ANSWERS

Answer 1
Analyzing the data at a first glance, we can detect un unexpected behaviour of the sensors. In particular we can notice that in given hours some monitors emit two readings for a single chemical while no reading is provided for an other chemical; furthermore one of the two readings given for a chemical is very high with respect to the other one.

In particular, by inspecting the line charts, we discovered that the chemicals involved in this strange behaviour are AGOC-3A and Methylosmolene. From the charts presented above, analazying the green line associated to Monitor 3, it is clear that whenever a monitor seems to be turned OFF with respect to the chemical Methylosmolene, during the same hour it gives two very different readings of chemical AGOC-3A, represented by the orthogonality and the length of peaks in the first chart. We detected this behaviour in all months.

Answer 2
Now we turn our attention to the chemical release itself by summing all the readings given by all monitors for a given hour, in order to have the total chemical emission; furthermore we arrange this derived data in heatmaps, organized by days on Y axis and by hours on X axis.

By inspecting these heatmaps, we noticed that for the chemicals Appluimonia and Chlorodinine we have an almost uniform release patterns; while for the chemicals AGOC-3A and Methylosmolene we can detect a particular pattern that follows from the result we have explained in the first answer. It is clear that the two chemicals release trends are complementary, meaning that when one presents a series of high measurements the other shows much lower readings and the other way around. In addition, we have noticed that the inversion of the trends between this two chemicals happens in two moments during the day:
- around 6:00, when Methylosmolene stops emitting high values while AGOC-3A starts;
- between 18:00 and 21:00, when AGOC-3A stops emitting high values while Methylosmolene starts;
Once again, this behaviour has been found in every month but not every single day.

Answer 3
In order to identify which factory emits which chemical, we combined readings data for chemicals releases and wind measurement; in particular, we derived the wind measurement for the given missing hours by taking the two consecutive existing wind measurements whose hours are closest to them and considering the wind direction moving uniformly from the first hour to the second one, deriving the value we are interested in; as regards the speed, we simply considered the mean between those two consecutive wind measurements.
In order to identify which factory to assign to monitors readings in a given hour, we take the angle between the wind direction and the direction joining the monitor and the factory and assign that reading to that factory for which this angle is minimum. Considering the wind speed, we computed the exact hour when that value of chemical was released by simply taking the hour of reading, subtracting to it the time needed to go from factory to monitor with the wind speed and, finally, rounding this value to an exact hour.
This process is described in the image below, where factories are represented by red crosses and monitors by black circles.

After that, we chose to represent by months our just derived data, through four heatmaps depicting the factories emissions of a given chemical. In such a way we can easily detect which factory emits the most of a given chemical and also discover any uncommon pattern of release.

From the analysis of the heatmaps we presented above, we can get the following results:
- for Appluimonia, Kasios and Roadrunner emits the most of chemical, immediately followed by Indigo; Radiance's emissions are much more toned down with respect to the other factories.
- for Chlorodinine, we have a similar trend as we have for Appluimonia; in particular the most of emissions are given by Kasios, Roadrunner and Indigo, while Radiance give a very little contribution.
- for AGOC-3A and Methylosmolene we have to analyze them together, because their release are related through a complementary relationship, as we noticed in the answers above. In particular this particular phenomenon is reproduced in the emissions of Roadrunner, Indigo and Radiance, while Kasios has a much more linear emissions for this particular chemicals. By a careful reading of the chemicals documentation, we can understand that Methylosmolene is much more toxic with respect to AGOC-3A; for this reason we suspect that those three industries tampered with sensors in order to maintain low emissions of Methylosmolene by transforming this emissions in AGOC-3A emissions; this could be the reason for which we have two readings of AGOC-3A in a single hour and none for Methylosmolene. Furthermore we can notice that the hours of shift of this complementarity behaviour could almost coincide with the beginning and end of factories employees working hours, possibly implying that some employees are in charge of tampering sensors before starting to work and fixing them up before coming back home.