PRESENT BY : MUHAMMAD FARIDZUL ADLI BIN ZAKARIA

PRESENT BY : MUHAMMAD FARIDZUL ADLI BIN ZAKARIA

Ahmad Mubin Wahab1 and Md. Latifur Rahman Sarker1, 2,*

1 Department of Geoinformation, Universiti Teknologi Malaysia,

Malaysia2 Department of Geography and Environmental Studies,

University of Rajshahi, Bangladesh.

*Corresponding author: sarker@utm.my

1.0 – INTRODUCTION

PM 2.5

PM 10

Atmospheric aerosol is a

suspension of liquid and

solid particles, with radii

varying from a few nm to

larger than 100 µm, in air.

Anthropogenic

Natural

Sources

WHAT IS

AEROSOL?

Sizes

Heart disease and stroke80%

Chronic obstructive pulmonary disease

14%

Lung cancer6%

0%

PREMATURE DEATH

1 - Human health Problems

asthma

hay fever

pulmonary inflammation

respiratory symptoms

Cardiovasculardiseases1 – PM enters to

respiratory system 2/3 – PM 10

trapped in

respiratory system

4 – PM 2.5 penetrates

deep into lungs

AEROSOL EFFECTS

2 - Visibility Degradation

Due to the extinction of light

when the light passing through

the atmosphere.

3 - Climate Change

Direct Effects

Indirect Effects

AEROSOL EFFECTS

Ground-based measurements Airborne-based measurements

Aerosol Robotic

Network

(AERONET)

Microstops II

Sunphotometer

Shipboard

measurement

Balloon Aircraft

Remote Sensing Satellite

Wide coverage Temporal resolution

Good spatial

information

Requires high spatial and temporal

resolution of data because of the short

life span of aerosol (7 to 10 days).

AEROSOL

MEASUREMENT

SATELLITE AEROSOL RETRIEVAL MECHANISM

Rayleigh reflectance

(𝝆𝐑𝐚𝐲) + Aerosol

reflectance

Surface Reflectance (𝝆𝒔𝒖𝒓𝒇)

Top of Atmosphere

Reflectance (𝝆𝐓𝐎𝐀)

𝝆𝐓𝐎𝐀 = 𝝆𝐀𝐞𝐫 + 𝝆𝐑𝐚𝐲 + 𝝆𝒔𝒖𝒓𝒇

The key factor of the aerosol retrieval is to estimate surface reflectance

that attempts to differentiate the aerosol signal from surface.

𝝆𝐀𝐞𝐫 = 𝝆𝐓𝐎𝐀 − 𝝆𝐑𝐚𝐲 − 𝝆𝒔𝒖𝒓𝒇

PROBLEM & SIGNIFICANT

MODIS Local Scale Aerosol

Low spatial resolution (10 km)

Lots of missing pixels

No real-time data available

High Resolution (500 m)

Real-time data available

Good spatial distribution

Based on the local

aerosol model

To compare the potential of two different

AOT algorithms,

To determine which technique can provide

effective aerosol retrieval estimation.

STUDY AREA

One of the most densely populated area.

7 million people living in 1104 km2 of land areas.

Availability of Long-term

Ground data measurement

(AERONET station).

Several studies have already

been conducted.

One of the most polluted

urban areas in the world.

Availability of Long-term

Ground data measurement

(AERONET station).

Several studies have already

been conducted.

One of the most polluted

urban areas in the world.

Why Hong Kong?

DATA USED

MOD02HKM MOD03 MOD09GA

Aerosol Robotic

Network

(AERONET)

• MOD02HKM - swath data with calibrated radiance at 500m.

• MOD03 - Geolocation data (geodetic coordinates, ground elevation, solar zenith angle, solar azimuth angle, satellite zenith angle and satellite azimuth angle).

• MOD09GA - Land surface reflectance product at 500m.

• MOD05 - Total Water Vapour content.

• MOD07 - Total Ozone Content.

