| FLUXES | SST |
WINDS |
| Evaluation of
Satellite Products Satellite sensors provide data that
can be mapped onto a regular spatial and temporal
grid. The data may be simply interpolated, merged with
other observations or assimilated into a dynamical
model. Products using the same observations may differ
substantially in both resolution and accuracy.
|
CLIMODE mooring location |
| Here we present a method for evaluating products with a focus on the accuracy of anomalies with respect to research-quality field observations. The observations used for these comparison are from the CLIMODE mooring in the Gulf Stream in the North Atlantic and from the KEO mooring in the Kuroshio Extension region of the North Pacific. | KEO mooring location |
 |
Taylor Diagram To compare a time series with observations ("truth") we use a Taylor diagram (2001), which includes the correlation and the typical magnitude of anomalies relative to "truth." The mean is first removed from each time series. The Taylor diagram is a plot in polar coordinates where the angle represents correlation r, given by theta = cos-1 r, and the standard deviation of the anomalies relative to the standard deviation of "truth" is the radius R. In the Taylor diagram the normalized error of the time series is the distance of the colored dot to the red dot (perfect result) on the x-axis. The time series corresponding to the green dot has a correlation greater than 0.95. The anomaly magnitudes are smaller than those of the data ("truth") so the dot is inside the R=1 circle. A second time series, corresponding to the black dot has a lower correlation and larger magnitudes than "truth." The first time series has a smaller normalized error, that is, the green dot is closer to the red dot than is the black dot. |