Mathematical Problems of Computer Science 49, 49–57, 2018.

Cloud Service for Analysis and Interactive

Visualization of Weather Data in Armenia

Hayk A. Grigoryan and Rita M. Abrahamyan

Institute for Informatics and Automation Problems of NAS RA
Armenian State Hydrometeorological and Monitoring Service

e-mail: hayk.grigoryan.a@gmail.com

Abstract

The Lesser Caucasus Mountains are crossing through the territory of Armenia,
creating vast differences in altitude, terrain, temperature and precipitation in provinces
and towns. Even Armenia’s lowlands are 500 to 1500m above sea level. Armenias
highlands extend up to Aragats mountain at 4090m where, 75% of the territory is
above 1000m, 50% is above 2000m, and 3.4% is above 3000m. This paper presents a
cloud service with interactive visualization and analytical capabilities for weather data
in Armenia by integrating the two existing infrastructures for observational data and
numerical weather prediction.

The weather data used in the platform consist of near-surface atmospheric elements
including air temperature, relative humidity, pressure, wind and precipitation. The
visualization and analitycs have been implemented for 2m air temperature. Cloud
service provides the Armenian State Hydrometeorological and Monitoring Service with
analytical capabilities to make a comparative analysis between the observation data
and the results of a numerical weather prediction model for per station and region for
a given period.

Keywords: Cloud service, Weather data, Observational data, Data analysis, Nu-
merical weather prediction, WRF, Spatial OLAP.

1. Introduction

Armenia occupies the north-eastern part of Armenian plateau and central part of Lesser
Caucasus range (latitude 38.51’ to 41.18’ North, longitude 43.29’ to 46.37’ East), with the
area of about 30 000 sq.km. The geographical location of Armenia and complex mountainous
relief have led to the diversity of natural conditions across the country. Armenia is on the
northern edge of the subtropical zone, in latitudes characterized by an arid and continental
climate. Due to the mountainous relief, different climatic zones exist, and the weather
may have high spatial gradients. High fluctuations in annual and daily temperatures are
typical for the Armenian climate. The presence of six climatic zones from dry subtropical
to rigorous high mountainous and from everlasting snowcaps to warm humid subtropical
forests and humid semi-desert steppes make additional challenges on weather forecasting
and climate prediction for the Armenian State Hydrometeorological and Monitoring Service
(AHMS).

49



50 Cloud Service for Analysis and Interactive Visualization of Weather Data in Armenia

The meteorological data, received from 47 meteorological stations, among which only
four stations provide historical data and monthly updates to the Global Climate Observing
System Surface Network and three meteorological stations provide synoptic data are being
used in the gridded analysis data set, which serves as an input for the global atmospheric
models to produce weather forecasts at the global scale.

The observation data received from the meteorological stations and data received from
a global model serve as inputs and outputs to the high-resolution numerical weather pre-
diction models to produce outputs of temperature, precipitation, and other meteorologi-
cal elements from the ground to the top of the atmosphere [1, 2]. The high-performance
computational (HPC) resources of the Armenian e-infrastructure are used to better resolve
mesoscale weather events, and hence to give reasonably accurate forecasts in a short range
[3]. The limitations in such forecasts lie in the availability of initial conditions at model
resolutions.

The article aims to present the weather data interactive web-based visualization and an-
alytical platform1. The platform has been developed for the weather data in Armenia by
integrating the three existing platforms for observational data and numerical weather pre-
diction. The platform provides a way to compare the output of forecasting model with the
observation data gathered from different stations for a chosen frame of time. The suggested
platform is essential for a wide range of applications, such as urban area management, sus-
tainable development and nature protection, regional and local planning, agriculture, forestry
and fisheries, health, civil protection, infrastructure, transport and mobility or tourism.

The remainder of this paper is divided into the following sections: Section 2 introduces
the infrastructure, Section 3 represents the discussions and analyses and finally Section 4 is
the conclusion.

2. Cloud Service Framework

The suggested cloud service consists of 5 primary layers (see Fig. 1). The base layer provides
HPC and data resources, which is particularly important for digital models [4]. The resources
of the Armenian e-infrastructure are utilized, which is a complex national IT infrastructure
consisting of both communication and distributed computing infrastructures.

Fig 1. The framework of the platform.

