Entry
Name: Qianxin-Yu-MC2
VAST
Challenge 2019
Mini-Challenge 2
Team Members:
Zhengyan Yu, Qianxin, yuzhengyan@qianxin.com
PRIMARY
Lingjun He, Qianxin, helingjun01@qianxin.com
Zhaokang Yuan, Qianxin, yuanzhaokang@qianxin.com
Student Team: No
Tools Used:
Echarts.js
D3.js
Turf.js
ZRender.js
Javascript
Python
R
Google Chrome
Radiation Measurement Dashboard, designed and
developed by the team
Approximately how many hours were spent
working on this submission in total?
240 hours
May we post your submission in the Visual
Analytics Benchmark Repository after VAST Challenge 2019 is complete? Yes
Video
Questions
Your task, as supported by visual analytics that you apply,
is to help St. Himark's emergency management team
combine data from the government-operated stationary monitors with data from
citizen-operated mobile sensors to help them better understand conditions in
the city and identify likely locations that will require further monitoring,
cleanup, or even evacuation. Will data from citizen scientists clarify the
situation or make it more uncertain? Use visual analytics to develop responses
to the questions below. Novel visualizations of uncertainty are especially
interesting for this mini-challenge.
1.Visualize radiation measurements over time from both static and
mobile sensors to identify areas where radiation over background is detected.
Characterize changes over time. Limit your response to 6 images and 500 words.
On 6th, the overall range of the average radiation measurement is from 0 to 50. The radiation of the city seems to be normal and the difference of the measurements of each region (presented as squares) is little. It's worth mentioning that the radiation measurement near Always Safe Loop is as low as in other regions in the city.
The average radiation measurement on 4/6
On 7th, the overall range of the average radiation measurement is from 0 to 59. The radiation measurement is similar to 6th.
The average radiation measurement on 4/7
On 8th, the overall range of the average radiation measurement is from 0 to 558. The radiation measurement is obviously greater than the previous days. The greatest radiation measurement show on the conjunction of Jade Bridge and the city (the average radiation measurement of the region is around 557), and the regions near Always Safe Loop(the average radiation measurement of the region is around 373). Except for the two regions, there is no other region having unusual high radiation measurement.
The average radiation measurement on 4/8
On 9th, the overall range of the average radiation measurement is from 0 to 1316. The radiation measurement is greater than the previous days. The region which has the greatest average radiation measurement is near the left side of Wilson Forest Hwy, and the value of the measurement is around 1316. At the right side of Wilson Forest Hwy, the average radiation measurement is 245.
On the down right side of the city, there is one region which has unusual high radiation measurement and the measurement value is 141.
The regions near Always Safe Loop are also higher than the usual radiation measurement. The range of these regions is roughly from 70 to 100.
The average radiation measurement on 4/9
On 10th, the overall range of the average radiation measurement is from 0 to 1400. The radiation measurement on 10th is the greatest in the five days. The pattern of the radiation measurement on 10th is similar to 9th.
The regions of Wilson Forest Hwy still have high radiation measurement. The greatest measurement value on 10th is also on the left side of Wilson Forest Hwy and the measurement value is around 1399. The right side of Wilson Forest Hwy has the second greatest radiation measurement which value is around 307.
On the bottom right of the city, there is one region having high radiation measurement and the value is around 195.
The radiation measurement around Always Safe Loop is still higher than in other regions with normal radiation measurement. The measurement of these regions near Always Safe Loop is around 70 to 80.
The average radiation measurement on 4/10
2.Use visual analytics to represent and analyze uncertainty in the
measurement of radiation across the city.
a. Compare uncertainty of
the static sensors to the mobile sensors. What anomalies can you see? Are there
sensors that are too uncertain to trust?
b. Which regions of the city
have greater uncertainty of radiation measurement? Use visual analytics to
explain your rationale.
c. What effects do you see in
the sensor readings after the earthquake and other major events? What effect do
these events have on uncertainty?
Limit your responses to 12 images and 1000 words.
The
definition of uncertainty and anomaly
We define the uncertainty as a normal range of radiation measurement.
Generally speaking, the radiation measurement is positive and it should have a data entry every five seconds. Also, the radiation measurement is not a constant value, it is going up and down continuously. In addition, the range is also affected by many other uncertain reasons, e.g. the quality of the sensors.
