Upload
eustacia-freeman
View
212
Download
0
Tags:
Embed Size (px)
Citation preview
Snow effect and practical questions of how to take it into
accountKorotkov V.*, Berkova M.*, Basalayev M.*, Belov A.*, Eroshenko
E.*, Yudachin K.* and Yanke V.*
*Institute of Terrestrial Magnetism, Ionosphere and Radio Wave Propagation RAS (IZMIRAN), Moscow, Russia
Our Goals: • to exclude the snow cover thickness effect from
the neutron component data • to compare the results of manual data correction
with the results of automatic correction
1076
Snow effect
)/exp( LxNN icorii
/icori NN )/exp( Lxi (1)where
Count rates of the based station S and the station with snow N
NM64
IGY
The basic detector choice
MCRL
Jungfraujoch
The method of the snow effect exclusion
1/)1(1/
1 iB
i
B
coricor
i N
N
N
N
Basing on (1) variations relative to the base period corrected on the snow effect and expressed by way of the measured variations can be entered as:
(2)
For the definition of the variations corrected on the snow effect by the measured variations it is necessary to estimate the efficiency.For that we’ll get the data of the detector registering nearly the same variations as the detector with snow.
corS If this condition apply to some average time interval it is possible to write:
1/
1 BB N
N
S
S 1
1
/
/
S
B
B
SS
NN
or (3)
Data averaging and filtration
Approximation by polynomials has been done for a polynomial of enough high degree m, where n - a number of hour in a month.
m
i
iin nay
0
A moving average prime filter in spite of its simplicity is optimal for the majority of tasks. The moving average filter equation is put down as
m
miinin xcy
- the constant weight factor)12/(1 mcior in the recursive form after the first step of calculations
mnmnnn xxyy 11
where
The Gaussian high-cut filter
The Gaussian high-cut filter is a moving average filter in which the Gaussian function is applied as a weight function. It is realized as
in
m
miin xcy
The weight factors are preset as Cici /)/exp( 22
and the normalizing factor as
m
mi
iC )/exp( 22
The value of the distribution dispersion defines the required width of the distribution.
Count rate variations of the based station S (Moscow) and MCRL for the March 2009
Automatic data correction – blue curvemanual data correction - dotted curve
Magadan station
We can estimate the effective snow cover thickness as
Distorted by the snow effect and corrected data during November 2008 – April 2009. Black curve for Gaussian smoothing, grey curve – polynominal smoothing. The based station – Moscow.
)/ln( icori NNLx
The efficiency (top curve) and the snow cover thickness (bottom). November 2008 – April 2009.
ESOI station
ESOI station
Distorted by the snow effect and corrected data of ESOI station during August 2007 - July 2008. The based station is Rome.
Efficiency (top curve) and the snow cover thickness (bottom curve) for the ESOI station during August 2007 - July 2008.
ESOI station
On the top - the automatic (black curve) and the manual (red curve) corrections for March 2009 for the ESOI station. On the bottom - the uncorrected data with snow effect.
IGY
NM64
Jungfraujoch
Jungfraujoch station
Jungfraujoch station
Distorted by the snow effect and corrected data of the Jungfraujoch station (3nm64) during January - December 2008. The based stations are Jungfraujoch - 18IGY (black curve) and Rome (grey curve).
Efficiency (top curve) and the snow cover thickness (bottom curve) for the Jungfraujoch station (3nm64) during January 2008 - December 2008.
Conclusions
• The method of exclusion of the snow cover thickness effect from the data basing on the comparison of the variations of tested and based stations was tested for some stations:
Magadan, ESOI and Jungfraujoch. • For this purpose the high-cut filters has been applied. While selecting
the filter degree the compromise choice between spasmodic signal change and its sufficient slow change with several days period had to be done. By default the effective width of the filter is about 24 hours.
• The snow effect can be excluded from the data with the accuracy is about 0.3-0.4%. The ideal case is when the detectors are identical and are in the same point.
• The described method allows not only to exclude the snow effect but also to make an estimate of the effective snow cover thickness.
• The method can be applied in real time too if use one-sided filters.
• We did not find out any advantages of the Gaussian filter over the
moving average filter.