1Cloud service for analysis of numerical weather prediction model accuracy using visualization technique:
http://meteo.grid.am



H. Grigoryan and R. Abrahamyan 51

The datasets layer combines two types of data platforms for further analysis:

− Model output: outputs of weather prediction models;

− Observation data: meteorological stations observations, as a base to investigate the
deviation values with other model outputs.

The Data management layer provides intelligent tools to transfer raw data to data an-
alytics layer. The Integrated Rule-Oriented Data System (iRODS) provides a middleware
between the physical data storage systems and the user interface [5].

As soon as data reaches data analytics layer, it is processed, and only several indexes
are left from the enormous amount of starting crude information. Finally, the top layer and
final destination of the already processed data is the visualization layer, where the result
lists are changed to more easy to understand graphs or tables. Besides, the advantages of
Google Maps are used to map these indexes with their real location on the map.

2.1 Data Gathering
Observational datasets provide the observed meteorological data from different locations
acquired from SYNOP (surface synoptic observations) messages issued by official ground
weather stations. SYNOP reports are regularly sent every three hours, which consist of
groups of numbers describing general weather information of the weather station, such as
temperature, barometric pressure and visibility at a weather station, and so on.

The numerical weather prediction models are initialized using NCEP (National Centers
for Environmental Prediction) Global Forecast System analysis and forecasts at 0.5 deg
horizontal resolution [6]. Information created during pre-processing and simulations of the
model is in the Lambert conformal projection, which is appropriate for mid-latitude domains.

The model setup (see Fig. 2) comprises a parent D01 domain (with a common center
located at longitude 44.7, latitude 40.0), partly covering Europe, all the Caucasus, parts of
the Central Asia and the Middle East and the nest domain d02, covering the entire region
of Armenia.

2.2 Numerical Weather Prediction Model
The mesoscale Weather Research and Forecasting (WRF) model [9, 10], which is adapted to
the territory of Armenia, is used in operational weather forecasting. Designed to serve both
research and operational needs, it has grown to offer a spectrum of options and capabilities
for a wide range of applications. The D01 domain with a common center located at longitude
44.7, latitude 40. has horizontal resolution 18-km with 202x202 grid points , the d2 nest with
9770 grid points has 6-km horizontal grid increment . The weather forecasts are performed
on a daily basis, using the following 1-way nesting strategy. The model uses vertical 31
eta levels and the geographic data resolution is 30 seconds. The model was initialized with
the initial and boundary conditions of Global Forecast System (GFS) at 0000 UTC (local
time on 04:00) for 31-day period, namely 1-31 January 2016. The applied version of WRF-
ARW model includes the WRF Single Moment 6-class scheme for cloud microphysics with
ice, snow and graupel processes. Sub-grid parameterization of deep and shallow convection
is based on the KainFritsch scheme applying mass flux approach with downdrafts and CAPE
removal time scale (Kain and Fritsch 1993). In this study the MellorYamadaJanjic (MYJ)



52 Cloud Service for Analysis and Interactive Visualization of Weather Data in Armenia

Fig 2. Parent (D01) and nested (d02) domains used in the model.

PBL scheme with Eta similarity surface layer was used (Mellor and Yamada 1982). The rapid
radiative transfer model (RRTM) scheme was used for longwave radiation, while shortwave
radiation processes were represented by the Dudhia scheme. Due to the limited computing
resources, we have considered the inner domain with 6 km resolution, and the range of every
run was 24 hours starting always at 0000 UTC of each day.

Observable 2-metres long temperatures from 42 operational stations in Armenia are used
to study the accuracy of forecasting air temperature by the model WRF for January 2016
It must be noted that the inversion was observed in Armenia frequently in January 2016
causing low temperatures in inversion affected regions. The surface inversion , trapped by
cold air near the ground, causes fog below the inversion layer .

2.3 Data Management

Everyday rapid growth of data and need to analyze these data pushed the development of
analytical processing tools. OLAP (Online analytical processing) is a model for accessing
multidimensional data in data warehouses [11]. Data cubes and OLAP session are main
concepts in OLAP. A data cube is a collection of facts and dimensions organizing the data
of a data warehouse according to different analysis axes and aggregation measures. OLAP
provides a set of operations (such as drill-down and slice-and-dice) that transform one mul-
tidimensional query into another, which provide high querying. OLAP queries are formu-
lated as sequences called OLAP sessions. For analyzing data with spatial and georeferenced
components the Spatial OLAP (SOLAP) technology is used, which allows rapid and easy
navigation within spatial databases and that offers many levels of information granularity,
many themes, many epochs and many display modes synchronized or not: maps, tables and
diagrams [12]. It allows a tight integration of GIS and OLAP systems. A SOLAP system
supports three types of spatial dimensions: the non-geometric spatial dimensions, the geo-
metric spatial dimensions and the mixed spatial dimensions. During an OLAP session, the



H. Grigoryan and R. Abrahamyan 53

user analyzes the results of a query and, contingent upon the particular information, applies
an activity to determine a new query that will give a superior comprehension of information.

