Yuh-Lurng Chung, Chaur-Tzuhn Chen

1
Study on applying MODIS image into drought
indicator analysis in Taiwan

• Yuh-Lurng Chung, Chaur-Tzuhn Chen
• Chen-Ni Hsi , Shih-Ming Liu
• 2004.11.04

2
Introduction
When drought occurs, due to insufficient
water supply, the variation/change of leaves in
aridity can be sensed by spectral reflection of
multi-temporal satellites.
3

• Overseas researches of employing
  satellite images to efficiently forecast and
  manage drought has achieved great outcomes.

4
This research employ MODIS images to
select sensitive thermal bands suitable for
monitoring drought. And by using bands for the
calculation of all kinds of vegetation indices,
it is expected to find proper and practical
indices for drought monitoring, which could be
used for future management and determination of
drought disaster.
Introduction
5
Using bands for the calculation of all kinds
of vegetation indices, it is expected to find
proper and practical indices for drought
monitoring, which could be used for future
management and determination of drought disaster.
6
Research Data/Information and Methodology
1. Study Area
Include whole Taiwan area of 19 districts.
2. Required Data/Information

• Precipitation Data

Information concerned includes the records of all
rainfall stations from the January of 1991 to the
March of 2004 in the entire Taiwan area.
7
(No Transcript)
8

• MODIS Images

Thermal Bands of MODIS Image
Seven MODIS Bands for Monitoring Earths Surface
Band Band Width (µm) Central Wavelength (µm) Required Ne?T (K) Primary use
20 3.660-3.841 3.7882 0.05 O,L
21 3.929-3.989 3.9921 2.00 Fire,volcano
22 3.929-3.989 3.9719 0.07 A,L
23 4.020-4.080 4.0567 0.07 A,L
29 8.400-8.700 8.5288 0.05 L
31 10.78-11.28 11.0186 0.05 A,L
32 11.77-12.27 12.0325 0.05 A,L
Aatmospheric studies, Lland studies,
Ooceanstudies
9
By utilizing the information of surface
regression through Kriging Model, our approach
then can get drought indicator and drought amount
of this area. Two images of the dry season
(January 25, 2004) and wet season (June 30, 2004)
are accordingly chosen for further analysis.
10
Research Methodology
Data from rainfall stations
Select clear MODIS images without cloud
Rainfall of continuous 30 days
Threshold of Each County and City
Preprocessing of MODIS images
Cumulative rainfall of all
Locate sample grassland areas
NO
Drought Amountgt130mm
Use MODIS images of dry and wet seasons to select
thermal bands for drought monitor
YES
Calculate all indices select some for
preliminary analysis
Climate Drought Indices
Specify Ranges of Dry and Wet Seasons Using
Climate Drought Indices
Choose index for drought
11
Discussions on Applying Drought Indices to
Drought Monitor

• Normalized Thermal Index (NTI)

• Normalized Difference Vegetation Index (NDVI)

12

• Normalized Difference Water Index (NDWI)

• Shortwave Infrared Water Stress Index (SIWSI)

