i Detection of Smoke and Dust Aerosols Using Multi-sensor Satellite ...

More documents

Recommendations

Info

is up to 1 km. By comparing with MODIS true color image, the core part of smoke/dust pixels are identified correctly except some missing pixels which have relatively low intensity. Since the quality of MODIS AOD products is not good enough for validating products in sample areas, measurements of multiple other sensors operated in the A-train orbit are selected for quantitative validation. Validated with OMI UVAI product, most smoke plumes are detected accurately. One main factor impacting the accuracy of algorithm may attribute to the spatial resolution difference between two sensors. The validation of dust aerosol monitoring with CALIPSO VFM product is also performed. More than 90% dust aerosol pixels are identified correctly in the select case. Although there are some pixels misclassified, most of them concentrate at the edge of dust storm with light dust aerosol loading. The algorithm also works well in the areas close to cloud or mixed with cloud. On the other hand, the deficiency of detecting small smoke aerosol by MODIS is found by comparing with ASTER and Hyperion measurements. Since both CALIPSO and Aqua MODIS are operated in the A-train orbit with similar local equatorial crossing time, a simple approach is developed to detect dust aerosols combing their measurements. The MODIS BTD (12, 11) is an effective index that can be used for differentiating dust aerosol and cloud in both day and night time. After spatial registration, the overlapped areas are divided into two categories according to MODIS BTD (12, 11) values. With the comparison between measurements of two sensors, if a aerosol layer is labeled as cloud in CALIPSO VFM product it located in the area having positive BTD (12, 11μm) values, it is corrected and identified as dust aerosol. 121
Based on the detection results, the vertical and horizontal information about several dust aerosols occurred in 2008 spring season are summarized. The MODIS spatial characterization with the SRCA and ground target approach is introduced and the impact of BBR shift, or mis-registration, on MODIS L1B measurements and dust aerosol detection is assessed. The impacts on L1B are generally small and are different in both directions due to the strong bow-tie effect in scan directions. The largest relative error of all RSBs in Aqua MODIS is less than 0.1% in the select case, which is negligible in real applications. The theoretical analysis shows that the influence of mis-registration on dust detection is larger than on L1B measurements. The influence on dust aerosol detection results is small at the homogenous or semi-homogenous areas but relative large at the mixed areas. The increase of correlation coefficient demonstrates that the quality of science data product can be improved by shifting a pixel in track direction. This study provides valuable information for the sensors without spatial characterization capability and also for the specification design of future sensors. 8.2 Limitations This dissertation is executed in some limited condition due to the data availability in time and space. In the absence of suitable laboratory and field data, such as the spectral curves of the smoke, some results are thus mainly based on the model simulation and limited available datasets. Statistical analysis of the training data is a primary ways in this research to achieve the desired spectral features of the smoke/dust. Although the 122
Page 1 and 2:
Detection of Smoke and Dust Aerosol
Page 3 and 4:
DEDICATION This is dedicated to my
Page 5 and 6:
TABLE OF CONTENTS Page LIST OF TABL
Page 7 and 8:
LIST OF TABLES Table Page Table 1.1
