Prediction of floods in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 222

may be difficult to find catchment characteristics that
are suitable for regression methods.

by the lines representing comparative studies. This clear
trend is in line with the Level 1 assessment. Arid regions
tend to be more heterogeneous than humid regions and
runoff processes are more non-linear. There is also a
decrease in performance with the air temperature (T A ).
In the catchments analysed, the warmer catchments tend
to be those with the larger aridity index, so this is
consistent with the dependence on aridity. There is a
slight increase in performance with elevation but, in
contrast to aridity and air temperature, the biases do not
change much with elevation. In the studies examined
here, the highest elevation catchments are influenced by
snowmelt, so there is a tendency for the flood predictions
to improve if snowmelt is involved in the flood gener-
ation processes.
The results stratified by catchment area (fourth panel,
Figures 9.28 and 9.29 ) indicate a clear increase in per-
formance (decrease of ANE) with increasing catchment
area for all methods. The increasing performance with
catchment size is likely related to two factors. The first
is related to the data availability. As the catchment size
increases the likelihood that gauged subcatchments
are available as donor stations increases. This will lead
to more reliable transfer of the flood characteristics. Add-
itionally, for larger catchments, there are aggregation
effects on the flood generating processes, so floods tend
to be less flashy and therefore easier to predict. None of
themethodsisbiasedinrelation to catchment area (fourth
panel, Figure 9.29 ).

- Runoff prediction performance increases clearly with

number of stations in a region, highlighting the need
for a dense stream gauge network for predicting
floods in ungauged basins.

9.5.2 Level 2 assessment
The Level 1 synthesis of existing studies ( Table A9.1 )
clearly showed that many studies only report summary
statistics of regionalisation performance and/or catch-
ment characteristics, which hampers detailed attribution
of the performance and inter-study comparison of results.
The objective of the Level 2 synthesis is to examine and
explain the performance of the regionalisation methods
in greater detail. Five authors from studies included in
the Level 1 assessment provided detailed information
about climate and catchment characteristics in a consist-
ent way and reported the regionalisation performance for
each catchment ( Table A9.2 ). This data set combines
data from 1640 catchments, three groups of regionalisa-
tion methods and four catchment characteristics. The
regionalisation methods are regression, index methods
and geostatistics. The catchment characteristics are arid-
ity (potential evaporation by mean annual precipitation),
mean annual air temperature, mean elevation and catch-
ment area. The performance was assessed based on the
100-year floods. The normalised error (NE) and the
absolute normalised error (ANE) were used as runoff
prediction performance indicators (Table 2.2). The NE
highlights biases in the methods, while the ANE is a
measure of the overall performance. Note that the ANE
is an error measure, so it has been plotted downwards on
the vertical axis to make it comparable with the perform-
ance measures, i.e., higher up in the plot indicates better
performance. For comparison with the other runoff sig-
natures in Chapter 12 ,theR 2 of the 100-year flood
quantile were calculated for all methods in each study
separately, considering only those sites with at least 40
years of data. The 25% and 75% quantiles of these R 2 are
0.53 and 0.70, respectively.

Which method performs best?
Figure 9.30 summarises the runoff prediction perform-
ance of different regionalisation approaches, stratified by
the aridity index. The top, middle and bottom panels show
the performance for all catchments in Table A9.2 ,and
catchments with an aridity index below and above 1,
respectively. Analysis of the overall performance of the
three methods shows that performance is similar for geo-
statistics and index methods, which have a slightly better
performance than the regression methods. For humid
catchments, again, the performance of geostatistics is
slightly better than index methods, and the performance
of the regression methods is slightly lower. For dry catch-
ments, however, the index methods perform significantly
worse than the other two methods. The low performance
of the index flood methods in arid regions may be related
to the underlying assumption of using the same non-
dimensional flood frequency curve (i.e., growth curve)
in the entire region. Arid regions may be spatially more
heterogeneous, leading to lower performance. More
importantly, most arid catchments are strongly biased in
that the predictions overestimate the 100-year floods
( Figure 9.29 , top centre). The median normalised error is

To what extent does runoff prediction performance depend
on climate and catchment characteristics?
The assessment of the ANE and NE error measures with
respect to the four climate and catchment characteristics
is presented in Figures 9.28 and 9.29 , respectively. The
lines indicate the median runoff prediction performance
of catchments belonging to the same study. The top panel
shows that the errors, ANE and NE, clearly increase with
increasing aridity, i.e., there is a decrease in performance
with aridity for all three methods. This is also supported