In most cases, the radiation measurement falls in a certain range. For example, the normal radiation measurement is 20 cpm ± 10 cpm. The range of the measurement (positive 10 and negative 10) is uncertainty. We use quartiles to define the reasonable range of radiation measurement. Thus, uncertainty equals to 4 times of IQR(interquartile range).
However, in some unusual cases, uncontrolled issues happen and cause the sensors to detect the unusual value of radiation, which we call it anomalies. For examples, the leak of coolant, and the malfunction of sensors.
Some anomalies are acceptable because they are caused by reasonable reasons. For examples, the leak of coolant causes the greater radiation measurement of some mobile sensors. Nevertheless, some anomalies are not acceptable because they are caused by unknown/unreasonable reasons, for example, the noises hidden in the radiation measurement.
a.
Compare uncertainty of the static sensors to the mobile sensors. What anomalies
can you see? Are there sensors that are too uncertain to trust?
(1)Noises:
unreasonable high radiation values
Satic sensors generally have fewer noises than mobile sensors.
Select all mobile and static sensors and randomly
choose an hour. (Due to the limitation
of the web browser, it is impossible to plot all of the data points, thus we
have to reduce the amount of data points.)
By exploring all of the sensors and the data points, we found that the noise values of the most sensors are from 200 to 2500. However, there is one noise of sensor no.12 which has an extremely high value. The value is close to 60,000. We believe this noise should be removed from the datasets.
(2)Constant radiation values
In normal cases, all of the radiation measurements have a variation. However, some mobile sensors start to detect constant radiation values at a specific time. These mobile sensors are no.1, 23, 26, 35 and 47. We believe, the constant values of these sensors are not trustworthy.
(3)Missing radiation values
Every sensor should have a data entry every five seconds. We found that many mobile sensors have missing radiation measurements probably due to their quality. Take mobile sensor no.5 as an example.
The blue dots mean the radiation measurements have positive values and the red dots means the measurements have zero values. When the blue dots and red dots are overlapping, it generally means this sensor is in the city and it has missing radiation measurements.
(There is a period that only red dots are shown. It is because this sensor is outside of the city.)
As for static sensors, their quality is much higher than mobile sensors so most of the static sensors do not have missing radiation measurements. However, there is one exception. Static sensor no.15 has missing radiation measurements during a period of time. We infer that static sensor no.15 is broken during that time.
(4)Negative radiation values
The values of radiation measurements should be positive (or zero). However, there are two static sensors have one negative value individually: static sensor no.4 (marked as 1 below) and no.14 (marked as 2 below).
We assert that the negative values of radiation measurements are not trustworthy.
(5)Leak of radiation
We will elaborate it later in the c part of this question.
b.
Which regions of the city have greater uncertainty of radiation measurement?
Use visual analytics to explain your rationale.
The uncertainty of each region changes every day. From 6th to 7th, the regions with greater uncertainty because many mobile sensors pass by these regions. From 8th to 10th, the regions with greater uncertainty are mainly because there are unusual radiation measurements.
4/6-4/7：During 6th-7th, the daily uncertainty patterns are very similar. The inner parts of the city have greater uncertainty than other regions of the city. More mobiles go across these regions then result in greater uncertainty. The values of the uncertainty are not high. Some regions have lesser uncertainty because few cars go to those regions, e.g. Broadview.
4/8: There is one region which has greater radiation measurement and uncertainty than other regions. It is at the conjunction of Jade Bridge and the city (marked as "1" below). The uncertainty is mainly from mobile sensor no.10. This sensor's range of radiation measurements is quite wide in this region.
4/9: There are two regions with greater uncertainty. First, Wilson Forest Hwy (marked as "1"). There are three squares on Wilson Forest Hwy. The right square has the greatest uncertainty. There are some mobile sensors moving back and forth. Almost every mobile sensor detect increasingly radiation when they are getting close the end of Wilson Forest Hwy. That's the reason why this region has the greatest uncertainty. The region marked with "2" has the second greatest uncertainty value. The uncertainty is mainly from mobile sensor no.20 which pass by the region and has higher radiation measurement at the same time.
4/10: On 10th, it has similar uncertainty pattern as on 9th.
c.
What effects do you see in the sensor readings after the earthquake and other
major events? What effect do these events have on uncertainty?
