GFI/BCCR Seminar: On physical interpretation of errors in ocean surface current analysis

Rick Danielson (NERSC):

On physical interpretation of errors in ocean surface current analysis

Abstract
This characterization of the GlobCurrent ocean surface current analysis applies methods calibration and validation across a broad range of physical conditions. One of the apparent ironies of a relatively complete observational analysis is its paucity of independent observations, which makes it difficult to characterize the analysis over a range of conditions. This requirement of independence may not be so dire, however, if we first postulate that the impact of observations on an analysis is generally local, as is the propagation of errors in space and time. 

Our approach exploits temporal extrapolation of the GlobCurrent analysis in both directions (T-symmetry), while avoiding a central extrapolation time window. This permits a calibration and validation by triple collocation using only two datasets (yielding many more collocations than before). First, scatterplots of extrapolated analysis versus in situ observations are interpreted physically. We seek evidence of analysis performance improvement (as expected) using the calibrations that can be derived, including across an entire current speed range from the low (0.1 ms-1) and high (1 ms-1) ends. Finally, we compare the efficacy of functional dependence on either current speed, distance to the coast, or bathymetry. Current speed appears to yield the most enlightening single-parameter dependence of triple collocation calibration and performance.