Theoretical and Applied Climatology (2023) 153:1411–1424

https://doi.org/10.1007/s00704-023-04525-x

RESEARCH

Frequency analysis of annual maximum daily precipitation

in northeastern Algeria: mapping and implications under climate
variability
Mohammed Benaini1,2 · Mohammed Achite1 · M. G. Mostofa Amin3 · Vijay P. Singh4

Received: 6 February 2023 / Accepted: 3 June 2023 / Published online: 29 June 2023
© The Author(s), under exclusive licence to Springer-Verlag GmbH Austria, part of Springer Nature 2023

Abstract
Precipitation in northeastern Algeria is irregular, often exhibiting exceptional values. Extreme precipitation events, such
as annual maximum daily precipitation (AMDP), can cause major disasters. Changes in global climate have increased the
frequency and intensity of AMDP in many areas. This article examines the distribution and quantile mapping of AMDP,
using the data from 180 precipitation stations covering eight watersheds in northeastern Algeria between 1970 and 2014.
The watersheds, from east to west, are Coastal Algiers, Soummam, El Hoddna, Coastal Constantine, Kebir Rhumel, High
Plateaus Constantine, Seybouse, and Medjerdah. By applying different goodness-of-fit tests, the chi-square (χ2), Kolmogo-
rov–Smirnov (KS), Anderson–Darling test (AD), and log-likelihood ratio test, the best-fit distribution function for frequency
analysis of AMDP was determined to be the generalized extreme value (GEV), with a 95% confidence level. The maximum
value of AMDP for the data series generally varied between 60 and 150 mm, with higher values in the humid zone and
lower values in the arid zone. The validity of the data used in this study was assessed through a series of tests, including
the Wald–Wolfowitz test for independence, the Kendall test for stationarity, and the Wilcoxon test for homogeneity. Results
of these tests demonstrated that a majority of the data met the conditions for independence, stationarity, and homogeneity.
Results of goodness-of-fit test and graphical fit provided further support for the appropriateness of the GEV probability
distribution function as the best fit for the data. Application of various goodness-of-fit tests for the first time on a significant
number of stations in northeastern Algeria provided a better understanding of the uncertainties in frequency analysis and
helped identify the most suitable distribution function for modeling AMDP in different climate regimes. The findings high-
light the high-risk zones for extreme precipitation events and emphasize the need for increased investment in water resource
infrastructure and management in order to address the challenges posed by climate change and population growth. This is
particularly important in ensuring water security for agricultural and urban needs in northeastern Algeria.

1 Introduction

There has been a significant variability in the climate regime

* Mohammed Benaini
over the past four decades, which has had an impact on water
mohammed.benaini@univ-dbkm.dz
resources through seasonal lag, droughts, irregular and
1
Faculty of Nature and Life Sciences, Laboratory of Water & extreme precipitation, and flooding (Noto and La Loggia
Environment, Hassiba Ben Bouali University of Chlef, B.P 2008; Goula et al. 2012; Hussain and Lee 2013; Garijo
78C, Ouled Fares, 02180 Chlef, Algeria and Mediero 2018; Hussain et al. 2021). These impacts are
2
Department of Ecology and Environment, Faculty of Nature mainly caused by variations in precipitation, particularly
and Life and Earth Sciences, Khemis Miliana University, extreme precipitation. The annual maximum daily precipita-
44225 Ain Defla, Algeria
3
tion (AMDP) is characterized by high intensity, which can
Department of Irrigation and Water Management, cause natural disasters, such as flooding (Katz and Brown
Bangladesh, Agricultural University, Mymensingh 2202,
Bangladesh 1992; Sen Roy and Balling 2004; Mailhot et al. 2010; Rousta
4 et al. 2016; Tabari 2020). Events like AMDP can seriously
Department of Biological and Agricultural Engineering
& Zachry Department of Civil Engineering, Texas A&M affect human life, economy, ecosystems, and agriculture
University, College Station, TX 77843‑2117, USA (Leclerc and Ouarda 2007; Raj and Azeez 2010; Murray and

13
Vol.:(0123456789)
1412 M. Benaini et al.

Ebi 2012; Teegavarapu 2012; Patel and Shete 2015; Pombo The main contribution of this study is to investigate the fre-
and de Oliveira 2015; Santos et al. 2017; Gao et al. 2018). quency and magnitude of AMDP across northeastern Algeria,
Climatic changes have led to an increase in the frequency a region known for its heavy precipitation and important water
and intensity of extreme hydrological events, particularly resources in Algeria. The study aims to determine the amount
AMDP. Flood disasters continually break previous records of precipitation and its frequency, return period with a proba-
in terms of intensity or area inundation, suggesting that the bilistic approach, in order to better understand and predict the
frequency and magnitude of floods will increase significantly impact of extreme hydrological events on the region.
in the future (Beniston et al. 2007; Villarini et al. 2009; Hira- This study is also the first to apply three different good-
bayashi et al. 2021; Sarfaraz et al. 2021; Chen et al. 2021). ness-of-fit methods for frequency analysis of AMDP in
In order to reduce the consequences of natural disasters, Northeast Algeria, namely, the chi-square (χ2), Kolmogo-
statistical methods, such as frequency analysis, are applied rov–Smirnov (KS), and Anderson–Darling (AD) tests. The
to predict their exceedance probabilities in future. Frequency application of these different goodness-of-fit tests provides a
analysis leads to the estimation of frequency at which a better understanding of the uncertainties in frequency analy-
given event will occur (Stedinger et al. 1993; Smithers and sis and helps identify the most appropriate distribution func-
Schulze 2001; Zhang et al. 2021). Dalrymple (1960) intro- tion for modeling AMDP under different climate regimes.
duced a regional frequency analysis method, which provides Results showed that the generalized extreme value
a framework for identifying hazards associated with extreme (GEV) distribution was the best fit for AMDP data series
events, where the concept of regionalization is the temporal under all three tests, indicating that the GEV distribution
variation at the spatial scale of a given region (Dalrymple is suitable for modeling AMDP in this region. However,
1960; Vogel and McMartin 1991; Stedinger et al. 1993; with the impact of climate change, the properties of the
Malekinezhad and Zare-Garizi 2014). Return periods for the GEV distribution may change, and it is necessary to assess
characterization of future events were based on the assump- the reliability of frequency analysis under the new climatic
tion that the future will be similar to the past and will be conditions. Results of this study will provide insights into
generated from the same physical processes as those of the the behavior of AMDP in northeastern Algeria and will
past (Frich et al. 2002; Meylan et al. 2012; Shao et al. 2017; contribute to the development of more accurate prediction
Kousar et al. 2020; Siddique and Palmer 2020). Further, models for extreme hydrological events. These findings
many statistical approaches are based on stationarity, imply- can be applied in hydrological design and management
ing temporal invariance in hydrologic frequency analysis. practices in order to mitigate the consequences of natural
However, climate change has cast doubts on this assumption, disasters, such as flooding.
based on the knowledge that an acceleration of changes in The next section describes the study area and the theo-
hydrologic events is now being anticipated (Katz et al. 2002; retical background and frequency analysis. Data are then
Cunderlik and Ouarda 2009; Salas and Obeysekera 2014; presented and treated in Section 3 including the discussion
Liu et al. 2015; Chen et al. 2017; Nasri et al. 2017; Chen of the results.
et al. 2018).
This study focuses on the eastern region of northern Algeria,
known for its heavy precipitation and important water resources 2 Materials and methods
compared to the predominantly semi-arid western region
(Mebarki 2005). In order to study the AMDP characteristics, 2.1 Study area
it is essential to determine the amount of precipitation and its
frequency (return period with a probabilistic approach) (Taibi The northeastern region of Algeria was selected as the study
et al. 2015; Rousseau et al. 2014). Various studies associated area, which represents half of the northern Algerian region
with statistical analysis using different techniques have been with an area of 83,377 ­km2, according to the surfaces given
carried out around the world (e.g., Hailegeorgis et al. 2013; by the National Agency for Water Resources (ANRH 2005).
Onibon et al. 2004; Rousta et al. 2016; Chen et al. 2017; Pombo The northeastern region of Algeria is divided into eight
and de Oliveira 2015; Beck et al. 2015; Keggenhoff et al. 2014; watersheds, namely, Coastal Algiers, Soummam, El Hoddna,
Ahammed et al. 2014; Nyatuame and Agodzo 2017) and also in Coastal Constantine, Kebir Rhumel, Constantine High Pla-
Algeria (e.g., Benhattab et al. 2011; Habibi et al. 2012; Benab- teaus, Seybouse, and the Medjerdah watersheds (Fig. 1). This
desselam and Amarchi 2013; Benhattab et al. 2014; Meddi and diversified geographical study area extends between 03°12’
Toumi 2015; Dad and Benabdesselam 2018; Boucefiane and W and 08°39’ E longitude and between 37°05’ N and 34°23’
Meddi 2020). All of the studies mentioned above, which were S latitude. The study area is bounded on the north by the
applied in the North of Algeria, have shown the dominance Mediterranean Sea, on the South by Chott Melrhir watershed,
of generalized extreme value (GEV) distribution as the most in the west by the following watersheds Isser, Chlef, and Zah-
appropriate regional distribution to model AMDP. rez, and in the east by Algerian–Tunisian border.