Page 9 and 10:
Figure 6.2 The UVAI images and the
Page 11 and 12:
VFM Vertical Feature Mask VIS Visib
Page 13 and 14:
are validated not only visually wit
Page 15 and 16:
1.1 Importance of Studying Smoke an
Page 17 and 18:
1.3 Objectives and Scopes The main
Page 19 and 20:
originalities, and discussions of f
Page 21 and 22:
1.6 Principal Results The principle
Page 23 and 24:
CHAPTER 2 REVIEW OF APPROACHES FOR
Page 25 and 26:
cooperatively: the reflectance at 0
Page 27 and 28:
more information about dust charact
Page 29 and 30:
CHAPTER 3 SMOKE AEROSOL DETECTION W
Page 31 and 32:
with a two-side scan mirror which r
Page 33 and 34:
3.1.2 MODIS data The 55 0 scan angl
Page 35 and 36:
Sun Satellite Figure 3.2: Basic rad
Page 37 and 38:
1 0 /2 b ( 0 0, 0; , ) k1 e
Page 39 and 40:
3.2.3 Training data collection Surf
Page 41 and 42:
Reflectance Reflectance Response cu
Page 43 and 44:
Figure 3.6: The normalized ratio of
Page 45 and 46:
all pixels over both land and ocean
Page 47 and 48:
NDDI 1 0.8 0.6 0.4 0.2 0 -0.2 Figur
Page 49 and 50:
MODIS L1B measurements Pixels over
Page 51 and 52:
is clear that most of smoke plumes
Page 53 and 54:
a b c b Figure 3.10: Smoke images o
Page 55:
3.4.2 California 2007 fire The 2007
Page 58 and 59:
The multi-spectral approach is test
Page 60 and 61:
4.1 Physical Principle of Dust Aero
Page 62 and 63:
Aqua 2006 102-04/12 04:15 Terra 200
Page 64 and 65:
The dust has a high reflectivity at
Page 66 and 67:
0.55 0.5 0.45 0.4 0.35 0.3 0.25 0.2
Page 68 and 69:
7 a Dust pixel 0.5 b 6 Cloud pixel
Page 70 and 71:
4.3 Results Three dust events occur
Page 72 and 73:
c c Figure 4.6: Dust storm over TAK
Page 74:
4.4 Chapter Summary In this chapter
Page 77 and 78:
operated in the A-train orbit with
Page 79 and 80:
5.1.2 CALIPSO measurements and VFM
Page 81 and 82:
Altitude (km) 20 15 10 5 UTC: 2006-
Page 83 and 84: After spatial registration, the ove
Page 85 and 86: Figure 5.4(a) The MODIS true color
Page 87 and 88: Altitude (km) BTD (K) 20 15 10 15 4
Page 89 and 90: in the same A-train orbit with simi
Page 91 and 92: Aqua spacecrafts are operated in th
Page 93 and 94: the Fig. 6.1a. For easy comparison,
Page 95 and 96: esolutions, the number of smoke pix
Page 97 and 98: 6.1.4 Validation of dust monitoring
Page 99 and 100: Figure 6.3 Validation of dust image
Page 101 and 102: eflectance in red band. Therefore,
Page 103 and 104: Altitude (km) 20 15 10 5 UTC: 2006-
Page 105 and 106: image at the same location. The rea
Page 107 and 108: accuracy could be increased further
Page 109 and 110: FPAs and cold FPAs (because of diff
Page 111 and 112: development of future remote sensin
Page 113 and 114: Figure 7.3: SRCA spatial mode retic
Page 115 and 116: 7.1.3.2 Date Source MODIS L1B data
Page 117 and 118: characterization. Only the measurem
Page 119 and 120: these bands, including reflective s
Page 121 and 122: different spectral bands are slight
Page 123 and 124: SZA (Degree) SZA (Degree) 65 60 55
Page 125 and 126: 7.2.1.2 Real Impact on L1B Measurem
Page 127 and 128: 7.2.2.1 Theoretical impact analysis
Page 129 and 130: spatial characterization, the bigge
Page 131 and 132: Difference NDDI with spatial correc
Page 133: 8.1 Conclusions CHAPTER 8 CONCLUSIO
Page 137 and 138: 3) Validating aerosol detection qua
Page 139 and 140: REFERENCES Ackerman, S. A., 1989, U
Page 141 and 142: imagery, Atmospheric Environment, v
Page 143 and 144: Salomonson, V. V., Privette, J. L.,
Page 145 and 146: Martinez, A., and Gros, G., 2007-10
Page 147 and 148: Wang W., Qu, J. J., Liu, Y., Hao, X
Page 149: CURRICULUM VITAE Yong Xie is a Ph.D
show all

i Detection of Smoke and Dust Aerosols Using Multi-sensor Satellite ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?