Entry Name:
SMU-Teo-MC2
VAST
Challenge 2019
Mini-Challenge 2
Team Members:
Teo Yaling,
Singapore Management University, yaling.teo.2018@mitb.smu.edu.sg PRIMARY
Dr. Kam Tin Seong, Singapore Management
University, tskam@smu.edu.sg (Advisor)
Student Team: Yes
Tools Used:
Python (data
integration)
Tableau
Link to Tableau
dashboard:
Approximately how many
hours were spent working on this submission in total?
132
May we post your submission
in the Visual Analytics Benchmark Repository after VAST Challenge 2019 is
complete? Yes
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.
Radiation measurements from both static and mobile
sensors can be broadly separated into baseline and elevated values as shown in Figure 1. Both sensor types exhibited
similar dense baseline values (blue) at the bottom of their charts. Elevated
values (orange) are defined as values above baseline.
Figure 1
From “Static Sensor Readings Across Time” in Figure
2, the baseline values of static sensors were not consistent across all 5
days. From 6 April to 8 April (mid-day), the baseline readings (blue) exhibited
an almost uniform minimum value of 10 cpm and a
maximum value of 20 cpm. There was a single outlier
point at 6.24 cpm. From 8 April (mid-day) to 10 April
(end-of-day), the minimum value of baseline readings started to descend towards
0 cpm. There were 2 negative outlier points at -0.35
and -20.24 cpm. While the minimum baseline values
exhibited a decreasing trend, there was an increase in baseline readings
(green) of static sensors from 20 cpm to 35 cpm. Since an earthquake has occurred in St. Himark, it can be inferred that the changes in baseline
radiation readings is likely due to the earthquake. Therefore, the baseline
values of static sensors were split into 2 categories, “Pre-Quake Baseline” and
“Post-Quake Baseline”.
From “Mobile Sensor Readings Across Time” in Figure
2, the baseline values of mobile sensors (blue) exhibit
an almost uniform minimum value of 0 cpm and a
maximum value of 67 cpm from 6 April to 10 April.
Figure 2
From both static and mobile elevated readings in Figure
3,
both depict that there was increased radiation above background between 8 April
(mid-day) to 10 April (end-of-day). The static sensors' elevated values
(orange), illustrated by “Static Sensor Readings Across Time” in Figure
3, showed minimal elevated readings between 6 April to
8 April (mid-day). Between 8 April
(mid-day) to 10 April (end-of-day), there were short spikes of elevated values
at multiple time points. Above these spikes, there was a scatter of elevated
values (200 to >1000 cpm) between 9 April
(mid-day) to 10 April (end-of-day).
In contrast, the elevated values of mobile sensors
(orange) exhibited some noise throughout all 5 days. Notably, there were 2
dense columns of elevated readings at approximately 8 April 16:30 and 9 April
19:00 respectively. Above both columns, there were groups of evevated values at around 2000cpm. Near the lower range of
elevated values (100 to 200 cpm), there was another
dense area between 9 April (mid-day) to 10 April (mid-day).
While the static sensors seem to paint a picture that
increased radiation may have occurred after a singular event, the mobile
sensors provided another perspective that there were 2 waves of increased
radiation in St. Himark.
Figure 3
Out of the elevated readings, there was an extreme
outlier reading of 57,345 cpm on 9 April 02:43:25 as
shown in Figure
4.
Since this data point far exceeded the remaining data points, it is likely that
this reading is unrelated to any radiation event and may be due to device
malfunction.
Figure 4
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.
2a.
Uncertainty can be defined as:
· Wide spread in the range of elevated values
· Irregular patterns in location of elevated values
· Different in readings amongst sensors within the same neighbourhood
Overall, mobile sensors have a wider spread (longer
interquartile range) in their elevated values than static sensors (Figure
5).
Figure
5
Outliers:
Sensor-id M12 contributed to the 2 highest values
amongst the mobile sensors. The highest value is the extreme outlier mentioned
earlier that lasted for only 5 seconds. The readings before and after this
outlier were 10.34 cpm and 9.11 cpm
respectively. The next highest value at 2419 cpm
occurred about an hour after at the same location.
Figure 6
Sensor-id S4 and S14 contributed to the 2 negative
values shown in Figure
2. Both values cannot be trusted as the reading cannot be negative.
Missing data:
Sensor-id S15 had a nearly 2-day gap between the last
reading captured on 8 April 22:06:55 (15.86 cpm) and
the next reading captured on 10 April 20:45:30 (14:44 cpm).
It is likely that this sensor was damaged when the power plant was damaged
after the earthquake.
