[Back]

S. Zwieback, K. Scipal, W. Dorigo, W. Wagner:
"Structural and statistical properties of the collocation technique for error characterization";
Nonlinear Processes in Geophysics, 19 (2012), 12 pages.

The validation of geophysical data sets (e.g. derived
from models, exploration techniques or remote sensing)
presents a formidable challenge as all products are inherently
different and subject to errors. The collocation technique
permits the retrieval of the error variances of different
data sources without the need to specify one data set as
a reference. In addition calibration constants can be determined
to account for biases and different dynamic ranges.
The method is frequently applied to the study and comparison
of remote sensing, in-situ and modelled data, particularly
in hydrology and oceanography. Previous studies have
almost exclusively focussed on the validation of three data
sources; in this paper it is shown how the technique generalizes
to an arbitrary number of data sets. It turns out that only
parts of the covariance structure can be resolved by the collocation
technique, thus emphasizing the necessity of expert
knowledge for the correct validation of geophysical products.
Furthermore the bias and error variance of the estimators are
derived with particular emphasis on the assumptions necessary
for establishing those characteristics. Important properties
of the method, such as the structural deficiencies, dependence
of the accuracy on the number of measurements and
the impact of violated assumptions, are illustrated by application
to simulated data.

http://dx.doi.org/10.5194/npg-19-69-2012

http://www.nonlin-processes-geophys.net/19/69/2012/npg-19-69-2012.pdf