Post 5 – Part 1. Bore dda Frindiau!
How to not get killed by an ice-cream
Yes, you have read correctly, ice-creams kill. Do not
let yourself be deceived by their sugary and innocent appearance. Ice-creams
are well known for their ability to kill thousands through their secret weapon:
drowning. Sceptic? Take a look at the graphs! Not only they are vicious
killers, they also cause fires in the woods and destroy our forests. There is
enough evidence to forbid ice-creams from your neighbourhood FOREVER!
Or maybe not?
Welcome to this new post on our ferry surveys data. This
week we are going to talk about seasonality, and I thought it would be a great
opportunity to introduce the infamous… confounding factors!!
And what does this have to do ice-creams? Ice-creams, of
course, do not kill nor generate fires, however as shown in the graphs, there
is a clear correlation between both. And here comes the confounding factor:
temperature. If we include temperature in the equation, we can make more sense
of these graphs. As the summer comes, and the temperature increases, the
ice-creams sales also increase. It is during this summer period, when the
affluence to swimming pools and beaches is at its highest, increasing the
probability of drowning. And the same occurs with the fires, which are directly
related to higher temperatures, but not so much to ice-creams. Another way of
saying this is “Correlation does not imply causation”. And there are thousands of examples. This is
something that is not easy to avoid when analysing data and we, at Sea Trust,
have encountered the same challenge.
In our case we wanted to use our data to investigate if the
cetaceans that we are studying presented seasonal patterns. So we analysed the number
of sightings per season as shown in the plot. For all the species there is
a common pattern, the maximum number of sightings occurs during the summer
season whilst the lowest numbers are recorded during the winter.
This can indeed be telling us something about the migratory behaviours,
however, there are some confounding factors that we had to scrutinize, for
example the number of surveys! The highest number of surveys takes place during
the summer, in contrast, the number of surveys that we are able to carry out
decreases significantly during the winter. So are we really seeing more
porpoises in summer? Are common dolphins less
present in the area during the winter?
Now that we are aware of this, we could think of visualizing
the number of sightings per survey. Interestingly, the shape of these
plots is now substantially different. The number of
common dolphins during the winter is higher than what the previous plot
suggested; and the new metrics seem to suggest that the highest
number of porpoises per survey is recorded during the spring rather than the
summer.
Even though through these graphs (number
of sightings per survey) we are getting a more realistic picture of what is
going on out there, we are still under the influence of other factors that we
need to account for. The sea state,
the temperature, the wind, etc., these are all variables that interact with the
epipelagic zone, but also with our survey conditions and thus, with our ability
to detect cetaceans. During the winter, the adverse weather conditions make it
more difficult to carry out as many surveys as during the spring and summer.
Likewise, these conditions interfere directly with the number of cetaceans that
can be detected during surveys, for example the number of sightings per survey
decrease sharply when the sea state is over 3 (Beaufort scale). And finally,
the number of observers that are available during the winter season tends to be
lower.
Luckily, as we explained during our first ferry post, this
data-set is particularly valuable because we have been able to maintain the
surveys under favourable weather conditions and the number of observers per
survey all year round. Meaning that most of these surveys took place with a
consistent average number of 3 observers and with good visibility and low sea
state values (<=3). At the same time, as we have such a comprehensive data-set,
we can use this data to understand how survey conditions influence the results
and derive some correction factors.
For this piece of analysis, we were able to remove surveys
in which the minimum favourable conditions were not met. This means, the
results are not as influenced by external factors and thus we can make
inferences with more confidence. Nonetheless, caution is always taken, as there
are other factors that are out of our control, such as the fatigue experienced
by the observers, which, due to the freezing temperatures. tend to be higher
during the winter.
To be continued …
Hasta la vista Cristina!