• MOD021KM – Channel 26 (cirrus reflectance).

• Additionally, MODIS aerosol level 2 collection 005 (MOD04 L2 C005) was used to compare with our result.

• AERONET Level 1.5 data was used for the validation.

2.0 – METHODOLOGY

OVERALL METHODOLOGY

AEROSOL REFLECTANCE (𝜬𝐀𝐞𝐫)

TOA

REFLECTANCE

RAYLEIGH

REFLECTANCE

SURFACE

REFLECTANCE

TOTAL

TRANSMISSION OF

WATER VAPOUR

TOTAL

TRANSMISSION OF

OZONE GAS

𝜬𝐀𝐞𝐫 𝛌,𝜽𝒔,𝜽𝒗,𝝓 =

𝝆𝐓𝐎𝐀 𝛌,𝜽𝒔,𝜽𝒗,𝝓

𝑻𝒈 𝑴,𝑼𝑶𝒈 𝑻𝑶𝟑 𝑴,𝑼𝑶𝟑

− 𝝆𝐑𝐚𝐲 𝛌,𝜽𝒔,𝜽𝒗,𝝓 –𝑻𝒂𝒕𝒎 𝜽𝒔,𝜽𝒗 𝝆𝐬 𝛌,𝜽𝒔,𝜽𝒗,𝝓 𝑻𝑯𝟐𝑶

𝒃 𝑴,𝑼𝑯𝟐𝑶𝟏 − 𝝆𝐬 𝛌,𝜽𝒔,𝜽𝒗,𝝓 𝝆𝑯𝒆𝒎

𝑻𝑯𝟐𝑶𝒂 𝑴,

𝑼𝑯𝟐𝑶𝟐

TOTAL

TRANSMISSION OF

OTHER GAS

TOTAL

ATMOSPHERIC

TRANSMISSION

HEMISPHERIC

REFLECTANCE

TOA REFLECTANCE

𝒅 =𝟏

(𝟏+𝟎.𝟎𝟑𝟑𝐜𝐨𝐬(𝑫𝑶𝒀𝟐𝝅

𝟑𝟔𝟓)

satellite receives TOA spectral radiance 𝐿𝑇𝑂𝐴 𝜆 was normalized to the

solar illumination condition for each wavelength to generate TOAspectral reflectance using the equation as follows:

Band Wavelength (µm) ESUN (Wm-2 μm-1)

1 0.646 1596

2 0.855 974.7

3 0.466 2017

4 0.553 1850

5 1.243 463.1

6 1.632 232.9

7 2.119 92.67𝒅 is earth-sun distance can

be calculated as following:

𝒅 is earth-sun distance can

be calculated as following:

𝝆𝑻𝑶𝑨 𝝀 =𝝅𝑳𝑻𝑶𝑨 𝝀 𝒅

𝟐

𝑬𝒔𝒖𝒏𝝀 ∗ 𝒄𝒐𝒔𝜽𝒔

Source : MODIS Science Team

DOY – Julian daysDOY – Julian days

𝜽𝒔 is solar zenith angle,𝜽𝒔 is solar zenith angle,

𝑬𝟎 is extraterrestrial solar

irradiance,

𝑬𝟎 is extraterrestrial solar

irradiance,

where, 𝑳𝑻𝑶𝑨 𝝀 is TOA

spectral radiance obtained

from MOD02HKM data.

where, 𝑳𝑻𝑶𝑨 𝝀 is TOA

spectral radiance obtained

from MOD02HKM data.

