Geoscience Reference
In-Depth Information
Fig. 4 Mean absolute error
(MAE; top panel), mean bias
error (MBE; middle panel) and
model performance index (d
r
;
bottom panel) for 14 GCMs
across the PADB for 1961-1990
Table 2 Spearman's rank correlation coefficient q between the pan-Arctic temperature and precipitation
values of MAE, MBE and d
r
MAE
MBE
d
r
q
-0.13
0.41
-0.13
basins is difficult if one wants to have consistently good model performance for both of
these parameters.
Model performance also varies considerably between the studied basins. The large
Mackenzie and Yenisey basins exhibit consistently good d
r
values for temperature, while
the d
r
of some models is considerably smaller for the Ob and Lena basins (not shown). In
contrast, temperature d
r
values for the relatively smaller and neighboring basins of Pechora
and Severnaya Dvina are poor for most models, indicating difficulties with temperature
simulations west of the Ural Mountains. This may be due to the relatively smaller size of
these basins, or the presence of natural low-frequency variation in the region (Hurrell and
van Loon
1997
). Precipitation d
r
values are overall lower, with the lowest values for the
smaller Olenek and Severnaya Dvina basins, followed by the medium-sized Yukon basin,
and the highest for the large Ob, medium-sized Nelson and small Khatanga basins.
Furthermore, results from an analysis of bases for prioritization monitoring under
conditions of climate change (Bring and Destouni
2013
) indicate that the basins with the
highest recently observed deviations of temperature and precipitation are not the same
basins that have the highest projected changes in future climate (rank correlations of basin
orders are close to zero, or negative; Table
3
). Therefore, prioritizing monitoring to basins
