Economic Computation and Economic Cybernetics Studies and Research, Vol.

58, Issue 3/2024

Alexandru ISAIC-MANIU, PhD

[email protected]
Centre of Industry and Services Economy, National Institute of Economic Research,
The Romanian Academy, Romania

Emilia GOGU, PhD

[email protected]
Bucharest University of Economic Studies, Romania

Irina-Maria DRAGAN, PhD (corresponding author)

[email protected]
Bucharest University of Economic Studies, Romania

Florentina CONSTANTIN, PhD

[email protected]
Bucharest University of Economic Studies, Romania

The Risk of Earthquakes in Romania – A Statistical Point

of View
Abstract. From a seismic point of view, Romania is dominated by the events in the Vrancea
area. There was only one major earthquake that occurred in the last 300 years with its
epicentre outside this area. A seismic risk assessment involves the estimation of the
probability of damage and losses resulting from potential future earthquakes. This damage
and loss might occur to buildings, infrastructure, people, or even the environment. Our study
starts with the inventorying and statistical processing of information regarding seismic
events grouped by intensity classes. The connection between the depth of the focal point and
the intensity of earthquakes, the frequency of destructive earthquakes, as well as the
statistical distribution of significant earthquakes, are subject to analysis. The statistical
models of behaviour on different intensity classes of the magnitude are being tested.
Calculations are made regarding the statistical characteristics of strong earthquakes in the
Vrancea seismogenic area in the Carpathian Mountains - Romania, as well as the
distribution of significant earthquakes over 50-year timespans. We've found a strong
correlation between the depth of the epicentre and the intensity of the earthquake. For
the destructive earthquakes, simulations were carried, using the proper function. At intervals
of 50 years each, the chances of recording less than three events are moderate, between one
and ten are higher, and more than ten are low.

Keywords: seismic risk, statistical model, modelling, validation, statistical prediction,

statistical distributions.

JEL Classification: C46, C53, C15.

DOI: 10.24818/18423264/58.3.24.01
© 2024 The Authors. Published by Editura ASE. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin

1. Introduction

From a geological point of view, Europe confirms high seismic risk areas such
as Italy, Turkey, Iceland, Serbia, Bulgaria, Greece, and Romania. The seismic
intensity zones are marked by a colour code. In Romania, the Eastern Region of the
country (Figure 1: Note: Red dots – epicentres of superficial (crustal) earthquakes,
Black dots – epicentres of intermediate earthquakes in the Vrancea area) stands out
in the area of the curvature of the Carpathian Mountains. Clusters of black dots on
the map represent the frequencies of earthquakes. A region with a high concentration
of earthquakes can be distinguished, which is known as the Vrancea area. Depending
on the depth, three categories of earthquakes are mainly registered in Romania:
superficial - where the hypocentre is located at a depth between 0 and 5 km from the
earth’s surface; crustal or ordinary - where the hypocentre is located at a depth
between 5 and 30 km from the earth's surface, reaching up to 60 km in the Vrancea
area; and intermediate, with the hypocentre placed at a depth ranging between 60-70
km and 100-220 km from the earth’s surface, and these are specific to the Vrancea
area. The Vrancea seismogenic region is located at continental convergence, placed
at the contact point of three tectonic plates: the East European plate, the Intra - Alpina
and Moesica subplates.

Figure 1. Seismic map of Romania

Source: E. S. Georgescu, I. G. Craifaleanu. Hazard, vulnerability and seismic risk
in Romania, ECBR, October 2021.

In addition to the Vrancea area, the following areas are relatively active in
Romania: the Fagaras - Câmpulung area, where there was only one major seismic
event (1916, with Mw 6.4 degrees) in the modern period, the Danubian area in the
vicinity of the Danube river, with events of lower intensity (less than 6 Mw). Other
seismicity areas in Romania: Banat Zone, Crisana-Maramures, Barlad,
Predobrogeana, and Intramoesica Zones, which are distinct zones but with a low
frequency of seismic events of significant magnitude (INFP, 2022).

6 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

A zonation of the continent from a seismic point of view is developed based on

a classification of the areas using the average values of the peak ground acceleration
transforming the data from the Global Seismic Hazard Assessment Project - GSHAP
into hazard classes (UNDRR, 2022). The activity area of the earthquakes that affect
continental Europe is called the “Mediterranean and Trans-Asian” area, and most of
the earthquakes in this area have focal points aligned along the mountain ranges, a
situation that was also confirmed in Romania.

