Content uploaded by Kate Wright
Author content
All content in this area was uploaded by Kate Wright on Jul 08, 2022
Content may be subject to copyright.
Data to Decision
uncertainties in environmental data science
This new uncertainty typology shows the sources of uncertainty along the end-to-end journey from data to decision. Based on multidisciplinary theories and data collected from
focus groups and interviews, it shows different types of uncertainty experienced by environmental data scientists. From the processes that are causing, or are predicted to cause,
environmental change; the likelihood of the change; through to how communication can affect stakeholder perception and their judgement. All these are used to assess the risk
of action, or inaction, and impact the final decision.
Kate Wright, Bran Knowles, Gordon Blair
k.wright@lancaster.ac.uk
Quotes from Focus Groups with members of DSNE in July 2021 and Interviews with EDS experts conducted between April 2019 and March 2022
"uncertainty is trying to
quantify the unknown
variation"
"model parameters
come from
human decisions"
"observations
are quite sparse
over Antarctica"
"if I’ve tried to
use uncertainty in
a...less academic
setting...it is
essentially interpreted
as inaccuracy"
"terminology around
uncertainty is confusing...people
might interpret it in different
ways, so the biggest barrier to
collaboration is
misunderstandings"
"you can't wait
till you're certain
to make decisions"
Scientific Uncertainty is usually
divided into:
•Aleatory -used to describe natural
variability (environmental and
behavioural) and is quantified using
statistical techniques.
•Epistemic - due to limited
knowledge, and so, in theory, can
be reduced with further studies.
"one thing I come up
against is units...where
data has been
converted ... there's a
level of uncertainty
around whether or not
that's been done
correctly"
"data scientists will spend
90% of the time cleaning the
data, and then 10% actually
doing the analysis. And it's
just because .. whoever's
collected the data hasn't
really thought about what
you, who is analysingit, what
you're actually going to do
with it"
"it's a nightmare,
people send data
with no metadata
and don't
understand why it's
important"
"it's not always one
person recording data
for years and years, it
changes hands,
different people
record it and different
people can record it
slightly differently" "for me it's really
important to keep in mind
the kind of potential
impacts of somebody
making the wrong decision
based on, you know, poor
quality information or
information that hasn't
been communicated
properly"
"messaging and
uncertainty and so ... a
particular message is
portrayed that it's not
quite in line with what
we think the evidence
shows"
"doubt your own
research.. and so I think
I saw uncertainty in my
research is being a
bigger thing than other
people might have
done"
"confidence and uncertainty - I
guess I'm seeing it sort of similar
thing. You know, the more
uncertain you are, the less
confident you are".
"the hardest
bit is communicating
the uncertainty"
"uncertainty is not the
same as error, sometimes
people use that
interchangeably and that's
not very helpful because
uncertainty is more than
error"