Figure 7
There were missing values in the mobile sensor
readings. Figure
8 shows sensors that have travelled to neighbourhood
marked as “Null” (blue). These are points on the bridges or highway or along
boundaries between neighbourhoods. Boxed up in orange
are 5 sensors, M6, M18, M20, M23, M34, that have no data after a certain point,
and their last readings are within St. Himark and the
sensors have not left the city. These sensors may have been damaged, or the
mobile signal has been disconnected due to the earthquake. Boxed up in blue are
11 sensors M21, M22, M24, M25, M27, M28, M29, M30, M45, M48, M49. Their last
data point was captured on the bridges or highway. These sensors may have left
the city.
Figure 8
Unusually high values:
Figure
9’s “Range of Sensor Readings” shows the range of elevated readings of 9
static sensors across all 5 days. It is interesting to observe that the 3
static sensors (S11 at Broadview, S6 at Southwest, S1 at Palace Hills) which
are located furthest from the power plant had the highest median values. In contrast, static sensors S12 in Old Town,
S13 and S15 in Safe Town, had lower median values. The highest value of the
static sensors (boxed in blue) is contributed by S9 at Old Town.
Figure 9
Figure
10 shows that Sensor-id M21, M22, M24, M25, M27, M28, M29, M45 had the
highest median elevated readings (5 days combined) amongst 50 mobile sensors.
These sensors have short box plots with median values at approximately 1500 cpm. The high readings were captured in Wilson Forest.
Sensor-id M10 at Old Town had the next highest median values. Notably, the
entire length of the box plot (whiskers inclusive) is the longest amongst the
mobile sensors. It is interesting to observe that both Wilson Forest and Old
Town are located away from the power plant.
Figure 10
2b.
The neighbourhoods with
greater uncertainty are Wilson Forest and Old Town. Figure
11 shows that Wilson Forest has the highest median elevated readings,
followed by Old Town. Safe Town is also another area of uncertainty. Its
relatively low median elevated readings, and short interquartile range is
contrary to expectations that it will be the most affected in the event of a
nuclear power plant damage.
Figure 11
Figure
12 shows the mobile sensors with readings captured at Wilson Forest.
Readings from Wilson Forest are relatively uncommon amongst the mobile sensors.
Prior to 9 and10 April, only M14, M26, M48 were at Wilson Forest area and their
readings were at baseline levels. M26 was near the entrance to Wilson Forest
Highway from 6 to 8 April but did not travel on the highway. Only M21, M22,
M24, M25, M27, M28, M29, M30, and M45 had left St.Himark via the Wilson Forest Highway and returned
via the same route. Of these 9 sensors, only M30 did not capture the high
radiation readings.
Figure 12
Figure
13 below illustrates the uncertainty of the readings within Old Town. On 7
April, the interquartile range of static sensors at Old Town is wider than the
mobile sensors. The median values of the static sensors are much higher than
that of mobile sensors. This is contrary to expectation as mobile sensors tend
to have a higher reading. On 8 April, as expected, the interquartile range of
mobile sensors is wider than that of static sensors. The box plot for Old Town
is also the longest. The median value of the elevated readings is the highest
for Old Town. On 9 April, both sensor types have comparable readings.
On 8 April, the mobile sensors at Safe Town had a
higher median reading than static sensors. On 9 April, both sensor types had
comparable readings. There was little variability in both sensor types as shown
in the short boxplots. On 9 April, in contrast to the readings picked up at
Wilson Forest, the median readings in Safe Town were less than 100 cpm.
Figure 13
2c.
There are 2 waves of radiation that occurred at St. Himark over the 5-day period. The first wave took place at
8 April 16:30 and the town that reflected the highest readings and greatest
uncertainty is Old Town. The second wave took place at 9 April 19:00 at Wilson
Forest. Both locations are located away from Safe Town. These effects further
confounded the uncertainty of the situation. From the 2 waves of high
radiation, it is plausible to suspect that the earthquake took place on 8 April
and there was a second aftershock that took place on the 9 April. If Always
Safe was damaged in the earthquake, one would expect that the immediate
vicinity will receive the highest radiation reading, followed by neighbouring towns such as Old Town, Easton, East Parton, Cheddarford, and Pepper Mill. Contrary to this, the higher
readings on 8 April were mostly from Old Town (Figure
14). The second wave corresponded to points at Wilson Forest (Figure
10).
Figure 14
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
3a.
Potential locations of contamination: Old Town and
Wilson Forest
Figure 15 is an animated image that shows the events from 8 April 00:02. The
purpose of the animation is to capture the path taken by the sensors within Old
Town (M5, M7, M8, M9, M10, M11, M12, M46, M47). Figure 16
is a static image that shows the path taken by the sensors from 6 to 10 April.