13
Frequency analysis of annual maximum daily precipitation in northeastern Algeria: mapping… 1413

Fig. 1  Study area location on the watershed map in northeastern Algeria

From north to south, the relief of eastern Algeria is char- Chott El Hodna watershed and Constantine High Plateaus
acterized by a parallel arrangement of two main mountain watershed, the areas of weakness in the center are marked by
ranges. In the North, the Atlas Tellien is the first hydrologi- a string of Chotts and Sebkhas, where the wadis come down
cal barrier to cloud masses generating rain. The maritime from the foothills. The Mediterranean hydrographic network
massifs isolate the coastal plains in several sub-regions of extends widely on the high plains, in their northern margins
variable width (80 km to 190 km), which succeed from east (Wadi Sebaou of the Coastal Algiers watershed, wadi El
to west. Going further south, the second mountain range Kebir of the Kebir Rhumel watershed), and especially in
“Saharan Atlas” makes the climate drier compared to the the east (wadi Medjerdah) and the west (wadi Boussellam
north, where vast plains extend as far as the eye can see. second tributary of wadi Soummam).
The two atlases come closer together towards the east of The Mediterranean climate is dominant on the coastal
Algeria and tend to merge. Vast plains and highlands are region of northern Algeria, where precipitation is very heavy
interspersed between the Tellian Atlas and the Saharan in the eastern part and can exceed 1000 mm/year on the
Atlas. The average altitude in the High Plateau region ranges maritime mountains. In the Highlands region, between the
between 1000 and 1200 m with the exception of the Chott Atlas Tellien and the Atlas Saharian, precipitation decreases
El Hoddna watershed where the altitude decreases to around and generally varies between 200 and 800 mm/year. Going
400 m (Beaudet and Marre 1988; Laborde 1998; Mebarki further south, following a strong latitudinal gradient, on the
2005; Meddi and Toumi 2015). Saharan foothills, precipitation becomes scarce and in some
According to Mebarki (2002, 2005) and Benaini (2015), cases does not exceed 100 mm/year (Mebarki 2002; Meddi
the hydrographic network in the north of Algeria is com- and Toumi 2015; Kondratieva and Amarchi 2015).
posed of permanent and dense wadis in the east, compared to According to the National Agency for Water Resources
the west where the semi-aridity is dominant or in the south (ANRH), Algeria has 17 major watersheds that are moni-
known by the Saharian climate and the wadis are often tem- tored through a network of stations located throughout
porary and endorheic. Often one finds flow in endorheic the country to collect hydrometeorological data (ANRH
basins of the interior converging in places with low altitudes, 2005). The selected AMDP data were obtained from the
forming salt lakes, called “Chott” or “Sebkha.” Like the archives of the National Agency for Hydraulic Resources

13
1414 M. Benaini et al.

(ANRH). The selected study period (1970–2014) corre- 2.2 Frequency analysis

sponds to a consistent series for most of the precipitation
stations and is long enough for statistical treatment. The Frequency analysis involves the estimation of quantiles of hydro-
description of precipitation stations in each watershed is logic extreme events, e.g., flow, flood, and AMDP. The main
provided in Table 1. The distribution of stations (Fig. 2) objective is to select an adequate distribution to describe the
is relatively uniform and covers the entire study area, hav- AMDP data series and thus obtain risk estimates of satisfactory
ing at least two stations per sub-basin, except for a few accuracy for the problems at hand (Stedinger et al. 1993; Coles
isolated locations. 2001; Gado et al. 2021). Prior to conducting the analysis, it is
The numbering of precipitations series (Table 1) has essential to confirm that the time series data satisfy the assump-
been introduced to facilitate the interpretation of results tions of independence, stationarity, and homogeneity. These con-
for each watershed. ditions were evaluated using established statistical tests, including

Table 1  Numbering of Watershed Code Area ­(km2) Number of Number of Stations number
pluviometric stations by sub-basins stations
watershed
Coastal Algiers of the East 02 3 950 09 15 S1–S15
Coastal Constantine 03 11 451 18 35 S16–S51
Chott El Hoddna 05 25 843 24 34 S52–S85
Constantine High Plateaus 07 9 578 07 19 S86–S104
Kebir Rhumel 10 8 815 07 20 S105–S124
Medjerdah 12 7 870 05 17 S125–S141
Seybouse 14 6 745 06 16 S142–S156
Soummam 15 9 125 10 24 S157–S180
Total 83 377 86 180 S1–S180

Fig. 2  Locations of selected rain gauges on a digital elevation model (DEM) of northeastern Algeria