We summarize 6 types of effects after the earthquakes and other major events. The first, second and sixth effects make the sensors more uncertain:
(1)More noises
More noises are detected. We use first order difference equation to filter out the noises (i.e. the red dots) from the data. Take mobile sensor no.3 as an example:
(2)Leak of radiation
Some mobile sensors detect usual radiation at the same place and same time. We infer it is the leak of radiation. For example, mobile sensor no.21, 22, 24, 25, 27, 28, 29 and 45:
(3)Some sensors are malfunctioned
Mobile sensor no.1, 23, 26, 35 and 47 have constant radiation measurements (as stated in 2.a.2 constant radiation values). The visualization shows where and when those sensors just detect constant measurements. We infer those sensors are malfunctioned due to the earthquakes because of the curve shape connected by the sensors below:
(4)Missing values are increasing
From 4/8 8:30 and/or 4/9 15:00, many mobile sensors start to have increasing missing values which is probably caused by the earthquakes. Take mobile sensor no.4 as an example:
(5)Increasing uncertainty
Use the visualization above as an example, many sensors start to have wider range of measurements (i.e. uncertainty) from 4/8 8:30.
(6)Decreasing range of mobile movements
On 4/9, the spread of mobile sensors is remarkably decreasing.
3.Given the uncertainty you observed in question 2, are the
radiation measurements reliable enough to locate areas of concern?
a. Highlight potential
locations of contamination, including the locations of contaminated cars.
Should St. Himark officials be worried about
contaminated cars moving around the city?
b. Estimate how many cars may
have been contaminated when coolant leaked from the Always Safe plant. Use
visual analysis of radiation measurements to determine if any have left the
area.
c. Indicated where you would
deploy more sensors to improve radiation monitoring in the city. Would you
recommend more static sensors or more mobile sensors or both? Use your
visualization of radiation measurement uncertainty to justify your
recommendation.
Limit your responses to 10 images and 1000 words
Most of the sensors are reliable, we can use the
radiation measurements to locate the areas of concerns.
a
Highlight potential locations of contamination, including the locations of
contaminated cars. Should St. Himark officials be
worried about contaminated cars moving around the city?
Potential locations of
contamination and the locations of contaminated cars
(1)Jade
Bridge (the part outside of the city)
In a specific period, some mobiles detect
increasing radiation when passing through Jabe
Bridge. When these mobiles are leaving Jade Bridge, the detected radiation is
decreasing. We infer that the regions around Jade Bridge outside of the city
are contaminated.
The mobiles which pass through Jade Bridge are
no.5, 7, 8, 9, 10, 11, 12, 14, 36, 46 and 49. After the earthquakes, these
mobiles detect unusual increasing radiation when passing through Jade Bridge.
To be more specific, the time is from 4/8 21:19 to 4/9 6:58. In addition, these
sensors cannot detect unusual radiation starting after 4/10 20:41. We infer
that the leak of radiation around the regions is cleaned up already.
Take mobile sensors no.5, 7, 8 and 9 as examples:
(2)The region close to Jade Bridge
As stated in question 2, there is one region
(marked as 1 on the below visualization) close to Jade Bridge has unusual
radiation measurements on 4/8. By exploring the data, we found that mobile
sensor no.10 is the main reason. From 4/8 16:40 to 22:03, this sensor has
unusual radiation measurements and stay in that region constantly.
(3)The region close to Always Safe Loop
The region close to Always Safe Loop (marked as 2
on the above visualization) has unusual radiation measurements on 4/8. The main
reason is that mobile sensor no.9 stays in the region for a long time and
detects high radiation during the same time.
(4)The regions around Wilson Forest Hwy
On 4/9, the regions around Wilson Forest Hwy have
unusual radiation measurements (marked as 1). The pattern is very similar to
Jade Bridge on 4/8. Certain mobiles sensors detect increasing radiation when
they pass through the highway (on the right side) and stay in a region (the
left side) when detecting high radiation. The specific time of starting to
detect high radiation on the left side of the highway is around 4/9 19:00.
(5)The regions on the bottom right side of the city
The region marked as 2 on the above visualization
has unusual radiation measurements on 4/9 as well. Mobile sensor no.20 is the
main reason causing these high radiation measurements. Around 4/9 15:00, this
sensor starts to detect high radiation on that region, and the sensor stays
there until the measurements are missing(around 4/10
13:00).
Should
St. Himark officials be worried about contaminated
cars moving around the city?
They should not worry about the contaminated cars
moving around the city. Those mobiles seem to deal with the leak of radiation.
When they start to detect high radiation, they stay in those regions for a long
time. When they start to move away, the radiation measurements decrease
remarkably.
