Visual Analysis of the Air Pollution Problem in Hong Kong

Huamin Qu1, Wing-Yi Chan1, Anbang Xu3, Kai-Lun Chung1, Kai-Hon Lau2, Ping Guo3

IEEE Transactions on Visualization and Computer Graphics (TVCG), vol. 13, no. 6, pp. 1408-1415, Nov.-Dec. 2007

(Proceedings of IEEE Visualization and Information Visualization 2007)


1 Department of Computer Science and Engineering, The Hong Kong Unviersity of Science and Techology (HKUST)

2 Institute for the Environment, The Hong Kong University of Science and Techology (HKUST)

3 Department of Computer Science, Beijing Normal University (BNU)


full paper vis'07 presentation slides demo on youtube

HKICT Awards 2007:
Best Innovation and Research (College & Undergraduates) Silver Award
Certificate of Merit - Best Innovation Award

Comments from Judging Panel

The project provides strong evidence about the source of pollutants and conditions of weather that cause pollution. The proposed underlying analysis model is sophisticated and the visual representation is very easy to understand for both scientists and the public.

Related Articles




We present a comprehensive system for weather data visualization. Weather data are multivariate and contain vector fields formed by wind speed and direction. Several well-established visualization techniques such as parallel coordinates and polar systems are integrated into our system. We also develop various novel methods, including circular pixel bar charts embedded into polar systems, enhanced parallel coordinates with S-shape axis, and weighted complete graphs. Our system was used to analyze the air pollution problem in Hong Kong and some interesting patterns have been found.


fig1 fig2 fig3a
Figure 1download Figure 2download Figure 3adownload
fig3b fig3c fig4a
Figure 3bdownload Figure 3cdownload Figure 4adownload
fig4b fig5a fig5b
Figure 4bdownload Figure 5adownload Figure 5bdownload
fig6a fig6b fig6c
Figue 6adownload Figue 6bdownload Figue 6cdownload
fig7a fig7b fig7c
Figue 7adownload Figue 7bdownload Figue 7cdownload
fig8 fig9 fig10a
Figure 8download Figure 9download Figue 10adownload
fig10b fig10c fig11
Figure 10bdownload Figure 10cdownload Figure 11download
fig12a fig12b fig12c
Figure 12adownload Figure 12bdownload Figure 12cdownload
fig13 fig14 fig15
Figure 13download Figure 14download Figure 15download
fig16a fig16b fig17a
Figure 16adownload Figure 16bdownload Figure 17adownload
Figure 17bdownload    
Figure 1

Hong Kong's air pollution problem. The spectacular harbor view has been increasingly crippled by a massive haze.

BBC Chinese, September 8 2002
CNN, June 2 2006
Time Asia, May 8 2006

Figure 2

Locations of different air quality monitoring stations shown as dots in 18 districts of Hong Kong.

Figure 3

Traditional polar system: (a) Encoding scheme; (b) Mapping radius without preserving the area; (c) Area-preserving polar system. Blue denotes low values and red denotes high values. The color bar is also used for other figures.

Figure 4

Polar system with circular pixel bar: (a) A sector selected by a user and a circular bar chart embedded into the sector; (b) Blending of circular pixel bar for data falling in the sector against one for its complement.

Figure 5

Comparing circular pixel bars with rectangular ones: (a) A polar system with multiple circular pixel bars; (b) Conventional pixel bars for the sectors. The overall patterns are preserved in the sector for comparison, and in-depth numerical analysis may be performed on the supplement rectangular pixel bars.

Figure 6

Polar system with time information: (a) x-position, y-position, and color of the sector indicate the month of observation, amount of SO2, and temperature, respectively; (b) the x-position now represents the day in which the entry was recorded; (c) the y-position now encodes the day and the x-position encodes the month.

Figure 7

Different layouts of parallel coordinates: (a) Traditional layout; (b) Circular layout; (c) S-style layout.

Figure 8

Enhanced parallel coordinates with S-shape axis to encode wind direction and scatterplot to reveal bi-variate relationship between neighboring axes.

Figure 9

Weighted complete graph: (a) Layout of weighted complete graph with node size encoding the accumulated correlation coefficients and edge encoding the correlation between two nodes. Edges with small weights are removed for clarity. (b) Parallel coordinates with a user-chosen axis order based on the weighted complete graph; (c) Parallel coordinates with a random axis order.

Figure 10

Detecting the correlation between the Air Pollution Index (API) and other attributes when API is high: (a) Initial polar system with color denoting API value. The northwest sector is chosen, plotting RSP against solar radiation. (b) Plotting RSP against SO2 instead, high API value (red pixels) are not found when SO2 is high, revealing SO2 contributed little to API. (c) Y-position now becomes O3 clearly correlated with API. For (b) and (c), suspicious clusters (blue clusters behind green ones) are shown.

Figure 11

Detecting correlations of the same set of data by Parallel Coordinates, with color denoting API value.

Figure 12

(a) Tracking the possible internal and external pollution sources through nine stations in the past three years. Pixel color represents the amount of SO2 recorded in each individual station. (b)-(c) The detailed plots for station Tung Chung and Kwai Chung respectively.

Figure 13

Comparing two stations, Kwai Chung and Tung Chung with parallel coordinates using color to represent wind direction. Clusters of wind direction records are found in Tung Chung station but not in Kwai Chung.

Figure 14

Visualizing time-series data. Each row is comprised of polar plots for different stations, namely Tung Chung, Yuen Long and Mong Kok, in different periods of time from March 2004 to March 2007 at intervals of six months. The pixel color denotes the Air Pollution Index (API).

Figure 15

Time-series polar plots for Kwai Chung station focusing on the impact of local pollution from the southwest direction. X-position, y-position, and color of the sector encode day, SO2, and API accordingly. Prominent red pixels are mainly seen in year 2004.

Figure 16

3-year time-series data of Yuen Long district that are constrained to a range of wind speed and direction by sector selection: (a) Weighted complete graph for each year with edge width encoding the correlation strength; (b) Parallel Coordinates with a time axis. Color also denotes time value for clarity.

Figure 17

Using a weighted complete graph as a visual aid in exploring dimension correlations for year 2006 data of the Yuen Long station: (a) By arranging more correlated attributes together, positive and negative correlations are clearly shown in the parallel coordinates; (b) Users can also plot the attributes demonstrating interesting relationships in the weighted complete graph as the embedded pixel bar in the polar system.