13
Frequency analysis of annual maximum daily precipitation in northeastern Algeria: mapping… 1415

the Wald–Wolfowitz test for independence, the Kendall test for specifically used for the description of tail-risk values (Stedinger
stationarity, and the Wilcoxon test for homogeneity (Ondo et al. et al. 1993; Norbiato et al. 2007; Maity 2022). The lognormal
1997; Hamed and Rao 2000; Yang et al. 2010). distribution with two parameters is a logarithmic transformation
Several distribution functions can be done for frequency analy- following a normal distribution. The LN2 distribution has two
sis of extreme hydrological events, often without taking into con- parameters, location and scale, mostly applicable for hydrologic
sideration the physical process of events (Singh and Strupczewski variables like monthly rainfall depth and river discharge volumes,
2002; El Adlouni et al. 2008). Stedinger et al. (1993) and Hamed which are symmetrically distributed. The lognormal distribution
and Rao (2000) assessed the widely used probability distribution is used to determine the extremes of variables at monthly and
functions (Table 2) for hydrologic extreme events (AMDP), such annual scales (Stedinger 1980; Ojha et al. 2008; Maity 2022;
as Gumbel (GM) (Salinas et al. 2014), lognormal with two param- Rizwan et al. 2023). The three-parameter lognormal distribu-
eters (LN2) (Yuan et al. 2018), lognormal with three parameters tion is an extension of the two-parameter lognormal distribu-
(LN3) (Malamud and Turcotte 2006), and Pearson type 3 (P3) tion, developed to better meet the needs of various other fields.
(Alam et al. 2018). The generalized extreme value (GEV) distri- Although the LN3 model has been shown to provide good fits for
bution is commonly adopted to model extreme precipitation data, some datasets, its use remains controversial. Estimating param-
as it has been shown to provide a good fit at various locations eters for the LN3 distribution can be challenging due to the inclu-
worldwide. Notable examples include Greece (Koutsoyiannis and sion of the location parameter, which creates technical difficul-
Baloutsos 2000), Panama (Shamir et al. 2013), Malaysia (Zakaria ties, particularly for interval estimation (Sangal and Biswas 1970;
and Shabri 2013), Czech Republic (Rulfová et al. 2016), India Stedinger 1980; Naghavi et al. 1993; Yurekli et al. 2009; Chen
(Bajirao 2021), Mexico (Flowers-Cano and Ortiz-Gómez 2021), and Miao 2012). However, it is worth noting that the lognormal
Qatar (Mamoon and Rahman 2016), and Tunisia (Merzougui and distribution with three parameters may not always be the best
Zghibi 2020). In this paper, we applied the GEV distribution to fit for all datasets, and alternative models should be considered,
our data and assessed the goodness of fit using conventional tests, particularly in the case of AMDP. Pearson type III is a three-
including the chi-square (χ2), Kolmogorov–Smirnov (KS), and parameter distribution from the seven different types of family of
Anderson–Darling (AD) tests. We also employed adjustment cri- Pearson distributions. It is sometimes called the three-parameter
teria to balance model complexity and fit. The parameters of all Gamma distribution. The P3 distribution is commonly used in
distribution function were estimated using the maximum likeli- hydrology and hydroclimatology for frequency analysis, with
hood estimation (MLE) method. three parameters: location, scale, and shape (Kite 1975; Sted-
In this study, we will generally adopt a number of probabil- inger 1980; Maity 2022). The Gumbel distribution, also known
ity distributions belonging to the normal, gamma, and extreme as extreme value type I distribution, is a limiting probability
value families of distributions. The normal family of distributions distribution used to model the maximum or minimum values
consists of LN2 and LN3, while the gamma family of distribu- from a sample of independent, identically distributed random
tions consists of P3. Likewise, GEV and Gumbel, also known as variables, as the size of the sample increases. The distribution is
extreme value type 1 distributions, are members of the extreme characterized by two parameters: location and scale. In hydrol-
value family of distributions. The generalized extreme value ogy, the Gumbel distribution is widely used to model extreme
(GEV) distribution is commonly used to model the extreme val- precipitation events or river discharge volumes. It is also used in
ues of random variable datasets, particularly in hydrology, such the frequency analysis of floods (Kite 1975; Haan 1994; Beirlant
as the cases of extreme precipitation, floods, and droughts. The et al. 2006). In summary, different distribution functions are used
GEV distribution has three parameters, location, scale, and shape, to model extreme precipitation events based on the nature of the
data and the research question. The choice of distribution func-
tion is based on the assumptions and equations of the theory of
Table 2  Probability density functions (PDF) probability.
PDF Equation Variate and parameter ranges
2.3 Goodness‑of‑fit test
{ [ ]1∕k }
GEV k(x−u) α>0
FX (x) = exp − 1 − u + ∝k ≤ x ≤ ∞ if k < 0
𝛼 To assess the goodness of fit between our AMDP data series
∞ < x ≤ u + ∝k if k > 0
{ } and theoretical distribution functions, we applied the most
GM −∞ < x < ∞, α > 0
F(x) = exp −exp(− x−u
𝛼
) widely used tests, including the chi-square (χ2), Kolmogo-
LN2
� � �2 � X>0 rov–Smirnov (KS), and Anderson–Darling (AD) tests. These
1
F(x) = √ exp − 21 ln(x)−b
a tests were applied at a significance level of 0.05, with the null
ax 2𝜋

LN3 F(x) = 1
�
[ln(x−y)−b]
�
m<x hypothesis (H0) being that the AMDP data series follows
√ exp −
(x−m)a 2𝜋
[ ]
2a
the specified distribution and the alternative hypothesis (HA)
P3 ( x−m
b−1
m ≤ x if x > 0
being that it does not (Cunnane 1989; Haan 1994; Kumar
)
F(x) = a
exp − x−m
|a|Γ(b) a x ≤ m if a < 0
et al. 2017; Bajirao 2021; Maity 2022). By using multiple

13
1416 M. Benaini et al.

∑n � � � � ��
goodness-of-fit tests, we were able to better evaluate the
i=1 (2i − 1) lnF Xi + ln(1 − F Xn−i+1 )
appropriateness of different distribution functions for mod-
2
A = −n (3)
n
eling AMDP under various climate regimes.
The null hypothesis is accepted if the calculated value of
2.3.1 Chi‑squared test (χ2) A2 does not exceed the critical value that is 2.4922 at α = 0.05.

The chi-squared test is one of the most commonly used 2.3.4 Adjustment criteria
goodness-of-fit tests for extreme precipitations. It is
designed to measure the level of agreement between a theo- In order to choose the best fit, different approaches facili-
retical probability distribution and an empirical distribution tate the choice of validation (Hebal and Remini 2011). The
based on a random sample (Melesse et al. 2010; Maity 2022; information criteria, Bayesian (BIC; Schwarz 1978) and
Rizwan et al. 2023). χ2 for random variable is defined as Akaike (AIC; Akaike 1974), were used in addition to the
χ2 test and graphical visual examination to provide a rela-
∑ (X0 − XE )2
𝜒2 = (1) tive measure of model fit for selecting between two or more
XE distribution functions with equal χ2 for a given data series.
Slight preference was given to the convenience in choosing
where X0 is the observed value (AMDP) and XE is the value
the preferred model. Both criteria allow a classification of
estimated with different distribution functions. Typically, the
statistical models to be constructed, taking into account the
probability distribution that yields the lowest value of χ2 is
principle of parsimony (Sienz et al. 2010; Bella et al. 2020;
considered as the most suitable probability distribution for
Flowers‑Cano and Ortiz‑Gómez 2021). If ̂  is the maxi-
the recorded range of the given data (Kumar et al. 2017;
mized log-likelihood for a model containing p parameters
Sabarish et al. 2017; Yuan et al. 2018; Bajirao 2021).
and n sample size, the criteria are defined as
The null hypothesis is accepted if the calculated χ 2 is
less than the critical value 𝜒𝛼2 , where the critical value 𝜒𝛼2 is
BIC = −2̂
 + plog(n) (4)
calculated according to α = 0.05 and the degree of freedom
ν = m − p − 1, knowing that p is the number of parameters
used in fitting the distribution. AIC = −2̂
 + 2p (5)

2.3.2 Kolmogorov–Smirnov test (KS) AIC is designed to be good at selecting the “least bad”
model in a list that does not contain the true model. In con-
The Kolmogorov–Smirnov test is an efficient goodness-of-fit trast, BIC is known for selecting (possible under-fitted) mod-
test. It is a non-parametric test used to determine the maxi- els with a lower dimension than AIC. The most accurate
mum absolute difference between a given sample (x1, x2,…, adjustment for each of the two criteria is the lower value of
xn) and a specific theoretical distribution function F (Slakter its coefficient (Burnham and Anderson 2002; Bousquet and
1965; Justel et al. 1997; Zhang and Wu 2002; Schervish and Bernardara 2021).
DeGroot 2012). It is defined as
2.4 Quantile estimation and mapping
Dn = max||Fn (x) − F(x)|| (2)
After selecting the most appropriate adjustment distribu-
The null hypothesis is accepted if the maximum absolute
tion function for each precipitation series, quantiles (i.e.,
difference (Dmax) is less than the critical value Dα, where the
the values corresponding to specific probabilities) were
critical value Dα of our study is 0.2027 for α = 0.05.
calculated using the fitted distribution. Estimating and
interpreting extreme value models in terms of quantiles
2.3.3 Anderson–Darling test (AD)
or return levels are often more convenient than focusing
on individual parameter values (Gumbel 1941; Hosking
The Anderson and Darling goodness-of-fit test was developed
and Wallis 1987; Hamed and Rao 2000; Coles 2001; Katz
by Anderson and Darling (1952) as a modification of the KS
et al. 2002). In order to estimate the upper tail behavior
test to reduce the stochastic progress errors and give more
of a common distribution function F of a series of inde-
weight to the tails than the KS test. This test makes use of the
pendent random variables x1, x2,..., xn, the p quantile of
particular distribution to determine the critical values and com-
the annual maximum distribution can be calculated using
pares the fit of an observed and theoretical cumulative distribu-
the following formula:
tion function F. Xi refers to the ordered data (Stephens 1974;
Laio 2004; Melesse et al. 2010; Shin et al. 2012; Maity 2022). [ ]
qq = 𝜇 + 𝜎 { −log(1 − p)} −𝜉 − 1 ∕𝜉 (6)
The Anderson–Darling statistic (A2) is calculated using Eq. (3):