2. Literature review

The characterisation of the seismogenic zones in Romania is presented

(Radulian et al., 2000) by highlighting the distinctiveness of earthquakes in the
Vrancea region and the characterization of the major fault system located on the
eastern edge of the Pannonian Basin and the Carpathian orogen, up to the intra-
Carpathian region. The seismological distinctiveness of the Vrancea region in the
southeast of the Carpathian Mountains is well-known for strong earthquakes of
medium depth (Bokelmann and Rodler, 2013), and there are geodynamic models to
describe the processes of wave transmission in the lithosphere. Radulian (2015)
develops the same problem of specific nature of the Vrancea earthquakes,
highlighting the distinctive elements of the Vrancea area, as a focus of earthquakes
or as a “earthquake nest”, of intermediate depth, with similarities around the world
such as Bucaramanga (Colombia) or Hindu Kush (Afghanistan). In Romania, in the
last years, the studies on earthquakes produced an increase in the awareness of the
population and authorities. After the two major earthquakes that occurred in
Romania, there is not much information about traditional buildings that suffered
complete collapse or major damage (Dutu et al., 2018).
An extension of the analysis of the seismic area in Romania, towards the
seismogenic zone of the Black Sea and its echoes in the continental crust in the
North-West is carried out by Beșuțiu and Zugrăvescu (2004). The recorded data are
processed, and the mathematical models used allow a more accurate description of
the geopotential fields and their evolution in the Black Sea area. Bala et al. (2021)
establish the patterns of seismicity in the Vrancea area of the Eastern Carpathians in
Romania, with a specific intermediate depth ranging between 55 – 105 km and 105
– 180 km, determined by the configuration of tectonic stresses. A detailed analysis
based on observations of major earthquakes in the Pannonian Basin and the
Curvature Carpathians area in Romania is performed by Grünthal and Stromeyer
(1992). A goal of researchers in this field is that of mathematical modelling of
earthquakes, the method employed by Ogata (1988) using stochastic models to
describe the origin times and magnitude of earthquakes, in the case of Japan. Faenza
et al. (2007) determine the dependences in time and space between earthquakes,
obtained through Monte Carlo simulations on the case of a region in the area of the
Lower Rhine - Germany. The topic of statistical modelling of the seismic activity
generated by hydrocarbon extraction through hydraulic fracturing in western Canada
is developed by Kothari et al. (2020). A catalogue of zonal earthquakes specific to

Vol. 58, Issue 3/2024 7

Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin

the continental shelf to establish the recurrence of major earthquakes is valorised by

Ali and Akkoyunlu (2022) for an active zone in eastern Turkey in the area of the
Anatolian Plate and the intersection of the Arabian Plate and the Eurasian Plate. For
Romania, Popescu et al. (2018) prepared a detailed catalogue with the definition of
the seismogenic zones in the area of the Carpathian Mountains and their
surroundings, over an observation interval between 1998 and 2012, for which 259
intermediate depth seismic events and 90 crustal seismic events were analysed. A
complete, homogeneous and accessible catalogue with input data to facilitate the
calculation of the seismic hazard, by developing and completing previous versions
made at the National Institute for Earth Physics, is compiled by Oncescu, et al.
(1999). Constantinescu et al. (1976) made a description of some earthquakes and
their causes, which was continued later by Sandi et al. (2005). Using the data
sequence of Vrancea earthquakes over the interval (1100–1973) and by decomposing
them by types of events, according to the nature of the faults and the cyclicity of the
events, Purcaru (1979) made predictions for the future. Therefore, for the major
earthquake of 1977, the author made, in 2005, the prediction 1980 ± 13 years and
the following ones between 2030 and 2040, respectively, 2070 and 2090.

3. Statistical processing of data on seismic events

3.1 General matters

Primary information on seismic phenomena is extracted from the catalogue

provided by the National Institute for Earth Physics. Considering that the
earthquake-generating focus in Romania is the Vrancea area, and they occur in the
subcrustal lithosphere, at the bend of the Eastern Carpathians, and significant
earthquakes outside the area are extremely rare, in this phase of the study, given the
extended horizon of observation which includes the range of the years 984 - 2022,
and the large number of recorded events, the statistical treatment will be done on the
entire database.
The processed database includes 31,537 earthquakes identified in the Romanian
space (longitude: min 43.5941 – max. 48.23 and Latitude: min 20.1 – max. 29.84),
in the period 1 January 1984 – 31 March 2022.
It is a known fact that the occurrence of major seismic phenomena is a “rare
event” from a statistical point of view. Statistical studies of earthquakes typically
start from the fact that rare events are well-modelled by the law of exponents – if
one considers the sequence of time intervals separating the events, or Poisson's law
if the frequency of earthquakes is modelled as an event described by a discrete
variable. The ease of using these two distribution laws, distinct of nature, consists in
the fact that they are defined by the same parameter, describing the same
phenomenon: the behaviour of a system over time, from a continuous or discrete
point of view. Such a study carried out on seismic phenomena in Romania, covering
the time period 1400-1977, did not confirm the hypothesis of an exponential
behaviour, the confirmed model being the bi-parametric Weibull model (Vodă and

8 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

Isaic-Maniu, 1983). An extension of the horizon of observation (1100-2004) but also

of the spectrum of the intensity of the monitored earthquakes, was treated in Dragan
and Isaic (2011). In the material presented, information from the technical literature
was used, respectively, other approved sources such as those of the National Institute
for Earth Physics, mainly the ROMPLUS catalogue (INFP-2022), the scientific
product of the specialised institute.