2.4 User Interface and Visualization

The implemented service consists of 3 main intelligent blocks, and the data flows between
these pieces. The first block is the Database. PostgreSQL with its PostGIS extension is used
to store data sets with geometrical information. The following block is implementing data
analytic logic. The special type of OLAP approach is used (Spatial OLAP) to process the
required data. As a query language for OLAP, the MultiDimensional eXpressions (MDX)
[13] is used to interact and perform tasks with multidimensional databases (OLAP Cubes).
Afterwards, the handled information is exchanged to User Interface where user can make
different graphs, tables and see the output on the map (see Fig. 3).

Fig 3. Web-based visualization and analytical platform consist of the following sections:

1- query form, 2 - temperature chart, 3 - coefficients table

The main screen includes many parts, which are marked with red numbers.

⇒ Query form - user can pick out the station, start and end dates and the period (by
default it takes all hours from 0 to 21) for requesting the wished records for plot. After
the plot action, the charts and the table will be updated with the corresponding values;

⇒ Temperature - shows the observation and model temperature information lines corre-
spondingly for each period;

⇒ Coefficients table - shows the RMSE, BIAS and R correlation coefficients.

Platform provides the ability to transfer new information to the database by utilizing API
endpoints. At present, only the administrator of the platform has access to these activities.
For observation data, SYNOP documents (the records are in ascii format) must be used,
which will be parsed utilizing a script written in JavaScript language. For WRF model
outputs, which are prroduced in netCDF file format, the Python script was made for finding
relating values based on the stations information stored in the database (see Fig. 4).



54 Cloud Service for Analysis and Interactive Visualization of Weather Data in Armenia

Fig. 4. Data workflow consists of API endpoint, database, SOLAP and user interface steps.

3. Evaluation of Results

Various statistical and object-oriented methods are used in the suggested web-based ana-
lytical platform to investigate the characteristics of model-forecast, which is important for
giving helpful guidance to end-users.

As a case study, the 2 metre temperature has been investigated using the observational
and regional high-resolution WRF model data for January, 2016. For the studied period, the
RMSE, BIAS and R correlation coefficients are calculated for the observational and model-
forecast data for 42 observation points distributed in the territory of Armenia (see Table
1).

Table 1: Mean estimates of verification of temperature forecasts for several meteo stations per region.



H. Grigoryan and R. Abrahamyan 55

The table shows that the average difference between the forecast data and observations
is about 4-50C. For all the stations studied, an acceptable correlation coefficient has been
obtained, which allows one to judge the adequacy of the model.

The analyses, which have been carried out for 42 stations (some high-altitude stations
are not considered), show that the model data for almost all stations are overstated. RMSE
values are about 1.6-2.50C for the stations located at altitudes above 1500m-2000m, 4.6-
9.80C for the stations below 1000m. It means that the model predicts well the temperature
values at a height of 2m for stations located above 1500m and gives unsatisfactory results
for stations below 1500m. The worst results are obtained for stations below 1000m, such
as stations located in Meghri, Kapan, Ijevan, Bagratashen, and the valley of Syunik and
Tavush region.

The January 2016 temperature for Ararat Valley and Yerevan is also projected unsatis-
factorily. In Ararat Valley RMSE 5.6-6.40C, the worst result was obtained for Ararat station.
In Yerevan, a good result was obtained for Yerevan-Arabkir station, which is located at an
altitude of 1113m, and the worst result for Yerevan-Zvartnots. The correlation coefficient
varies from 0.80-0.97 for mountain and foothill areas, as for valleys it is from 0.7 to 0.25 and
from -0.7 to -0.31.

