Visual Analysis of the Air Pollution Visual Analysis of the Air - - PowerPoint PPT Presentation

CHAN Wing Yi, Winnie [Represented by MAK Wai Ho, Wallace]

Visual Analysis of the Air Pollution Visual Analysis of the Air Pollution Problem in Hong Kong Problem in Hong Kong

Hong Kong ICT Awards 2007: Best Innovation and Research Award ICT07/ IR/ CU-18

Department of Computer S cience and Engineering The Hong Kong University of S cience and Technology (HKUS T)

2

Preface

• This is the work of a final year thesis

(research option of final year proj ect)

• The research paper will appear in

IEEE Transactions on Visualization and

Computer Graphics (TVCG).

Visual Analysis of the Air Pollution Problem in Hong Kong

Huamin Qu, Wing-Yi Chan, Anbang Xu, Kai-Lun Chung, Kai-Hon Lau, Ping Guo

IEEE Transactions on Visualization and Computer Graphics (TVCG), vol.13, no. 6, Nov.-Dec. 2007 (Proceedings of IEEE Visualization/ Information Visualization 2007)

3

Outline

• Introduction

▫ Background ▫ Uniqueness of Air Quality Data

• Visualization Techniques
• Experimental Results
• Conclusions and Future Work

4

Introduction

• Visualization

▫ Presents data in pictorial form ▫ Visualizes the underlying data

effectively

• Visual analysis

▫ Is a visual way for data mining

and decision making

▫ Performs analysis on the

visualization result

5

Hong Kong Air Pollution Problem

• Hong Kong air quality is

decreasing tremendously

• Air pollution problem

becomes one of the biggest social issues

• Causes are still unknown

▫ Many hypotheses are

proposed without any formal proof yet

The spectacular harbor view has been increasingly crippled by massive haze.

6

Institute for the Environment of HKUST

• Maintain a comprehensive database on Hong Kong

air quality data

• Cannot obtain convincing results for high-level

correlations with mathematical techniques

• Demand visualization techniques for analysis

7

Uniqueness of Air Quality Data

• Time-series (hourly-based)
• Inherited geographic

information

• Multi-dimensional

(typically >10 attributes)

• Important vector field –

wind speed and direction

• 1. Precipitation
• 2. Wind Direction
• 3. Air Temperature
• 4. Wind S

peed

• 5. Dew Point
• 6. Relative Humidity
• 7. S

ea Level Pressure

• 8. Respirable S

uspended Particulates (RS P)

• 9. Nitrogen dioxide (NO2)
• 10. S

ulphur dioxide (S O2)

• 11. Ozone (O3)
• 12. Carbon monoxide (CO)
• 13. S
• lar Radiation
• 14. Air Pollution Index (API)
• 15. Contributed Pollutant to API

(S pans more than 10 years)

8

Outline

• Introduction
• Visualization Techniques

▫ Polar S

ystem

▫ Parallel Coordinates ▫ Weighted Complete Graph

• Experimental Results
• Conclusions and Future Work

9

Outline

• Introduction
• Visualization Techniques

▫ Polar S

ystem

▫ Parallel Coordinates ▫ Weighted Complete Graph

• Experimental Results
• Conclusions and Future Work

10

Polar System

• Is a common vector

representation

• Is heavily applied by

domain scientists in environmental field

Distance from center Wind S peed Angle from the north Wind Direction Color S calar Attribute

very strong south wind high attribute value weak southwest wind low attribute value

11

Circular Pixel Bars

• Users select a sector to plot the inside-sector data

(i.e. of certain wind direction and speed)

• The corresponding wind direction and wind speed is
• bvious for rapid comparisons between sectors

12

Outline

• Introduction
• Visualization Techniques

▫ Polar S

ystem

▫ Parallel Coordinates ▫ Weighted Complete Graph

• Experimental Results
• Conclusions and Future Work

13

Parallel Coordinates

• Parallel Coordinates are well-established visualization

tool for multi-dimensional data

• Each parallel vertical axis represents an attribute
• A data item is plotted by a polygonal line intersecting

each axis at the respective attribute data value

14

S-Shape Axis for Vector

Traditional layout (not intuitive) Circular layout (lots of overlapping) S-style layout An example

15

Outline

• Introduction
• Visualization Techniques

▫ Polar S

ystem

▫ Parallel Coordinates ▫ Weighted Complete Graph

• Experimental Results
• Conclusions and Future Work

16

Weighted Complete Graph

• It is used for exploring overall relationship

among all data dimensions

• Each node represents one data dimension
• Distance between nodes encodes

their correlation

A B C

correlated not really correlated

17

Outline

• Introduction
• Visualization Techniques
• Experimental Results

▫ Correlation Detection ▫ S

imilarities and Differences

▫ Time-S

eries Trend

• Conclusions and Future Work

18

[solar radiation]

Correlation Detection

• RS

P is correlated with S O2 and O3, but not solar radiation

• High API value (red pixels) are not found when S

O2 is high, inferring that S O2 contributed little to API

• API is strongly correlated with O3 which is known to experts
• S
• me suspicious clusters are shown in [S

O2] and [O3] - a blue cluster is seen behind a green one

[SO2] [O3]

Color = Air Pollution Index (API)

19

Similarities and Differences (1)

• The Hong Kong society mostly weighs external

pollution factors more ▫ Pollutants blown in from factories on the Pearl

River Delta at the northwest of Hong Kong

• Local pollution is often ignored

▫ Power plants ▫ Vehicles and vessels

20

• High S

O2 for most stations:

▫ S

trong wind

▫ Northwest wind ▫ External S

• urces
• High S

O2 for Kwai Chung:

▫ All wind speed ▫ S

• uthwest wind

▫ Internal sources likely due

to cargo ships at Kwai Tsing Container Terminals

Similarities and Differences (2)

9 stat ions of 3 years data Color represents amount of S O2

21

Time-Series Trend for Tung Chung

• 2004 and 2005 plots are more similar
• In 2006 plot, temperature varies dramatically

22

Positive Feedback from Users

• Domain scientists found that

the polar system with embedded pixel bar offers easy navigation to explore the data interactively

• Parallel coordinates show

the general relationship for them to compare different data-sets rapidly

• Weighted complete graph

provides correlation

• verview that is useful for

initiating an analysis

23

Outline

• Introduction
• Visualization Techniques
• Experimental Results
• Conclusions and Future Work

24

Conclusions

• Comprehensive System

▫ The first attempt designed for air quality analysis

• Novel Techniques

▫ Polar system with circular pixel bars: scalar + vector ▫ Enhanced parallel coordinates: vector + time axes ▫ Weighted complete graph: correlation overview

• Significant Application

▫ Analyzed Hong Kong air pollution problem ▫ Revealed known findings effectively ▫ Detected unknown patterns by domain scientists

25

Future Work

• Continue as a long-term proj ect with ENVF
• Make the system available to the public on Web
• Incorporate new datasets for further exploration
• Add animations and other visual aids

Thank You!

The End The End

27

Q & A

Polar system with embedded circular pixel bars Weighted complete graph Enhanced parallel coordinates with S

• shape vector axis