Answers

VAST 2019 MC2: Citizen Science to the Rescue

Overview

Data Cleaning

Dashboard

Question & Answers

Question 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.

The chart above shows the average radiation of the mobile sensors (blue circles) and static sensors (yellow circles) by Neighbourhood. The red line is the 75cpm mark whereby any value beyond that point is abnormal.

As seen from above, from the 9th to 10th, Wilson Forest has spikes in the radiation readings with an average value of 1349cpm on the 9th and 1388cpm on the 10th while the rest are under 54 and 72 on the 9th and 10th respectively.

On the 8th, most neighbourhoods have increasing number of sensor ids' readings above 75cpm from 5Am-8AM onwards, excluding Wilson Forest, Oak Willow and Chapparal.

On the 9th, most neighbourhoods continued to have many sensor ids' readings above 75cpm, excluding Oak Willow and Chapparal.

On the 10th, once again, most neighbourhoods have a significant number of sensor ids' readings above 75cpm, excluding Terrapin Springs, Oak Willow and Chapparal.

From the 6th to the 7th, there is not much difference in the level of the radiation level, with a maximum of 15.4% increase in the average radiation level in Terrapin Springs. From the 7th to the 8th, there is a visible increase in the radiation level in old town whereby it increased by 126.1%. Many neighbourhoods' average readings also increased on 8 April. Safe Town's readings has increased significantly by 43%. There is a high likelihood that the earthquake occured on the 8th. Further analysis will be shown in question 2 with regards to the earthquake.

Question 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?

As seen above, On the 9th of April, there is an outlier sensor id that has a value of 57,345 at 02:00 belonging to M12 while the rest are below 3K. This shows that sensor id M12 may not be reliable.

The following sensors have missing data during the highlighted periods:

1. M5
2. M9
3. M14
4. M18
5. M20
6. M21
7. M22
8. M23
9. M24
10. M25
11. M27
12. M28
13. M29
14. M30
15. M34
16. M45
17. M48
18. M49
19. S15

b. Which regions of the city have greater uncertainty of radiation measurement? Use visual analytics to explain your rationale.

The chart above is broken down by days, whereby the table shows the difference in value of the average radiation value of the respective day as compared to the previous day. Red represents an increase in the radiation value while green represents a drop in the radiation value. As seen above, on 9th April, Wilson Forest has an increase of 3792% as compared to the previous’ day readings while the other cities only have a maximum of 57% increase on the 9th of April, which is a drastic difference.
background is detected. Characterize changes over time.

As seen from above, only Wilson Forest has a spike in radiation level whereby the radiation level peaked at 2193 on 9th April whereby there was a sudden spike that lasted till 10th April. As compared to the other neighbourhoods, the average radiation level of Wilson Forest on the 9th is 1349cpm while the rest are 54cpm and below. On the 10th, the average radiation level of Wilson Forest is 1388cpm while the rest are 72cpm and below. This is odd as the power plant is far away from Wilson Forest. We are unable to conclude whether the readings are reliable as well as only mobile sensor readings are recorded and there were no static sensors.

Also, as seen from the gaps in the data, Chapparal, Eastton, Oak Willow, Pepper Mill, Scenic Vista, Southton, Terrapin Springs and Wilson Forest have missing data during certain timings. There are also no static sensor readings and thus, there is insufficient data to analyze throughout the days for the respective neighbourhoods.

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?

As seen above, there is a significant increase in the radiation level in most neighbourhoods. Thus, I suspect that the earthquake happened around sometime on the 8th of April.

As the nuclear plant is in Safe Town, I decided to further analyse the radiation level in Safe Town. Based on the static sensor, the radiation level spiked at 4:20AM to 439cpm. More spikes followed later during the day as well.

By looking at the mobile sensors, we can see that there the radiation level has a spike at near 4am, at 3:42AM. The next spike was at 9:04, which a series of spikes followed afterwards.

By looking at the other neighbourhoods, many of them have a spike in the radiation level on the 8th from 5AM to 8AM onwards. Thus, I deduced that the earthquake happened around 4:20AM on the 8th of April. The radiation then started to disperse during later part of the day. As seen from above, the number of sensors with radiation level above 75cpm has increased significantly.

Question 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?

The above the static sensor readings. As highlighted above, most of the static sensors in the respective neighbourhoods have many spikes above 75cpm that followed after the occurrence of the earthquake.

Based on the findings above, the potential locations of contamination on the 8th are:

1. Broadview
2. Palace Hills
3. Southwest
4. Cheddarford
5. Downtown
6. Old Town
7. Safe Town

The potential locations of contamination on the 9th and 10th are:

1. Cheddarford
2. Downtown
3. Old Town
4. Safe Town

If we were to look at the mobile sensor readings, the first spike that followed after the static reading above is at 9:02AM with a reading of 934cpm. Thus, if any cars are contaminated, it would most likely be from 9AM onwards.

The above shows the number of mobile sensors(cars) in the respective neighbourhoods from 9AM onwards with a sensor reading of 75cpm and above. As seen above, the locations of potentially contaminated cars:

1. Downtown - 13 cars
2. Easton - 13 cars
3. Old Town - 10 cars
4. West Parton - 10 cars
5. Wilson Forest - 9 cars
6. Scenic Vista - 8 cars
7. Weston - 8 cars
8. Southwest - 7 cars
9. East Parton - 8 cars
10. Cheddarford - 5 cars
11. Northwest - 4 cars
12. Broadview - 4 cars
13. Southton - 4 cars
14. Palace Hill - 3 cars
15. Pepper Mill - 2 cars
16. Terrapin Springs - 2 cars
17. Oak Willow - 2 cars
18. Chapparal - 1 car

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.