From all that has been described above, it can be concluded that the WRF model with the
described configuration, for a cold period of time, gives a positive forecast for a variable air
temperature of 2 m for the mountain and foothill regions of the republic (Shirak, Kotayk,
Gegharkunik, Lori, mountain and foothill areas of Aragatsotn) and at the same time an
unsatisfactory forecast for the Valley of Syunik, Tavush, Ararat Valley and Yerevan.

Analyzing all the factors that make up the temperature, we see that there were fogs in
the valley during the selected period, which was the reason for low temperatures. Therefore
we can conclude that the model WRF is not predicted by low temperatures of occurrence of
the lowland, effect of the fog at surface inversion. Such a result does not satisfy, therefore, it
is necessary to improve the accuracy of the model data, through a proper tuning, it would
be useful to test the impact of resolution given, because the Armenian terrain has a rather
complex orography and is characterized by several land-category types.

4. Conclusion

The suggested platform enables to integrate the already available observational and model-
forecast informations and use these sources for studies and analyzes in a web-based visual-
ization environment. The interactive comparison charts for 2m air temperature allows to
visually analyze and gather the information about model accuracy. It enables to adjust
the forecasting results with additional methods by implementing statistical analyses and
provides a fairly high result in cases where the model’s sensitivity is low.

It is planned to enhance the functionality of the platform by adding a new source of
satellite image processing, the land surface temperature, which will give an opportunity to
analyze locations where the meteorological stations do not exist. It will include also new
visualization tools of various formats, such as to analyze and compare other near-surface
atmospheric elements. Different nowcasting methodologies based on artificial intelligence
will be implemented for the development of a hazardous hydro-meteorological phenomena
alarm system.



56 Cloud Service for Analysis and Interactive Visualization of Weather Data in Armenia

References

[1] A. Gevorgyan, “Summertime wind climate in Yerevan: valley wind systems”, Climate
Dynamics, vol. 48, no. 5–6, pp. 1827–1840, 2017.

[2] A. Gevorgyan, H. Melkonyan, R. Abrahamyan, Z. Petrosyan, A. Shachnazaryan, H.
Astsatryan, V.Sahakyan and Yu. Shoukourian, A Persistent Surface Inversion Event
in Armenia as Simulated by WRF Model, in IEEE Proceedings of the International
Conference on Computer Science and Information Technologies, CSIT’2015, pp. 105–
110, 2015.

[3] H. Astsatryan, V. Sahakyan, Y. Shoukourian, P.-H. Cros, M. Dayde, J. Dongarra and
P. Oster, “Strengthening Compute and Data intensive Capacities of Armenia”, in
IEEE Proceedings of 14th RoEduNet International Conference - Networking in Educa-

tion and Research, NER’2015, pp. 28–33, 2015.

[4] H. Astsatryan, Yu. Shoukourian and V. Sahakyan, “The ArmCluster Project: Brief In-
troduction”, in Proceedings of the International Conference on Parallel and Distributed
Processing Techniques and Applications, pp. 1291–1295. 2004.

[5] M. Hedges, T. Blanke and A. Hasan, Rule-based curation and preservation of data:
A data Grid approach using iRODS , Future Gener. Comput. Syst, vol. 25, no. 4, pp.
446–452, 2009.

[6] W. S. Jefrey , T. M. Hamill, X. S. Yucheng and Z. Toth, Ensemble data assimilation
with the ncep global forecast system,Monthly Weather Review, vol. 136, pp. 463–482,
2008.

[7] J. A. Sobrino, J. C. Jimenez-Munoz and L. Paolini, Land surface temperature retrieval
from Landsat TM 5, Remote Sensing of Environment, vol. 90, pp. 434–440, 2004.

[8] M. Neteler, M. Bowman M. Landa and M. Metz, Grass gis: A multi-purpose open
source gis, Environmental Modelling & Software, vol. 31, pp. 124–130, 2011.

[9] W. C. Skamarock and J. B. Klem, A time-split non-hydrostatic atmospheric model
for weather research and forecasting applications, Computational Physics, vol. 227, no.
7, pp. 3465–3485, 2008.

[10] J.G. Powers, J.B. Klemp, et. al, The weather research and forecasting model:
Overview, system efforts, and future directions, Bulletin of the American Meteoro-
logical Society, vol. 98, no. 8, pp. 1717–1737, 2017.

[11] S. Chaudhuri and U. Dayal, An Overview of Data Warehousing and OLAP Technol-
ogy, SIGMOD Record, vol. 26, no. 1, 1996.

