Geography Reference
In-Depth Information
Thiessen polygons
Inverse distance
12
12
10
10
8
8
6
6
4
4
2
2
0
0
-2
-2
-4
-4
-6
-6
-8
12
-8
-4
-4
5
10
5
-2
-2
4
4
3
3
0
0
2
2
1
1
0
0
8
2
2
-1
-1
-2
-2
-3
-3
PC1
PC2
PC1
6
PC2
Ordinary kriging
Universal kriging
4
2
12
12
0
10
10
8
8
6
6
4
4
2
2
0
0
-2
-2
-4
-4
-6
-6
-8
-8
-4
-4
5
5
-2
-2
4
4
3
3
0
2
0
2
1
1
0
0
2
2
-1
-1
-2
-2
-3
-3
PC2
PC1
PC2
PC1
Figure 7.15. Spatial interpolation in the principal component space for the estimation of a quantile of the FDC (Q 97 low flow). Representation
of the surfaces generated in the space of catchment characteristics by four different techniques for the River Aterno at L ' Aquila, Italy.
From Castiglioni et al.( 2009 ).
technique for interpolating long-term dimensionless FDCs
in the physiographical space. The horizontal coordinates
are the first and second canonical variables of the set of
available catchment characteristics (i.e., two-dimensional
physiographic space), whereas the vertical coordinate is
the runoff duration (see Figure 7.15 ).
parametric) with empirical estimates of long-term FDCs
based upon a few years of observation (one, two and five
years). This resampling experiment shows that predictions
of long-term FDCs based upon five years of observed
runoff largely outperform the best performing regional
model, while one or two years of daily runoff are generally
sufficient to obtain predictions of FDCs that are as accurate
as the ones retrieved through regionalisation. This result
highlights the value associated with observed runoff, even
for short series. As a general remark, the regional estima-
tion of the FDC for ungauged basins should be performed,
where possible, by applying different approaches and
selecting the estimated FDC only after a scrupulous analy-
sis of the suitability of each regional model to the particular
sub-region where the ungauged site is located (Castellarin
et al., 2004a ). Also, it is generally not advisable to rely
7.3.4 Estimation from short records
Castellarin et al.( 2004a ) presented a series of resampling
experiments to assess the sensitivity of empirical FDCs to
the sample length (see Chapter 11 , Case studies, for more
details). The analysis considers a number of river basins
with long series of daily runoff, and compares the cross-
validated FDCs predicted with three different regional
approaches (i.e., quantile regression, parametric and non-
Search WWH ::




Custom Search