Raindrop Size Distribution and Terminal Velocity For Rainfall Erosivity Studies. A Review

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Journal of Hydrology 576 (2019) 210–228

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Review papers

Raindrop size distribution and terminal velocity for rainfall erosivity studies. T
A review

Maria A. Seriob, Francesco G. Carolloa, Vito Ferrob,
a
Department of Agricultural, Food and Forestry Sciences, University of Palermo, Italy
b
Department of Earth and Marine Science, University of Palermo, Italy

A R T I C LE I N FO A B S T R A C T

This manuscript was handled by Marco Borga, The knowledge of the rainfall drop size distribution (DSD) at the land surface is essential for understanding
Editor-in-Chief, with the assistance of Sergio M. precipitation mechanisms affecting soil erosion processes. Rainfall erosivity is defined as the potential of rain to
Vicente-Serrano, Associate Editor cause erosion and it can be evaluated by rainfall kinetic power, which is determined by DSD and raindrop
Keywords: terminal velocity.
Rainfall erosivity This paper firstly deals with the raindrop terminal velocity estimate. Then the most widely used DSD are
Rainfall kinetic power reviewed highlighting the difference between the raindrop size distribution per unit volume of air and that per
Rainfall intensity unit area and time. The reliability of the available kinetic power-rainfall intensity relationships and their ap-
Raindrop size distribution
plication in several part of the world is discussed, highlighting that the use of rainfall intensity is not sufficient to
Seasonality
determine the rainfall kinetic power everywhere. Finally, the influence of seasonality on both raindrop size
distribution and rainfall energy characteristics is investigated using DSD measurements carried out by an optical
disdrometer placed at Palermo experimental area.

1. Introduction (seasonal and yearly variation) (Yin et al., 2017). The quantitative ex-
pression of energy per unit of rainfall developed by Wischmeier and
The acceleration of soil erosion process through anthropogenic Smith (1978) was based on the work of Laws and Parsons (1943) and
perturbation has severe impacts on soil and environmental quality. Soil Gunn and Kinzer (1949).
erosion reduces the long term soil productivity and it exposes subsoil, The rainfall erosivity factor R, proposed in the Universal Soil Loss
which has often poor qualities for crop establishment and growth. Equation (USLE) and in its Revised (RUSLE) version, is recognized as
Water soil erosion is a process of detachment and transport of soil one of the best indicators for modelling the erosive potential of a
particles due to rainfall and runoff and it is one the main causes of rainstorm (Renard et al., 1991). Regional erosivity maps are used to
landform modelling on earth's surface. At the plot and field scale, when identify areas with high potential rainfall erosivity, thus with high risk
interrill erosion occurs, runoff is a factor affecting the transportability of severe soil erosion (Ferro et al., 1991; Aronica and Ferro, 1997; Ferro
of soil material detached by raindrop impact while runoff is able to et al., 1999; Bonilla and Vidal, 2011; Klik and Konecny, 2013; Panagos
cause both detachment and transport when rill erosion takes place. et al., 2015a,b, 2016a).
Rainfall erosivity is the main property affecting erosion processes in- At the plot and field scales soil loss by water erosion can be related
volving both detachment of soil particles and the subsequent transport to rainfall by different precipitation properties, i.e. rainfall intensity, I
of the detached particles away from the site of detachment (Ferro et al., (mm/h), rainfall momentum, M (N/m2), and rainfall kinetic energy per
1991; Mannaerts and Gabriels, 2000; Salles and Poesen, 2000). unit time and area, named kinetic power, Pn (J/m2 h) (Van Dijk et al.,
According to Wischmeier and Smith (1978), rainfall erosivity is 2002; Carollo and Ferro, 2015; Carollo et al., 2016, 2017, 2018). The
commonly expressed as the product of two factors, the rainfall energy E rainfall momentum, M, represents the pressure or force per unit area,
and the maximum continuous 30-min intensity I30 during the individual provoking the mechanical stress, and thus might be expected to be re-
storm (Nearing et al., 2017). Wischmeier (1959), analyzing approxi- lated to the breakdown of soil aggregates (Rose, 1960). In other words
mately 8000 plot-years of rainfall, runoff and soil loss data, confirmed when the raindrop reaches soil, exerts a knock on the ground area, due
the suitability of EI30 index for fallow and continuous row crop plots. to the charge of momentum amount. This phenomenon provokes the
The suitability was also positively tested at different temporal scales split of the bonds between soil particles, making them available to be


Corresponding author.
E-mail address: vito.ferro@unipa.it (V. Ferro).

https://doi.org/10.1016/j.jhydrol.2019.06.040
Received 6 February 2019; Received in revised form 3 June 2019; Accepted 14 June 2019
Available online 15 June 2019
0022-1694/ © 2019 Elsevier B.V. All rights reserved.
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