13
Frequency analysis of annual maximum daily precipitation in northeastern Algeria: mapping… 1417

Parameters μ, σ (> 0), and ξ are dependent on the tail 3 Results and discussion
behavior of F. This formula is often used in engineering to
determine the return level associated with a {-log(1 -p)}-1- 3.1 Data statistics
year return period (Coles and Tawn 1996).
The calculated AMDP for each return period was spa- The statistics of AMDP recorded during 1970–2014 at
tially distributed over the entire study area using the inverse the 180 stations of the study area are illustrated in Fig. 3
distance weighting (IDW) interpolation method which is and Fig. 4. The maximum value of AMDP, for each series,
most frequently used for spatial interpolation. Dirks et al. generally varied between 60 and 150 mm. For the region
(1998) compared IDW with other methods, and they noted where the climate is arid (South of El Hoddna watershed),
that IDW estimation error was lower (Ly et al. 2011). They the AMDP was very low ranging between 25 and 55 mm.
recommended the use of IDW for interpolation of high For regions with high precipitation (e.g., Coastal Constan-
spatial density networks. Additionally, it has been shown tine watershed), the AMDP was greater than 150 mm and
to perform well in situations where the data are spatially may exceed 200 mm. The minimum values of AMDP of
correlated and exhibit smooth trends. This has been dem- the humid climate were higher than the minimum values of
onstrated in numerous studies, including those by Szolgay AMDP of the semi-arid or arid climate.
et al. (2009), Rahman et al. (2013), and Yang et al. (2015), The average AMDP values provided an overall idea about
which have successfully applied IDW to estimate rainfall at the variation of AMDP among the stations (Fig. 3). The
ungauged sites. coastal watersheds, Coastal Constantine and Coastal Algiers

Fig. 3  Average, maximum,

and minimum values of annual 250
Average (mm) Max (mm) Min (mm)
maximum daily precipitation
(1970–2014) for different sta-
tions 200
Precipitation (mm)

150

100

50

0
0 20 40 60 80 100 120 140 160 180
Station code number

Fig. 4  Standard deviation, coef-

ficient skewness, and coefficient 40.0 5.0
Standard Deviation (mm) Coef. Skewness Coef. Variation
variation of annual maximum
daily precipitation (1970–2014) 35.0 4.0
Coef. Skewness & Coef. Variation

30.0
Standard deviation (%)

3.0
25.0
2.0
20.0
1.0
15.0
0.0
10.0

5.0 -1.0

0.0 -2.0
0 20 40 60 80 100 120 140 160 180
Station code number

13
1418 M. Benaini et al.

of the Est, and the coastal part of other exoreic watersheds, empirical P-value. On the other hand, LN2 and LN3 distri-
Soummam (S179–S180), Kebir Rhumel (S119–S123), and bution function overestimated the AMDP values.
Seybouse (S152–S156) always had higher average AMDP The best or the suitable distribution function for 95% of
(50–100 mm) than the interior watersheds: Chott El Hoddna, data series was the GEV distribution. In addition, graphi-
Constantine High Plateaus, and Medjerdah with < 50 mm of cal assessment confirmed that there was an adjustment with
average AMDP. Going south, towards the watersheds (e.g., GEV distribution for all precipitation series, as shown for
Chott El Hoddna, Medjerdah) with increasingly drier cli- sample stations, i.e., Bouati Mahmoud (S42), Draa El Hadjar
mate, the standard deviation becomes smaller, as there are (S54), Baiou (S93), and Gastel (S138) (Fig. 5).
no large values of the AMDP, so the fluctuation around the According to the graph in Fig. 5, the GEV distribution
mean decreases (Fig. 4). The values of the standard devia- provided a good fit as confirmed by the results of differ-
tion considerably vary among the stations because of the ent goodness-of-fit tests. For the sample stations of Bouati
distinct weather conditions among the watersheds. Mahmoud (S42), the values of χ2, D, and A2 were 1.2, 0.071,
The skewness coefficient for most of the AMDP series and 0.071, respectively, with respective probabilities of 0.95,
(90%) was positive, which means that the distribution 0.96, and 0.27. For the station of Gastel (S138), the values of
extended to the right for these series (Fig. 4). The high val- χ2, D, and A2 were 2.0, 0.077, and 0.16, respectively, with
ues of the coefficient of variation oscillating between 1 and respective probabilities of 0.86, 0.94, and 0.092.
2 confirmed the large variation of the annual maximum daily The results of goodness-of-fit tests confirmed the graphi-
precipitation (Fig. 4). cal analysis of each data series (Table 3). The results of
AIC and BIC adjustment criteria obtained for different dis-
3.2 Independence, stationarity, and homogeneity tribution functions for some sample stations analyzed are
analysis presented in Table 4. To facilitate the discussion and com-
parison, the stations with the same χ2 value for two or three
The purpose of Table 3 was to present the results of verifica- distribution functions were selected and tabulated (Table 4).
tion tests of independence, stationarity, and homogeneity for The goodness-of-fit criteria tests (AIC and BIC) obtained
the AMDP series used in the frequency analysis of precipi- for all the data series confirmed the results of χ2 test and
tation with different significance levels (P). The tests were graphical analysis. Table 4 displays a representative station
accepted at the 5% significance level. The results indicated each watershed with an identical χ2 value for the number of
that the majority of the series (84.44%) were independent, functions being two or more. The GEV distribution showed
83.33% were stationary, and 87.22% were homogeneous at a clear dominance compared to the other distributions, and
the 5% significance level. The remaining AMDP series were this was due to the low value of the adjustment criteria. For
independent, stationary, and homogeneous at the 1% sig- example, the station of Boghni (S12) had two distribution
nificance level. These results demonstrate that the majority functions with the same χ2 value (2.8); the lowest values
of the AMDP series met the necessary conditions for use in of AIC and BIC were, respectively, 401.9 and 396.4 for the
frequency analysis, ensuring the reliability and accuracy of GEV distribution. Also for the station of Ouenza (S135),
the subsequent calculations. there were three distributions with the same χ2 value (7.6),
and the lowest values of AIC and BIC were, respectively,
3.3 Goodness‑of‑fit test 452.6 and 347.2 for the GEV distribution.

In addition to goodness-of-fit tests, graphical visual exami- 3.4 Quantile estimation and mapping
nation was used to identify the best-fit probability distribu-
tion function and the number of series that are fitted by each After applying various goodness-of-fit tests to determine
one. Based on the best strong empirical probability value the best-fitted probability distribution for each data series,
(P-value) and the test parameter values for the χ2, KS, and the AMDP quantiles were computed and mapped using
AD tests at significance level of 0.05, the results show that the selected distributions, including GEV, GM, LN2, LN3,
the GM and P3 function predicted the data with a strong and P3. The spatial variability of precipitation amounts

Table 3  Verification results Tests Independence Stationarity Homogeneity

of independence, stationarity,
and homogeneity probabilities P (%) No (%) No (%) No (%)
by number and percentage of
stations P>5 152 84.44 150 83.33 157 87.22
1<P<5 23 12.78 22 12.22 16 8.89
P<1 5 2.78 8 4.44 7 3.89

13
Frequency analysis of annual maximum daily precipitation in northeastern Algeria: mapping… 1419

Fig. 5  Graphical comparison of annual maximum daily precipitation series between 1970 and 2014 adjustments to statistical distribution func-
tion for Bouati Mahmoud station (S42), Draa El Hadjar (S54), Baiou (93) and Gastel station (S138)

