the snow, which is impervious for microwave
radiation ant thus indiscernible from snow-free
terrain. At present, it is no more than one meteorological observation station per EASE-Grid
cell of 25 Â 25 km size on the territory of the
Russian Federation, which can lead to poor representation of snow cover characteristics due to
“bad” placement of observation within a cell
bound.
From the result of completed study (Volckak
et al. 2016) and based on the obtained values of
root mean square error and Pearson correlation
coefficient (RMSE = 24.9 mm, r = 0.39 ±
0.01), we have come to the conclusion that the
best SWE retrieval performance has the artificial
neural network which takes the results of SSM/I
measurements of brightness temperatures for
channels at 19.35, 37.0 and 85.5 GHz for both
orthogonal polarizations with distinct differentiation of terrain cover features within EASE-Grid
cells as input data. It is shown that the low
obtained values of correlation coefficient (lower
than 0.5) in contrast with the published result of
studies over relatively small territories come
from specific features of the completed experiments which are the significant diversity of climatic and terrain characteristics and irregular
meteorological observation station coverage of
the studied territory.
It must be noted that the forest cover density,
the forest cover type and the air temperature at
the ground level were not taken into consideration in the presented research. There is a probability that taking into account the forest cover
density will improve the accuracy of the SWE
value retrieval, because the vegetation cover
significantly scatters the microwave radiation,
and on the other hand, the snow cover characteristics depend on the forest cover type. The
consideration of the air temperature at the ground
level will help to detect the thaw conditions more
precisely. To further improve the accuracy of the
snow cover retrieval, it seems to be practical to
employ the artificial neural network ensemble,
which is trained on the samples collected from
the set of territories organized by climatic and
land cover similarity.
18.4 Estimating How the Snow
Water Equivalent Dynamics
Influences the Water
Discharge in a River Flow
Control Point
An open access archive of the snow water
equivalent dynamics data was used as a source
data for the research. The archive contains
observations carried out in bounds of the
GlobSnow project on the Northern Hemisphere
of Earth and is covering years of the 1979–2011
range. Archive data are obtained by combining land meteorological station observations
with SMMR and SSM/I sensors measurements
(Luojus et al. 2010). Measurements of the satellite passive sensors are performed on daily basis,
while snow-gauging surveys have 5–10 days
interval. The archive itself is a set of matrices of
721 Â 721 size, with elements containing values
of averaged thickness of the snow cover water
equivalent in millimeters. Area of each cell is
equal to 625 km
2 .
A preselection of the matrix cells belonging to
the Pripyat river watershed was done, with an
outlet near the Mozyr hydrological station, based
on data from the 1979 to 2004 years range, and
the volume of water in the snow for each day
from November to March was calculated for this
territory.
Primary evaluation of the calculation results
and their connection to the flood specifics was
done by measuring average daily water discharge
of the Pripyat river, on the Mozyr hydrological
station.
Typical case can be seen in Fig. 18.5, which
illustrates the dynamics of 2003–2004 years and
for 2004–2005 years. Winter of 2003–2004 was
marked by a strong thaw, which results in two
peaks as shown in Fig. 5a. Next year had no
significant winter thaw, which is clearly seen by
the only peak in Fig. 5c.
Therefore, the approach to evaluate dynamics
of the snow cover water storage illustrated by
Fig. 18.5 allows in carrying out the qualitative
prediction of the spring water discharge dynamics at specific hydrological station. Meanwhile,
242
A. A. Volchak et al.
radiation ant thus indiscernible from snow-free
terrain. At present, it is no more than one meteorological observation station per EASE-Grid
cell of 25 Â 25 km size on the territory of the
Russian Federation, which can lead to poor representation of snow cover characteristics due to
“bad” placement of observation within a cell
bound.
From the result of completed study (Volckak
et al. 2016) and based on the obtained values of
root mean square error and Pearson correlation
coefficient (RMSE = 24.9 mm, r = 0.39 ±
0.01), we have come to the conclusion that the
best SWE retrieval performance has the artificial
neural network which takes the results of SSM/I
measurements of brightness temperatures for
channels at 19.35, 37.0 and 85.5 GHz for both
orthogonal polarizations with distinct differentiation of terrain cover features within EASE-Grid
cells as input data. It is shown that the low
obtained values of correlation coefficient (lower
than 0.5) in contrast with the published result of
studies over relatively small territories come
from specific features of the completed experiments which are the significant diversity of climatic and terrain characteristics and irregular
meteorological observation station coverage of
the studied territory.
It must be noted that the forest cover density,
the forest cover type and the air temperature at
the ground level were not taken into consideration in the presented research. There is a probability that taking into account the forest cover
density will improve the accuracy of the SWE
value retrieval, because the vegetation cover
significantly scatters the microwave radiation,
and on the other hand, the snow cover characteristics depend on the forest cover type. The
consideration of the air temperature at the ground
level will help to detect the thaw conditions more
precisely. To further improve the accuracy of the
snow cover retrieval, it seems to be practical to
employ the artificial neural network ensemble,
which is trained on the samples collected from
the set of territories organized by climatic and
land cover similarity.
18.4 Estimating How the Snow
Water Equivalent Dynamics
Influences the Water
Discharge in a River Flow
Control Point
An open access archive of the snow water
equivalent dynamics data was used as a source
data for the research. The archive contains
observations carried out in bounds of the
GlobSnow project on the Northern Hemisphere
of Earth and is covering years of the 1979–2011
range. Archive data are obtained by combining land meteorological station observations
with SMMR and SSM/I sensors measurements
(Luojus et al. 2010). Measurements of the satellite passive sensors are performed on daily basis,
while snow-gauging surveys have 5–10 days
interval. The archive itself is a set of matrices of
721 Â 721 size, with elements containing values
of averaged thickness of the snow cover water
equivalent in millimeters. Area of each cell is
equal to 625 km
2 .
A preselection of the matrix cells belonging to
the Pripyat river watershed was done, with an
outlet near the Mozyr hydrological station, based
on data from the 1979 to 2004 years range, and
the volume of water in the snow for each day
from November to March was calculated for this
territory.
Primary evaluation of the calculation results
and their connection to the flood specifics was
done by measuring average daily water discharge
of the Pripyat river, on the Mozyr hydrological
station.
Typical case can be seen in Fig. 18.5, which
illustrates the dynamics of 2003–2004 years and
for 2004–2005 years. Winter of 2003–2004 was
marked by a strong thaw, which results in two
peaks as shown in Fig. 5a. Next year had no
significant winter thaw, which is clearly seen by
the only peak in Fig. 5c.
Therefore, the approach to evaluate dynamics
of the snow cover water storage illustrated by
Fig. 18.5 allows in carrying out the qualitative
prediction of the spring water discharge dynamics at specific hydrological station. Meanwhile,
242
A. A. Volchak et al.