All sensors except for M47, had left Old Town via the Jade Bridge on the morning
of 9 April. The cars returned just after the second wave of radiation high had
occurred. The 2 waves of high radiation events are marked by the dotted lines
on “Sensor Movements Across Time”.
Figure 15
Figure 16
Figure
17
is an animated image that shows the events from 8 April 10:00. The purpose of
the animation is to capture the path taken by these sensors out of Wilson
Forest Highway and back into St. Himark. M30 is
different from the remaining 8 sensors (M21, M22, M24, M25, M27, M28, M29, M45). On the morning of 9 April,
M30’s path originated from Pepper Mill, on Hilmark
Ave that is bordering Pepper Mill and Wilson Forest, before turning onto the
Wilson Forest Highway. The 8 sensors started from either Broadview and Scenic
Vista before going on Wilson Forest Highway. All 9 cars left St. Himark for the day and returned only in the evening. At
18:00, M29 and M30 returned. M30 turned right towards Pepper Mill. M29 remained
at the junction of Wilson Forest Highway and Himark
Ave. By 20:00, the other cars returned, capturing the high radiation at the
same junction. It is interesting to note that the 8 sensors remained at the
same spot for hours, until they reversed out of St. Himark
on the next morning (10 April 06:00 to 07:00). It is uncertain if there was a
blockage at the point and these cars are unable to pass through on 9 April, or
the sensors were disconnected from the network and continued to register the
last location until the next morning. Figure 18 is a static image that shows the path taken by the sensors from 6 to 10
April.
Figure 17
Figure 18
Officials should be worried at the locations of high
radiation that are not within Safe Town. However, there seems to be no obvious
signs of transfer of contamination across contaminated cars.
3b.
The assumption is that each mobile sensor is mounted
onto 1 car.
Contaminated cars from Always Safe power plant
leakage:
It was mentioned that after the earthquake, the
damaged nuclear power plant had an event of coolant leakage that contaminated
employees' cars. It was not revealed when the coolant leakage had occurred.
Based on the earthquake and the spike in radiation readings, the coolant
leakage is likely to occur from 8 April 16:30. During the leakage, these cars
must be at Safe Town during that time frame. To identify these cars, mobile
sensors that have visited Safe Town between 8 April and 10 April have been
highlighted (dark brown).
The likely suspects are:
· Sensors that are at Safe Town just before and up to 8 April 16:30 are
M9, M39, M43. M43 is likely an employee of the power plant due to working hours
spent at Safe Town.
· Sensors that are at Safe Town after 8 April 16:30 are M10 and M14. M10
may have picked up the radiation from Old Town.
· M15 was at Safe Town after 8 April 16:30 for a long period, with some
short trips to other neighbourhoods. This car is
likely to have picked up the coolant as well.
· M13 is the only sensor that was continuously at Safe Town throughout the
day and night from 8 to 9 April. The driver may be living and working at Safe
Town. This car is most likely to have picked up the coolant due to the
prolonged exposure. Readings were intermittent between high of 600 to 1000 cpm and baseline levels.
The blue highlighted areas represent the bridges and
highways leading out of St. Himark, as well as some
areas on the boundary between neighbourhoods. Out of
the sensors mentioned above, only M9 and M14 have left St. Himark.
Contaminated cars from an unknown source:
The 8 mobile sensors, M21, M22, M24, M25, M27, M28,
M29, and M45, had picked up unusually high readings at the junction of Wilson
Forest Highway and Himark Ave into Scenic Vista, on
their return from the mainland. As illustrated by Figure
19, these cars, except for M22 who had a brief visit to Safe Town on 8
April, had no trips to Safe Town. It is unlikely that they have been contaminated
by the power plant’s coolant leakage.
Figure 19
There are a few possibilities:
· The coolant had leaked into the soil at Safe Town, spread through the neighbouring Pepper Mill, Terrapin Springs, Wilson Forest
and contaminated the area entrance/exit of Wilson Forest Highway into Scenic
Vista. Since readings for Pepper Mill and Terrapin Springs were not as high, it
could be that the coolant had accumulated, and contaminated underground
drainage systems and the exit point was somewhere near the highway.
· The cars have picked up radiation from the mainland. If St. Himark had experienced an earthquake that came in 2 waves,
it is likely that the neighbouring mainland would
have experienced an earthquake of similar scale. Of course, it is not known if
there is a nuclear power plant on the mainland.
3c.
I would deploy more of both sensor types. For static
sensors, I would advocate for each neighbourhood to
have at least 1 sensor. There are 12 neighbourhoods
without static sensors are Northwest, Weston, Easton, Southton,
West Parton, East Parton, Oak Willow, Pepper Mill, Chapparal,
Terrapin Springs, Scenic Vista, and Wilson Forest. These areas are highlighted
in yellow in Figure
20.