Table 4  AIC and BIC criteria for function adjustment classification showed consistent patterns across different return periods,
Station code Station name χ2 Distribution BIC AIC
as illustrated in Fig. 6. The northern part of the study area,
Function particularly the coastal strip along the Mediterranean Sea,
exhibited higher quantile values compared to other regions.
S12 Boghni 2.8 GEV 401.9 396.4 Among the monitored stations, Erraguene (S26) exhibited
LN3 401.9 396.5
the highest quantile values. For return periods of 5, 10, 20,
S27 Col de Fedoules 9.6 GEV 438.7 435.5
and 50 years, the corresponding quantiles were 136 mm,
GM 441.9 436.5
159 mm, 181 mm, and 209 mm, respectively. However, for
S83 Ain Melh 6.8 GEV 324.8 321.2
longer return periods of 100 and 1000 years, the station of
GM 328.5 323.1
Beni Aziz (S104) recorded the largest quantiles, measuring
LN3 328.5 323.1
274 mm and 1094 mm, respectively.
S85 Ain Azel 5.6 GEV 360.5 355.1
Notably, the highlands region served as a transitional zone
LN3 360.9 355.5
between the highest quantile values observed in the humid
S110 Boumalek M.C 2.4 GEV 352.6 347.2
climate of the north and the lowest quantile values in the
LN3 352.7 347.2
arid climate of the south. The variation of quantiles among
P3 352.7 347.2
the stations and for different return periods was as follows:
S135 Ouenza 7.6 GEV 352.6 347.2
Bordj Bouarreridj station (S67) recorded 42 mm for a return
LN3 352.7 347.2
period of 5 years, Timgad station (S95) recorded 50 mm for
P3 352.7 347.3
a return period of 10 years, Ain Azal station (S85) recorded
S152 Nechmeya 5.6 GEV 391.5 386.1
67 mm for a return period of 20 years, Tebesssa station
LN3 391.6 386.2
(S131) recorded 85 mm for a return period of 50 years, and
P3 391.6 386.2
Constantine station (S112) recorded 118 mm and 174 mm
S177 Zemmorah 0.4 GEV 370.9 365.4
for return periods of 100 and 1000 years, respectively. Mov-
LN3 370.9 365.5
ing towards the south, specifically in the extreme southwest

13
1420 M. Benaini et al.

Fig. 6  Quantile mapping of annual maximum daily precipitation for all the return periods (T = 05, 10, 20, 50, 100, and 1000 years) in northeast-
ern Algeria

of the Chott El Hoddna watershed, the quantiles followed expected to intensify due to the interplay between increased
the climatic pattern and decreased for all return periods. The AMDP magnitude (flooding) and heightened stress on water
lowest quantile values were estimated for the stations of Oued resources caused by climate change impacts, such as vari-
Chair (S84) and Ain Rich (S82), located in the far south of ations in precipitation patterns, surface water availability,
the El Hoddna watershed, ranging from 18 mm for a return groundwater resources, and recharge rates. Based on the
period of 5 years to 54 mm for a return period of 1000 years. findings of this study, it is recommended to prioritize addi-
The spatial and temporal distribution of precipitation tional investments in water resources infrastructure and
exhibited considerable heterogeneity, and this complexity is management to ensure water security for various sectors

13
Frequency analysis of annual maximum daily precipitation in northeastern Algeria: mapping… 1421

including agriculture and urban needs. Addressing these Declarations

challenges will be particularly demanding considering the
anticipated population growth and the associated demands Conflict of interest The authors declare no competing interests.
for water, agriculture, and food production. Future planning
and strategic actions are essential to effectively manage and
sustainably meet these growing water-related demands. References
Ahammed F, Hewa GA, Argue RJ (2014) Variability of annual daily
maximum rainfall of Dhaka, Bangladesh. Atmos Res 137:176–
4 Conclusions 182. https://​doi.​org/​10.​1016/j.​atmos​res.​2013.​10.​013
Akaike H (1974) A new look at the statistical model identification.
Based on comprehensive analysis and the findings pre- IEEE Transact Auto Control 19(6):716–723. https://​doi.​org/​10.​
sented in this study, it can be concluded that the general- 1109/​TAC.​1974.​11007​05
Alam MA, Emura K, Farnham C, Yuan J (2018) Best-fit probability
ized extreme value (GEV) distribution provides a well-fitted distributions and return periods for maximum monthly rainfall
model for approximately 95% of the annual maximum daily in Bangladesh. Clim 6:1–16. https://​doi.​org/​10.​3390/​cli60​10009
precipitation (AMDP) data series in northeastern Algeria. Anderson TW, Darling DA (1952) Asymptotic theory of certain “good-
The AMDP values ranged from 60 to 150 mm, with higher ness-of-fit” criteria based on stochastic processes. Ann Math Stat
23:193–212
magnitudes observed in humid zones and lower values in ANRH (National Agency for Water Resources) (2005) Map of the
arid regions. The majority of data series, accounting for over hydroclimatological network and water quality monitoring in
83%, demonstrated independence, stationarity, and homoge- Algeria [Map]. Scale 1:500,000. Algiers, Algeria: ANRH
neity, indicating a favorable level of data quality. The agree- Bajirao TS (2021) Comparative performance of different probability
distribution functions for maximum rainfall estimation at differ-
ment between the goodness-of-fit tests and the graphical ent time scales. Arab J Geosci 14:2138. https://​doi.​org/​10.​1007/​
analysis further validated the suitability of the GEV distri- s12517-​021-​08580-4
bution for modeling AMDP in the region. The spatial vari- Beaudet G, Marre A (1988) Géomorphologie du Tell oriental algérien.
ability of precipitation amounts exhibited consistent patterns Méditerranée, Troisième Série, Tome 63:68–70
Beck F, Bárdossy A, Seidel J, Müller T, Sanchis EF, Hauser A (2015)
across different return periods, providing a reliable represen- Statistical analysis of sub-daily precipitation extremes in Singa-
tation of the regional hydrological characteristics. Notably, pore. J Hydrol: Region Stud 3:337–358. https://d​ oi.o​ rg/1​ 0.1​ 016/j.​
the climate variability within the watersheds significantly ejrh.​2015.​02.​001
influenced the AMDP values, with humid regions facing a Beirlant J, Goegebeur Y, Segers J, Teugels JL (2004) Statistics of
extremes: theory and applications, vol 558. John Wiley &Sons,
higher risk of experiencing extreme precipitation events in Ltd., Wiltshire, UK
the future compared to drier regions. These findings high- Bella N, Dridi H, Kalla M (2020) Statistical modeling of annual maxi-
light the importance of considering climate change as a criti- mum precipitation in Oued El Gourzi Watershed. Algeria Appl
cal issue for future generations. The frequency analysis of Water Sci 10:94. https://​doi.​org/​10.​1007/​s13201-​020-​1175-6
Benabdesselam T, Amarchi H (2013) Approche régionale pour
AMDP can serve as a valuable tool in identifying hydrologi- l’estimation des précipitations journalieres extremes du nord est
cal risk zones and guiding the planning, construction, and Algérien [Regional approach for estimating extreme daily rainfall
maintenance of hydraulic projects in northeastern Algeria. in northeastern Algeria]. Courrier Du Savoir 17:175–184
Proactive measures, including enhanced water resources Benaini M (2015) Etude de la variabilité spatio-temporelle des
précipitations dans le Nord-Est d’Algérie [Study of the spatio-
management, infrastructure development, and mitigation temporal variability of precipitation in the North-East of Alge-
strategies, are essential to address the challenges posed by ria]. Université de hassiba Benbouali de Chlef, Mémoire de
climate change and ensure the long-term sustainability and Magistère, p 185
resilience of the region’s water systems. By implementing Benhattab K, Bouvier C, Meddi M (2011) Regional analysis for the
estimation of low-frequency daily rainfalls in Cheliff catch-
these measures, northeastern Algeria can effectively manage ment, Algeria. From prediction to prevention of hydrological
its water resources, meet the needs of a growing population, risk in Mediterranean countries. Proceedings of the MED-
and safeguard agricultural and urban requirements. FRIEND International Workshop on Hydrological Extremes
held at University of Calabria (Italy). pp 55–65
Benhattab K, Bouvier C, Meddi M (2014) Regional frequency anal-
Author contribution MB designed the research, performed the experi- ysis of maximal daily annual rainfalls in Cheliff catchment.
ments and data processing, wrote the manuscript, and contributed to the Algeria Revue Des Sciences De L’eau 27(3):189–203
analysis and discussion of results. MA provided strategic guidance and Beniston M, Stephenson DB, Christensen OB et al (2007) Future
oversight of the project. MGMA conducted a thorough revision of the extreme events in European climate: an exploration of regional
manuscript and made important editorial suggestions. VPS supervised climate model projections. Clim Change 81:71–95. https://​doi.​
the work, verified the methods used, and provided valuable feedback org/​10.​1007/​s10584-​006-​9226-z
on the results. All authors contributed to the critical evaluation and Boucefiane A, Meddi M (2020) Regional growth curves and extreme
shaping of the research, analysis, and manuscript. precipitation events estimation in the steppe area of northwest-
ern Algeria. Atmósfera 32(4):287–303. https://​d oi.​o rg/​1 0.​
Data availability Not applicable. 20937/​atm.​2019.​32.​04.​03

13
1422 M. Benaini et al.