3.2 Statistical description of earthquake

The 31537 seismic events recorded over the entire observation horizon are
distributed by intensity classes (Figure 2).

Histogram
20000 120,00%
16814
100,00%
15000
80,00%
9373
10000 Frequency 60,00%
4082 40,00%
5000
199 747 214 20,00%
71 37
0 0,00%
sub 1,00 – 2,00 – 3,00 – 4,00 – 5,00 – 6,00 – peste
0,99 1,99 2,99 3,99 4,99 5,99 6,99 7,00
Mw

Figure 2. Distribution of earthquakes recorded in the period 984-2022

Source: Authors’ processing.

Considering that in the period before the 19th century, the data usually show
major earthquakes, for a comparison over time, it is logical to select only the peak
values. The trend of the evolution of the annual earthquakes of maximum intensity
starting from the year 984, turns out to be linear (Figure 3). The graph allowed the
identification of subperiods with different trends, for which chronological trend
equations were tested and validated (Table 1).
The recorded data presented two depth measurement units in Km (25,216 events
in the interval 0-220 km) and in f - 6327 events in the interval 0-218.4f (f- Fathom =
= 1.8288 m).
The strongest seismic activity on the territory of Romania is concentrated at
intermediate depths (60 - 200 km), in a cooler lithospheric body, in gravitational
descent, orientated almost vertically (INFP, 2022). High activity was observed in
two depth ranges - between 80 and 100 km, and between 120 and 160 km,
respectively. Strong earthquakes in the 20th century occurred in both segments: the
1977 (Mw 7.4) and 1990 (Mw 6.9) earthquakes in the upper segment, and the 1940
(Mw 7.7) and 1986 (Mw 7.1) events in the lower segment.

Vol. 58, Issue 3/2024 9

Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin
9 7,9
7,37,5
7,3
7,37,5 7,3 7,5 7,7
7,5
7,5
7,3
7,3 7,5 7,7 7,4 7,1
8 7,1
7,1
7,1
7,1
7,1 7,1
7,1
7,1
7,1
7,1 7,17,3 7,1 7,3 6,86,8 7,1
7,17,1
6,8
6,8 7,1 7,1 6,9
6,5 6,5
6,5
6,5
6,5
6,5
6,4 6,56,56,5
6,5 6,6 6,56,5
6,5 6,5
6,5 6,5 6,6 6,7
6,66,3 6,4 6,6 6,86,3
7 6,2 5,96 6,1 6,16,1
5,9 6,2
5,9
5,9
5,9
5,9
5,96,15,9
5,9 6 5,9
5,7 5,96,1
5,96,1
5,9
5,7 66666,2
6,1
5,9 65,965,7
5,9
5,8 5,9
5,8 5,9
5,8
5,8 66 5,8 6
6 5,3 5,3 5,25,6
5,35 5,55,65,6 5,3
55 5,3 5,4 5,5
5,4
5,4
5 5,3 5,4 5,6
5,65,3
5,5 5,65,5
5,5
5,4
555,35,2
5,1 5
5 5,3 5,6 5,6 5,35,5
5,5
5,3
5,2
5,1 5 5
5,45,6
5,5
5,4
5,2
4,7 4,9 4,7 4,9 4,7
4,74,84,7 4,9 4,94,8
4,8
4,5 4,5 4,3 4,7 4,74,2 4,5
4,5
4,4
4,3 4,3
4,7
4,6
4,4 4,7
4,7
4,6
4,4
4,3
4,7
4,7
4,7
4,6
4,4
4,4
4,3 4,34,3
4,2
5 4,1 4,1 3,94,1
4,1 4,14 4 4,1 4,1
Mw

3,5 3,5 3,8

4 2,93 3,2
3
2
Yt = - 0.0081 t + 6.5818
1
0
984
1126
1327
1517
1569
1599
1623
1679
1738
1786
1798
1821
1830
1843
1862
1876
1887
1895
1901
1907
1913
1919
1927
1935
1941
1947
1953
1959
1965
1971
1977
1983
1989
1995
2001
2007
2013
2019
Ani
Figure 3. Evolution of maximum intensity earthquakes in the period 984-2022
Source: Authors’ processing.

Up to 100 km, almost 75% of the earthquakes recorded on the observation

horizon 984-2022 occurred, and up to 160 km depth, 99.6% of the events occurred.
The distribution of earthquakes according to depth shows two modal values,
recording peak values between 10 and 20 km, respectively between 130-150 km
(Figure 4). The big difference between the mean and the median depth indicates a
strong asymmetry, also generated by the two modal values.

Table 1. Chronological equations for sub-periods of maximum earthquakes

Sub-period (years) Trend equation
984 - 2022 Yt = -0.0081 t + 6.5818
984 -1499 Yt = 0.01670 t + 6.8559
1500 -1699 Yt = -0.0284 t + 6.8168
1700-1799 Yt = -0.0532 t + 6.5904
1800-1899 Yt = -0.0065 t + 5.753
1900-1999 Yt = -0.0065 t + 5.753
1922-2021 Yt = -0.0109 t + 5.8104
Source: Authors’ processing.

