BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:British Antarctic Survey's Natural Complexity: Dat
a and Theory in Dialogue
SUMMARY:Unnatural Complexity: Towards Achievable Aims for
Complicated Simulation Models in Applications with
Data (Weather-like) and those without (Climate-li
ke) - Leonard A. Smith\, London School of Economic
s\, London\, U.K.
DTSTART;TZID=Europe/London:20070813T140000
DTEND;TZID=Europe/London:20070813T144500
UID:TALK7793AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/7793
DESCRIPTION:Complexity in Nature is seen in the behaviour of s
ystems often viewed as having many interacting par
ts\, while the success of scientific models is mos
t apparent in systems which have been isolated as
much as possible. Natural complexity flows from th
e seamless interaction of these parts as a single
whole. While state-of-the-art scientific simulati
on models are\, arguably\, unnaturally complex. R
ather than aiming to simplify either the system st
udied or the mathematical framework employed\, var
ious important bits of a natural system are modell
ed explicitly. These components are then bolted to
gether\, along with an assortment of parameterisat
ions to handle those other known aspects of the sy
stem which are not "resolved explicitly".\n\nHow c
an these simulation models be gainfully employed t
o address questions of decision-support in high-pr
iority areas of Earth system science? And how migh
t the dialogue between model and observation be im
proved to speed the advancement of both? Can we fi
nd a more coherent approach interpreting data in a
field often characterised by what it is not (non-
linear\, non-Brownian\, non- Gaussian tails\, ...)
? In particular\, how might we develop faith that
these unnaturally complex models can extrapolate b
eyond observed conditions to provide a "comprehens
ive\, physically consistent\, prudent projection"
of future likely conditions? Does Bayesian emulati
on provide a decision-relevant alternative to naiv
e realism? Do our simulation models have a well-de
fined mathematical target? Do we have grounds to
believe that unnaturally complex models will be st
ructurally robust?\n\nThese questions will be addr
essed within the framework of nonlinear dynamical
systems. While modern computing power allows us mu
ch more complicated models\, similar issues concer
ned Maxwell and Poincare. We will ask what achieva
ble aims we might target given the mathematical na
ture of our models\; weather-like applications whe
re we have a historical archive of forecast-observ
ation pairs will be contrasted with climate-like a
pplications where by definition no such archive ex
ists. Resource allocation (in terms of experimenta
l design) for model improvement will be contrasted
with that for decision support. Our ability to f
orecast\, and in particular to construct decision-
relevant probability forecasts from ensembles of s
imulations\, will be examined. Probabilistic Simi
larity and Shadowing are suggested as primary tool
s to evaluate the decision-support relevance of en
semble simulations with imperfect models. Applica
tions in weather forecasting and climate modelling
are contrasted with those involving Near-Earth Ob
jects to suggest necessary conditions for deployin
g probability forecasts (as such) from imperfect m
odels in practice. This appears to be somewhat mo
re difficult than using state-of-the-art simulatio
n models to improve our understanding of natural c
omplex systems.\n
LOCATION:Law Faculty\, Cambridge
CONTACT:Nick Watkins
END:VEVENT
END:VCALENDAR