Bousquet N, Bernardara P (2021) Extreme value theory with appli- Gumbel EJ (1941) The return period of flood flows. The Annals Math-
cations to natural hazards. Springer International Publishing ematical Statistics 12(2):163–190. https://d​ oi.o​ rg/1​ 0.1​ 214/a​ oms/​
Burnham KP, Anderson D (2002) Model selection and multi-model 11777​31747
inference. Springer, New York Haan CT (1994) Statistical methods in hydrology. Affiliated East-West
Chen PC, Wang YH, You GJY, Wei CC (2017) Comparison of meth- Press Pvt. Ltd., New Delhi, India
ods for non-stationary hydrologic frequency analysis: case Habibi B, Meddi M, Boucefiane A (2012) Analyse fréquentielle des
study using annual maximum daily precipitation in Taiwan. pluies journalières maximales Cas du Bassin Chott-Chergui.
J Hydrol 545:197–211. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2016.​ Nature & Technologie 8:41–48
12.​001 Hailegeorgis TT, Thorolfsson ST, Alfredsen K (2013) Regional frequency
Chen L, Singh VP, Huang K (2018) Bayesian technique for the selec- analysis of extreme precipitation with consideration of uncertainties
tion of probability distributions for frequency analyses of hydro- to update IDF curves for the city of Trondheim. J Hydrol 498:305–
meteorological extremes. Entropy 20(2):117. https://​doi.​org/​10.​ 318. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2013.​06.​019
3390/​e2002​0117 Hamed K, Rao AR (eds) (2019) Flood frequency analysis. CRC press
Chen MZ, Papadikis K, Jun CY (2021) An investigation on the non-sta- Hebal A, Remini B (2011) Choice of the most appropriate frequency
tionarity of flood frequency across the UK. J Hydrol 597:126309. model for estimating extreme flood values (case of northern
https://​doi.​org/​10.​1016/j.​jhydr​ol.​2021.​126309 Algeria. Can J Civ Eng 38(8):881–892. https://​doi.​org/​10.​1139/​
Chen Z, & Miao F (2012) Interval and point estimators for the loca- l11-​067
tion parameter of the three-parameter lognormal distribution. Hirabayashi Y, Alifu H, Yamazaki D et al (2021) Anthropogenic cli-
J Quality Reliabil Engineering, 2012. https://​doi.​org/​10.​1155/​ mate change has changed frequency of past flood during 2010–
2012/​897106 2013. Prog Earth Planet Sci 8(36):1–9. https://​doi.​org/​10.​1186/​
Coles SG, Tawn JA (1996) A Bayesian analysis of extreme rainfall s40645-​021-​00431-w
data. J R Stat Soc: Ser C: Appl Stat 45(4):463–478. https://​doi.​ Hosking JRM, Wallis JR (1987) Parameter and quantile estimation for
org/​10.​2307/​29860​68 generalized Pareto distribution. Technometrics 29(3):339–349.
Coles S, Bawa J, Trenner L, Dorazio P (2001) An introduction to sta- https://​doi.​org/​10.​1080/​00401​706.​1987.​10488​243
tistical modeling of extreme values, vol 208. Springer, London, Hussain MS, Lee S (2013) The regional and the seasonal variability of
p 208 extreme precipitation trends in Pakistan. Asia-Pacific J Atmos
Cunderlik JM, Ouarda TBMJ (2009) Trends in the timing and magni- Sci 49:421–441. https://​doi.​org/​10.​1007/​s13143-​013-​0039-5
tude of floods in Canada. J Hydrol 375:471–480. https://​doi.​org/​ Hussain Z, Khan IR, Nisar M et al (2021) Frequency analysis of annual
10.​1016/j.​jhydr​ol.​2009.​06.​050 maximum rainfall series of fifteen meteorological observatories
Dad S, Benabdesselam T (2018) Regional frequency analysis of of Sindh. Pakistan Arab J Geosci 14(749):1–9. https://​doi.​org/​
extreme precipitation in northeastern Algeria. J Water Land Dev 10.​1007/​s12517-​021-​06981-z
39:27–37. https://​doi.​org/​10.​2478/​jwld-​2018-​0056 Justel A, Peña D, Zamar R (1997) A multivariate Kolmogorov-Smirnov
Dalrymple T (1960) Flood-frequency analyses (No. 1543). US Govern- test of goodness of fit. Statist Probab Lett 35(3):251–259. https://​
ment Printing Office doi.​org/​10.​1016/​S0167-​7152(97)​00020-5
Dirks KN, Hay JE, Stow CD, Harris D (1998) High-resolution studies Katz RW, Brown BG (1992) Extreme events in a changing climate:
of rainfall on Norfolk Island: Part II: Interpolation of rainfall variability is more important than averages. Clim Change
data. J Hydrol 208(3–4):187–193. https://d​ oi.o​ rg/1​ 0.1​ 016/S​ 0022-​ 21(3):289–302. https://​doi.​org/​10.​1007/​BF001​39728
1694(98)​00155-3 Katz RW, Parlange MB, Naveau P (2002) Statistics of extremes in
El Adlouni S, Bobée B, Ouarda T (2008) On the tails of extreme event hydrology. Adv Water Resour 25(1287):1304. https://​doi.​org/​10.​
distributions. J Hydrol 355:16–33. https://d​ oi.o​ rg/1​ 0.1​ 016/j.j​ hydr​ 1016/​S0309-​1708(02)​00056-8
ol.​2008.​02.​011 Keggenhoff I, Elizbarashvili M, Amiri-Farahani A, King L (2014)
Flowers-Cano RS, Ortiz-Gómez R (2021) Comparison of four methods Trends in daily temperature and precipitation extremes over
to select the best probability distribution for frequency analysis Georgia, 1971–2010. Weather Clim Extrem 4:75–85. https://​
of annual maximum precipitation using Monte Carlo simulations. doi.​org/​10.​1016/j.​wace.​2014.​05.​001
Theor Appl Climatol 145:1177–1192. https://​doi.​org/​10.​1007/​ Kite GW (1975) Confidence limits for design events. Water Resour Res
s00704-​021-​03683-0 11(1):48–53. https://​doi.​org/​10.​1029/​WR011​i001p​00048
Frich P, Alexander LV, Della-Marta P, Gleason B et al (2002) Observed Kondratieva T, Amarchi H (2015) Régionalisation des précipitations
coherent changes in climatic extremes during the second half of journalières extrêmes : cas de la région située au Nord-Est de
the twentieth century. Clim Res 19(3):193–212. https://​doi.​org/​ l’Algérie [Regionalization of extreme daily precipitation: case
10.​3354/​cr019​193 of the region located in the North-East of Algeria]. Hydrol Sci
Gado TA, Salama AM, Zeidan BA (2021) Selection of the best prob- J 60(3):1–10. https://​doi.​org/​10.​1080/​02626​667.​2014.​988154
ability models for daily annual maximum rainfalls in Egypt. Kousar S, Khan AR, Ul Hassan M, Noreen Z, Bhatti SH (2020) Some
Theor Appl Climatol 144(3):1267–1284. https://d​ oi.o​ rg/1​ 0.1​ 007/​ best-fit probability distributions for at-site flood frequency anal-
s00704-​021-​03594-0 ysis of the Ume River. J Flood Risk Management 13:e12640.
Gao L, Huang J, Chen XW, Chen Y, Liu MB (2018) Contributions https://​doi.​org/​10.​1111/​jfr3.​12640
of natural climate changes and human activities to the trend of Koutsoyiannis D, Baloutsos G (2000) Analysis of a long record of
extreme precipitation. Atmos Res 205:60–69. https://​doi.​org/​10.​ annual maximum rainfall in Athens, Greece, and design rainfall
1016/j.​atmos​res.​2018.​02.​006 inferences. Nat Hazards 22:29–48. https://​doi.​org/​10.​1023/A:​
Garijo C, Mediero L (2018) Influence of climate change on flood 10080​01312​219
magnitude and seasonality in the Arga River catchment in Kumar V, Shanu, Jahangeer (2017) Statistical distribution of rainfall
Spain. Acta Geophys 66:769–790. https://​doi.​org/​10.​1007/​ in Uttarakhand, India. Appl Water Sci 7:4765–4776. https://​doi.​
s11600-​018-​0143-0 org/​10.​1007/​s13201-​017-​0586-5
Goula BTA, Soro EG, Kouassi W, Srohourou B (2012) Tendances Laio F (2004) Cramer–von Mises and Anderson‐Darling goodness of
et ruptures au niveau des pluies journalières extrêmes en Côte fit tests for extreme value distributions with unknown parameters.
d’Ivoire (Afrique de l’Ouest). Hydrol Sci J 57(6):1067–1080. Water Resources Research, 40(9). https://d​ oi.o​ rg/1​ 0.1​ 029/2​ 004W​
https://​doi.​org/​10.​1080/​02626​667.​2012.​692880 R0032​04

