PUB in practice: case studies - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 280

for this is that the relation between runoff depth and
geographical characteristics is very complicated and it
is not always possible to use one regression function to
show these complex relations.
(4) The more detailed information obtained by this
method on the spatial distribution of runoff is very
useful for river basin management, such as the selec-
tion of reforestation areas, reservoir sites etc.
(5) One potential merit of the regression model is that it
may be extended and used to predict runoff depth in
ungauged basins with similar geographical condi-
tions, whereas interpolation methods do not have
this function. Because of the geographical differ-
ence between the north and the south in the study
area, we used two different equations even in the
same basin. This shows that it is very difficult to
assess the geographical similarity between various
areas, and it requires more effort and study to find a
feasible method to decide whether an area

50 km

Figure 11.8. Revision based on vegetation depth.

s runoff
depth can be estimated by the runoff depth equation
of another area.

'

the revised contours were improved. The standard devi-
ation of the relative residuals dropped further to 11.7%
and the absolute standard deviation dropped from 18.0
mm/yr to 16.5 mm/yr. Stepwise regression, however,
indicated that the vegetation is strongly related to eleva-
tion and that it did not add much to what the elevation
already provides. Moreover, in the valleys where crops
are grown, the correlation between vegetation and runoff
depth is not present.

11.4 AN INDEX APPROACH TO
MAPPING ANNUAL RUNOFF IN A
SIBERIAN CATCHMENT, RUSSIA
l. m. korytny, e. a. ilyichyova and
b. gartsman

The issue from societal and hydrological perspectives
By combining hydrology and geomorphology we can
obtain a uniform principle to analyse structures of valleys
and river networks. An innovation in the analysis of river
systems is the introduction of entropy characteristics,
which have high information content and are closely
associated with hydrographic characteristics of river
systems. (Gartsman et al., 1976 ). The entropy character-
istics take into account the number of elements in the
structure and their distribution, and the interrelationships
of these elements, which permit them to be used as
informative parameters, closely associated with the
functional characteristics of a river system. These
parameters take account of the structural characteristics
of river systems, such as hierarchy, ordering and co-
subordination. This new approach is highly relevant for
PUB, since it can derive mean annual discharge estimates
on the basis of the physical characteristics of the basin
structures. Such maps are of high societal value when
planning water resources development in poorly gauged
or ungauged basins.

Discussion
In this study, we succeeded in improving the precision of
the mapping of annual runoff depth, using DEM and
related geographic data. By comparing various models
we reached the following conclusions:

(1) The mapping of annual runoff depth making use of
spatially distributed data appears to be better than the
empirical mapping still in use. This study shows that
using GIS software to map runoff depth can not only
improve the efficiency, but can also give a more reli-
able result.
(2) Geographical variables, such as elevation and vegeta-
tion cover, can be used to improve the precision of
mapping of runoff depth.
(3) Using multiple linear regression techniques, we also
created a regression model that relates runoff depth to
altitude and distance to the source of water vapour.
This regression model can achieve a similar precision
to that obtained using spatial interpolation methods.
But it fails to greatly improve the precision. The reason