transport by overland flow (Carollo et al., 2018). during the years several methods (stain method, flour pellet method; oil
During years several researchers (Rose, 1960; Hudson, 1971; Abd immersion method; photographic method; raindrop spectrograph,
Elbasit et al., 2010; Sanchez-Moreno et al., 2012; Lim et al., 2015) have acoustic disdrometer; optical disdrometer; micro rain radar) have been
investigated to understand if the raindrop impact is more dependent on developed for measuring the drop size distribution of the precipitation
M than Pn, or if these two variables are correlated with each other. Rose at the ground and for assessing and monitoring the rainfall erosivity
(1960) stated that rainfall momentum is a slightly better predictor for (Wiesner, 1895; Bentley, 1904; Lenard, 1904; Mache, 1904; Defant,
soil detachment than kinetic power. For natural rainfall, Hudson (1971) 1905; Fuchs and Petrjanoff, 1937; Laws, 1941; Laws and Parsons, 1943;
demonstrated that momentum and kinetic power present a very similar Gucker, 1949; Bowen and Davidson, 1951; Blanchard, 1952; Courshee
relationship with rainfall intensity and the only question is which and Byass, 1953; Mason and Ramanadham, 1953; Mikirov, 1957;
variable can be more easily and accurately measured (Van Dijk et al., Hudson, 1963; Joss and Waldvogel, 1967; Mutchler, 1967; Hall, 1970;
2002). Other studies (Brodie and Rosewell, 2007; Abd Elbasit et al., Mutchler and Hansen, 1970; Nawaby, 1970; Mutchler, 1971; Mutchler
2010; Sanchez-Moreno et al., 2012), using rainfall simulators, showed and Larson, 1971; Kohl, 1974; Carter et al., 1974; Asseline and
that both kinetic energy and rainfall momentum can be used to predict Valentine, 1978; Donnadieu, 1980; Quinn, 1981; Roels, 1981; Lowe,
soil splash. In particular Brodie and Rosewell (2007), carrying out an 1982; McCooll, 1982; Eigel and Moore, 1983; Hauser et al., 1984;
analysis on rainfall intensity- erosive indices relationships, found that Illingworth and Stevens, 1987; Navas et al., 1990; Kincaid et al., 1996;
the rainfall momentum and Pn could be interchangeable with each Tokay and Short, 1996; Grossklaus et al., 1998; Campos, 1999; Salles
other for the estimation of the rainfall erosivity in soil erosion processes and Poesen, 1999; Löffler-Mang and Joss, 2000; Nešpor et al., 2000;
(Abd Elbasit et al., 2010; Sanchez-Moreno et al., 2012). Using mea- Nystuen, 2001; Reddy and Kozu, 2003; Herngren, 2005; Lanzinger
surements of rainfall Drop Size Distribution (DSD) and raindrop term- et al., 2006; Arnaez et al., 2007; Egodawatta, 2007; Schönhuber et al.,
inal velocity carried out by a laser optical disdrometer (Parsivel), Lim 2007; Salvador et al., 2009; Pérez-Latorre et al., 2010; Asante, 2011;
et al. (2015) stated that the rainfall momentum is the best predictor of Maahn and Kollias, 2012; Parsakhoo et al., 2012; Friedrich et al., 2013;
rainfall erosivity in Korea. Carollo et al. (2018) using measurements of Carollo and Ferro, 2015; Kathiravelu et al., 2016; Carollo et al., 2016).
DSDs carried out by an optical disdrometer installed in two different During the years several analytical forms of the raindrop size
Mediterranean sites (Palermo- Italy, El Telauret-Spain) suggested that parameterization have been proposed as lognormal (LN) (Feingold and
M and Pn referred to the unit volume of rainfall are related each other Levin, 1986), Weibull (W) (Weibull, 1951), exponential (MP) (Marshall
and thus both indices can be used to express the erosive power of the and Palmer, 1948) and gamma (U) (Ulbrich, 1983) distribution. How-
precipitation. ever, at the best of our knowledge, for meteorological and soil erosion
The rainfall kinetic power results from the kinetic power of single purposes, the most widely applied analytical form is the gamma dis-
raindrops that constitute precipitation and varies with the size, velocity tribution (Khrgian et al., 1952; Levin, 1961; Sulakvelidze, 1969;
and impact frequency of the raindrops impacting an area in unit time. Sulakvelidze and Dadali, 1971; Uijlenhoet and Stricker, 1999, Carollo
The rainfall kinetic power is determined by relating drop size dis- et al., 2016).
tribution (DSD) to raindrop terminal velocity and rainfall intensity The knowledge of the DSD allows (i) to understand how the rainfall
(Assouline, 2009). is made up, (ii) to calculate the rainfall kinetic power, Pn, and so (iii) to
The force applied to soil by raindrop impact is related to drop size, characterize energetically the precipitation which plays a fundamental
in fact a very small raindrop striking the soil at a low impact velocity role in soil erosion (Assouline and Mualem, 1989).
exerts very low force on the soil and causes very little erosion regardless The storm kinetic energy is the total value of the kinetic energy
of rainfall intensity (Toy et al., 2002). possessed by all raindrops reaching a unit area of the earth's surface
The measured rainfall DSD at the ground level is the result of a during the rainfall event. Devices that can measure storm kinetic energy
series of physical phenomena (i.e. raindrop collision, coalescence, break are not as commonly used as devices that measure rainfall intensity.
up) that influence raindrop formation and its evolution during its This circumstance favored the use of empirical relationships between
falling process (Assouline and Mualem, 1989). Depending on the cli- rainfall intensity and rainfall kinetic power to estimate the storm ki-
mate, specific location, season, the dominant microphysical rainfall netic energies needed in the prediction of soil loss by empirical soil
processes may be different and then lead to different DSDs (Hachani erosion models. To date many researchers proposed empirical re-
et al., 2017). The rain formation is typically classified microphysically lationships, having different mathematical forms (polynomial, ex-
how warm or cold process. Warm rain formation involves the growth of ponential, logarithmical, and power type) for estimating kinetic power
droplets via collision and different models (List et al., 1987; Hu and by rainfall intensity (Table 1). Van Dijk et al. (2002) critically reviewed
Srivastava, 1995) showed that collision-coalescence and breakup pro- published studies of rainfall intensity- kinetic power relationships,
cesses result in an equilibrium shape to DSD regardless of overall evaluating their applicability for predictive purposes.
raindrop concentration (Munchak et al., 2012). Cold rain formation Since the rainfall kinetic power can be calculated by DSD mea-
occurs with the melting frozen hydrometeors such as snow or hail. surements combined with measured fall velocity or empirical laws
These frozen particles are larger than the raindrops out of which warm linking terminal fall velocity and drop diameter (Salles et al., 2002),
rain forms and melt into correspondingly larger rain drops. Falling this paper firstly deals with the raindrop terminal velocity estimate.
larger raindrops breakup reducing their size, although this process Then the most widely used DSD models are reviewed, highlighting also
depends on the depth of the above-freezing layer and the initial DSD. the difference between the raindrop size distribution per unit volume of
Munchak et al. (2012) highlighted also that besides formation and in- air and the DSD per unit area and time. A review of the reliability of the
ternal processes, external processes such as evaporation and size sorting most commonly used relationships to estimate rainfall kinetic power in
can also influence the DSD. Evaporation preferentially acts on small several part of the world is also presented and discussed. Finally, the
drops, thereby increasing the median volume diameter of the dis- influence of seasonality on DSD and energy characteristics of the pre-
tribution, i.e. the diameter that divides the distribution in two parts of cipitation is also investigated using raindrops size distributions mea-
equal volume. In addiction, the influence of size sorting by wind shear sured by an optical disdrometer at Palermo experimental site.
and turbulence on the DSD depends on the particular situation and may
act to increase or decrease the median volume drop size (Munchak 2. Raindrop terminal velocity
et al., 2012).
Wiesner (1895) was a pioneer in measuring the DSD of natural The raindrop impact velocity depends on both the falling height of
rainfall arriving at the earth’s surface, using a piece of absorbent paper each raindrop, H, and its diameter, D (Davies, 1942; Spilhaus, 1948;
dusted with a water-soluble dye (Van Dijk et al., 2002). However Gunn and Kinzer, 1949; Best, 1950; Kessler and Wilks, 1968; Lui and

211
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Table 1
Pn-I relationships available in literature.
Reference Site Pn-I Relationship Range of rainfall intensity (mm/h)

Hudson (1965) Zimbabwe 29.86 (I-4.29) n.a


Carter et al. (1974) USA 11.32 I + 0.5546 I2- 0.5009 × 10-2 × I-3 + 0.126 × 10-4 I4 1–250
McGregor and Mutchler (1976) USA [27.3 + 21.68 exp(-0.048I)-41.26 exp(-0.072I)]I n.a.
Wischmeier and Smith (1978) USA (11.87 + 8.73 log I)I 0.25–151.8
Kinnell (1981) Zimbabwe 29.2[1–0.89 exp(-0.048I)]I 19–229
Kinnell (1981) Zimbabwe (9.705 + 9.258 log I)I 19–229
Kinnell (1981) Zimbabwe 29.863(I-4.287) 19–229
Park et al. (1980) USA 21.1 I1.156 n.a.
Zanchi and Torri (1980) Italy (9.81 + 11.25 log I)I n.a.
Kinnell (1981) Florida 29.3[1–0.28 exp(-0.018I)]I 1.9–309
Kinnell (1981) Florida (17.124 + 5.229 log I)I 1.9–309
Kinnell (1981) Florida 30.132(I-5.484) 1.9–309
Bollinne et al. (1984) Belgium 12.32 × I + 0.56 × I2 0.27–38.6
Tracy et al. (1984) Arizona 2.1 exp[0.0766 I0.175-2.118 I n.a.
Rosewell (1986) Australia 29[1–0.59 exp(-0.04 I)]I 1–145.9
Rosewell (1986) Australia 26.4[1–0.67 exp(-0.035 I)]I 1–161.2
Rosewell (1986) Australia 24.48 (I-1.253) n.a.
Rosewell (1986) Australia 24.8 (I-1.292) n.a.
Brown and Foster (1987) USA 29[1–0.72exp(-0.05I)]I 0–250
Onaga et al. (1988) Japan (9.81 + 10.6 log I)I n.a.
Brandt (1990) USA (8.95 + 8.44 log I)I n.a
Renard et al. (1991) USA 29[1–0.72 exp(-0.05 I)]I n.a.
Sempere-Torres et al., 1992 France 34 I-190 20–100
Smith and De Veaux, 1992 USA 13 I1.21 n.a.
Coutinho and Tomás, 1995 Portugal 35.9[1–0.56 exp(-0.034I)]I 0.8–103
Cerro et al., 1998 Spain 38.4[1–0.54 exp(-0.029I)]I 1–150
Jayawardena and Rezaur, 2000 China 36.8[1–0.69 exp(-0.038I)]I 0–150
Steiner and Smith, 2000 USA 11 I1.25 n.a.
Usòn and Ramos, 2001 Spain 23.4 I-18 < 20
Van Dijk et al., 2002 Universal 28.3[1–0.52 exp(-0.042 I)]I n.a.
Fornis et al., 2005 Philippines 30.8[1–0.05 exp(-0.03 I)]I 2.8–142
Petan et al., 2010 Slovenia (Koseze) 29.8[1–0.60 exp(-0.07 I)]I 0.1–288
Petan et al., 2010 Slovenia (Kozjane) 31.9[1–0.60 exp(-0.055 I)]I 0.1–220
Sanchez-Moreno et al. (2012) Cape Verde 35[1–0.79 exp(-0.03 I)]I 0–157
Sanchez-Moreno et al. (2012) Cape Verde 8.4 I0.30 0–157
Sanchez-Moreno et al. (2012) Cape Verde (10.09 + 12 log I)I 0–157
Lim et al. (2015) Korea 25.75[1–0.54 exp(-0.05 I)]I 0.1–142

