ICL-DSI-MC2

Entry Name: "ICL-DSI-MC2"

VAST Challenge 2019
Mini-Challenge 2

Team Members:

James Scott-Brown, Data Science Institute, Imperial College London, james@jamesscottbrown.com PRIMARY

Student Team: NO

Tools Used:

D3.js (for creating interactive visualizations)

Python (for initial data preprocessing/reformatting), with Pandas and Flask

Papa Parse CSV parsing library for JavaScript

three.js JavaScript 3D library

Approximately how many hours were spent working on this submission in total?

20 (?).

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2019 is complete? YES

Video

https://vimeo.com/347860184

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.

Static sensors

Examining the measurements from the static sensors, it seems that:

measurements from sensor 12 increase after the first earthquake on Wednesday afternoon, and remain elevates for about 5 hours
measurements form sensor 5 then increase, and remain elevated for about 8 hours
measurements from sensor 9, 11, and 12 increase after the second major shake on Thursday afternoon, and remain elevated until the end of the recorded time period at midnight on Friday evening

There are also transient spikes in the measurements from the static sensors, some of which coincide with the onset of the earthquakes.

Plots of measurements from each static sensor. The top subplot shows 5-minute moving averages of measurements from all sensors, both mobile (M-) and static (M-). The subplots below show measurements from a single sensor: circular marks represent individual measurements, and lines represent the moving-median. Red marks indicate a value that is outside the displayed range (0-100 cpm).

Mobile sensors

The picture from the moving sensors is more complex, as they have more noise and missing data.

The most significant feature that can be identified is the contamination of vehicles carrying mobile sensors, which occurs on Thursday evening. This contamination seemed to originate from Scenic Vista (neighborhood 8).

Plot of measurements from all sensors, on a common vertical time axis that increases downwards. Measurements are colored with a Virdis colormap: high radioactivity measurements are colored green/yellow and positioned to the left; low measurements are colored blue and positioned to the right. Red marks indicate a value that is outside the displayed range (0-100 cpm). A pattern of continuous red marks, suggesting contamination of a vehicle carrying a sensor, can be seen for several mobile sensors beginning on Friday evening.

Map showing radioactivity measurements near the onset of contamination,at the time represented by the horizontal line in the plot above. Square marks represent static sensors; circular marks represent mobile sensors. Measured values are redundantly encoded as marker area and color; red marks indicate a value that is outside of the scale (0-100 cpm). Contaminated vehicles can be seen in the Scenic Vista neighborhood.

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.

There are a number of interesting data quality issues apparent in the sensor readings:

Noise: the mobile sensors seem to have more noise, in the sense that there is more fluctuation between successive measurements: when measurements are plotted as a scatterplot, the measurements for the mobile sensors are much more spread out.

The measurements from mobile sensors (left, with labels beginning with ‘M-’ written in black) seem to be noisier than those from the static sensors (right, with labels beginning with ‘S-’ written in blue)

Measurement gaps: the static sensors provided a continuous stream of measurements (with the exception of a gap in measurements reported from sensor 6). In contrast, only a small proportion of the mobile sensors operated without interruption.

There are many interruptions in the measurements from mobile sensors (left), but only a single interruption to measurements from the static sensors (right)

Contamination: several of the mobile sensors had measurements that remained very high (above 100 cpm) for a prolonged period, apparently because the cars on which they mounted were themselves contaminated. Whilst contaminated, a mobile sensor was essentially reporting its own contamination status rather than measuring anything about the environment.

Red marks indicate a value that is outside the displayed range (0-100 cpm); a pattern of continuous red marks, suggesting contamination of a vehicle carrying a sensor, can be seen for several mobile sensors beginning on Friday evening. No such direct detector contamination occurs for the static sensors.

Isolated outliers: both the static and mobile sensors have a relatively small number of isolated measurements that are >100 cpm, and far higher than the surrounding measurements (measurements >100 cpm also occur at the peak of spikes for static sensors, as discussed below, and for contaminated mobile sensors, as discussed above)
Spikes in measurements: most of the fixed sensors experienced short spikes in measured activity, which spanned many measurements. Interestingly, several of these spikes occurred nearly simultaneously at several stations and may be due to seismic activity, whilst others apparently affected only a single sensor. Similar spikes are not really noticeable in the measurements from the mobile sensors, but might be masked by the presence of more noise and other artifacts.

The first major shake on Wednesday afternoon coincides with a spike in reported activity for static sensors 4, 12, 13 and 15 These increase uncertainty, as the ‘true’ value of radioactivity is not known for these times; in retrospect, activity during these times could be estimated by interpolating between the measurements before and after the spike, but while the spike is occurring, it would be difficult to determine whether or not this spike was masking an increase in radioactivity level due to an earthquake-triggered event.

Interestingly, the second major shake on Thursday afternoon does not seem to cause such a spike: instead it cause a step-up in measurements from sensors 9 and 12.

The vertical lines highlight the time of the first major quake (on Wednesday morning); this quake seems to have caused brief spikes in the measurements from static sensors 4/12/13/14/15.

The vertical lines highlight the time of the second major shake (on Thursday afternoon); this quake seems to have caused a step-up in measurements from sensors 9 and 12, rather than spikes in measurements.

Mobile sensor 18 has an interesting pattern: it is mostly not providing measurements; when it starts supplying measurements these ramp up to a high level, remain high (with a lot of fluctuations), and then ramp down before stopping.

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

Overall confidence in measurements

The highest levels of reported radioactivity are from cars in Scenic Vista (neighborhood 8); the levels from cars contaminated here are high enough to clearly stand out. However, it is much harder to be sure about the levels of radioactivity elsewhere on the island.

Mobile sensors 12, 19, 20, 21, 24, 25, 27, 28, 29, 32 and 45 appear to have been contaminated. It is difficult to extrapolate from this to make a statement about the total number of cars contaminated, but this is 22% of the cars with sensors.

Whilst contaminated, these cars seem to have mostly remained in the vicinity of where they became contaminated. However, some may have left the island along the Wilson Forest Highway, and potentially spread radioactive contamination to the mainland; this is a potential cause for concern.

Space-time trajectories of the cars suspected to have been contaminated (mobile sensors 12, 19, 20, 21, 24, 25, 27, 28, 29, 32 and 45). Color encodes measured radioactivity using a Viridis colormap (blue is low activity, yellow is high activity). Some of these cars seem to be using the Wilson Forest Highway (far-right corner), which leads off the island to the mainland

Static vs mobile sensors

For the reasons given in answer to question 2, static sensors seem to be a more reliable source of measurements.

However, there are other considerations that should be taken into account:

whilst contamination of mobile sensors interferes with the measurement of background radiation levels, it does provide direct information about vehicle contamination
static and mobile sensors may have different costs, and it may be possible to obtain more sensors of whichever type is cheaper if there is a fixed budget available
analysis of measurements from a fixed sensor is much easier: it is easy to identify a change point, whereas a change in a measurement from a moving sensor will in general have contributions from both changing time and location.

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 static sensors seem to provide more reliable measurements, but the mobile sensor network provides a greater number of sensors, and as these sensors move around the island they provide measurements that are more dispersed than would be provided by the same number of static sensors. They are also able to provide direct information about vehicle contamination.

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.

For simplicity and speed of implementation, I treated the dataset as a static collection of measurements, but in a real disaster-response scenario it would probably be preferable for a tool to ingest a real-time stream of measurements.

However, the same visual representations could be applied to streaming data: handling a stream would require a change in how the data is processed, but not necessarily in how it is presented.