The average depth is 45.6 km, the maximum depth was 218.4 km, and the
median 15.4 km, which indicates a strong asymmetry, the coefficient of kurtosis
recording the value - 0.776, with a strongly flattened distribution, the coefficient of
skewness having the value 0.936, far from the standard value of the normal
distribution where k = 3.
To establish the connection between the average magnitude and depth, the 1633
values were organised two-dimensionally (Figure 5), and the image of the cluster of
dots in the vicinity of the regression line suggests a close connection between the
two variables.
The estimated regression equation 𝑌̂𝑥 = 0.01𝑥 + 1.8877 was validated by the
usual tests, being statistically confirmed. The value of the correlation ratio (0.85)
indicates a strong link between the depth of the epicentre and the magnitude value.

10 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

The average depth is 45.6 km with a standard deviation of 51.6. The histogram
image indicates a bimodal distribution, with a first peak value in the 10-20 km depth
zone and a second one in the 130-140 km zone.

9000 120,00% 6,00

8000
Yx = 0.01 x + 1.8877
100,00% 5,00 R² = 0.7362
7000
6000 80,00% 4,00

Mw
5000 3,00
Frequency 60,00%
4000
3000 40,00% 2,00
2000 1,00
20,00%
1000
0 0,00% 0,00
20
40
60
80
0

100
120
140
160
180
200
220

KM 0,00 100,00 km 200,00 300,00

Figure 4. Earthquakes by depth in Km Figure 5. Dependence between mean

magnitude and depth
Source: Authors’ processing.

The coefficient of determination 0.86 shows the proportion in which the

magnitude is dependent on the depth, and the Fisher (F) and Student (t) tests have
significant statistical values and confirm the validity of the estimated equation.
In the last two centuries, in distribution of major earthquakes (over 5 Mw), there
is an increase in frequency due to the existence of rigorous records and also the
development of techniques for measuring magnitude.
The peak of events was recorded in the interval 1900-1950 with 126 events; in
the interval 1350-1400 there was no significant event recorded; also the acceleration
of events starts with the year 1700, the coefficient of kurtosis of 9.72 and skewness
of 3.055.

4. Fitting the Distribution

4.1 Testing distribution of focal points depth

The distribution was tested in order to identify the distribution law, and
following the advancement of several hypotheses, based on successively applied
tests, the General Extreme Value (Johnson et al., 1994) was validated, which belongs
to the family of extreme distributions that also includes Gumbel, Fréchet and
Weibull.
The estimated parameters have the following values: k = 0.55895, 𝜎 = 440.49,
µ= 301.7, where: k – shape parameter µ- location parameter, 𝜎 - scale parameter.
Statistical analysis:
Domain
(𝑥−µ)
1+𝑘  0 for k ≠ 0

(1)

Vol. 58, Issue 3/2024 11

 Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin

−∞ < 𝑥 < +∞ for k = 0

probability density function (pdf) of the Generalized Extreme Value (GEV):
PDF
1 1
1
(𝑥) = exp(−(1 + 𝑘𝑧)−𝑘 (1 + 𝑘𝑧)−1−𝑘 for k ≠ 0 (2)

1
𝑓(𝑥) = exp( −𝑧 − exp(−𝑧) for k = 0

CDF
1
𝐹 (𝑥) = exp(−(1 + 𝑘𝑧)− 𝑘 for k ≠ 0 (3)
𝐹(𝑥) = exp( −exp(−𝑧)) for k = 0
𝑥−𝜇
where: 𝑧 ≡
𝜎
Thus, these functions are defined for all x when k = 0, for all x≥μ – σ/k when k
> 0, and for all x≤μ – σ/k when k< 0. Outside this domain, we can consider f(x) = 0,
and F(x) = 0 when k> 0 and F(x) = 1 when k< 0.

Hazard Function Probability Density Function Cumulative Distribution Function

0.001 1
9E-4 0.9
0.0012
8E-4 0.8
0.001 7E-4 0.7
0.6
6E-4
F(x)
8E-4 0.5
h(x)

f(x)

5E-4
6E-4 0.4
4E-4
0.3
4E-4 3E-4
0.2
2E-4 0.1
2E-4
1E-4 0
0 0 0 1000 2000 3000
0 1000 2000 3000 0 1000 2000 3000 x
x x
Gen. Extreme Value (0.55895; 440.49; 301.7)
Gen. Extreme Value (0.55895; 440.49; 301.7) Gen. Extreme Value (0.55895; 440.49; 301.7)

Figure 6. The h(x)-hazard function, f(x)-probability density function and F(x)

Source: Authors’ processing.