𝑷𝑹𝒂𝒚 𝝀 =𝝉𝑹𝒂𝒚 𝝀 . 𝝆𝑹𝒂𝒚

𝟒(𝒄𝒐𝒔𝜽𝒔. 𝒄𝒐𝒔𝜽𝒗)

RAYLEIGH REFLECTANCE (𝑷𝑹𝒂𝒚)

where, 𝑐𝑜𝑠𝜃𝑠 is cosine solar zenith

angle, and 𝑐𝑜𝑠𝜃𝑣 is cosine sensor

zenith angle.

where, 𝑐𝑜𝑠𝜃𝑠 is cosine solar zenith

angle, and 𝑐𝑜𝑠𝜃𝑣 is cosine sensor

zenith angle. 𝝉𝐑𝐚𝐲 𝛌 = 𝒂. 𝛌− 𝒃+𝒄𝛌+ 𝐝 𝛌 . 𝐞𝐱𝐩 − 𝒛 𝟖. 𝟓

Constant 0.2 – 0.5 µm > 0.5 µm

a 3.01577 x 10-28 4.01061 x 10-28

b 3.55212 3.99668

c 1.35579 1.10298 x 10-3

d 0.11563 2.71393 x 10-2

)𝜸 = 𝜹 (𝟐 − 𝜹Ɵ is scattering

phase angle

Ɵ is scattering

phase angle

Ɵ = 𝒄𝒐𝒔−𝟏(−𝒄𝒐𝒔𝜽𝒔𝒄𝒐𝒔𝜽𝒗 + 𝒔𝒊𝒏𝜽𝒔𝒔𝒊𝒏𝜽𝒗𝝓

Wavelength (µm) 𝛅 𝛄

0.466 0.02899 0.01471

0.553 0.02842 0.01442

0.646 0.02786 0.01413

Source : Butcholtz, 1995

Source : Butcholtz, 1995

𝜹 is depolarization factor𝜹 is depolarization factor

𝒛 is elevation𝒛 is elevation 𝒂, 𝒃, 𝒄, 𝑎𝑛𝑑 𝒅 𝑖𝑠 𝑅𝑎𝑦𝑙𝑒𝑖𝑔ℎ𝑠𝑐𝑎𝑡𝑡𝑒𝑟𝑖𝑛𝑔 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡𝒂, 𝒃, 𝒄, 𝑎𝑛𝑑 𝒅 𝑖𝑠 𝑅𝑎𝑦𝑙𝑒𝑖𝑔ℎ𝑠𝑐𝑎𝑡𝑡𝑒𝑟𝑖𝑛𝑔 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡

𝝉𝑹𝒂𝒚 𝝀 is Rayleigh optical depth𝝉𝑹𝒂𝒚 𝝀 is Rayleigh optical depth

𝝆𝑹𝒂𝒚 =𝟑

𝟒 𝟏 + 𝟐𝜸[ 𝟏 + 𝟑𝜸 + 𝟏 − 𝜸 𝒄𝒐𝒔 𝟐Ɵ

𝝆𝑹𝒂𝒚 is Rayleigh phase function𝝆𝑹𝒂𝒚 is Rayleigh phase function

Total Atmospheric Transmission (𝑻𝒂𝒕𝒎)

𝑻𝒂𝒕𝒎(𝜽𝒔,𝜽𝒗) = 𝑻𝒂𝒕𝒎(𝜽𝒔). 𝑻𝒂𝒕𝒎 𝜽𝒗𝑻𝒂𝒕𝒎(𝜽) = 𝑻𝑹𝒂𝒚(𝛉) . 𝑻𝒂𝒆𝒓(𝛉)

𝑻𝑹𝒂𝒚(𝛉) = 𝒆𝒙𝒑(−𝜷𝑹𝒂𝒚 . 𝝉𝑹𝒂𝒚 . (𝟏/𝒄𝒐𝒔𝜽)) 𝑻𝒂𝒆𝒓(𝛉) = 𝒆𝒙𝒑(−𝜷𝒂𝒆𝒓. 𝝉𝒂𝒆𝒓 . (𝟏/𝒄𝒐𝒔𝜽))

