Adequacy criterion A is also convenient to
calculate another criterion that characterizes the
model sensitivity to the natural variations of
environmental factors (Kirsta and Puzanov
2015).
Any developed model should be validated
through independent data not used for its development and identification. Such a verification is
easily performed via excluding a part of river
streamflow data from SAM and comparing this
part as a testing dataset with the calculated pattern newly resulted from SAM. Then we exclude
the next part of the data and again repeat the
procedure of SAM. With more than a tenfold
excess of the observed data over the model
parameters’ number, the elimination of some
data has a negligible effect on the parameter
values defined during SAM. Our experience of
SAM execution (Kirsta 2006a; Kirsta and Kirsta
2014) shows that, generally, the RMS discrepancy (quadratic residual) in testing comparisons
does not differ from the primarily derived one.
Thus, the model verification on the independent
testing dataset in SAM turns in a formal
procedure.
7.3 Materials
SAM was based on observations of mid-size and
small river streamflow and concentrations of
seven HCR components. The latter includes three
nitrogen mineral forms (NO
À
2 ; NO
À
3 ; and NH
þ
4 );
phosphates (PO
3À
4 ); ions (cations of Ca, Mg, Na,
K, and anions of hydrocarbonates, sulfates,
chlorides) close in content to water mineralization; total dissolved iron (Fe); suspended matter.
Ion concentration also accounts for ions of Fe,
nitrates, etc., if their content exceeds
0.1 mg/dm
3 . The observations were made by the
Hydrometeorological Service of the former
USSR and the Russian Federation in the AltaiSayan mountain country in 1951–2003. We also
used the data on monthly precipitation, average
monthly temperature, landscape structure of river
basins, area and altitude of landscapes, and some
other cartographic characteristics. A total of 34
river gauges with parallel observations of
streamflow and substance concentration were
studied in SAM. As indicated above, a great
number of river gauges analyzed simultaneously
in SAM allow to characterize the patterns of
WR/HCR formation, which are the same for the
whole territory.
Taking into account the streamflow annual
dynamics, we specified four hydrological
periods/seasons: the first (winter low water—
XII–III months), the second (spring–summer
flood—IV–VI), the third (summer low water—
VII–VIII) and the fourth (autumn low water with
possible flood in case of heavy rains—IX–XI).
The data on daily streamflow observations at
each of 34 river gauges as well as concentrations
of seven HCR components were averaged by
seasons; thereafter only four season-average
values of streamflow and analyte concentration
for each year were employed in SAM. On average, the streamflow made up approximately 10,
140, 50, 30 m
3 /s with standard deviation of
streamflow time series for each basin 33, 27, 41,
40% for seasons 1, 2, 3, 4, correspondingly. In
turn, the average annual analyte concentrations
for NO
À
2 ; NO
À
3 ; NH
þ
4 , PO
3À
4 , ions, total dissolved Fe, and suspended matter were about
0.01, 0.4, 0.2, 0.02, 200, 0.1, 20 mg/dm
3 ,
respectively.
Typification of landscapes of the Altai-Sayan
mountain country, which reflects the conditions
of WR formation including altitudinal-belt and
structural-layering heterogeneity of the territory,
was made to account for the landscape structure
of river basins and for spatial separation of
different types of a hydrological regime (Kirsta
et al. 2011). A total of 12 typological geosystem
groups including one extra for aquatic landscapes of the small area were selected
(Table 7.1).
We established the boundaries of 34 basins
(which have river gauges as outlets; see Fig. 7.4)
and calculated areas and mean altitudes of
geosystem groups in each basin. Basin square
ranged from 177 to 21,000 km
2 . Two maps
“Landscapes of Altai (Altai Krai and Altai
Republic)” (Chernykh and Samoilova 2011) and
7 System-Analytical Modeling of Water Quality …
85
calculate another criterion that characterizes the
model sensitivity to the natural variations of
environmental factors (Kirsta and Puzanov
2015).
Any developed model should be validated
through independent data not used for its development and identification. Such a verification is
easily performed via excluding a part of river
streamflow data from SAM and comparing this
part as a testing dataset with the calculated pattern newly resulted from SAM. Then we exclude
the next part of the data and again repeat the
procedure of SAM. With more than a tenfold
excess of the observed data over the model
parameters’ number, the elimination of some
data has a negligible effect on the parameter
values defined during SAM. Our experience of
SAM execution (Kirsta 2006a; Kirsta and Kirsta
2014) shows that, generally, the RMS discrepancy (quadratic residual) in testing comparisons
does not differ from the primarily derived one.
Thus, the model verification on the independent
testing dataset in SAM turns in a formal
procedure.
7.3 Materials
SAM was based on observations of mid-size and
small river streamflow and concentrations of
seven HCR components. The latter includes three
nitrogen mineral forms (NO
À
2 ; NO
À
3 ; and NH
þ
4 );
phosphates (PO
3À
4 ); ions (cations of Ca, Mg, Na,
K, and anions of hydrocarbonates, sulfates,
chlorides) close in content to water mineralization; total dissolved iron (Fe); suspended matter.
Ion concentration also accounts for ions of Fe,
nitrates, etc., if their content exceeds
0.1 mg/dm
3 . The observations were made by the
Hydrometeorological Service of the former
USSR and the Russian Federation in the AltaiSayan mountain country in 1951–2003. We also
used the data on monthly precipitation, average
monthly temperature, landscape structure of river
basins, area and altitude of landscapes, and some
other cartographic characteristics. A total of 34
river gauges with parallel observations of
streamflow and substance concentration were
studied in SAM. As indicated above, a great
number of river gauges analyzed simultaneously
in SAM allow to characterize the patterns of
WR/HCR formation, which are the same for the
whole territory.
Taking into account the streamflow annual
dynamics, we specified four hydrological
periods/seasons: the first (winter low water—
XII–III months), the second (spring–summer
flood—IV–VI), the third (summer low water—
VII–VIII) and the fourth (autumn low water with
possible flood in case of heavy rains—IX–XI).
The data on daily streamflow observations at
each of 34 river gauges as well as concentrations
of seven HCR components were averaged by
seasons; thereafter only four season-average
values of streamflow and analyte concentration
for each year were employed in SAM. On average, the streamflow made up approximately 10,
140, 50, 30 m
3 /s with standard deviation of
streamflow time series for each basin 33, 27, 41,
40% for seasons 1, 2, 3, 4, correspondingly. In
turn, the average annual analyte concentrations
for NO
À
2 ; NO
À
3 ; NH
þ
4 , PO
3À
4 , ions, total dissolved Fe, and suspended matter were about
0.01, 0.4, 0.2, 0.02, 200, 0.1, 20 mg/dm
3 ,
respectively.
Typification of landscapes of the Altai-Sayan
mountain country, which reflects the conditions
of WR formation including altitudinal-belt and
structural-layering heterogeneity of the territory,
was made to account for the landscape structure
of river basins and for spatial separation of
different types of a hydrological regime (Kirsta
et al. 2011). A total of 12 typological geosystem
groups including one extra for aquatic landscapes of the small area were selected
(Table 7.1).
We established the boundaries of 34 basins
(which have river gauges as outlets; see Fig. 7.4)
and calculated areas and mean altitudes of
geosystem groups in each basin. Basin square
ranged from 177 to 21,000 km
2 . Two maps
“Landscapes of Altai (Altai Krai and Altai
Republic)” (Chernykh and Samoilova 2011) and
7 System-Analytical Modeling of Water Quality …
85