The validation tests confirmed the statistical hypothesis regarding the

distribution of GEV (Generalized Extreme Value) earthquake frequencies depending
on the depth of their occurrence, at all frequently used significance levels α = 0.2,
0.1, 0.05, 0.02, 0.01, for all the three tests used: Kolmogorov-Smirnov (calculated
statistics 0.13677), Anderson-Darling (0.51626), respectively, Pearson-Fisher with
the statistics 0.0389. The comparison with the p-value (0.73276 for the Kolmogorov
test, respectively, 0.98074 for the Chi Square test) reconfirms the GEV type
distribution. The images of the probability density f(x), CDF, respectively, hazard
function h(x) for the estimated parameter values (Figure 6). It arises from the
calculations as follows: f(0) = 8.1435E-4, F(0) = 8.9821E-4 , and h(0) = 8.9821E-4,
and for x =10 the values of the three indicators are: f= 0.0045, F=0.9726, h= 0.16436.

4.2 Characterisation of earthquakes with significant magnitude

The 26 groups of earthquakes of 5, 6 and more than 7 Mw, with a step of

0.1 Mw (Figure 7). The frequency of earthquakes decreases as the magnitude
increases, with a few exceptions such as the 6.5 Mw earthquakes, with a frequency
of 24, or the 18 earthquakes of 7.1 magnitude. The average frequency of an event

12 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

over 5 degrees is 12.39, with a variation of 166.5, respectively, a standard deviation

of 12.9. The distribution is strongly leptokurtic (Kurtosis = 6.57), respectively,
asymmetric (Skewness=2.20).

70
60
60

50
frequency

40
31
30 26 24
18 20 16 18
20 13 13 15
10 11
7 8 7
10 5 4 5
2 3 1 1 1 2 1
0
5.0

7.4
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
6.0
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9
7.1
7.3

7.5
7.7
7.9
Mw
Figure 7. Distribution of major earthquakes (over 5 Mw)
recorded in the period 984-2022
Source: Authors’ processing.

The testing of the distribution law was carried out by successively applying the
three tests, thus the statistic of the Kolmogorov-Smirnov test with the value of
Dn=0.1414 is lower than the theoretical thresholds at different significance levels, as
follows: α=10% (0.2332), α=5% (0.2591), respectively α=1% (0.31064), so the
previously advanced hypothesis of the Pareto type 2 distribution is confirmed.
The Andersen - Darling test also confirms the hypothesis raised, since the
calculated value A2 = 0.2613 falls below the critical thresholds 1.9286 at the 10%
significance level, 2.5018 at the 5%, threshold or 3.907 at the 1% threshold.
Finally, the Chi-Squared χ2 (Pearson-Fisher) test, with the statistic 0.0347,
which is below the critical thresholds at 10% (4.605), at 5% (5.99), but also at 1%
(9.21) also confirms the hypothesis of the Pareto type 2 distribution.
The estimated parameters of the distribution are:
α - continuous shape parameter (𝛼 = 46.514),
respectively
β- continuous scale parameter (𝛽 = 563.67).
After having tested different variants, the Pareto type 2 distribution was
considered valid, with the probability density:
𝛼𝛽𝛼
𝑓(𝑥) = (4)
(𝛽+𝑥)𝛼+1
which, for the value x=0, leads to f(x)=0.08252, and for x=10, f(x)=0,
respectively, the cumulative distribution function takes the following form:

Vol. 58, Issue 3/2024 13

 Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin
𝛽𝛼
F(x) = 1 − (5)
(𝛽+𝑥)𝛼
domain: 0 ≤ 𝑥 ≤ + ∞, with the following restrictions on the parameters: 𝛼 >
0, 𝛽 > 0. The specific indicators have the following values:
𝛽
mean = 12.385, for 𝛼 > 1, and
𝛼−1
𝛽2 𝛼
variance: = 166.5 , for 𝛼 > 2
(𝛼−2 )( 𝛼−1)2
The standard variation has the value 12.9, the coefficient of variation 1.0222, and
the symmetry of the distribution is characterised by a Skewness of 1.0222,
respectively, the curvature by the Kurtosis coefficient=7.1484. The coefficient of
variation had a value of 1.042. The first decile has the value 1, quartile 1 is 2.75, and
Q3 is equal to 18 (Figure 8).

Probability Density Function Cumulative Distribution Function

0.08 1
0.072 0.9
0.064 0.8
0.056 0.7
0.048 0.6
f(x)

F(x)

0.04
0.5
0.032
0.4
0.024
0.3
0.016
0.008 0.2
0 0.1
0 50 100 0
x 0 50 100
x
Pareto 2 (46.514; 563.67)
Pareto 2 (46.514; 563.67)

Figure 8. The probability density and the distribution function

in case of over 5 Mw earthquakes
Source: Authors’ processing.

