GEOS 597e
Spatiotemporal Data Analysis Workshop
Prework 3: Know your data
This week we'll have our first look at the data we are going to
use to study empirical orthogonal function analysis, and we'll estimate
the temporal autocovariance of sea surface temperature anomaly by
calculating the temporal covariance matrix of the gridded dataset.
Last updated 9/13/06. To be
completed prior to class session Weds., Sept. 13th.
Introduction:
Since the mid-19th century, volunteer
observing ships and research vessels have made and reported
observations of marine meteorological and atmospheric variables.
In the 1980s and 1990s, the Global Ocean Surface Temperature
Atlas (GOSTA) project produced from this 'raw' data a gridded dataset
for use in climatological studies. This was -- and is -- not an easy task, for the heterogeneous nature of the dataset
makes corrections for known
and suspected random and systematic error difficult to know
precisely. Nevertheless, many of us in the climate research
community would like to use this dataset to address a variety of
important research questions. Over the next several weeks we will
analyze gridded sea surface temperature anomalies, a dataset known as
British Meteorological Office Historical Sea Surface Temperatures
(MOHSST5).
Reading:
- Bottomley et al.
(1990),
Introduction.
- Optional but fun: spend some time viewing the data using a fairly
intuitive graphical browser here.
Reading questions:
- What scientific question would you like to ask of this dataset?
- What are the potential uncertainties in analyzing this (or any of
this class of) long-term gridded datasets? Your list should be
generalizable to other data sets you may find yourself analyzing in the
future.
- Suppose I have two variables, X and Y, which both vary in time
(i.e. X
and Y are time series). I have a realization (loosely speaking, a
finite set of observations) of each time series
variable, x and y respectively, which are column
vectors of length nt.
Look up and write down the equation to calculate the
covariance estimate c of x and y, using matrix notation (see Homework 2 for clues).
Products to hand in:
- In 1-2 paragraphs, please concisely write down your scientific
question and
list of
uncertainties. Be prepared to discuss in class whether you think
this
dataset is sufficient for the study of the scientific question you have
in mind, given its inherent uncertainties.
- Your equation for the covariance of x and y. Also,
please answer the question: why is c
an estimate of the true
covariance C
between X and Y?
Back
to Schedule/Syllabus.