me parece que a lo que se refiere Fraus es a que sobre las medidas, que hacen desde hace tiempo, la noaa por ejemplo, aplican factores de corrección, islas de calor, cambio de instrumental y/o ubicación, filtros y desestimaciones,....
por lo que el valor resultante, aparte del error sistemático de medida, es diferente al absoluto de las estaciones,
vamos, que hay unos ajustes dinámicos que no sabemos
Ajustes C.R.U sobre Sabiñánigo, entre otros.
ANN variant results for temperature predictand indices (Figs. 4.20, 4.22, and
4.24) show a high degree of spatial coherence, and the Gaussian variant does not show
the difference between eastern and western skill seen in Figure 4.15. Skill is generally
very high across the majority of the basin for all indices except HWDI, which displays
only moderate or high skill in similar locations (northern and eastern Spain, southern
France, eastern Italy, and south eastern Greece) for all variants. These regions are the
same as those for which moderate or poor skill is found with the OSR variants, as
described above. Temperature predictand index performance for ANN variants differs
from that for OSR variants in one other respect,
two stations show consistently
negligible skill across all temperature indices and variants: Nis in Yugoslavia and
Sabiñanigo in northern Spain.
The ANN method is clearly not appropriate for these
stations, and greater skill for all temperature indices can be found using
the ROSR
variant. Otherwise the most effective method (in terms of skill) for all winter
temperature indices,
is the Linear ANN variant.REFERENCIA :
http://www.cru.uea.ac.uk/cru/pubs/thesis/2006-harding/MCE.4.pdfSALUDOS.