𝜷𝑹𝒂𝒚 =

𝒊=𝟏

𝟓

𝒃𝒊𝑹𝒂𝒚. (𝟏/𝒄𝒐𝒔𝜽)−(𝒊−𝟏) 𝜷𝑨𝒆𝒓 =

𝒊=𝟏

𝟓

𝒃𝒊𝒂𝒆𝒓. (𝟏/𝒄𝒐𝒔𝜽)−(𝒊−𝟏)

Coefficient Rayleigh Aerosol

𝒃𝟏 -0.44408 0.01176

𝒃𝟐 4.49481 1.01682

𝒃𝟑 -9.71368 -2.32949

𝒃𝟒 9.49795 2.11831

𝒃𝟓 -3.42016 -0.71737

Total Rayleigh Transmission (𝑻𝑹𝒂𝒚(𝛉) ) Total Aerosol Transmission (𝑻𝒂𝒆𝒓(𝛉) )

Source : Hoyningen-Huene et al., 2007

SURFACE REFLECTANCE (𝜌𝑠)

An improvement of DDV techniques (more robust)

Empirical relationship (nonlinear relationship) between

visible channel and SWIR channel.

Calibrated by refining atmospheric correction algorithm

(6SV code).

An improvement of DDV techniques (more robust)

Empirical relationship (nonlinear relationship) between

visible channel and SWIR channel.

Calibrated by refining atmospheric correction algorithm

(6SV code).

𝐌𝐎𝐃𝟎𝟗𝐆𝐀

https://lpdaac.usgs.gov/dataset_discovery/modis/modis_products_table/mod09ga

𝑼𝑶𝟑 − the total ozone content (obtained from the MOD07 level 2).

𝑴− air mass factor (𝑴 =1/𝒄𝒐𝒔𝜽).

𝒌𝑶𝟑 − weighting coefficient of ozone gases (derived from 6SV code).

𝑻𝑶𝟑(𝑴,𝑼𝑶𝟑) = 𝒆−𝑴𝒌

𝑶𝟑𝑼𝑶𝟑

𝑼𝑯𝟐𝑶 − total water vapour content (obtained from MOD05 level 2) .

𝑴−air mass factor (𝑴 =1/𝒄𝒐𝒔𝜽). .

𝒌𝑯𝟐𝑶𝟏 , 𝒌𝑯𝟐𝑶

𝟐 , and 𝒌𝑯𝟐𝑶𝟑 − weighting coefficients of water vapour (derived from 6SV code)

Total transmission of other gases (𝑪𝑶𝟐 𝒂𝒏𝒅 𝑵𝟐𝑶)

• Only for the wavelength at 2.119 µm.

• Obtained directly from 6SV code using the standard atmosphere model.

Wavelength (µm) Gas Absorption Effect

0.466 O3

0.553 O3

0.646 O3 and 𝐻2𝑂

2.119 𝐻2𝑂, CO2 and N𝟐O

Total Gaseous Transmission

Total transmission of ozone gas (𝑇𝑂3)

Total gaseous transmission of water vapour (𝑇𝐻2𝑂)

𝑻𝑯𝟐𝑶 𝑴,𝑼𝑯𝟐𝑶 = 𝒆𝒙𝒑[𝒌𝑯𝟐𝑶𝟏 𝑴𝑼𝑯𝟐𝑶 + 𝒌𝑯𝟐𝑶

𝟐 𝑳𝒐𝒈(𝑴𝑼𝑯𝟐𝑶) + 𝒌𝑯𝟐𝑶𝟑 𝑴𝑼𝑯𝟐𝑶𝑳𝒐𝒈(𝑴𝑼𝑯𝟐𝑶)]

Hemispheric reflectance

𝛕𝐚𝐭𝐦 is atmospheric optical depth (𝛕𝐑𝐚𝐲 + 𝛕𝐚𝐞𝐫).

𝐛𝐢 is polynomial coefficients of hemispheric reflectance.

𝛕𝐚𝐭𝐦 is atmospheric optical depth (𝛕𝐑𝐚𝐲 + 𝛕𝐚𝐞𝐫).