Orville, 1969; Sekhon and Srivastava, 1971; Atlas and Ulbrich, 1977; measured values of the drag coefficient rather questionable (Gunn and
Uplinger, 1981; Leone and Pica, 1993; Grosh, 1996; Ferro, 2001). For a Kinzer, 1949; Ferro, 2001).
drop having a diameter equal to D, the falling raindrop velocity grows At date the relationship between drop size and terminal velocity is
with H until a threshold value equal to 20 m, after which becomes usually described by an exponential or power equation. Moreover all
constant and it is named rainfall terminal velocity, V(D), expressed in m these relationships are calibrated by laboratory data collected for free
s−1. In particular the terminal velocity is reached when there is an falling water drops in still air.
equilibrium between the gravitational force and the raindrop aero- Using data from Gunn and Kinzer (1949), Atlas and Ulbrich (1977)
dynamic resistance. proposed the following power law equation to estimate raindrop
Actually, few terminal velocity measurements of natural raindrops terminal velocity:
are available in literature (Beard, 1976; Jayawardena and Rezaur,
2000) and the relationships between terminal velocity-raindrop dia- V (D) = a0 D a1 (1)
meter were based on measurements of simulated single raindrops that
fall in stagnant air (Lenard, 1904; Schmidt, 1909; Laws, 1941; Gunn in which D is expressed in cm and V in m s−1, and a0, a1 are coefficients
and Kinzer, 1949; Blanchard, 1950; Beard, 1976; Epema and Riezebos, equal to 17.67 and 0.67 respectively. Atlas and Ulbrich (1977) also
1983; Jayawardena and Rezaur, 2000). demonstrated that a power law provides a close fit to the data of Gunn
Theoretically the terminal velocity of a rigid and not deformed and Kinzer (1949) in the range 0.05 ≤ D ≤ 0.50 cm. The power law
sphere having a given diameter and density, that falls freely through a proposed by Sekhon and Srivastava (1971), substituting in Eq. (1) the
medium, i.e. air, having an own density, can be calculated setting the values of a0 = 14.2 and a1 = 0.5, allows to have a reasonable accuracy
gravitational force contribution equal to aerodynamic drag one that the of terminal velocity estimate for drops in the range 0.1–0.3 cm, but
particle meets during its motion of fall (Best, 1950; Park et al., 1983). A overestimates the fall velocity of drop > 0.3 cm and smaller than
large amount of work has been done in wind tunnels for determining 0.1 cm (Van Boxel, 1998). The power law relations, proposed by several
the drag coefficient for rigid and not deformed spheres. However, authors (Spilhaus, 1948; Kessler and Wilks, 1968; Lui and Orville, 1969;
considering natural rainfall a water droplet, falling freely through the Sekhon and Srivastava, 1971; Atlas and Ulbrich, 1977; Grosh, 1996) are
air is deformed by the aerodynamic forces and has been observed to widely applied, even if according to Van Boxel (1998) can be “mostly
vibrate strongly and spin, thus producing definite departures from inaccurate”.
spherical symmetry. For large drops, vibrations and deformations fre- For describing the variation of the terminal velocity with drop
quently break up the droplet if it falls sufficiently far. This turbulence diameter, Best (1950) used measurements of terminal velocities carried
together with the inevitable deformations of a not rigid sphere make the out by Davies (1942) and proposed the following exponential re-
applicability of the theoretical calculation of the terminal velocity with lationship:

212
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Π1 = g α D β ρ ε V (D) (6)
in which α, β, and ε are constant parameters. Substituting in the Eq. (6)
the measurement units of g, D and ρ, the following relationship is ob-
tained:
0 = mα s−2α mβkg ε s−2β m−4ε m s−1 (7)
The α, β, and ε values can be determined solving the following
system of equations:
α + β − 4ε + 1 = 0 (8a)

− 2α + 2ε − 1 = 0 (8b)

ε=0 (8c)
The solution of the system (α = 1/2, β = 1/2, ε = 0) allows to de-
duce the following dimensionless group:
V (D)
Π1 =
gD (9)
Using a similar procedure, the other dimensionless groups are de-
duced:
Fig. 1. Relationship between Π1, Π2 groupings using measurements carried out
H
by Laws (1941) (modified by Ferro, 2001) (D = drop diameter, V = falling Π2 =
D (10)
velocity, H = fall height).
νc
Π3 =
1.147 g1/2D3/2 (11)
D
V (D) = 9.43 ⎧1 − exp ⎡−⎛ ⎞ ⎤⎫
⎨ ⎢
⎣ ⎝ 17.7 ⎠ ⎥
⎦⎬ (2) Va
⎩ ⎭ Π4 =
−1 g1/2D1/2 (12)
where V(D) is the terminal velocity in m s , D is the drop diameter
expressed in cm. Eq. (2) adequately represents Davies’s experimental Considering the Eqs. (11) and (12) the following relationship can be
data for diameters ranging from 0.069 to 0.595 cm (Best, 1950). deduced:
Other researchers (Uplinger, 1981; Leone and Pica, 1993) proposed
Π4 V g1/2 D3/2 V D
the following exponential relationship relating terminal velocity- rain- = 1/2 a 1/2 = a
Π3 g D νc νc (13)
drop diameter:
in which νc is the cinematic viscosity. Finally the Eq. (5) can be
V (D) = α′ D exp(−β′ D) (3)
expressed as follows:
in which the constants α'and β', are equal to 48.74 and 1.95 (Uplinger,
V (D) H VD
1981). Leone and Pica (1993) using raindrop terminal velocity mea- =f⎛ , a ⎞
⎜ ⎟

surements carried out in laboratory by Beard (1976), Laws (1941) and gD ⎝ D νc ⎠ (14)
Gunn and Kinzer (1949) estimated α' = 49 and β' = 2. In stagnant air condition (Va = 0) Eq. (14) can be expressed using
Taking into account that for D < 0.10 cm the drop has a slow the Π1 and Π2 groups. Fig. 1 shows the experimental rainfall velocity
terminal velocity and small mass producing a negligible kinetic power and raindrop measurements carried out by Laws (1941), in the plot
and for D > 0.55 cm the raindrops are unstable and tend to break up H / D , V (D)/ g D . This figure highlights a clear similitude relation for
before reaching their terminal velocity (Pruppacher and Pitter, 1971), all measurements carried out using H ≤ 8 m. However, despite the
Leone and Pica (1993) considered only raindrops having a diameter condition of the perfect similitude between the relationships Π1 = f(Π2)
varying in the range 0.10–0.55 cm. for H ≤ 8 m, this relationship cannot be applied for the estimation of
Cerro et al. (1998) used Eq. (3) with the coefficients proposed by the natural raindrop terminal velocity, for which H must be almost
Uplinger (1981) to terminal velocity estimates in Spain, and found that equal to 20 m.
the measured values of raindrop terminal velocity were lower than Ferro (2001), considering that the terminal velocity increases for
those predicted by Uplinger’s Eq. (3). D < 0.56 cm until which becomes constant, used the measurement
Ferro (2001) suggested that for a raindrop having diameter D and carried out by many researchers (Laws, 1941; Gunn and Kinzer, 1949;
falling with a velocity, V(D), from a fall height, H, through a medium Blanchard, 1950; Beard, 1976; Epema and Riezebos, 1983;
with density ρ and viscosity, νc, and having a velocity Va, this phe- Jayawardena and Rezaur, 2000) for deducing the following relation-
nomenon can be expressed by the following functional relationship: ship (Fig. 2):
F (V (D), D , H , g , νc , ρ , Va) = 0 (4) V (D) = Vmax [1 − exp(−an D)] (15)
−1
where F is the functional symbol and g is the acceleration due to in which Vmax (m s ) and an (cm) parameters vary with H (m)
gravity. Since Eq. (4) describes a physical phenomenon that is in- (Table 2). For calculating raindrop terminal velocity (H = 20 m) Vmax
dependent on the choice of the measurement units of the considered and an, respectively equal to 9.5 m s−1 and 6 cm, have to be applied.
variables, the Π-theorem of dimensional analysis can be applied for This equation was used by Carollo and Ferro (2015), Carollo et al.
expressing Eq. (4) by four dimensionless groups Π1, Π2, Π3, Π4: (2016) and Serio (2017) for estimating the kinetic power by raindrop
size distribution (DSD) in Italy.
Φ (Π1, Π2 , Π3 , Π 4) = 0 (5)
Based on the investigations of Beard and Pruppacher (1969) and
where Φ is a functional symbol. Choosing as independent variables Beard (1976), Van Dijk et al. (2002) used a cubic polynomial function
from dimensional point of view, g, D and ρ, the Π1 grouping assumes the to describe raindrop terminal velocity, V (m s−1) as function of D (mm),
following form: for drop size of 0.1–7.0 mm:

213
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

1989).
The raindrop size distribution (DSD) is defined as the function
stating the expected number of drops with diameters between D and
D + dD. The study of the DSD formation and evolution allowed to
distinguish two different raindrop size distributions referred to unitary
volume of air or to unit area and time.
The distribution Nv(D)dD represents the number of raindrops with
diameters between D and D + dD (cm) per unit volume of air (m3), in
which D is expressed in cm and volume in m3 and the units of Nv(D)
become cm−1 m−3. According to this definition, the DSD refers to the
spatial distribution of raindrops in the air (which governs the raindrop
concentration) and to the probability distribution of raindrop sizes in
the air.
The second form of the raindrop size distribution, N(D)dD, re-
presents the expected number of raindrops with diameters between D
and D + dD (cm) arriving at a surface per unit area (m2) and per unit
time (s). Where D is expressed in cm, area in m2 and time in s, and so
the units N(D) become cm−1 m−2 s−1.
For meteorological studies, the size distribution of raindrops com-
monly refers to the number of drops present in unit volume of air, Nv(D)
Fig. 2. Comparison between the measurements of terminal velocity and drop
dD while for hydrological studies, the DSD usually refers to the number
diameter and Eq. (15) (modified by Ferro, 2001).
of droplets N(D)dD, that reaches a unit horizontal area during a unit
time.
Table 2 Uijlenhoet and Stricker (1999) suggest that Nv(D) yields to know the
Vmax and an parameter values of Eq. (15) versus rain falling height, static properties of the raindrop population, i.e. the spatial size dis-
H.
tribution of the raindrops in a volume of air (which governs the rain-
H [m] Vmax an drop concentration) and the probability distribution of their sizes. In
addition to these static properties, N(D) involves the dynamic proper-
0.5 3.1 14
0.75 3.75 13
ties of the raindrop population as well its velocity distribution.
1 4.31 12 It should be noted that since the study of the DSD has for the most
1.5 5.2 10 part been the work of the meteorologists, they considered the number
2 5.9 9 of drops present in unit volume of air, Nv(D)dD, even though N(D)dD is
2.5 6.46 8
the distribution which is actually measured in most cases (Uijlenhoet
3 6.94 8
4 7.69 7 and Stricker, 1999).
5 8.2 7 Considering a raindrop, constituted by drops that fall freely in air,
6 8.55 7 i.e. neglecting the effects of wind, turbulence and raindrop interaction,
8 9.01 7 the two size distributions are related by the following relationship
20 9.55 6
(Austin, 1987; Hall and Calder, 1993; Uijlenhoet and Stricker, 1999;
Carollo and Ferro, 2015):
V (D) = 0.0000561 D3 − 0.00912 D 2 + 0.503 D − 0.254 (16) N (D) dD = V (D) Nv (D) dD (17)
Angulo-Martínez et al. (2016) using raindrop terminal velocity in which V(D) is the terminal velocity of the drop having a diameter D.
measurements carried out in Spain by an Optical Spectro Pluviometer, For studying accurately, the energetic characteristics of raindrops
found that Eq. (16) “provided a better fit for smaller drop that reach the ground area, the number, N(D)dD, of raindrops with
(0.01 < D < 0.07 cm), whereas” Eq. (1), “provided a better fit for mid- diameters between D and D + dD arriving at a unit area per unit time
size drops (0.1 < D < 0.3 cm). The Uplinger (1981) equation performed has to be determined.
similarly to” Eq. (16), “but it overestimated the raindrop terminal velocity Measurements of raindrop size distributions at the ground carried
values for the smallest drops”. out with different techniques (i.e. photographic method, acoustic, im-
In conclusion, Eq. (15) is capable to reproduce the physical pact and optical disdrometer, micro rain radar) suggested that averaged
boundary condition for D ≥ 0.55 cm (the raindrops are unstable and raindrop size distributions can generally be parameterized as unimodal,
tend to break up) and the agreement with the available measurements positively skewed distributions characterized by few parameters
(Fig. 2) states its best effectiveness in describing the drop velocity to (Uijlenhoet and Stricker, 1999).
drop size relationship. The most widely applied analytical forms of the raindrop size
parameterization are lognormal (LN), Weibull (W), exponential (MP)
3. Rain drop size distributions and gamma (U) distribution.

3.1. Types of DSD 3.2. The lognormal distribution (LN)

The raindrop size distribution (DSD) is an interesting rainfall Several authors (Mueller and Sims, 1966, 1967; Levin, 1971;
property for its scientific implication as well as its engineering aspect Bradley and Stow, 1974; Markowitz, 1976; Bezdek and Solomon, 1983;
(Assouline and Mualem, 1989). The DSD formation is not fully under- Feingold and Levin, 1986; Ochou et al., 2007) proposed a lognormal
stood because of the random and complex nature of the physical phe- distribution (LN) for describing the rain drop size distribution Nv(D) in
nomena that govern the aggregation and disaggregation of the rain- an unit volume of air in different climatic conditions:
drops. However the knowledge of DSD is essential to test theoretical
Nv (D) 1
models and for estimating the rainfall kinetic energy which plays a p (D) = = exp[−ln2 (D / Dm)/2ln2 σ(D) ]
NT 2π ln σ(D) D (18)
fundamental role in soil sealing and erosion (Assouline and Mualem,

214
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

in which p(D) is the probability density function of the drop diameter, 3.3. The Weibull distribution (W)
D, NT is the total number of the drops per unit volume of air, Dm, and
σ(D), represent the geometric mean diameter (cm) and the standard Langmuir (1948) considered the raindrop formation as a chain re-
deviation of the raindrop diameter (cm) respectively. action process in which raindrops seem to grow up to a point where
According to Feingold and Levin (1986) the parameters of log- turbulence, raindrop collision or coalescence cause breakup into two or
normal distribution can be calculated by the following relationships: smaller drops, which in turn, repeat the same process.
∞ Weibull (1951) described this phenomenon assuming to have a
NT = ∫0 N (D) dD (19) chain consisting of n links, the chain as a whole fail if any one of its
links fails. Therefore, denoting as P(f) the probability of failure of the
ln Dm = ln¯D (20) chain, the nonfailure probability of the chain (1- P(f)) is equal to the
probability of the simultaneous nonfailure of all the links (Weibull,
ln2 σ(D) = (ln D −¯ ln Dm )2 (21) 1951), thus:

in which ln¯D is the arithmetic mean of the natural logarithm of the P (f ) = 1 − exp−n f (x ) (26)
diameter, D, and (ln D −¯ ln Dm )2 is the mean of the variable where f(x) is the function of any load, x, applied to a single link. The
(ln D − ln Dm )2 . analogy of the chain with raindrop is due to the break up mechanism. In
Notice that for large values of σ(D) and small values of Dm, the other words, the probability of break up of a chain constituted by n
lognormal function has the concave upward shape in much the same links is equal to the probability of raindrop break up into n parts of it.
way that the gamma distribution does when μ < 0 (Feingold and Sekine and Lind (1982) stated that, if in a unit volume of air a drop
Levin, 1986). breaks up and a series of chain reaction processes occur, the probability
The rainfall intensity can be estimated by the following equation P(D) that raindrop diameter is less than D can be written as:
(Feingold and Levin, 1986):
δ
D ⎡ D v⎤
I = 7.1 × 10−3 NT Dm3.67 exp (6.73 ln2 σ(D) ) (22) 1− ∫0 p (D) dD=1 − P (D) = exp ⎢−⎛ ⎞ ⎥
ω
⎜ ⎟