Figure 20
For static sensors, I would advocate the St. Himark community to encourage more residents of Palace
Hills, Northwest, Oak Willow, Pepper Mill, Terrapin Springs, Broadview to
install mobile sensors on their vehicles. These areas are highlighted in yellow
in Figure 21.
Figure 21
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.
The state of radiation measurements at the
end of 10 April is as follows:
·
Baseline
values of both static and mobile sensors are elevated.
·
The
performance of S15 is uncertain. It is unknown if S15 was shifted to a new
location within the vicinity of the power plant after the power plant was
damaged.
·
5 mobile
sensors (M6, M18, M20, M23, M34) had discontinued.
·
11 sensors
(M21, M22, M24, M25, M27, M28, M29, M30, M45, M48, M49) have left the city.
Some may have been out for the day.
·
Reduced
mobile sensor network post-earthquake and radiation events. Figure 22 shows the reduction in mobile sensor position from 8
April to 10 April. Network reduction is evident in the Southern and Eastern
parts of the city.
Figure 22
The strengths and weaknesses of each approach are listed in the
following table.
|
Sensor Type |
Strengths |
Weaknesses |
|
Mobile |
·
Sensitive
to changes in radiation present in the environment and was able to pick up the
2 waves of radiation (Figure 1). ·
Reachability,
sensors can traverse roads, walkways, enter buildings, etc. ·
Cost efficient
to increase depth of network. Mobile sensors are relatively inexpensive to
build and mount on cars or bicycles. ·
Each
sensor is not limited to a single neighbourhood.
Depending on the range of movement of its user, each sensor can pick up readings
across multiple neighbourhoods and can contribute
to improving the stability and accuracy of the overall readings. |
·
Unstable
(Figure 4) with extreme outlier values. ·
Quality and
comparability across different mobile units are unknown ·
Unknown
if the sensors are of the same brand, or entirely homemade. ·
Subject
to availability. When the vehicles leave St. Himark
for the mainland perhaps for work, a day’s reading is lost for that sensor (Figure 8, Figure 22). ·
Subject
to availability of the cellphone network. If the network is down in the event
of an earthquake, no data can be posted online (Figure 8, Figure 22). |
|
Static |
·
Stable,
tends to result in shorter interquartile range, or fewer outliers (Figure 5, Figure 13) ·
Installed
by centralized body – St. Himark officials. ·
May have
regular maintenance and calibration. |
·
Limited reachability.
1 neighbourhood = 1 sensor. The existing static
sensor network covers barely half of St. Himark (Figure 20). ·
Reduced sensitivity
to changes in radiation levels. While mobile sensors are reflecting that St. Himark may be in danger of excessive radiation, the static
sensors seem to depict a mild radiation event (Figure 1). ·
It is odd
that the minimum values of the baseline decreased after the earthquake (Figure 2) |
Short-term measures for St. Himark:
1.
Due to the
uncertainty of the elevated readings at Wilson Forest Highway, immediate
monitoring at that area is essential. Since there is a lack of an official
static sensor in the South-Eastern area of St. Himark,
a temporary mobile or static sensor can be placed at the entrance/exit of
Wilson Forest Highway to monitor the readings of that area temporarily.
2.
Due to the
consistency of 8 mobile sensors reporting the same high radiation readings, St.
Himark officials should close the entrance/exit of
Wilson Forest Highway into Scenic Vista. During the closure, they could
investigate the surrounding area to identify the source of radiation and
de-contaminate the area, if necessary.
Long-term measures for St. Himark:
1.
Given the
earthquake and its damage to Always Safe power plant, it is likely that
radiation readings will continue to be elevated. Static sensors should be
installed in neighbourhoods that do not have an
existing one.
2.
St. Himark officials should schedule routine static sensor
calibration and maintenance.
3.
The St. Himark community could encourage more citizens to install
mobile sensors on their vehicles, to collect more readings and improve the variability
in the readings. The community could organize routine mobile sensor calibration
sessions so that members can ensure that their devices can capture radiation in
the environment accurately.
4.
Having
both static and mobile sensors within each neighbourhood
provides a good comparison between the reliability of the static and mobile
sensor network. It serves to provide a good alternative source of data as well
as an alternative source of "truth" for the St. Himark
community.
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.
The data for this challenge was approached
predominantly as a static collection of data so that trends and patterns in
historical data can be identified. At times, the data was analyzed as a dynamic
stream to visualize the movement of multiple mobile sensors across St. Himark (Figure 15, Figure 17). The dashboard was built to analyze the data in both
static and dynamic form. The dashboard was built using unaggregated radiation
readings at 5 seconds interval. The dashboard can take in a live stream of data,
since each new data point can be added to each of the charts.