when local information alone (method 2) is
not sufficient.
In order to take into account seasonal effects, a moving window of +/– 45 days centered on the
target day was considered for the selection of analogues, so that in an archive made of Y years,
each target day was at most associated to Y x 91 potential analogues. The window size was
arbitrarily defined and not optimized. A similar window
/media/vedurstofan/utgafa/skyrslur/2013/VI_2013_008.pdf