⎣ ⎝ v⎠ ⎦ (27)
The median volume diameter, D0, that is the diameter that divides
the distribution in two parts of equal volume, is given by (Feingold and where p(D) is the probability density function of diameter D; δv is the
Levin, 1986): shape parameter responsible for the skew of the distribution and ωv
(mm) is the scale parameter.
D0 = Dm exp (3 ln2 σ(D) ) (23) The Weibull shape parameter, δv, is a pure number (i.e. di-
mensionless) and is also named Weibull slope. Different values of the
which underlines that D0, is a scale and shape parameter, as it depends
shape parameter can have marked effects on the behavior of the dis-
on Dm and σ(D). Feingold and Levin (1986), using DSD measurements
tribution. When δv = 1, the P(D) of the Weibull distribution reduces to
carried out at Hadera in Israel by an optical disdrometer, found that:
that of the exponential law proposed by Marshall and Palmer (1948).
When δv = 2, the P(D) of the Weibull distribution reduces to Rayleig
(i) although the value of σ(D) varies considerably for low rain rate, for
distribution (Cohen, 1965). Changing the scale parameter, ωv, it does
I > 5 mm/h is quasi constant and equal to 1.43. This result implies
stretch out the existing shape. In particular with δv constant, the in-
that the ratio D0/Dm = KD is constant and equal to 1.47. According
crease of ωv yields to have the distribution stretched to the right in the
to Markowitz (1976), using Laws and Parsons data, for
graph. Instead decreasing ωv results in the graph being shrunk to the
2.5 < I < 152.4 mm/h the σ(D) parameter remains constant and
left (towards zero).
equal to 1.39;
Applying logarithms to both members of Eq. (27), it follows:
(ii) NT increases with rainfall intensity as follows:
ln{ −ln[1 − P (D))]} = δ v [ln(D) − ln(ω v )] (28)
NT = 172 I 0.22 (24)
Assuming Nv(D), that is the number of the raindrops in an unit
in agreement with Gagin (1980) that proposed an increase NT with I for volume of air, equal to P(D) a linear equation is obtained to present the
convective clouds in Israel; parameters of the Weibull distribution:

(iii) Dm parameter increases with I according to the equation: X = ln(D) (29)

Dm = 0.75 I 0.21 (25)

Feingold and Levin (1986) found that lognormal distribution can be


{
Y = ln −ln ⎡1 −

∫0
D
Nv (D) dD⎤
⎦ } (30)
fitted to observed DSDs in Israel, better than gamma or exponential By differentiating Eq. (27) and considering the total number of the
distribution. They reported also that the lognormal distribution has the drops, NWv, the raindrop size distribution, Nv(D), is obtained (Sekine
advantage that the statistical parameters (Dm, σ(D) and NT) have a and Lind, 1982):
physical meaning. In particular the variations in the size parameter Dm
δv−1 δ
are indicative of the relative importance of coalescence and breakup δv ⎛ D ⎞ ⎡ D v⎤
Nv (D) = NWv ⎜ ⎟ exp ⎢−⎛ ⎞ ⎥
⎜ ⎟

processes, and the parameter σ(D) reflects the effect that these processes ωv ⎝ ωv ⎠ ω
⎣ ⎝ v⎠ ⎦ (31)
have on the breadth of the spectrum.
Recently, Ochou et al. (2007) fitted the lognormal distribution to where NWv (m−3), δ v and ω v (mm) are the parameters of the dis-
DSDs detected at Abidjan (Ivory Coast), Dakar (Sénégal), Niamey tribution.
(Niger) and Boyélé (Congo) using JWD disdrometer. The authors sug- Sekine and Lind (1982), fitting Eq. (31) to the raindrop size dis-
gested that this distribution is preferred because its parameters have a tributions measured by Laws and Parsons (1943) in Washington, Sander
simple geometric interpretation and the moments can be written in the (1975) in Berlin and Wickerts (1982) in Stockholm, found that:
form of a product of three terms, each of them being a function of only
NWv = 1000 m−3 (32)
one of the three parameters.
At date, at the best of our knowledge, no relationship to estimate the while δ v and ω v parameters are dependent on I according to the fol-
rainfall kinetic power has been deduced for the lognormal distribution. lowing relationships:

215
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

δ v = 0.95 I 0.14 (33) (42) overestimates the probability function for small D* values
(D* < 2).
ω v = 0.26 I 0.44 (34) Jiang et al. (1997), using DSD measurements carried out in Tokyo,
Jiang et al. (1997), using raindrop size distributions detected by a suggested that for I = 1.88 mm/h, I = 14.32 mm/h, I = 46.40 mm/h
disdrometer installed in Tokyo, positively verified the reliability of the the Weibull distribution agrees with the experimental data better than
Weibull distribution and suggested that this theoretical probability the (U) and (MP) ones. Instead the Marshall and Palmer distribution
distribution, as the Ulbrich one, presents a peak, that tends to shift gives a good fit to the experimental data only for the lowest rainfall
toward large raindrop diameter with increasing rainfall intensity. Jiang intensity value (I = 1.88 mm/h).
et al. (1997), comparing the U, MP and W distributions with the ex- At date, at the best of our knowledge, no relationship to estimate the
perimental data, found that both U and W distribution give better fit to rainfall kinetic power has been deduced for the Weibull distribution.
data than MP one.
Assouline and Mualem (1989) suggested a new approach for 3.4. The Marshall and Palmer distribution (MP)
studying the mechanism of the forces that shape the DSDs through
coalescence and breakup of the falling drops along their pathway from Marshall and Palmer (1948) proposed a simple negative exponential
the cloud to the soil surface. The Authors hypothesized that the result of parameterization (MP) for the raindrop size distribution as a fit to filter
coalescence and breakup processes is equivalent to a series of frag- paper measurements of the DSDs for rainfall intensities between 1 and
mentations that are random in nature. Considering the assumptions 23 mm/h (Uijlenhoet and Stricker, 1999):
that: (i) the probability for a drop break up is proportional to its vo- Nv (D) = N0, v exp (−Λv D) (43)
lume; (ii) the probability that a break up of a drop of size w’ yields drops
of size V’ is independent of V’; (iii) V’ is uniquely related to raindrop where Nv(D), expressed in m−3 cm−1, is equal to the number of rain-
diameter, Assouline and Mualem (1989) proposed the following prob- drops per unit volume of air and per unit size interval having equivo-
ability function: lume spherical diameter D (cm). The parameters of the exponential
n
distribution N0,v (m−3 cm−1) and Λv (cm−1) represent the total number
F (d, I ) = 1 − exp−r D (35) of the raindrops and the scale parameter, respectively.
where n and r are the two parameters of the Weibull distribution, which Marshall and Palmer (1948) suggested that N0,v has a constant value
are estimated by the following equations: equal to 8 × 104 m−3 cm−1 and Λv varies with the rainfall rate, I (mm
h−1), according to the following power law:
DG = r −1/ n Γ(1 + 1/ n) (36)
Λv = 4.1 I 0.21 (44)
σ 2 = r −2/ n [Γ(1 + 2/n) − Γ 2 (1 + 1/n)] (37)
Ulbrich (1983) carrying out an analysis of drop size spectra of Laws
where Γ is the Gamma function. and Parsons (1943) revealed that these data can be also represented by
Using DG as the scaling factor, the following scaled diameter D* is the Marshall and Palmer distribution but N0,v varies with rainfall in-
obtained: tensity.
Many other researchers (Marshall and Palmer, 1948; Waldvogel,
D D
D∗ = = −1/ n 1974; Carbone and Nelson, 1978; Joss and Gori, 1978) suggested that
DG r Γ(1 + 1/ n) (38)
the exponential distribution is a very good approximation to the rain-
From Eq. (38) it follows: drop size distribution referred to natural rainfall.
Several authors (Gunn and Marshall, 1955; Mueller, 1966;
Γn (1 + 1/ n) D ∗n = r Dn (39)
Waldvogel, 1974; Joss and Gori, 1978; Sempere Torres et al., 1994) also
Substituting Eq. (39) into Eq. (35), with pointed out that the data analyzed by Marshall and Palmer (1948)
failed to include the smaller raindrops (< 1 mm) and in particular this
γ = Γn (1 + 1/ n) (40)
DSD tends to overestimate the number of both smallest drops (Gunn
the following dimensionless one-parameter distribution function is and Marshall, 1955; Mueller, 1966; Waldvogel, 1974; Sempere Torres
obtained: et al., 1994) and largest ones (Joss and Gori, 1978; Sempere Torres
et al., 1994).
F (D ∗ ) = 1 − exp−γ D ∗ n (41)
Ochou et al. (2007) carrying out an analysis about DSD types sug-
The power n = 3 of the diameter D derives from the assumption that gested that the MP distribution is widely used for describing the mid-
fragmentation probability is related to drop volume. If the theoretical latitude DSDs that are characterized by low to moderate intensity. In-
power n = 3 is adopted in Eqs. (40) and (41), considering the basing stead for tropical latitudes, where mean rainfall intensities is 6 to 7
assumption which relates fragmentation probability to the volume of times higher than at mid-latitudes (Sauvageot, 1994), Eq. (43) is not
the drop, Assouline and Mualem (1989) proposed the following rela- able to give a good DSD characterization. It was also observed that for I
tion: higher than 10–20 mm h−1 and for D larger than the modal value
(0.1–0.2 cm), the Λ parameter becomes almost constant (Pasqualucci,
F (D ∗ ) = 1 − exp−0.71 D∗ 3
(42)
1982; Sauvageot and Lacaux, 1995) in contrast with Eq. (44). The
which is a probability distribution function, F(D*), that is independent constant slope observed in heavy tropical rain expresses the evolution
of both intensity and site where precipitation occurs, named “universal of the DSD towards an equilibrium form in which the shape of DSDs
distribution” (Assouline and Mualem, 1989). In other words, the in- does not change any more (Ochou et al., 2007).
troduction of a relative raindrop diameter (Eq. (38)) transformed the
theoretical DSD into a universal relationship (Eq. (42)). Assouline and 3.5. The Ulbrich distribution (U)
Mualem (1989), using raindrop size distributions measured by Hudson
(1965) in Rhodesia and by Carter et al. (1974) at Baton Rouge and The gamma distribution (U) proposed by Ulbrich (1983), which is a
Holly Springs, observed that the measured distributions at each site are generalization of the MP one, has been applied by several authors for
overlapped to the universal distribution (Eq. (42)). This circumstance describing DSD measurements carried out using different measurement
confirms that the use of the scaled diameter D* involves that Eq. (42) is techniques (filter paper, photoelectric spectrometer, drop camera,
not site-specific. Assouline (2009) using DSDs measurements carried doppler radar, optical disdrometer) (Blanchard, 1953; Dingle and
out at Princeton (New Jersey) by Smith et al. (2009), found that Eq. Hardy, 1962; Muller, 1965; Caton, 1966; Joss and Gori, 1978; Gori and

216
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Fig. 4. The product ΛD0 as a function of ΛDmax for a gamma raindrop size
distribution having the form N(D) = N0 Dμ exp(-ΛD) (0 ≤ D ≤ Dmax). Each
curve is labeled with the value of the exponent μ to which it corresponds
(modified by Ulbrich, 1983).

D0 D max
2 ∫0 D3 N (D) dD = ∫0 D3 N (D) dD (48)
This relationship can be rewritten as follows:
2 Γ(4 + μ, Λ D0) = Γ (4 + μ, Λ Dmax ) (49)
Fig. 3. Example of the gamma raindrop size distribution for μv = -2, 0 and 2
and with liquid water content W = 1 g m−3 and median volume diameter where Γ is the gamma function. Fig. 4 shows the relationship between
D0 = 2 mm (modified by Ulbrich, 1983). ΛD0 and ΛDmax as defined by Eq. (49) for values of −2 ≤ μ ≤ 3. It
highlights that the value of ΛDmax at which ΛD0 is close to its limiting
value depends on μ. Ulbrich (1983) found that ΛD0, according to Eq.
Geotis, 1981; Chandrasekar and Bringi, 1987; Tokay and Short, 1996;
(49), can be assumed approximately equal to:
Ulbrich and Atlas, 1998; Brawn and Upton, 2008; Carollo and Ferro,
2015; Carollo et al., 2016). Λ D0 = 3.67 + μ (50)
Ulbrich’s distribution is largely used because of its appropriate form which is accurate to within 0.5% for all μ > −3.
for characterizing the distribution of the droplets in clouds (Khrgian For Rhodesian precipitations, Hudson (1971) represented the re-
et al., 1952), aerosols (Levin, 1961), and precipitation particles lationship between D0 and rainfall intensity by a curve that presents a
(Sulakvelidze, 1969; Sulakvelidze and Dadali, 1971). maximum point for I ≈ 80 mm/h even if D0 values are characterized by
This theoretical distribution has the following expression: a low variability (2.0 < D0 < 2.5 mm) in the explored range of I
Nv (D) dD = N0, v D μ v exp (−Λv D) dD (45) (25 < I < 200 mm/h).
Carter et al. (1974) aggregated the DSDs relieved in Louisiana and
in which μv, Λv and N0,v are the statistical parameters of the distribution. Mississippi for 13 intensity classes and found that D0 does not increase
In particular the exponent μv represents the shape parameter and it for rainfall intensities greater than about 65 mm/h.
can assume positive or negative values (Fig. 3). For positive values of μ v Many other researches (Laws and Parsons, 1943; Atlas, 1953;
the DSD is concave downward, has a narrow breadth and falls rapidly to Kelkar, 1959; Zanchi and Torri, 1980; Brandt, 1988; Jau-yau et al.,
zero as D → 0. For negative values of μv the DSD is concave upward on 2008) proposed a power law for describing the relationship D0-I im-
semi-logarithmic plot and has large breadth with increasing numbers of plying that D0 continues to increase indefinitely with I. This last result is
drops at both small and large diameters (Ulbrich, 1983). The N0,v in contrast with other researches which stated that a maximum median
parameter is expressed as m−3 cm−1-μ, and Λv is a scale parameter. For volume diameter value is reached at high rainfall intensities (usually
μv equal to zero, Eq. (45) coincides with the Marshall-Palmer distribu- above 70–100 mm/h), after which the D0 becomes constant (Kinnell,
tion. In agreement with Uijlenhoet and Stricker (1999), Carollo and 1981; Rosewell, 1986; Brown and Foster, 1987, Carollo et al., 2016) or
Ferro (2015) found that, combining Eq. (45), Eq. (17) and Eq. (1) the even decreases (Hudson, 1963; Baruah, 1973; Carter et al., 1974) (Van
following relationship can be obtained: Dijk et al., 2002).
N (D) dD = 17.67 D 0.67N0, v D μ v exp(−Λv D) dD When the parameters D0 and μ are known, Λ parameter can be es-
(46)
timated by the following equation:
Eq. (46) can be rewritten in the following form:
Λ = (3.67 + μ)/ D0 (51)
N (D) dD = N0 D μ exp(−Λ D) dD (47) −1
Rainfall intensity I (mm h ) can be calculated from the following
where N0 = 17.67 N0,v, μ = μv + 0.67, and Λ = Λv. Eq. (47) having the expression (Salles et al., 2002; Carollo and Ferro, 2015):
same mathematical shape of Eq. (45), is also reliable for estimating π ∞ π ∞

drop size distribution per unit area and unit time.