[12] S. Aissi, M. S. Gouider, T. Sboui and L.B. Said, Enhancing spatial data warehouse ex-
ploitation: a solap recommendation approach, In: Computer and Information Science,
Springer, pp. 131–147, 2016.

[13] M. Whitehorn, R. Zare and M. Pasumansky, Fast Track to MDX, Springer-Verlag
London, 2006, DOI: 10.1007/1-84628-182-2

Submitted 09.10.2017, accepted 22.01.2018.



H. Grigoryan and R. Abrahamyan 5 7

г۳ëï³ÝáõÙ »Õ³Ý³Ï³ÛÇÝ ïíÛ³ÉÝ»ñÇ Ñ³Ù³ñ í»ñÉáõÍáõÃÛ³Ý ¨
ÇÝï»ñ³ÏïÇí íǽáõ³Éǽ³ódzÛÇ ³Ùå³ÛÇÝ Í³é³ÛáõÃÛáõÝ

Ð. ¶ñÇ·áñÛ³Ý ¨ è. ²µñ³Ñ³ÙÛ³Ý

²Ù÷á÷áõÙ

öáùñ ÎáíϳëÛ³Ý É»éÝ»ñÁ ³ÝóÝáõÙ »Ý г۳ëï³ÝÇ ï³ñ³Íùáí, ëï»ÕÍ»Éáí ï»Õ³ÝùÇ
µ³ñÓñáõÃÛ³Ý, ç»ñÙ³ëïÇ׳ÝÇ ¨ ï»ÕáõÙÝ»ñÇ Ù»Í ï³ñµ»ñáõÃÛáõÝÝ»ñ: г۳ëï³ÝÇ
ÑáíÇïÝ»ñÁ ·ïÝíáõÙ »Ý ÍáíÇ Ù³Ï³ñ¹³ÏÇó 500-Çó 1500Ù µ³ñÓñáõÃÛ³Ý íñ³:
È»éݳßÕóݻñÁ ï³ñ³ÍíáõÙ »Ý ÙÇÝã ²ñ³·³Í É»éÁ` 4090Ù, áñï»Õ ï³ñ³ÍùÇ 75 % -Á
·»ñ³½³ÝóáõÙ ¿ 1000Ù-Á, 50 % -Á` 2000Ù-Çó µ³ñÓñ ¿, ÇëÏ 3.4 % -Á` 3000 Ù µ³ñÓñáõÃÛáõÝÇó:
²ß˳ï³ÝùÇ Ýå³ï³ÏÝ ¿ Ý»ñϳ۳óÝ»É Ð³Û³ëï³ÝáõÙ »Õ³Ý³Ï³ÛÇÝ ïíÛ³ÉÝ»ñÇ
ѳٳñ í»ñÉáõÍáõÃÛ³Ý ¨ ÇÝï»ñ³ÏïÇí íǽáõ³Éǽ³ódzÛÇ ³Ùå³ÛÇÝ Í³é³ÛáõÃÛáõÝ,
áñÁ µ³Õϳó³Í ¿ ó³Ù³ù³ÛÇÝ Ï³Û³ÝÝ»ñÇ ¹Çï³ñÏáõÙÝ»ñÇ ¨ »Õ³Ý³ÏÇ Ãí³ÛÇÝ
ϳÝ˳ï»ëÙ³Ý Ùá¹»ÉÇ »Ýóϳéáõóí³ÍùÝ»ñÇó:

