64
Chapter 8
been partly corrected with ISBA. Over Spain, the reduction of the
temperature bias is important (the error is reduced by a factor 2). Clearly,
this impact is related to the improvement of the vegetation map. In the old
scheme, the fraction of vegetation over Spain was too high, resulting in a
forecast which was too cold and wet.
With the AVHRR classification and the calibration of ISBA over Spain
(particularly for dry soils, see Giordani et al. 1996), the surface flux
treatment has been improved, leading to acceptable score over the Iberian
Peninsula. The improvement over France is not as good because the old
system was already well calibrated for that domain. Overall, the general
quality of the forecast of quantities at 2 m was improved in this experiment
for the European domain for which the AVHRR classification was available.