Reconstructing Historical Winds - Methodology
-
Divide the wind observations into a relatively well-sampled
recent interval and a more poorly sampled interval in the past.
Recent interval: 1945 - present
Past interval:
prior to 1945
-
Identify patterns of wind variations in the recent interval
using statistical pattern recognition techniques.
Examples of patterns include those
associated with El Nino,
the North Atlantic Oscillation and
other modes of climate
variability. Note that, because
there is noise in the wind
observations, some of the patterns
will not correspond to
climatologically important wind varibility.
We filter out
these "noisy" patterns.
-
Fit the sparse data from the past interval to the climate
patterns recognized from the recent data. For each time in the past,
estimate the relative importance of each climate pattern.
-
At each time in the past, compose the total wind field by
multiplying each climate pattern by its relative importance.
-
Estimate the uncertainty of the wind reconstruction for each
time in the past based on the distribution of observations and our understanding
of recent wind variability. Note that the uncertainty is largest
for times with little data.