𝐛𝐢 is polynomial coefficients of hemispheric reflectance.

𝝆𝑯𝒆𝒎 =

𝒊=𝟏

𝟒

𝒃𝒊 . 𝝉𝒂𝒕𝒎𝒊 Coefficient

Hemispheric

Reflectance

𝒃𝟏 0.33185

𝒃𝟐 -0.19653

𝒃𝟑 0.08935

𝒃𝟒 -0.01675

Source : Hoyningen-Huene et al., 2007

Integral of the bidirectional reflectance distribution function

(BRDF) over all viewing directions.

Crucial for surface function correction due to multiple scattering

effect.

Has a high influence on the bright surfaces, while less over low

surface reflectance.

Integral of the bidirectional reflectance distribution function

(BRDF) over all viewing directions.

Crucial for surface function correction due to multiple scattering

effect.

Has a high influence on the bright surfaces, while less over low

surface reflectance.

LOCAL AEROSOL MODEL CHARACTERIZATION

Identify number of cluster (k)Identify number of cluster (k)

VRC methodVRC method Ward’s methodWard’s method

Clustering Analysis

K-means clustering analysis

Local Aerosol Model

K-means

clustering

ANOVA Tables

Sum of F-test

values (𝑉𝑅𝐶𝑘)

𝝎𝒌 = 𝑽𝑹𝑪𝒌+𝟏 − 𝑽𝑹𝑪𝒌 − 𝑽𝑹𝑪𝒌 − 𝑽𝑹𝑪𝒌−𝟏

Number of cluster (k)

(smallest value of 𝜔𝑘)

Hierarchical

cluster analysis

Agglomerative

procedures

Ward’s method

Elbow rule

Number of cluster (k)-based on the number of

step has biggest jump.

AOT RETRIEVE USING SBDART CODE

MODIS Aerosol Reflectance

(0.466 µm, 0.553 µm, and 0.646 µm)

Local Aerosol Model parameters

SBDART code

Variables No. Parameters

Wavelength 30.466 µm, 0.553 µm,

and 0.646 µm

AOT at

0.55 µm9

0.0, 0.2, 0.4, 0.8,

1.4, 1.8, 2.2, 3.0,

and 5.0

SZA 9 0º ~ 80 º, Δ = 10 º

VZA 17 0º ~ 80 º, Δ = 5 º

PHI 18 0º ~ 170 º, Δ = 10 º

Aerosol

Model4

SSA, Qext, and g at

0.439 µm, 0.676 µm,

0.869 µm, and 1.02

µm.

TOA Reflectance as a function of AOT

Aerosol Reflectance as a function of AOT

Interpolation (Optimal spectral

shape-fitting technique)

No

AOT (0.466 µm, 0.553 µm, and 0.646 µm)

AOT at 0.55 µm

Yes

𝑥2 =1

𝑛

𝑖=1

𝑛𝜌𝐴𝑒𝑟𝑚 λ𝑖 − 𝜌𝐴𝑒𝑟

𝑐 λ𝑖𝜌𝐴𝑒𝑟𝑚 λ𝑖

2

𝑥2 =1

𝑛

𝑖=1

𝑛𝜌𝐴𝑒𝑟𝑚 λ𝑖 − 𝜌𝐴𝑒𝑟

𝑐 λ𝑖𝜌𝐴𝑒𝑟𝑚 λ𝑖

2

ρAer(λ) = ρTOA λ − ρRay λ

AOT RETRIEVE USING DIRECT RETRIEVAL

MODIS Aerosol Reflectance

(0.466 µm, 0.553 µm, and 0.646 µm)

Local Aerosol Model parameters

MIEV Code

Aerosol Phase Function as a

function of Scattering Angle

Interpolation(linear) with

MODIS scattering angle

AOT at 0.55 µm (model 1)

AOT at 0.55 µm (model 2)

AOT at 0.55 µm (model 3)

AOT at 0.55 µm (model 4)

Legendre coefficient 𝒑 𝛉 =

𝒏=𝟎

∞

𝟐𝒏 + 𝟏 . 𝒌𝒏. 𝑷𝒏 𝝁

𝝁 − cosine scattering angle.

