DustDection_liuyang_Final.ppt

A Thermal Index from MODIS Data for Dust Detection

Yang Liu, Ronggao LiuInstitute of Geographic Sciences and Natural Resources Research, CAS

2011.7.28 • Vancouver

Outline

Dust storm detection from satellite data

Method

Results

Summary

Dust Storms in Earth Research

With increasing land degradation and deforestation, dust storms spread in the world since the 20th century.

air pollution public health: respiratory diseases

global climate

biogeochemical cycle

Characterization of the dust properties and distribution at global scale helps to understand their roles on the Earth radiative budget and global biogeochemical cycle

Remote sensing in dust storm monitoring

Satellite remote sensing is advantageous in monitoring the global spatial and temporal variations of dust storms.

ultraviolet channels (Hsu et al., 1996; Chiapello et al., 1999)

visible, infrared channels (Miller, 2003; Hsu et al., 2006)

thermal infrared channels (Ackerman, 1997; Legrand et al., 1992; Schepanski et al., 2007; Walker et al., 2009)

Challenges in dust storm monitoring

Visible/ultraviolet band-based algorithm:

blue or ultraviolet bands are required for bright surface

Problems:

1) not reliable for nighttime;

2) no observations in these bands of many sensors, such as AVHRR (long-term dataset)

Thermal Infrared band-based algorithm:

Dust: brightness temperature difference (eg. BTD12,11, BTD8.6,11)

Problems:

the BTD is also related to other factors, such as land surface temperature and emissivity

Our work

Aim: develop an algorithm for dust storm monitor based on thermal infrared bands measurements of MODIS with consideration of the effects of LST on BTD.

Data MODIS brightness temperature in 11 (Band 31), 12 (Band 32), 8.6 (Band 29) Method Normalized BTD12,11, BTD8.6,11 with consideration of the effects of LST on BTD Results algorithm application over major land cover types over China

mµ mµ mµ

Outline

Dust storm detection from satellite data

Method

Results

Summary

Spectral response: dust, cloud and surfaceBTD12,11=BT12-BT11

Dust (BTD12,11>0)

Clear-sky surface(BTD12,11~0)

Cloud(BTD12,11<0)

BTD12,11 could separate dust from cloud, which affected by the LST

Spectral response: dust, cloud and surfaceBTD8.6,11=BT8.6-BT11

Dust/Cloud(BTD8.6,11>0)

Clear-sky surface(BTD8.6,11~0)

Bright surface(BTD8.6,11<0)

BTD8.6,11 helps to separate dust from bright surface

(1) Construction of pixel-by-pixel relationships of BT11-BT12 and BT8.6-BT11

The ratio of BT in 8.6 and 11, as well as 12 and 11 bands are mapped pixel-by-pixel based on clear-sky observations during 2000-2008;

(2) BTD normalization

The BTD12,11 and BTD8.6,11 are normalized to clear-sky condition using the pixel-based ratio relationship, and then used to calculate the difference (ΔBTD) with the observed BTD.

(3) Dust determination

Methods

11,6.811,12 BTDBTDDustIndex ∆+∆=

Methods

BTD12,11 and BTD8.6,11 over clear-sky surface

After normalization, BTD12,11 and BTD8.6,11 concentrated around 0

0

200000000

400000000

600000000

800000000

1000000000

-6 -4 -2 0 2 4

BTD(K)

Freq

uenc

y

NBTD29

NBTD32

BTD29

BTD32

Methods

DustIndex (DI) of dust, cloud and clear-sky surface

Dust: DI>0; Cloud: DI<0; Clear-sky surface: DI around 0

0

1

2

3

4

5

6

-10 0 10 20 30 40 50 60 70

DustIndex

Aer

osol

Thi

ckne

ss

Cloud

Dust

Clear-sky surface

Outline

Dust storm detection from satellite data

Method

Results

Summary

Results 1-Normalized BTD12,11

BT11=260K

BT11=280K

EstClearBTD 11,12

The BTD12,11 is related to surface temperature

Results 2-Normalized BTD8.6,11

EstClearBTD 11,6.8

BT11=260K

BT11=280K

The BTD8.6,11 is related to surface temperature

Results 3-performance over vegetated surface

a) Terra MODIS true-color image of dust storm over Northeast China on April 7, 2001; b) Dust detection results

DI could separate the airborne dust from clouds, vegetated surface and ocean

Results 4-performance over bright surface (1)

a) Terra MODIS true-color image shows dust storm in Taklimakan desert. b) Dust detection results

DI could separate airborne dust from sand and cloud over bright desert surface

Results 4-performance over bright surface (2)

c) Terra MODIS true-color image shows dust storm in Gobi desert;d) Dust detection results

DI could separate airborne dust from sand and cloud over bright desert surface

Results 5-with and without normalization (Day)

DOY97, 2001 over China

Without normalization

With normalization

Desert

Daytime

Algorithm could reduce the effects of bright surface on dust detection

Results 6-with and without normalization (Night)

NighttimeWithout normalization

With normalization

DOY97, 2001 over China

Algorithm could detect the airborn dust in nighttime

Dust storm detection from satellite data

Method

Results

Summary

Outline

Summary

A new algorithm has been developed to detect dust based

on satellite thermal infrared imagery;

here uses the brightness temperature of three thermal infrared

channels of MODIS, including 8.6, 11 and 12

The algorithm considers the effects of LST on BTD;

The algorithm could distinguish airborne dust from cloud

and land surface over bright and dark surface during daytime

and nighttime.

Thank you！