However, there is one exception, it is mobile
sensor no.20. This sensor starts to detect high radiation until it is not
working. Before this sensor loses the radiation measurement, the radiation it
detects is still high. The officials should pay attention to this sensor.
b
Estimate how many cars may have been contaminated when coolant leaked from the
Always Safe plant. Use visual analysis of radiation measurements to determine if
any have left the area.
There are 11 cars which may have been contaminated.
(1)The regions around Jade Bridge: mobile no.10
The sensor leaves the region where it detects unusual
radiation and moves inward the city. The no.10 sensor detects great radiation
when returning from Jade Bridge around 4/8 16:40. This sensor stays in that
region when detecting great radiation. When the sensor starts to move inward
the city, the level of radiation measurements already becomes normal.
(2)The regions around Wilson Forest Hwy: mobile no.21, 22,
24, 25, 27, 28, 29 and 45
Those sensors leave the region where they detect
great radiation and leave the city through Wilson Forest Hwy. Around 4/9 18:00,
those sensors start to detect high radiation measurements on the regions
showing below. Those sensors stay in the region (the red square below) until
4/10 6:00. When they start moving toward Wilson Forest Hwy, their radiation
decreases dramatically.
(3)The regions on the bottom right side of the city: mobile
no.20
This sensor moves to a certain region (the red
square below) for a while then starts to detect great radiation suddenly at
around 15:00. It keeps detecting the unusual radiation until this sensor is not
working and loses the radiation measurement at around 4/10 13:00. We cannot
tell if this sensor moves away to other regions because the sensor is not
working.
(4)The region close to Always Safe Loop: mobile no.9
From 13:22 to 16:47 on 4/8, mobile sensor no.9
detect great radiation near Always Safe Loop. During the high radiation
measurements, this sensor stays in that region without moving away. After the
radiation measurements become normal, this sensor moves to other regions.
c
Indicated where you would deploy more sensors to improve radiation monitoring
in the city. Would you recommend more static sensors or more mobile sensors or
both? Use your visualization of radiation measurement uncertainty to justify
your recommendation.
We would recommend deploying both the static and moblie sensors to improve radiation monitoring in the city
because these two types of sensors have their own pros and cons.
The
places we would deploy more static sensors
(1)The
right side of Jade Bridge
Those mobile sensors, no.5, 7, 8, 9, 10, 11, 12,
14, 36, 46 and 49, detect increasing radiation when getting through Jade Bridge
(the red square on the map below). The increasing radiation makes a greater
uncertainty of this region.
Take mobile sensor no.5 as an example:
(2)The regions around Wilson Forest Hwy
As stated in question two, the below visualization
shows that the regions around Wilson Forest Hwy have the greatest uncertainty
on 4/9. The region of the red square should especially have a static sensor.
(3)The region on the bottom right side of the city
As stated in question two, the above visualization
shows the region marked as 2 has greater uncertainty on 4/9.
(4)The blank parts of the city
During 4/9 and 4/10, the range of mobile movements
is remarkedly decreasing compared to previous days.
The parts without mobiles passing by should have static sensors.
The
places we would deploy more mobile sensors
(1)The
places with lesser mobile sensor
By selecting the time from 0:00 to 1:00 on 4/6 for
example, we can find out where all of the drivers of the mobiles live. We can
deploy more mobiles in the regions with lesser mobiles.
(2)The regions with lesser mobile movements and greater
uncertainty
On normal days, e.g. 4/6 and 4/7, we visualize the
counts of mobile movements on the map and then we can find there are some
regions with lesser mobile movements but the uncertainty is higher. We should
deploy more mobile sensor in those regions to reduce the uncertainty.
4.Summarize the state of
radiation measurements at the end of the available period. Use your novel
visualizations and analysis approaches to suggest a course of action for the
city. Use visual analytics to compare the static sensor network to the mobile
sensor network. What are the strengths and weaknesses of each approach? How do
they support each other? Limit your response to 6 images and 800 words.
a
Summarize the state of radiation measurements at the end of the available
period. Use your novel visualizations and analysis approaches to suggest a
course of action for the city.
Step 1 to 3:
(1)Identify anomaly: the regions with greater radiation
measurements and uncertainty
By using the grid map marked as 1 on the tool, we
identify the regions with unusual radiation measurements and uncertainty on the
bottom right side of the city. In addition, the movement area decreases.
