This form is a web page which was created in MS WORD and therefore can be easily edited that way

Entry Name: "VRVis-Matkovic-MC2"

VAST Challenge 2017

Mini-Challenge 2

Team Members:

Rainer Splechtna, VRVis Research Center, Vienna, Austria, Splechtna@VRVis.at PRIMARY

Silvana Podaras, VRVis Research Center, Vienna, Austria, podaras@vrvis.at

Michael Beham, VRVis Research Center, Vienna, Austria, Beham@VRVis.at

Denis Gracanin, Virginia Tech, Blacksburg, VA, USA, gracanin@vt.edu

Kresimir Matkovic, VRVis Research Center, Vienna, Austria, Matkovic@VRVis.at

Student Team: NO

Tools Used:

ComVis: a visual analytics research tool developed at VRVis Research Center.

Python scripts: data processing and particle tracer used for MC2.3.

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

45h

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2017 is complete? YES

Video

https://drive.google.com/open?id=0B1K_OL3S7HTCN05nQlhjdktTRWc

Questions

MC2.1 – Characterize the sensors’ performance and operation. Are they all working properly at all times? Can you detect any unexpected behaviors of the sensors through analyzing the readings they capture? Limit your response to no more than 9 images and 1000 words.

The sensors are not working at all times. There are no readings on 2 and 6 April; 2, 4 and 7 August; and 2 and 7 December at 12:00 am for almost all chemicals (Figure 1). These outages never happen on a Monday but more often on a Wednesday (Figure 2). The exceptions, i.e., sensors still working, are:

· 2 August:

o Sensor 3: Methylosmolene and AGOC-3A

· 7 December:

o Sensor 6: AGOC-3A

o Sensor 7: Appluimonia and AGOC-3A

o Sensor 8: Methylosmolene and AGOC-3A

Figure 1: Sensor outages at 12:00 am for all chemical readings. The scatter plot shows the days on which the outages occur. The histogram shows that not all sensors for all readings are affected.

Figure 2: Weekdays of sensor outages at 12:00 am (Order in histogram: Friday, Saturday, Sunday, Monday, Tuesday, Thursday).

The readings for Methylosmolene are compromised more often. The outages are mostly one or two consecutive readings (hours) long and occur only between 6:00 am and 9:00 pm (Figure 3). The longest outage is six hours long (Figure 5) and multiple outages can occur on one day (Figure 4, Figure 5, and Figure 6).

Figure 3: All sensor outages (top). Mythlosmolene reading outages not at 12:00 am (bottom).

Outage days in April per sensor:

· Sensor 3: 7, 17

· Sensor 4: 12, 17, 22

· Sensor 5: 1, 2, 6, 8, 12, 21, 22

· Sensor 6: 2, 6, 12, 17, 21, 25

· Sensor 7: 14, 15, 16, 19

· Sensor 8: 15, 16 (Figure 4)

· Sensor 9: 1, 2, 8, 21, 22

Figure 4: Example of dating of outages of Methylosmolene readings for sensor 8 in April.

Outage days in August per sensor:

· Sensor 1: 2

· Sensor 2: 1, 20

· Sensor 3: 1, 2, 13 (Figure 5), 20

· Sensor 4: 3, 10, 14, 16, 17, 21

· Sensor 5: 4, 5, 6, 10, 11, 12, 14, 16, 17, 18, 24, 25

· Sensor 6: 6, 8 (4h), 9, 11, 12, 14, 17, 18, 22, 23

· Sensor 9: 11, 14, 18, 22, 24

Figure 5: Longest outage of Methylosmolene reading for sensor 3 on 13 August.

Outage days in December per sensor:

· Sensor 1: 5

· Sensor 2: 5

· Sensor 3: 1, 5, 12

· Sensor 4: 5, 7, 12, 17 (4h), 18 (5h), 24

· Sensor 5: 7, 8, 14, 15, 21, 22

· Sensor 6: 2, 8, 9, 13, 14, 15, 16, 18, 19, 23

· Sensor 9: 1, 2, 8, 11 (5h, Figure 6), 15, 21, 22, 24

Figure 6: Four outages of Methylosmolene reading for sensor 9 on 11 December.

Sensor 4 shows an interesting behavior. All readings of the measured chemicals show an increased level over the course of the three months of data provided (see Figure 7). This could indicate a calibration or drifting issue.

Figure 7: The readings of all chemicals captured by sensor 4 increase systematically per month. (April: orange, August: purple, December: olive).

Sensor 3 shows much more noise in the readings than all other sensors (see Figure 8).

Figure 8: Readings of sensor 3 (top) are very noisy over the course of all three months. Readings of sensor 5 are shown for comparison (bottom). (April: orange, August: purple, December: olive).

MC2.2 – Now turn your attention to the chemicals themselves. Which chemicals are being detected by the sensor group? What patterns of chemical releases do you see, as being reported in the data?

Limit your response to no more than 6 images and 500 words.

AGOC-3A readings show just a few peaks at 14 and 15 April and 5 December (all for sensor 6).

Figure 9: Peaks of AGOC-3A readings.

Chlorodine readings show many peaks even when filtering out (noisy) sensor 3 (Figure 10).

Figure 10: Peaks of Chlorodine readings without sensor 3 readings.

Appluimonia readings also show many peaks even when filtering out (noisy) sensor 3.

Figure 11: Peaks of Appluimonia readings without sensor 3 readings.

Methylosmolene peaks (> 10.0) show a particular pattern. They occur only between 9:00pm and 6:00am (on 20 of 36 days where peaks occurred the sensors also had outages during the day, Figure 12).

Figure 12: Peaks of Methylosmolene excluding sensor 3. Most peaks are detected at sensor 6. Many peaks occur on days when also sensor outages occur.

MC2.3 – Which factories are responsible for which chemical releases? Carefully describe how you determined this using all the data you have available. For the factories you identified, describe any observed patterns of operation revealed in the data.

Limit your response to no more than 8 images and 1000 words.

We run a particle tracer written in Python (Figure 13) where all factories are emitters and the particles arriving at the sensor positions are counted and accumulated per hour. We use the given meteorological data to setup the vector field by which the particles are propelled. Hence by running this particle tracer we create an additional data set, i.e., transform the meteorological data set, so that we now know for each sensor the number of particles arriving at a particular sensor and originating from a particular factory at a particular hour. Additionally, the percentage of particles originating from a particular factory relative to the total number of particles arriving at the sensor over the course of an hour is calculated and used as the “exclusivity” score.

Figure 13: Visualization of the particle tracer results.

By comparing the peaks in the sensor readings with the number of particles arriving at the same time the sensor readings are taken and using the “exclusivity” score we can determine which factories are responsible for the emission of which chemical.

Kasio is responsible for the emission of Methylosmolene (Figure 14, Figure 15). Emissions only occur during night hours.

Radiance is responsible for the emission of AGOC-3A (Figure 16, Figure 17). There are only few peaks/emissions occurring.

Figure 14: Peaks in Methylosmolene correlate with peaks in particles/percentage of particles arriving from Kasio.

Figure 15: Peaks in Methylosmolene correlate with peaks in percentage of particles arriving from Kasio. Individual days are shown (one day in each image).

Figure 16: Peaks in AGOC-3A correlate with peaks in particles/percentage of particles arriving from Radiance.

Figure 17: Peaks in AGOC-3A correlate with peaks in particles/percentage of particles arriving from Radiance. Individual days are shown (two days in top image, one day in bottom image).