13
Results and Discussion
Characteristics of Taiwan Rainfall Data
Based on the rainfall data of 355
rainfall stations from 1992 to 2003, clearly
shows different standards (levels) of different
places in different periods. It also indicates
the relativity of drought definition due to
spatial and temporal factors.
14
Historical Curves of the first decile values of
Cumulative
15
Application of MODIS Images to Select Thermal
Bands for Drought Monitor
MODIS Images After Geometric Correction
Original MODIS Image
16
Extraction of Sample Sites of Grasslands
From land-use maps we query all natural
grasslands from the database of ArcGIS. And after
removing those sample cloud hovering, we mark
those sample sites on the extracted images of
natural grasslands without cloud covered.
Legend
Sample
Boundary
17
Select Thermal Bands of MODIS images for Drought
Monitor
Extracts data for the seven MODIS bands, and
compares the seven bands of dry season and wet
season to find out what are the real differences.
18
Differences of Thermal Infrared Band Values of
MODIS Images of Grasslands in Dry and Wet Seasons
Differences of Mean Thermal Infrared Bands of
Taiwan Grasslands MODIS Images in Dry and Wet
Seasons
19
The Calculation of Normalized Thermal Index (NTI)
The research done by Robert et al. (2002)
about monitoring volcano indicates that NTI value
are ranged between -0.850 -0.950 due to the
high surface temperature of the volcanic region.
NTI Image and Histogram In Dry Season
NTI Image and Histogram In Wet Season
20
Statistics of NTI Values of MODIS Images in Dry
and Wet Season, and T-test
MODIS NTI Images Sample Numbers NTI Average P(Tltt) Difference
In Dry Season 36600 -0.199 0 Distinct
In Wet Season 36600 0.058 0 Distinct
Note T-test with confidence interval 5
21
The Calculation of Normalized Difference
Vegetation Index (NDVI)
MODIS NDVI Images in Dry Season
MODIS NDVI Images in Wet Season
22
NDVI Differences of MODIS Grassland Images in Dry
and Wet Seasons
Statistics of NDVI Values of MODIS Images in Dry
and Wet Season, and T-test
MODIS NDVI Images Samples NDVI Average P(Tltt) Difference
In Dry Season 78 0.88 2.84E-12 significant
In Wet Season 78 0.94 2.84E-12 significant
23
The Calculation of Normalized Difference Water
Index (NDWI)
MODIS NDWI Images in Dry Season
MODIS NDWI Images in Wet Season
24
Difference of MODIS NDWI Images of Grasslands in
Dry and Wet Seasons
Statistics of NDWI Values of MODIS Images in Dry
and Wet Seasons, and T-test
MODIS NDWI Images Samples NDWI Average P(Tltt) Difference
In Dry Season 78 0.52 7.67E-08 significant
In Wet Season 78 0.63 7.67E-08 significant
25
Difference Between NDVI and NDWI of MODIS Images
of Grasslands in Dry and Wet Seasons
26
Correlation Matrix Between NDVI and NDWI in DS
and WS
NDVI in Dry Season NDVI in Wet Season NDWI in Dry Season NDWI in Wet Season
NDVI in DS 1.000
NDVI in WS 0.079 1.000
NDWI in DS 0.461 0.045 1.000
NDWI in WS 0.197 0.799 0.023 1.000
Distinct Correlation as distinction level if
0.01(). DS Dry Season WS Wet Season
27
Calculation of Shortwave Infrared Water Index
(SIWSI)
MODIS SIWSI Images of Taiwan in Dry Season
MODIS SIWSI Image of Taiwan in Wet Season
28
SIWSI Differences for Sample Sites of Grasslands
in Dry and Wet Seasons, based on MODIS Images
Statistics and T-test Table of SIWSI Values of
grasslands in Dry and Wet Seasons (MODIS Images)
MODIS SIWSI Images Number of Samples SIWSI Mean P(Tltt) differences
Dry Season 78 -0.29 0.3214 Insignificant
Wet Season 78 -0.31 0.3214 Insignificant
29
Conclusion
This research indicates MODIS images with 36
bands have substantial potential in drought
sensing. It is hereby possible to replace the
NOAA-AVHRR satellite images with MODIS images,
for more precise image data/information.
30
The MODIS Band 22 at the spatial resolution
of 1,000 m is the most sensitive thermal bands to
drought. And the NTI is the unique index of
sensing thermal energy only available in MODIS
images. Furthermore, this research hence utilizes
the important wave bands which are chosen from
the Band 22 to calculate NTI.
31
We can conclude that the NDVI, NDWI and
NTI are sensitive to the monitoring of surface
vegetation status, water content of vegetation
and surface temperature respectively. As a
result, they hereby have practical usages for
drought forecast and monitoring.
32
The End