I = 3.6
6
∫0 D3N (D) dD = 3.6 N0
6
∫0 D3 + μ exp(−ΛD) dD
(52)
Ulbrich (1983) demonstrated that the median volume diameter, D0,
where D is expressed in cm and N(D)dD in m−2 s−1.
that is the diameter that divides the distribution in two part of equal
Taking into account that (Olver, 1997):
volume, can be expressed uniquely in term of Λ. For μ ≠ 0, the re-
lationship between Λ, D0 and Dmax, i.e. the maximum diameter of the ∞ Γ(4 + μ)
DSD, is determined by the analytical definition of D0, as:
∫0 D3 + μ exp(−ΛD) dD =
Λ4 + μ (53)

217
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Eq. (52) becomes: −0.85 to 76.9, is generally positive (23756 DSD of the 23,967 mea-
sured 160 ones) and assumes a median value of about 5 while para-
π Γ(4 + μ)
I = 3.6 N0 meter Λ ranges from 3.3 to 591.5. Both parameters slightly decrease as
6 Λ4 + μ (54)
rainfall intensity increases and, in agreement with results of other au-
Eq. (54) links the DSD parameters to rainfall intensity and allows to thors (Zhang et al., 2003; Brawn and Upton, 2008), are correlated.
calculate N0 parameter as function of I and of the other two DSD
parameters:
4. Rainfall kinetic power relationships
Λ4 + μ
N0 = π I
3.6 6 Γ(4 + μ) (55) 4.1. Rainfall kinetic power – intensity relationships

The median drop diameter, D50, and the median volume drop dia- Rainfall kinetic power, Pn (J/m2 h), can be calculated by adding the
meter, D0, considering the Ulbrich’s distribution can be approximately contribution of each raindrop once its mass and terminal velocity are
calculated by the following relationships (Ulbrich, 1983; Uijlenhoet and known. Therefore, the energy characteristics of a rainfall event can be
Stricker, 1999, Carollo et al., 2016; Serio, 2017): indirectly measured if its DSD and the relationship between terminal
0.67 + μ drop velocity and drop diameter are known.
D50 =
Λ (56) At date many researchers proposed empirical relationships linking
kinetic power to rainfall intensity having different mathematical forms
3.67 + μ (polynomial exponential, logarithmical and power type (Table 1). The
D0 =
Λ (57)
most commonly used relationship for estimating Pn as function of
From Eqs. (56) and (57) the two Ulbrich’s parameters can be de- rainfall intensity is that proposed by Wischmeier and Smith (1978) and
duced: used in the Universal Soil Loss Equation (USLE).
According to Wischmeier and Smith (1978) the event rainfall-runoff
μ = D0 Λ − 3.67 (58)
erosivity factor, EI30, is the result of the product of the total rainfall
3 kinetic energy of the rainstorm per unit area, E (MJ ha−1), and I30
Λ= (mm h−1), which is the maximum 30-min rainfall intensity. E is due to
D0 − D50 (59)
the product of the rainfall kinetic energy per unit volume of rainfall, Pn/
Carollo and Ferro (2015), using 23,967 DSDs measured in Sicily in I, (J m−2 mm−1), and the rainfall depth referred to each raining period.
the period June 2006 – February 2012, verified the reliability of the For estimating Pn/I, Wischmeier and Smith (1978) proposed the fol-
Ulbrich’s distribution in Sicilian environments estimating μ and Λ by lowing relationship, having a logarithmic form, based on the mea-
both maximum likelihood method (ML) and momentum method (MM), surements of drop size distribution and terminal velocity observed at
imposing the coincidence between the mean and standard deviation of Washington, DC, by Laws and Parsons (1943):
the measured distribution with the theoretical statistical parameters
one. Fig. 5 shows, as an example for three rainfall intensity values (9.5, Pn (11.9 + 8.73 log I ) for I ⩽ It
=⎧
52.5 and 124.5 mm h−1), the fitting of the gamma distribution (U) to ⎨
I ⎩ (11.9 + 8.73 log It ) for I > It (60)
the DSDs detected in Sicily. This figure demonstrates the reliability of
the Ulbrich’s distribution in the Sicilian environment and its ability to in which It is the intensity threshold value that according to Wischmeier
reproduce the presence of large raindrops corresponding to high P(D) and Smith (1978) is equal to 76 mm h−1. According to Eq. (60) the ratio
values. Carollo and Ferro (2015) found that the parameter μ varies from Pn/I, increases for rainfall intensity value less than or equal to It and

Fig. 5. Comparison between Ulbrich’s distribution and DSDs detected at Palermo site (Sicily, South Italy).