13
Frequency analysis of annual maximum daily precipitation in northeastern Algeria: mapping… 1423

Leclerc M, Ouarda TB (2007) Non-stationary regional flood frequency Noto LV, La Loggia G (2008) Use of L-moments approach for regional
analysis at ungauged sites. J Hydrol 343:254–265. https://d​ oi.o​ rg/​ flood frequency analysis in Sicily, Italy. Water Resour Manag
10.​1016/j.​jhydr​ol.​2007.​06.​021 23:2207–2229. https://​doi.​org/​10.​1007/​s11269-​008-​9378-x
Liu D, Guo S, Lian Y, Xiong L, Chen X (2015) Climate-informed low- Nyatuame M, Agodzo S (2017) Analysis of extreme rainfall events
flow frequency analysis using nonstationary modelling. Hydrol (drought and flood) over Tordzie Watershed in the Volta Region
Process 29(9):2112–2124. https://​doi.​org/​10.​1002/​hyp.​10360 of Ghana. J Geosci Environ Prot 5:275–295. https://​doi.​org/​10.​
Ly S, Charles C, Degré A (2011) Geostatistical interpolation of 4236/​gep.​2017.​59019
daily rainfall at catchment scale: the use of several variogram Ojha CSP, Berndtsson R, Bhunya P (2008) Engineering hydrology.
models in the Ourthe and Ambleve catchments, Belgium. Oxford University Press, Oxford, p 440
Hydrol Earth Syst Sci 15:2259–2274. https://​doi.​org/​10.​5194/​ Ondo JC, Ouarda TB, Bobée B (1997) Bibliographic review of homo-
hess-​15-​2259-​2011 geneity and independence tests [Revue bibliographique des tests
Mailhot A, Kingumbi A, Talbot G, Poulin A (2010) Future changes d’homogénéité et d’indépendance]
in intensity and seasonal pattern of occurrence of daily and Onibon H, Ouarda TB, Rarbet M, St-Hilaire A, Bobee B, Bruneau P
multi-day annual maximum precipitation over Canada. J Hydrol (2004) Analyse fréquentielle régionale des précipitations jour-
388:173–185. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2010.​04.​038 nalières maximales annuelles au Québec. Canada Hydrol Sci J
Maity R, Maity (2018) Statistical methods in hydrology and hydrocli- 49(4):717–735. https://​doi.​org/​10.​1623/​hysj.​49.4.​717.​54421
matology, vol 555. Springer, Singapore Patel NR, Shete DT (2015) Analyzing precipitation using concentration
Malamud BD, Turcotte DL (2006) The applicability of power law fre- indices for North Gujarat Agro Climatic Zone, India. Aquatic
quency statistics to floods. J Hydrol 322:168–180. https://d​ oi.o​ rg/​ Procedia 4:917–924. https://d​ oi.o​ rg/1​ 0.1​ 016/j.a​ qpro.2​ 015.0​ 2.1​ 15
10.​1016/j.​jhydr​ol.​2005.​02.​032 Pombo S, de Oliveira RP (2015) Evaluation of extreme precipitation
Malekinezhad H, Zare-Garizi A (2014) Regional frequency analysis estimates from TRMM in Angola. J Hydrol 523:663–679. https://​
of daily rainfall extremes using L-moments approach. Atmósfera doi.​org/​10.​1016/j.​jhydr​ol.​2015.​02.​014
27(4):411–427. https://d​ oi.o​ rg/1​ 0.1​ 016/S
​ 0187-6​ 236(14)7​ 0039-6 Rahman MM, Sarkar S, Najafi MR, Rai RK (2013) Regional extreme
Mamoon A, Rahman A (2016) Selection of the best fit probability rainfall mapping for Bangladesh using L-moment technique. J
distribution in rainfall frequency analysis for Qatar. Nat Hazards Hydrol Eng 18(5):603–615. https://​doi.​org/​10.​1061/​(ASCE)​HE.​
Springer, Netherlands. 86(1):281–296. https://​doi.​org/​10.​1007/​ 1943-​5584.​00006​63
s11069-​016-​2687-0 Rizwan M, Anjum L, Mehmood Q et al (2023) Daily maximum rainfall
Mebarki A (2002) Contributions of river flows and hydrological bal- estimation by best-fit probability distribution in the source region
ances in the Eastern Algerian basins [Apports des cours d’eau of Indus River. Theor Appl Climatol 151:1171–1183. https://d​ oi.​
et bilans hydrologiques des bassins de l’est Algérien]. Bull des org/​10.​1007/​s00704-​022-​04334-8
Sci Géograph: 46–54 Rousseau AN, Klein IM, Freudiger D, Gagnon P, Frigon A, Ratté-
Mebarki A (2005) Hydrology of Eastern Algerian basins: water Fortin C (2014) Development of a methodology to evaluate
resources, development, and environment [Hydrologie des Bas- probable maximum precipitation (PMP) under changing climate
sins de l’est Algérien : Ressources en Eau, Aménagement et conditions: application to southern Quebec, Canada. J Hydrol
Environnement]. doctoral dissertation. University Salah-Men- 519:3094–3109. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2014.​10.​053
touri of Constantine Rousta I, Soltani M, Zhou W, Cheung H (2016) Analysis of extreme
Meddi M, Toumi S (2015) Spatial variability and cartography of maxi- precipitation events over Central Plateau of Iran. Amer J Clim
mum annual daily rainfall under different return periods in the Chan 05:297–313. https://​doi.​org/​10.​4236/​ajcc.​2016.​53024
North of Algeria. J Mt Sci 12:1403–1421. https://​doi.​org/​10.​ Rulfová Z, Buishand A, Roth M, Kyselý J (2016) A two-component
1007/​s11629-​014-​3084-3 generalized extreme value distribution for precipitation fre-
Melesse A, Abtew W, Dessalegne T, Wang X (2010) Low and high quency analysis. J Hydrol 534:659–668. https://​doi.​org/​10.​
flow analyses and wavelet application for characterization of the 1016/j.​jhydr​ol.​2016.​01.​03
Blue Nile River system. Hydrol Processes: an Int J 24(3):241– Sabarish RM, Narasimhan R, Chandhru AR, Suribabu CR, Sudharsan
252. https://​doi.​org/​10.​1002/​hyp.​7312 J, Nithiyanantham S (2017) Probability analysis for consecutive
Merzougui A, Zghibi A (2020) Characterization of homogeneous day maximum rainfall for Tiruchirapalli City (south India, Asia).
regions for regional frequency analysis of heavy daily precipita- Appl Water Sci 7:1033–1042
tion in central Tunisia. Arab J Geosci 13:1141. https://​doi.​org/​ Salas JD, Obeysekera J (2014) Revisiting the concepts of return period
10.​1007/​s12517-​020-​06151-7 and risk for nonstationary hydrologic extreme events. J Hydrol Eng
Meylan P, Favre AC, Musy A (2012) Predictive hydrology: a frequency 19(3):554–568. https://​doi.​org/​10.​1061/​(asce)​he.​1943-​5584.​00008​20
analysis approach. CRC Press Salinas JL, Castellarin A, Kohnová S, Kjeldsen TR (2014) Regional
Murray V, Ebi KL (2012) IPCC special report on managing the risks of parent flood frequency distributions in Europe–part 2: climate
extreme events and disasters to advance climate change adapta- and scale controls. Hydrol Earth Syst Sci 18(11):4391–4401.
tion (SREX). J Epidemiol Commun Health 66(9):759–760 https://​doi.​org/​10.​5194/​hess-​18-​4391-​2014
Naghavi B, Yu FX, Singh VP (1993) Comparative evaluation of fre- Sangal BP, Biswas AK (1970) The 3-parameter lognormal distribution
quency distributions for Louisiana extreme rainfall 1. JAWRA and its applications in hydrology. Water Resour Res 6(2):505–515.
J Am Water Res Ass 29(2):211–219. https://​doi.​org/​10.​1111/j.​ https://​doi.​org/​10.​1029/​WR006​i002p​00505
1752-​1688.​1993.​tb032​02.x Santos M, Fragoso M, Santos JA (2017) Regionalization and suscep-
Nasri B, Bouezmarni T, Ouarda TBMJ (2017) Non-stationary hydro- tibility assessment to daily precipitation extremes in mainland
logic frequency analysis using B-spline quantile regression. J Portugal. Appl Geog 86:128–138. https://​doi.​org/​10.​1016/j.​
Hydrol 554:532–544. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2017.​09.​ apgeog.​2017.​06.​020
035 Sarfaraz Q, Masood M, Shakir AS, Sarwar MK, Khan NM, Azhar
Norbiato D, Borga M, Sangat M, Zanon F (2007) Regional frequency AH (2021) Flood frequency analysis of River Swat using easyfit
analysis of extreme precipitation in the eastern Italian Alps and model & statistical approach. Pak J Engg Appl Sci 29:8–21
the August 29, 2003 flash flood. J Hydrol 345(3-4):149–166. Schervish MJ, DeGroot MH (2012) Probability and statistics. Pearson
https://​doi.​org/​10.​1016/j.​jhydr​ol.​2007.​07.​009 Education