Øß³Ïí³Í ѳٳϳñ·áõÙ û·ï³·áñÍíáÕ »Õ³Ý³Ï³ÛÇÝ ïíÛ³ÉÝ»ñÁ µ³Õϳó³Í »Ý
ٳϻñ¨áõÛÃÇÝ Ùáï ÙÃÝáÉáñï³ÛÇÝ µ³Õ³¹ñÇãÝ»ñÇó, Ý»ñ³éÛ³É û¹Ç ç»ñÙ³ëïÇ׳ÝÁ,
ѳñ³µ»ñ³Ï³Ý ËáݳíáõÃÛáõÝÁ, ×ÝßáõÙÁ, ù³ÙÇÝ ¨ ï»ÕáõÙÝ»ñÁ: ìǽáõ³Éǽ³óÇ³Ý ¨
í»ñÉáõÍáõÃÛáõÝÁ Çñ³Ï³Ý³óí»É »Ý 2Ù µ³ñÓñáõÃÛ³Ý û¹Ç ç»ñÙ³ëïÇ׳ÝÇ Ñ³Ù³ñ: ²Ùå³ÛÇÝ
ͳé³ÛáõÃÛáõÝÁ г۳ëï³ÝÇ å»ï³Ï³Ý Ñǹñáû¹»ñ¨áõóµ³Ý³Ï³Ý ¨ ÙáÝÇïáñÇÝ·Ç
ͳé³ÛáõÃÛ³ÝÁ Ñݳñ³íáñáõÃÛáõÝ ¿ ï³ÉÇë áñáß³ÏÇ Å³Ù³Ý³Ï³Ñ³ïí³ÍáõÙ Ûáõñ³ù³ÝãÛáõñ
ϳ۳ÝÇ ¨ ï³ñ³Í³ßñç³ÝÇ Ñ³Ù³ñ ϳï³ñ»É ¹Çï³ñÏáõÙÝ»ñ, ÇÝãå»ë ݳ¨ »Õ³Ý³ÏÇ
ϳÝ˳ï»ëÙ³Ý Ãí³ÛÇÝ Ùá¹»ÉÇ ³ñ¹ÛáõÝùÝ»ñÇ Ñ³Ù»Ù³ï³Ï³Ý í»ñÉáõÍáõÃÛáõÝÝ»ñ:

Îáëà÷íûé ñåðâèñ äëÿ àíàëèçà è èíòåðàêòèâíîé
âèçóàëèçàöèè äàííûõ î ïîãîäå â Àðìåíèè

À. Ãðèãîðÿí è Ð. Àáðààìÿí

Àííîòàöèÿ

Ìàëûå Êàâêàçñêèå ãîðû, ïåðåñåêàÿ òåððèòîðèþ Àðìåíèè, ñîçäàþò áîëüøèå
ïåðåïàäû âûñîò ìåñòíîñòåé, òåìïåðàòóð è îñàäêîâ. Ðàâíèíû Àðìåíèè íàõîäÿòñÿ
íà âûñîòå îò 500 äî 1500ì íàä óðîâíåì ìîðÿ. Àðìÿíñêîå íàãîðüå ïðîñòèðàåòñÿ
äî ãîðû Àðàãàö(âûñîòà 4090 ì), ãäå 75% òåððèòîðèè ïðåâûøàåò 1000ì, 50% -
ñâûøå 2000ì, à 3,4% - ñâûøå 3000ì. Öåëü äàííîé ðàáîòû ïðåäñòàâèòü îáëà÷íûé
ñåðâèñ äëÿ äàííûõ î ïîãîäå â Àðìåíè, ðàçðàáîòàííûé ïóòåì èíòåãðàöèè äâóõ
ñóùåñòâóþùèõ èíôðàñòðóêòóð - íàáëþäåíèÿ ñ íàçåìíûõ ñòàíöèé è ÷èñëåííàÿ
ìîäåëü ïðîãíîçèðîâàíèÿ ïîãîäû.

Ïîãîäíûå äàííûå, èñïîëüçóåìûå â ðàçðàáîòàííîé ñèñòåìå, ñîñòîÿò èç
àòìîñôåðíûõ ýëåìåíòîâ â ïðèçåìíîì ñëîå, âêëþ÷àÿ òåìïåðàòóðó âîçäóõà,
îòíîñèòåëüíóþ âëàæíîñòü, äàâëåíèå, âåòåð è îñàäêè. Àíàëèç è âèçóàëèçàöèÿ
ïðîâîäèëèñü äëÿ òåìïåðàòóðû âîçäóõà íà âûñîòå 2ì. Îáëà÷íûé ñåðâèñ
ïðåäîñòàâëÿåò âîçìîæíîñòü ãîñóäàðñòâåííîé ñëóæáå ïî ãèäðîìåòåîðîëîãèè è
ìîíèòîðèíãó Àðìåíèè â îïðåäåëåííûé ïåðèîä âðåìåíè äëÿ êàæäîé ñòàíöèè è
ìåñòíîñòè ïðîâåñòè ñðàâíèòåëüíûé àíàëèç íàáëþäàåìûõ äàííûõ ñ ðåçóëüòàòàìè
äàííûõ îò ÷èñëåííîé ìîäåëè.


	hg
	Abstract_Hayk_Grigoryan_abstract