𝒌𝒏 − n-th Legendre coefficient.

𝑷𝒏 − n-th order of Legendre polynomial.

AOT retrieval

𝜏𝑎𝑒𝑟 𝜆 =4𝜇𝑠𝜇𝑣𝑃𝑎𝑒𝑟 𝜆

𝜔𝑜𝑝 θ

Ref. ind. real and imaginary, and effective

radius at 0.439 µm, 0.676 µm, 0.869 µm, and

1.02 µm

3.0 – RESULT &

DISCUSSION

VALIDATION OF MODIS AOT 500 M USING AOT FROM

AERONET STATION

R = 0.48RMSE = 1.47

R = 0.48RMSE = 1.47

R = 0.86RMSE = 0.56

R = 0.86RMSE = 0.56

R = 0.89RMSE = 0.09

R = 0.89RMSE = 0.09

SBDARTSBDART

Direct Model -1Direct Model -1 Direct Model -2Direct Model -2

R = 0.74RMSE = 0.99

R = 0.74RMSE = 0.99

Direct Model -3Direct Model -3

R = 0.77RMSE = 0.81

R = 0.77RMSE = 0.81

Direct Model -4Direct Model -4

Low accuracy against AERONET

AOT.

The accuracy varies with local

aerosol models.

It is because an improper account to

molecular effects in RT calculation

(Kokhanovsky & de Leeuw, 2009).

Low accuracy against AERONET

AOT.

The accuracy varies with local

aerosol models.

It is because an improper account to

molecular effects in RT calculation

(Kokhanovsky & de Leeuw, 2009).

High accuracy against AERONET

AOT.

Provide AOT with better

performance and less error.

It is because of RT code has the

ability to solve the complexity of RT

equations with rigorous computation

in order to minimize substantial error

(Kokhanovsky and de Leeuw, 2009).

High accuracy against AERONET

AOT.

Provide AOT with better

performance and less error.

It is because of RT code has the

ability to solve the complexity of RT

equations with rigorous computation

in order to minimize substantial error

(Kokhanovsky and de Leeuw, 2009).

DISCUSSION

SBDART code Direct retrieval

MODIS AOT 500 M VS MODIS AOT PRODUCT

MODIS AOT 500 M VS AERONET AOTMODIS AOT PRODUCT VS AERONET AOT

R = 0.94RMSE = 0.09

R = 0.94RMSE = 0.09

R = 0.90RMSE = 0.11

R = 0.90RMSE = 0.11

AOT Spatial Distribution

Comparison of spatial distribution of MODIS AOT 500 m and MODIS AOT product

MODIS AOT 500 m

Good spatial information

and high spatial resolution

(500 m).

No missing pixels aredetected.

Poor spatial information and

lower spatial resolution (10

km).

lot of missing pixel especially

in urban and industrial areas.

Due to bright pixels was

discarded in the retrieval

algorithm.

MODIS AOT product (10 km)

4.0 – CONCLUSION

CONCLUSION

MODIS AOT generated from SBDART code (RT code)

agrees very well with the AOT from AERONET

measurement.

It showed better accuracy and small error compared to

MODIS AOT generated from direct approach.

Considering the reasonable accuracy, high spatial

resolution and good spatial distribution, it can be

concluded AOT is possible to be estimated from MODIS

500m using RT code.

MODIS AOT generated from SBDART code (RT code)

agrees very well with the AOT from AERONET

measurement.

It showed better accuracy and small error compared to

MODIS AOT generated from direct approach.

Considering the reasonable accuracy, high spatial

resolution and good spatial distribution, it can be

concluded AOT is possible to be estimated from MODIS

500m using RT code.