13
1424 M. Benaini et al.

Schwarz GE (1978) Estimating the dimension of a model. Ann Stat Taibi S, Meddi M, Mahé G (2015) Evolution des pluies extrêmes dans
6(2):461–464 le bassin du Chéliff (Algérie) au cours des 40 dernières années
Sen Roy S, Balling RC (2004) Trends in extreme daily precipitation 1971–2010. Proc IAHS 369:175–180. https://​doi.​org/​10.​5194/​
indices in India. Int J Climatol J R Meteorol Soc 24:457–466. piahs-​369-​175-​2015
https://​doi.​org/​10.​1002/​joc.​995 Teegavarapu R (2012) Floods in a changing climate: extreme precipi-
Shamir E, Georgakakos KP, Murphy MJ (2013) Frequency analysis tation (International Hydrology Series). Cambridge University
of the 7–8 December 2010 extreme precipitation in the Panama Press, Cambridge
Canal Watershed. J Hydrol 480:136–148. https://​doi.​org/​10.​ Villarini G, Serinaldi F, Smith JA, Krajewski WF (2009) On the sta-
1016/j.​jhydr​ol.​2012.​12.​010 tionarity of annual flood peaks in the continental United States
Shao Y, Wu J, Li M (2017) Study on quantile estimates of extreme pre- during the 20th century. Water Resour Res 45:W08417. https://​
cipitation and their spatiotemporal consistency adjustment over doi.​org/​10.​1029/​2008W​R0076​45
the Huaihe River basin. Theor Appl Climatol 127(1):495–511. Yang T, Xu CY, Shao QX, Chen X (2010) Regional flood frequency
https://​doi.​org/​10.​1007/​s00704-​016-​1940-5 and spatial patterns analysis in the Pearl River Delta region using
Shin H, Jung Y, Jeong C, Heo JH (2012) Assessment of modified L-moments approach. Stoch Environ Res Risk Assess 24(2):165–
Anderson-Darling test statistics for the generalized extreme value 182. https://​doi.​org/​10.​1007/​s00477-​009-​0308-0
and generalized logistic distributions. Stoch Env Res Risk Assess Yang X, Xie X, Liu DL, Ji F, Wang L (2015) Spatial interpolation
26:105–114. https://​doi.​org/​10.​1007/​s00477-​011-​0463-y of daily rainfall data for local climate impact assessment over
Siddique R, Palmer R (2020) Climate change impacts on local flood greater Sydney region. Advances in Meteorology 2015:1–12.
risks in the U.S. northeast: a case study on the Connecticut and https://​doi.​org/​10.​1155/​2015/​563629
Merrimack River basins. J Am Water Resour Assoc 57(1):75–95. Yuan J, Emura K, Farnham C, Alam MA (2018) Frequency analysis
https://​doi.​org/​10.​1111/​1752-​1688.​12886 of annual maximum hourly precipitation and determination of
Sienz F, Schneidereit A, Blender R, Fraedrich K, Lunkeit F (2010) best fit probability distribution for regions in Japan. Urban Clim
Extreme value statistics for North Atlantic cyclones. Tellus a: 24:276–286. https://​doi.​org/​10.​1016/j.​uclim.​2017.​07.​008
Dyn Meteorol and Oceanog 62(4):347–360. https://​doi.​org/​10.​ Yurekli K, Modarres R, Ozturk F (2009) Regional daily maximum rain-
1111/j.​1600-​0870.​2009.​00449.x fall estimation for Cekerek Watershed by L‐moments. Meteorol
Slakter MJ (1965) A Comparison of the Pearson Chi-Square and Kol- Applic J Forecast Pract Applic Train Tech Modell 16(4):435–
mogorov goodness-of-fit tests with respect to validity. J Am Stat 444. https://​doi.​org/​10.​1002/​met.​139
Assoc 60(311):854–858. https://​doi.​org/​10.​2307/​22832​51 Zakaria ZA, Shabri A (2013) Regional frequency analysis of extreme
Smithers JC, Schulze RE (2001) A methodology for the estimation of rainfalls using partial L moments method. Theor Appl Climatol
short duration design storms in South Africa using a regional 113:83–94. https://​doi.​org/​10.​1007/​s00704-​012-​0763-2
approach based on L-moments. J Hydrol 241(1–2):42–52. https://​ Zhang J, Wu Y (2002) Beta approximation to the distribution of Kol-
doi.​org/​10.​1016/​S0022-​1694(00)​00374-7 mogorov-Smirnov statistic. Ann Inst Stat Math 54:577–584.
Stedinger JR (1980) Fitting log normal distributions to hydrologic https://​doi.​org/​10.​1023/A:​10224​63111​224
data. Water Res Res 16(3):481–490. https://​doi.​org/​10.​1029/​ Zhang H, Chen L, Singh VP (2021) Flood frequency analysis using
wr016​i003p​00481 generalized distributions and entropy-based model selection
Stedinger JR, Vogel RM, Foufoula GE (1993) Frequency analysis of method. J Hydrol 595:125610. https://​doi.​org/​10.​1016/J.​JHYDR​
extreme events. In: Maidment D (ed), Handbook of hydrology. OL.​2020.​125610
New York: McGraw-Hill
Stephens MA (1974) EDF statistics for goodness of fit and some com- Publisher's note Springer Nature remains neutral with regard to
parisons. J Am Stat Assoc 69:730–737 jurisdictional claims in published maps and institutional affiliations.
Szolgay J, Parajka J, Kohnová S, Hlavčová K (2009) Comparison of
mapping approaches of design annual maximum daily precipi- Springer Nature or its licensor (e.g. a society or other partner) holds
tation. Atmos Res 92:289–307. https://​doi.​org/​10.​1016/j.​atmos​ exclusive rights to this article under a publishing agreement with the
res.​2009.​01.​009 author(s) or other rightsholder(s); author self-archiving of the accepted
Tabari H (2020) Climate change impact on flood and extreme precipi- manuscript version of this article is solely governed by the terms of
tation increases with water availability. Sci Rep 10(13768):10. such publishing agreement and applicable law.
https://​doi.​org/​10.​1038/​s41598-​020-​70816-2

13