Johan Lindström
Mathematical Statistics, Centre for Mathematical Sciences
Lund University
Lund, Sweden.

Title: Image Warping for Forecast Verification

Abstract: A method for forecast verification using image warping will be presented.
Image warping based on thin plate splines is used to deform forecast 
fields to match observation fields. The warping approach gives an estimate 
of the location errors in the forecast, as well as the residual amplitude 
errors. The statistical formulation of the image warping problem should 
also provide (approximate) confidence intervals for several features of 
the deformation; Such as global translation, global rotation and the 
reduction in the amplitude error that the warping deformation gives rise 
to, i.e.~how much better than the original forecast the deformed forecast 
is.

The talk will give a brief background to image warping, followed by a 
description of our verification method. Possible extensions, issues and 
limitations will also be discussed.