Random Error in Space-Time Bin Averages of Sea Surface Temperature
Observations from Ships
Abstract
Sea surface temperature (SST) observations made at ships are distributed
irregularly in space and time and are affected by systematic biases and
random errors. Such observations are often “binned”: split into
samples, contained within “bins” - grid boxes of a space-time grid
(1oX1o monthly bins are used here), and their statistics are computed.
Bin averages often serve as gridded representations of such data, thus
requiring reliable uncertainty estimates, which for ship observations
are particularly important because of their domination in the early
observational records. Here ship SST observations for 1992–2010 are
compared with an independent high-resolution satellite-based SST data
set. To remove systematic biases, seasonal means were subtracted from
the difference between bin-averaged data sets. In more than 66%(50%)
of locations with binned temporal coverage exceeding 50%(66%), the
magnitude of remaining anomalies agreed within 20%(10%) with random
error model estimates. Separate estimates for sampling and measurement
error components were obtained.