Regarding the distribution of earthquakes with a magnitude greater than 5 Mw,

over 50-year timespans, the tests applied led to the confirmation of the Pareto type 2
distribution law. The statistics of the tests applied had the following values: for the
Kolmogorov-Smirnov test 0.1453, for the Anderson-Darlin test 2.0446 and for Chi-
Squared the value of 0.36968, in all cases falling below the critical threshold allowed
for different values of the decision risk α.
The distribution parameters had the following results: shape parameter α =
1.2122, respectively scale parameter β= 5.6037; the distribution density f (x,1.2122,
5.6037) and the hazard function h (x,1.2122, 5.6037). The high probability (about
82%) of recording more than one earthquake can be observed with higher values.
The probability of registering a number of earthquakes between 1 and 10 is 53%,
and less than 10 is 71%. The chances of recording more than 5 earthquakes are

14 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

approximately 46%, but more than 10 earthquakes are only 28.9%, a probability that
decreases drastically as x increases, a situation that is also visible on the f (1.2122,
5.6037) and h (1.2122, 5.6037) curves, whose trend is accelerating downward.

4.3 Statistical characteristics of destructive earthquakes

We include in the analysis events with a magnitude of over 6 degrees,

earthquakes with a magnitude of over 6 degrees, earthquakes with devastating effects
on people’s lives and on infrastructure. The primary data of the 108 events are
organised as a frequency distribution series, starting with magnitude 6 and step 0.1.
The distribution is multimodal with clusters at 6 and 6.1 degrees, 26 earthquakes,
respectively, 24 earthquakes of 6.5 degrees, and 26 earthquakes of 7.1 to 7.3 degrees.
The skewness coefficient is 1.42, and the excess kurtosis is 1.41.
The average value of the frequencies is 6.75, in the range of 1 earthquake,
encountered in earthquakes with magnitudes of 6.7, 6.9, 7.4, respectively, 7.9 and 24
events, in the case of earthquakes with a magnitude of 6.5 Mw. The
variance/dispersion recorded the value of 47.8, with a standard deviation of 6.9, a
distortion of 1.42 (Skewness coefficient) and a fairly flat distribution (Excess
Kurtosis 1.314), compared to the normal distribution, where the value of the
coefficient is 3.
The average value of the earthquake intensity (6.84) is very close to the main
modal value 6.5 Mw, which has a frequency of 24. The secondary modal value is 7.1
with a frequency of 18. The median value was 6.75, Q1 = 6.325, and the third quartile
Q3 = 7.375. The deciles recorded the following values: D1= 6.07, respectively, D9 =
7.76, a value close to the maximum recorded earthquake of 7.9 Mw.
The frequency distribution was validated as of Log-Logistic type, while the
critical values for a wide range of significance levels, from 0.2 to 0.01, and the P-
value also confirm the initially formulated null hypothesis.
The Probability Density Function takes the following form:
 −1 −2
  x −    x −   
f (x ) =   1 +  
    
(6)
    
Cumulative Distribution Function (CDF)
 −1
    
F (x ) = 1 +    (7)
  x −  
 
The parameters have the meanings below:
 - continuous shape parameter ( β  0)
 - continuous scale parameter (  0)
 - continuous location parameter (  0 yields the two-parameter Log-
Logistic distribution).

Vol. 58, Issue 3/2024 15

 Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin

The domain of the variable:   x +  . and the estimated parameters have the
following values:
α = 1.4157, respectively β =3.5755, respectively γ = 0,
so that the bi-parametric model can be described as follows:
the probability density for this bi-parametric model is:

f (x :  ,  ) =
( /  )(x /  ) −1
1 + (x /  ) 2
(8)

where x  0,   0,   0 and CDF

( x /  ) = x

1
F ( x :,  ) = = (9)
1+ ( x /  ) 1+ ( x /  )
 
  + x

Hazard Function
Cumulative Distribution Function
Probability Density Function
1 0.18
0.2
0.9
0.16
0.8
0.14
0.7 0.15
0.12
0.6
h(x)
F(x)

0.1
f(x)
0.5 0.1
0.4 0.08
0.3 0.06
0.05
0.2 0.04
0.1 0.02
0 0 0
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
x x x
Log-Logistic (1.4157; 3.5755) Log-Logistic (1.4157; 3.5755) Log-Logistic (1.4157; 3.5755)

Figure 9. Functions f(x), F(x) and h(x) for devastating earthquakes over 6 Mw
Source: Authors’ processing.

The quantile function

1/ 
 p 
F ( p;  ,  ) =  
−1
 . (10)
1− p 
The log-logistic has been used as a simple model of the distribution of wealth
or income in economics, where the Gini coefficient is 1 /  (Kleiber and Kotz, 2003;
Al-Shomrani et al., 2016) it is known as the Fisk distribution. The log-logistic
distribution provides one parametric model for survival analysis. The survival
function is

S (t ) = 1 − F (t ) = 1 + (x /  )
 −1
, (11)
and so the hazard function is
f (t ) ( /  )(x /  )
 −1
h(t ) = = . (12)
S (t ) 1 + (x /  )


For various values of the variable x, the main statistical indicators of the Log-
logistic distribution, f(x), F(x), S(x) and of the intensity of events h(x) (Figure 9).

16 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

The data indicate that as the number of possible seismic events increases, the
intensity rate h(x) of the events decreases drastically (Table 2).

