERWDA, 2000) and sample areas for comparison with bottom index

and vegetation index.

Fig. 6 Example of bottom index image.

Fig. 7 Provisional bottom type map.

Fig. 8 Locations of the field survey displayed on Terra ASTER provisional bottom type map.

Locality shown in Fig. 7.

Table 2 Results of the field survey.

No. Longitude Latitude Habitat Substrata

Water

depth

(ft)

Water

depth

(m)

Notes BI*

1 54.4849 24.59254 Seaweed, shallow Sandy 9 2.7432

Water turbid, seaweed sp.

Padina minor (30% cover) 2.412(3)

2 54.48755 24.59254 Sandy bottom Sandy 14.5 4.4196

Water turbid, High current, Part

of dredged channel? 2.422(21)

3 54.48848 24.59754 Sw/Seagrass shallow Sandy 8.5 2.5908

Padina minor (10%) and

Halodule uninervis (40%) 2.465(31)

4 54.47671 24.5841 Mixed Sandy beach Sandy 0 0

Sand, dead marine shells and

gravel mixed beach 2.767(29)

5 54.50869 24.58402 Seagrass - Deep Sandy 21.8 6.64464 Halodule uninervis (60%) null

6 54.51655 24.58335 Seagrass shallow Sandy 5.4 1.64592

Halophila ovalis (10%),

Halodule uninervis (30%) 2.442(39)

7 54.5558 24.59457 Seagrass shallow Sandy 5.9 1.79832 Halodule uninervis ( 30%) 3.208(48)

8 54.55444 24.59495 Sheltered Intertidal flat Sandy 0 0 Algal mats at periphery 2.566(49)

9 54.47143 24.57531

Exposed compacted

tidal flats Sandy 0 0

Close to Ras Al Gurab, sparse

algal growth. 2.686(13)

10 54.49993 24.61121 Rocky bottom

Rocks and

Sand 7.5 2.286

Gravel, boulders and small sized

rocks 2.636(49)

*: 7 x 7 BIs are averaged. Bracketed number shows the number of non-null pixels in the 7 x 7 window.

No. 1 No. 4

No. 5 No. 6

No. 8 No. 9

Fig. 9 Photographs of survey localities.

(No photographs for localities No. 2, 3, 7 and 10)

Fig. 10 Example of extracted vegetation areas.

Red pixels have SAVI values larger than or equal to 0.09.

Fig. 11 SAVI image map superimposed on Terra ASTER VNIR false color composite with mangroves and salt

marshes from Natural Resources Map (ERWDA, 2000).

Fig. 12 Classification result of mangroves and halophytes.

A B

A (Height 2-6m) B (Height < 2m)

Fig. 13 Field photographs of mangroves. Locality shown in Fig. 12.

: Mangrove

: Salt Marsh

: Mangrove after Natural

Resources Map

: Salt Marsh after Natural

Resources Map