On the 8th of April, Safe Town has a spike in radiation at around 4:20am which led to the radiation level to be above 75cpm later. As mentioned early, this is the potential period of time whereby the earthquake took place as the power plant is located in Safe Town.

If we were to look at the mobile sensor readings, the first spike that followed after the static reading above is at 9:02AM with a reading of 934cpm. Thus, if any cars are contaminated, it would most likely be from 9AM onwards.

The above graph shows the mobile sensors that are in Safe Town on the 8th of April when the earthquake occurred (with reference to the previous graph), with sensor reading above 75cpm. As seen above, the following mobile sensor ids that are potentially contaminated are:

1. M9
2. M13
3. M14
4. M15
5. M21
6. M22
7. M32
8. M39
9. M43
10. M44

The above shows the neighbourhoods that the car has travelled to filtered by the potentially contaminated cars as identified above, after the earthquake occurred. Below are the neighbourhoods which the cars travelled to afterwards, causing the coolant to leak into the following neighbourhoods:

• M9 – Old Town
• M13 – Easton
• M14 – Cheddarford
• M15 – East Parton, Southwest
• M22 – Wilson Forest and Scenic Vista
• M32 – Cheddarford an East Parton
• M39 – Easton, Southton and Weston
• M43 – West Parton
• M44 – Weston, Easton, East Parton

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.

As mentioned earlier, the mobile sensors shows different readings patterns, and this may be because they came from other neighbourhoods that have higher readings. Secondly, there may be times that none of the cars that are tagged with the sensors passed by certain neighbourhoods throughout the day. One example can be seen from Chapparal whereby there is 0 readings on the 10th of April. Thirdly, the mobile sensors are inconsistent whereby there are certain times of the day that there are missing data. Thus, the readings may not be accurate and sensor sensors should be used instead. Currently, the following towns do not have static sensors:

1. Chapparal
2. Oak Willow
3. Pepper Mill
4. East Parton
5. Easton
6. Northwest
7. Oak Willow
8. Scenic Vista
9. Southton
10. Terrapin Springs
11. West Parton
12. Weston

Apart from deploying static sensors in these following neighbourhoods, I would also recommend deploying more than one static sensor to each neighbourhood. As seen from above, Static S15, which is located in Safe Town, has missing data from the 8th 10PM onwards till the 10th 8PM.

However, there is another static sensor, S13, that is deployed in Safe Town as well. Thus, even though there were missing data for S15, there is S13’s readings. In most neighbourhoods with static sensors, only have 1 each. Thus, there should be more than 1 static sensor being deployed in each neighbourhood.

Question 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?

Also, with static sensors, there will always be readings as there will always be a guarantee that there is a sensor in the respective neighbourhood. As seen above, there is no missing data for static sensors in the neighbourhood. However, even though mobile sensors maybe inconsistent, they provide data such as where the coolant is being leaked to (other neighbourhoods), and determine if the readings of one neighbourhood is due to the radiation exposure in that respective neighbourhood or is it due to the cars that travelled from the contaminated neighbourhoods.

With static sensors, the readings will be less bias as compared to mobile sensors as the readings would be purely based on the radiation from the place that they are deployed from. Comparing both Mobile and Static sensors, Mobile sensors tend to have missing data as compared to static sensors whereby there are 19/50 mobile sensors with missing data and 1/9 static sensors with missing data as seen above.

On the other hand, if we were to analyse with mobile sensor readings, there are missing data in Chapparal, Oak Willow, Scenic Vista and Terrapin Springs as there may be no cars with the tagged mobile sensor passing by the neighbourhoods during those respective timings without any data.

I would recommend using a combination of both static sensors and mobile sensors. I would deploy more static sensors to each neighbourhood whereby there would be at least 4-5 static sensors in each neighbourhood. The sensors will be placed at the North, South, East, West and Central of each neighbourhood. This would enable a more consistent reading across all neighbourhoods and cover each region of each neighbourhood as the readings may also differ due to the radiation levels of other neighbouring neighbourhoods. I would still use mobile sensors to detect any leakage of radiation to other neighbourhoods. With mobile sensors, we will then be able to know the reason of a high radiation level if there should be any high radiation level in any neighbourhoods.

Question 5

For this challenge, I analysed the data as a static collection as we are given one fixed dataset. Static collection enables us to analyse the accuracy of the data such as the certainty of the sensors. By doing a time series analytics, it would also enable us to compare the readings across the previous few days and perform comparison analysis. The above screenshot shows the average radiation level across 5 days by per hour. We are able to deduce when there is a spike in radiation level and compare the readings. For an example, the static sensor readings increased more on the 9th and 10th of April whereby more spike in the readings can be seen as compared to the previous 3 days.

However, in a real-life situation, a dynamic stream of data would be necessary to visualize changes in data with up-to-date data. In this case, it allows us to better visualize the movement of the radiation across the different neighbourhoods and toggle between different periods.