Table 2. Values of the Log-logistic distribution indicators f

x=0 x=1 x =5 x = 10 x = 25
f (x) 0 0.17187 0.06694 0.02171 0.00319
F(x) 0 0.14140 0.61650 0.81091 0.94009
S (x) 1.000 0.85860 0.38350 0.18909 0.05991
h (x) 0 0.20017 0.17455 0.11480 0.05323
Source: Authors’ processing.

5. Conclusions

The overwhelming majority of earthquakes in Romania are in the Vrancea area,

at the curvature of the Carpathian Mountains, in the last 300 years only one major
earthquake occurred with the epicentre outside this area, the 1916 earthquake. The
Vrancea seismogenic region is located at the continental convergence, placed at the
contact point of three tectonic plates: the East - European plate, the Intra - Alpina
and Moesica subplates and intermediate, and the hypocentre of most earthquakes is
located at a depth between 60-70 km and 100-220 km from the earth’s surface;
In addition to the Vrancea area, the following areas are relatively active in
Romania: the Fagaras - Câmpulung area, the Danubian area in the vicinity of the
Danube river, the Banat zone, the Crisana-Maramures, Barlad, Predobrogeana, and
Intramoesica Zones, which are distinct zones but with a low frequency of seismic
events of significant magnitude.
The analysis of the seismic activity is facilitated for Romania by the preparation
and permanent updating of a complete, homogeneous and accessible catalogue with
input data to facilitate the calculation of the seismic hazard, through the development
and completion of previous versions drawn up at the National Institute for Earth
Physics, a catalogue updated periodically.
Thus, total recorded events amount is 31537, of which 86.0% are up to a
magnitude of 4 degrees, 980 earthquakes were over 5 degrees, and 37 were
devastating with a magnitude of over 7 Mw degrees on the Richter scale. The year
in which the most earthquakes were recorded was 2015 with 2152 events, and the
month in which the highest frequency was recorded was October, respectively, the
22nd day. The trend of the evolution of maximum intensity earthquakes starting with
the year 984 is linear in the form 𝑌̂𝑡 = −0.0081𝑡 + 6.5818
A strong correlation (0.86) was recorded between the depth of the epicentre and
the intensity of the earthquake. The specificity of the statistical distribution was
tested for the depth of the focus, for the distribution of earthquakes of 5+,
respectively, those of 6-7 degrees, for which purpose three tests (Kolmogorov-
Smirnov, Andersen-Darling, Chi-Squared) were used, then finally choosing the
result where the agreement between the obtained results was the highest.
Simulations were carried out regarding the modification of the statistical
indicators of the distribution of destructive earthquakes (over 6 Mw) depending on

Vol. 58, Issue 3/2024 17

Alexandru Isaic-Maniu, Emilia Gogu, Irina-Maria Dragan, Florentina Constantin

different values of the x variable. At intervals of 50 years each, 322 events were
recorded starting with the year 984 throughout the entire analysis interval, the
chances of recording less than three events are about 40%, between one and ten 82%,
and more than ten only 29%.

Acknowledgments: This paper was partially supported by the project “Societal and
Economic Resilience within multi-hazards environment in Romania” funded by
European Un-ion—NextgenerationEU and Romanian Government, under National
Recovery and Resilience Plan for Romania, contract no.760050/ 23.05.2023, cod
PNRR-C9-I8-CF 267/ 29.11.2022, through the Romanian Ministry of Research,
Innovation and Digitalization, within Component 9, Investment I8.

References
[1] Al-Shomrani, A.A., Shawky, A.I., Arif, O.H. et al. (2016), Log-logistic distribution for
survival data analysis using MCMC. SpringerPlus 5, 1774. https://doi.org/10.1186/
s40064-016-3476-7.
[2] Ali, S.M., Akkoyunlu, M.F. (2022), Statistical analysis of earthquake catalogs for
seismic hazard studies around the Karliova Triple Junction (eastern Turkey). Journal
of African Earth Sciences, 186, 104436, https://doi.org/10.1016/
j.jafrearsci.2021.104436.
[3] Bala, A., Radulian, M., Toma-Danila, D. (2021), Present-day stress field pattern in the
Vrancea seismic zone (Romania) deduced from earthquake focal mechanism inversion.
Ann. Geophys. 64(6), 1-24, https://doi.org/10.4401/ag-8632.
[4] Beșuțiu, L., Zugrăvescu, D. (2004), Geophysical Considerations on the Black Sea
Opening And its Seismo-Tectonic Consequences. Rev. Roum. GÉOPHYSIQUE, 48,
3-13. http://www.geodin.ro/RRG/revue2004/Art%2001.pdf.
[5] Bokelmann, G., Rodler, F.A. (2014), Nature of the Vrancea seismic zone (Eastern
Carpathians) – New constraints from dispersion of first-arriving P-waves. Earth and
Planetary Science Letters, 390, 59-68, https://doi.org/10.1016/j.epsl.2013.12.034.
[6] Constantinescu, L., Constantinescu, P., Cornea, I., Lazarescu, V. (1976), Recent seismic
information on the lithosphere in Romania. Rev. Roum. Géophys., 20, 33-49.
[7] Dutu, A., Niste, M., Spatarelu, I., Dima, D.I., Kishiki, S. (2018), Seismic evaluation of
Romanian traditional buildings with timber frame and mud masonry infills by in-plane
static cyclic tests, Engineering Structures, 167, 655-670, https://doi.org/10.1016/
j.engstruct.2018.02.062.
[8] Dragan, I., Isaic-Maniu, A. (2011), The Risk Analysis of Seismic Activity Incidence in
România. Reliability: Theory & Applications, 2(3), 1-12. https://www.gnedenko.net/
Journal/2011/032011/RTA_3_2011-01.pdf.
[9] Faenza, L., Hainzl, S., Scherbaum, F., Beauval, C. (2007), Statistical analysis of time-
dependent earthquake occurrence and its impact on hazard in the low seismicity region
Lower Rhine Embayment. Geophysical Journal International, 171(2), 797-806.
https://doi.org/10.1111/j.1365-246X.2007.03564.x.

