This plot shows a test of our reconstruction of the east-west ("zonal") wind. This test examines the sensitivity of our methods to the geographic distribution of wind observations.
Our test takes a recent wind field measured by the satellite as "truth" and samples it using the network of ship observations for February 1930. Satellite observations of zonal wind are shown for February 1997 in the upper left panel. This wind field is sampled using the distribution of ship observations for February 1930 (upper center panel). Because the ship network was sparse in February 1930, the resulting picture of the wind has large gaps, particularly in the southern hemisphere and the Pacific Ocean (upper right panel).
We use the field in the upper right panel as input to
our wind reconstruction procedure. The results for three choices
of reconstruction methods are shown in the middle row. The bottom
row shows the difference between each of these reconstructed fields and
the full wind field observed by the satellite.
Note the good agreement between the original field in the upper left and the reconstructed fields in the middle row. The comparison is particularly good in the North Atlantic Ocean, where the network of ship observations was very good. The agreement is slightly worse, but still impressive in the Pacific Ocean, where there were fewer ship observations.
Results from tests like these gives us great confidence in our methods. Comparing various reconstruction procedures (e.g., center row) helps us decide how to fine tune the reconstruction procedure to obtain the best possible results with real data.