(2)Find out which mobile sensors are in the regions
By using timeline marked as 2 on the tool, we find
out the mobile sensors appear in the regions. The mobile sensors are no.21, 22,
24, 25, 27, 28, 29, 30, and 45.
(3)Loot at the relationship between radiation values and
time: higher level
We use the bubble chart marked as 3 on the tool to
compare those mobile sensors. We aggregate the measurements by hours. If the
circles are larger, it means the sensor has higher radiation measurements
during that time. By comparing those sensors, we find that most sensors have
similar patterns except mobile sensor no.30. Most sensors have greater
uncertainty and radiation measurements before their measurements are missing.
Step
4 to 6:
(4)Loot at the relationship between radiation values and
time: lower level
Instead of aggregating the data, we visualization
the data of every 5 second to explore the radiation measurements in details. We
filter the timespan to the time before the measurements are missing. Most of
the sensors have similar patterns except mobile no.30 which is more stable than
others. All of the sensors detect increasing radiation before they leave the
city through Wilson Forest Hwy.
(5)Look at the dynamic routes of all of the mobiles
Each circle on the map represents a mobile. We use
log2 to normalize the data to improve the readability of the visualization.
Most of the sensors are overlapping, except mobile
no.30. Mobile no.30 is far away from the region where other mobiles are
staying. From exploring this dynamic
map, we found that mobile no.30 does not stay in that region so it does not
detect the unusual radiation there.
(6)Make interpretation and record it
After exploring the data, we already have some
important findings. In order to get back to this analysis quickly next time, we
save the analysis and leave some comments. Other colleagues can read the
comments and keep analyzing the data.
b
Use visual analytics to compare the static sensor network to the mobile sensor
network. What are the strengths and weaknesses of each approach? How do they
support each other?
The strength and weakness
of mobile sensors and static sensors
Mobile sensors:
Strengths – can move around, can detect unusual
radiation issues (e.g. leak of radiation)
Weaknesses – lower quality (unstable measurements,
more noises), cannot monitor a region stably,
Static sensors:
Strengths – higher quality (stable measurements,
lesser noises), can monitor a region stably
Weaknesses – cannot move (the area of radiation
measurements is restricted)
Example 1: the strength of static sensors – lesser noises
Randomly select a timespan: 4/8 10:00 - 11:00.
Example 2: the weakness of mobile sensors – detect
different radiation
At the same time and same place, there two mobile
sensors detecting two different radiation values. It's probably because of the
lower quality of mobile sensors.
From around 4/7 19:00 to 20:00, mobile sensor no.13
and 32 are very close to each other on the map but they detect two different
radiation values.
Example 3: the strength of mobile sensors – detect
unusual radiation
From 4/8 to 4/10, mobile sensors detect potential
unusual radiation issues. It is not that obvious to see the patterns from
static sensors measurements.
How
do they support each other
Generally speaking, the uncertainty of mobile sensors is higher than the uncertainty of static sensors. Hence, static sensors can provide more stable radiation measurements than mobile sensors. Nevertheless, the mobility of static sensors is poor. When a region has an unusual radiation issue, static sensors usually cannot detect that. We need mobile sensors to detect unusual radiation issues. For example, the radiation leak on 4/8.
5.The data for this
challenge can be analyzed either as a static collection or as a dynamic stream
of data, as it would occur in a real emergency. Describe how you analyzed the data - as a
static collection or a stream. How do
you think this choice affected your analysis? Limit your response to 200 words
and 3 images.
Describe
how you analyzed the data - as a static collection or a stream
We analyzed the data as a static collection. After
we get the data, we do not maintain the database anymore. When we analyze the
data, we explore the data and find out unusual patterns. When we are familiar
with the data, we start to design the dashboard and integrate the
visualizations. When we operate the dashboard, we can freely choose any time
span:
We summarize the category of potential untrusted
anomaly. We want analysts to know the anomalies before they start to explore
the data. Because we use a static collection of data, we will not change the
category.
How
do you think this choice affected your analysis?
If we choose to analyze the data as a dynamic
stream, it will affect our analysis in three parts:
(1)The filter of time
We can only use the latest data and cannot store
historical data. The amounts of the data we can analyze will be much smaller.
(2)The management and
summary of unusual events
Without historical data, we probably will be more
sensitive to potential anomaly. For example, mobile sensor No.12 has an
extremely high radiation measurement on 4/9.
(3)Action
We can do immediate reaction to unusual events when
using dynamic stream of data. Using static collection of data can let us have
broad thinking on deploying sensors.