18 Vol. 58, Issue 3/2024

 The Risk of Earthquakes in Romania – a Statistical Point of View

[10] Grünthal, G., Stromeyer, D. (1992), The recent crustal stress field in central Europe:
Trajectories and finite element modeling, J. Geophys. Res., 97(B8), 11805-11820,
https://doi.org/10.1029/91JB01963.
[11] INFP, Institutul Național de Cercetare - Dezvoltare pentru Fizica Pământului (2022),
Catalogul Romplus Actualizat, https://www.infp.ro/index.php?i=romplus.
[12] Johnson, N.L., Kotz, S., Balakrishnan, N. (1994), Continuous Univariate Distributions,
Vol. 2, John Wiley, New York.
[13] Kleiber, C., Kotz, S. (2003), Statistical Size Distributions in Economics and Actuarial
Sciences, John Wiley, New York.
[14] Kothari, S., Shcherbakov, R., Atkinson, G. (2020), Statistical modeling and
characterization of induced seismicity within the Western Canada Sedimentary Basin.
Journal of Geophysical Research: Solid Earth, 125(12), https://doi.org/
10.1029/2020JB020606.
[15] Ogata, Y. (1988), Statistical Models for Earthquake Occurrences and Residual Analysis
for Point Processes. Journal of the American Statistical Association, 83(401), 9-27,
https://www.tandfonline.com/doi/abs/10.1080/01621459.1988.10478560.
[16] Oncescu, M.C., Marza, V.I., Rizescu, M., Popa, M. (1999), The Romanian Earthquake
Catalogue Between 984 – 1997. In: Wenzel, F., Lungu, D., Novak, O. (eds) Vrancea
Earthquakes: Tectonics, Hazard and Risk Mitigation. Advances in Natural and
Technological Hazards Research, vol 11, 43-47, Springer, Dordrecht,
https://doi.org/10.1007/978-94-011-4748-4_4.
[17] Popescu, E., Radulian, M., Bala, A., Toma-Danila, D. (2018), Earthquake mechanism
in the Vrancea subcrustal source and in the adjacent crustal seismogenic zones of the
South-Eastern Romania. Românian Reports in Physics 70(70), 1-16,
https://rrp.nipne.ro/2018/AN70704.pdf.
[18] Purcaru, G. (1979), The Vrancea, Romania, earthquake of March 4, 1977 — a quite
successful prediction. Physics of the Earth and Planetary Interiors, 18(4), 274-287,
https://doi.org/10.1016/0031-9201(79)90064-5.
[19] Radulian, M. (2015), Mechanisms of Earthquakes in Vrancea. In: Beer, M.,
Kougioumtzoglou, I.A., Patelli, E., Au, SK. (eds) Encyclopedia of Earthquake
Engineering, Springer, Berlin, Heidelberg, https://doi.org/10.1007/978-3-642-35344-
4_302.
[20] Radulian, M., Mândrescu, N., Panza, G., Popescu, E., Utale, A. (2000), Characterization
of Seismogenic Zones of Romania. Pure appl. Geophys, 157, 57-77,
https://doi.org/10.1007/PL00001100.
[21] Sandi, H., Borcia, I.S., Stancu, M. (2005), Analysis of attenuation for aecent Vrancea
intermediate depth earthquakes. Rev. Roum. GÉOPHYSIQUE, 49, 31-48,
http://www.geodin.ro/RRG/revue2005/Art.02.pdf.
[22] UNDRR - United Nations Office for Disaster Risk Reduction (2022), Europe:
earthquake hazard map, European Spatial Planning Observation Network,
https://www.preventionweb.net/publication/europe-earthquake-hazard-map.
[23] Vodă C., Isaic-Maniu, A. (1983), An application of the Weibull model in studying
seismic phenomena. Mathematics Gazette, IV (1-2), 68-73.

Vol. 58, Issue 3/2024 19