• Spatially generalized precipitation adequately
reflects the areal distribution of actual precipitation and provides more accurate river
runoff computations as compared with point
observations at some weather stations. Obviously, this conclusion holds significance in
any hydrological calculations, which use areal
meteorological characteristics.
• Model parameters correspond to average
meteorological and hydrological characteristics of the mountain country, which significantly vary from one river basin to another.
For a single basin, one can specify some
parameters by using long-term river observations of WR/HCR and thereby considerably
reduce the computation error.
The developed high-performance WR/HCR
models have a large number of parameters that
characterize the universal hydrophysical and
hydrochemical features of river basins throughout the Altai-Sayan mountain country. In contrast, traditional runoff/streamflow models
(Dingman and Sharma 1997; Moriasi et al. 2007;
Sene 2008; Agal'tseva et al. 2011) have much
fewer parameters. The reason is that bringing
every additional parameter to these models does
require special field investigations to determine
its value. Additional parameters require studying
climatic, orographic, hydrogeological and
hydrochemical characteristics of a river basin, a
soil-vegetation cover, a glacier, a hydrographic
structure of river network, etc., that is costly and
time-consuming. SAM does not face such a
problem because it uses the implicit information
contained in long data series on dynamics of the
characteristic under study, namely, WR and
HCR.
Most traditional runoff models are based on
the data of a single river basin that makes the
analysis and quantitative estimation of landscape
structure influence on runoff extremely difficult.
In fact, only one value of an area can be related
to each landscape in the basin. It is hardly possible to establish runoff dependence upon a
landscape area using one available value of an
area. It is also impossible to exclude the individual features of the studied basin from the
model that restricts the model applicability to
other territories. It is obvious that SAM is devoid
of both problems.
To apply the developed model package to any
river basin of the Altai-Sayan mountain country,
we just need landscape and topographic maps
supplemented with normalized monthly air temperature and precipitation spatially generalized
for the country. The calculated long-term
dynamics of normalized WR and HCR for each
hydrological season can be easily converted into
real runoff dynamics (m
3 /s and g/s). To do this,
the value of long-term mean runoff should be
determined via comparison of calculated normalized WR/HCR and the real one for 1–2 years
of observations. Subsequently, using the right
side of Eqs. (7.3 and 7.4), one can find the longterm dynamics of WR and HCR in m
3 /s and g/s
for each geosystem group (landscape) and the
whole watershed.
Using the presented model package, it is easy
to forecast the WR and HCR for the next
hydrological season, i.e. for 3–4 months ahead. It
demonstrates a significant improvement of forecast accuracy as compared with the long-term
predictions using the long-term mean value of
the observed river WR. In particular, the most
probable seasonal WR can be forecasted with a
twice-reduced variance as compared with the
similar forecast based solely on the observed
mean WR. The forecast of seasonal WR, for
example, is particularly important for the regulation of water releases from mountain hydroelectric plant reservoirs in spring–summer flood
season, during which the main volume of annual
WR enters the reservoir. To do prediction, the
actual values of normalized monthly air temperature and precipitation for the current season
(first summand in Eqs. (7.3 and 7.4)), and their
long-term mean values for the next one (second
summand in Eqs. (7.3 and 7.4)) are substituted in
the models. Incidentally, such a substitution is
not required for winter low water season (XII–III
months) because its WR has been already calculated from meteorological data of two previous
seasons. It was found that original and predictive
models have similar adequacy criterion A < 0.7
in Eq. (7.2) for all hydrological seasons. In
98
Y. Kirsta and A. Puzanov
reflects the areal distribution of actual precipitation and provides more accurate river
runoff computations as compared with point
observations at some weather stations. Obviously, this conclusion holds significance in
any hydrological calculations, which use areal
meteorological characteristics.
• Model parameters correspond to average
meteorological and hydrological characteristics of the mountain country, which significantly vary from one river basin to another.
For a single basin, one can specify some
parameters by using long-term river observations of WR/HCR and thereby considerably
reduce the computation error.
The developed high-performance WR/HCR
models have a large number of parameters that
characterize the universal hydrophysical and
hydrochemical features of river basins throughout the Altai-Sayan mountain country. In contrast, traditional runoff/streamflow models
(Dingman and Sharma 1997; Moriasi et al. 2007;
Sene 2008; Agal'tseva et al. 2011) have much
fewer parameters. The reason is that bringing
every additional parameter to these models does
require special field investigations to determine
its value. Additional parameters require studying
climatic, orographic, hydrogeological and
hydrochemical characteristics of a river basin, a
soil-vegetation cover, a glacier, a hydrographic
structure of river network, etc., that is costly and
time-consuming. SAM does not face such a
problem because it uses the implicit information
contained in long data series on dynamics of the
characteristic under study, namely, WR and
HCR.
Most traditional runoff models are based on
the data of a single river basin that makes the
analysis and quantitative estimation of landscape
structure influence on runoff extremely difficult.
In fact, only one value of an area can be related
to each landscape in the basin. It is hardly possible to establish runoff dependence upon a
landscape area using one available value of an
area. It is also impossible to exclude the individual features of the studied basin from the
model that restricts the model applicability to
other territories. It is obvious that SAM is devoid
of both problems.
To apply the developed model package to any
river basin of the Altai-Sayan mountain country,
we just need landscape and topographic maps
supplemented with normalized monthly air temperature and precipitation spatially generalized
for the country. The calculated long-term
dynamics of normalized WR and HCR for each
hydrological season can be easily converted into
real runoff dynamics (m
3 /s and g/s). To do this,
the value of long-term mean runoff should be
determined via comparison of calculated normalized WR/HCR and the real one for 1–2 years
of observations. Subsequently, using the right
side of Eqs. (7.3 and 7.4), one can find the longterm dynamics of WR and HCR in m
3 /s and g/s
for each geosystem group (landscape) and the
whole watershed.
Using the presented model package, it is easy
to forecast the WR and HCR for the next
hydrological season, i.e. for 3–4 months ahead. It
demonstrates a significant improvement of forecast accuracy as compared with the long-term
predictions using the long-term mean value of
the observed river WR. In particular, the most
probable seasonal WR can be forecasted with a
twice-reduced variance as compared with the
similar forecast based solely on the observed
mean WR. The forecast of seasonal WR, for
example, is particularly important for the regulation of water releases from mountain hydroelectric plant reservoirs in spring–summer flood
season, during which the main volume of annual
WR enters the reservoir. To do prediction, the
actual values of normalized monthly air temperature and precipitation for the current season
(first summand in Eqs. (7.3 and 7.4)), and their
long-term mean values for the next one (second
summand in Eqs. (7.3 and 7.4)) are substituted in
the models. Incidentally, such a substitution is
not required for winter low water season (XII–III
months) because its WR has been already calculated from meteorological data of two previous
seasons. It was found that original and predictive
models have similar adequacy criterion A < 0.7
in Eq. (7.2) for all hydrological seasons. In
98
Y. Kirsta and A. Puzanov