218
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

then it becomes constant for rainfall intensity greater than It. by raindrop size distribution measured in Korea, suggested that as I
Wischmeier and Smith (1978) justified this threshold value suggesting increases, the Pn/I value also increases until I is equal to 60 mm/h, and
that the median volume diameter, D0, does not continue to increase then it becomes constant. The calibrated exponential relationship pro-
when rainfall intensities exceed 76 mm h−1. posed by Kinnell (1981) (Eq. (61), was able to predict the kinetic power
For describing the same trend, Kinnell (1981) proposed the fol- per unit volume of rainfall. In particular they stated that for
lowing relationship: I < 30 mm/h the exponential model fitted the observed data well, but
it heavily underestimated the Pn/I values for I > 60 mm/h.
Pn
= a (1 − b exp (−c I )) Lobo and Bonilla (2015), developing a sensitivity analysis of kinetic
I (61)
energy-intensity relationships on rainfall erosivity using rainfall in-
where a, b and c are parameters. According to Eq. (61) Pn/I has a finite tensity values measured in central Chile, suggested that: (i) Eq. (60)
positive value at zero intensity and approaches, as well as Eq. (60), to provides statistically equivalent erosivity values to those that are
an asymptotic value (Pn/I = a) at high intensities. computed by Eq. (61) with parameter values suggested by McGregor
Brown and Foster (1987), using data from Brisbane and Gunnedah and Mutchler (1976); (ii) Eq. (60) predicts larger erosivity estimates
(Australia), Holly Springs (Mississippi), Miami (Florida), Washington than Eq. (61) for I < 5.5 mm/h, which is the maximum measured
and Zimbabwe, estimated a = 29 J m−2 mm−1, b = 0.72 and rainfall intensity in central Chile. Finally, Authors concluded that Eq.
c = 0.05 h mm−1. The asymptotic value a is practically coincident with (60) and Eq. (61) are highly sensitive to change in its regression
the one (Pn/I = 28.3 J m−2 mm−1) obtained by Eq. (61) for parameters. Thus, these equations should be calibrated based on local
I > 76 mm/h. precipitation data to generate reliable erosivity estimates.
Eq. (61) with the parameter values proposed by Brown and Foster Ciaccioni et al. (2016) carried out an analysis on rainfall erosivity in
(1987) is used in Revised Universal Soil Loss Equation (RUSLE) (Renard Slovenia using disdrometer data. In particular they concluded that: (i)
et al., 1997). In RUSLE2 Equation, Foster (2004) suggested Eq. (61) for at high values of I equations from Coutinho and Tomás (1995) and
estimating rainfall kinetic power, with the parameter a, b and c equal to Cerro et al. (1998) (Table 1) tend to slightly overestimate the measured
29 J m−2 mm−1, 0.72 and 0.082 h mm−1, respectively (McGregor kinetic power; (ii) the use of the equations suggested by Wischmeier
et al., 1995). and Smith (1978), Brandt (1990), Usòn and Ramos (2001) and Van Dijk
Kinnell (1987) reported that a parameter can be assumed equal to et al. (2002) (Table 1) tend to underestimate the rainfall kinetic power.
29 J m−2 mm−1 while b and c parameters are site-specific (Salles et al., Therefore Ciaccioni et al. (2016) concluded that, a general equation
2002). However, other researchers proposed different values of a, b and relating Pn and I is not suitable to estimate correctly the kinetic power
c parameters confirming their dependence on geographical location of the precipitation, but equations as Eq. (61) need to be calibrated
(Table 3). using at-site data.
McGregor et al. (1995) comparing Eqs. (60) and (61) verified that Angulo-Martínez et al. (2016) using DSDs detected in Spain found
for 1 mm/h < I < 35 mm/h, Pn/I values estimated by Eq. (61) with that the relationship between Pn/I and I is asymptotic and in particular
the coefficient values suggested by Brown and Foster (1987) are 12% Pn/I becomes nearly constant at an I value of about 65 mm/h. These
less than those predicted by Eq. (60). They concluded that the annual authors verified the reliability of Eq. (61) for Pn/I estimates using
erosivity predicted by Eq. (61) with parameter values suggested by parameter values proposed by Brown and Foster (1987). Angulo-
McGregor and Mutchler (1976) and Eq. (60) were almost identical, Martínez et al. (2016) concluded that none of these empirical expres-
whereas Pn/I values predicted by Eq. (61) with the parameter values sions can be used universally, since their applicability depends on the
proposed by Brown and Foster (1987) were about 8% lower (Yin et al., geographical and meteorological conditions of investigated sites.
2017). The underestimation of the Pn/I in the RUSLE model was also Carollo et al. (2017), using DSD detected at Palermo (Sicily) in the
reported in Australia (Yu, 1998a,b), Belgium (Verstraeten and Poesen, period June 2006-March 2014 and aggregated in intensity classes,
2006) and Peninsular Malaysia (Yin et al., 2017). found that the Wischmeier and Smith (1978) relationship is fully ap-
Fornis et al. (2005) using disdrometric measurements carried out in plicable for estimating the rainfall kinetic power in Sicilian environ-
Central Cebu, Philippine, calibrated both Eq. (60) and Eq. (61). They ment. In fact, Eq. (60) is able to reproduce the Pn/I values measured at
verified that the calibrated exponential relationship Eq. (61) Palermo experimental site for I < 40 mm/h even if it determines a
(a = 30.68, b = 0.05 and c = 0.03), is more reliable to predict the ki- slight overestimate for 2 < I < 15 mm h−1 and systematically over-
netic power per unit volume of rainfall than the equation proposed by estimates the kinetic power for I > 40 mm h−1. However this last
Wischmeier and Smith. overestimation can be avoid setting It in Eq. (60) equal to 40 mm/h. The
Lim et al. (2015), carrying out an analysis on rainfall kinetic power median volume diameter, D0, and Pn/I values measured at Palermo
show an increasing trend with rainfall intensity for I ≤ 40 mm/h after
Table 3 which does not more increase with I (Carollo et al., 2017). These results
Values of Eq. (61) parameters. represent a further confirmation of the applicability of the Wischmeier
Reference Site a b c
and Smith (1978) approach even if the threshold value of rainfall in-
tensity (It = 40 mm/h) resulted less than the one proposed by
Kinnell (1981) Zimbabwe 29.2 0.89 0.048 Wischmeier and Smith (1978) (It = 76 mm/h) (Carollo et al., 2017).
Kinnell (1981) Florida 29.3 0.28 0.018 Carollo et al. (2017) also tested the reliability of the Eq. (61) for
Rosewell (1986) Australia 29 0.59 0.04
estimating Pn/I at Palermo. In particular, they found that Eq. (61) with
Rosewell (1986) Australia 26.4 0.67 0.035
Brown and Foster (1987) USA 29 0.72 0.05 the parameters values proposed by Brown and Foster (1987), for
Renard et al. (1991) USA 29 0.72 0.05 I > 30 mm h−1, gives values slightly greater than the ones obtained by
Coutinho and Tomás (1995) Portugal 35.9 0.56 0.034 Eq. (60). Instead, Eq. (61) underestimates systematically Pn/I especially
Cerro et al. (1998) Spain 38.4 0.54 0.029
for the lowest values of I (I < 20 mm/h). However in the whole range
Jayawardena and Rezaur (2000) China 36.8 0.69 0.038
Van Dijk et al. (2002) Universal 28.3 0.52 0.042 of I, the shape of the curve corresponding to Eq. (61) describes the Pn/I
Fornis et al. (2005) Philippines 30.8 0.05 0.03 –I relationship better than Eq. (60) even if Eq. (61) needs to be cali-
Petan et al. (2010) Slovenia (Koseze) 29.8 0.6 0.07 brated.
Petan et al. (2010) Slovenia (Kozjane) 31.9 0.6 0.055 Other researchers (Park et al., 1980; Smith and De Veaux, 1992;
Sanchez-Moreno et al. (2012) Cape Verde 35 0.79 0.03
Uijlenhoet and Stricker, 1999; Steiner and Smith, 2000; Sanchez-
Lim et al. (2015) Korea 25.75 0.54 0.05
Angulo-Martínez et al. (2016) Spain 29 0.72 0.05 Moreno et al., 2012) verified the reliability of a power law relationship
for which the ratio Pn/I does not become constant for any value of I. In

219
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

other words, for high values of rainfall intensity, a power law re-
lationship implies a trend different from Eq. (60) and Eq. (61). Sanchez-
Moreno et al. (2012) suggested that the power law equation seems to
represent better than Eq. (60) and Eq. (61) the kinetic power of the
rainfall in Cape Verde, characterized by a sequence of extreme short
events with high intensities.
Salles et al. (2002), carrying out an overview of many empirical
relationships Pn-I showed that, for a fixed rainfall intensity, these re-
lationships yield to very different values of kinetic power.
Using measurements of rainfall intensity, drop size distribution and
drop size specific fall velocity carried out by optical laser disdrometers
in five experimental stations in Germany, Wilken et al. (2018) sug-
gested that the commonly used empirical Pn-I relationships (Table 1)
overestimate the measured rainfall kinetic power. Furthermore, in
agreement to Van Dijk et al. (2002) and Meshesha et al. (2014), Carollo
et al. (2018) stated that there is no a single empirical Pn-I equation that
can be recommended for any part of the world. According to Parsons
and Gadian (2000), Salles et al. (2002) concluded that a global para-
meter, as rainfall intensity I or median volume diameter D0, is not
sufficient to determine the rainfall erosivity since that kinetic power
measurements are also dependent on other parameters (rain type, al-
titude, climate and method of measurement).

4.2. Deducing rainfall kinetic power – intensity relationship by DSD

The kinetic power, Pn (J/m2 h), knowing both DSD and terminal
velocity, can be calculated as (Salles et al., 2002):
ρπ ∞
Pn = 10−6
12
∫0 [V (D)]2 D3N (D) dD
(62)
−3
where ρ is the water density (kg m ), V(D) is the terminal velocity of
the drop having a diameter D (m s−1) and N(D)dD is the number of
raindrops with diameters between D and D + dD arriving at a unit area
per unit time. The use of Eq. (62) requires the knowledge of the rain-
drop size distribution.
At date, at the best of our knowledge, no relationship to rainfall
kinetic power estimate has been deduced for the Weibull and
Lognormal distribution.
Applying the gamma distribution (Ulbrich, 1983), raindrop terminal
velocity estimated by Eq. (15) (Ferro, 2001), and using Eq. (62), Carollo Fig. 6. Comparison among all available pairs (I, Pn/I) (a), (I, D0) (b) and (D0,
and Ferro (2015) deduced theoretically the following expression: Pn/I) (c) measured in different geographical sites.

Pn 9. 52 4 + μ ⎡ 1 2 1 ⎤
= 10−6 ρΛ 4+μ
− + which clearly shows that the ratio Pn/I depends only on median volume
I 7.2 ⎢
⎣Λ (6 + Λ) 4 + μ (12 + Λ) 4 + μ ⎥
⎦ (63)
diameter. According to Eq. (64), if D0 assumes a constant value (when I
According to Eq. (63) the rainfall kinetic power can be determined if exceeded a threshold value), Pn/I becomes constant. In other words, Eq.
both the rainfall intensity and μ and Λ parameters of the DSD are (64) represents a theoretical confirmation of the Wischmeier and Smith
known. This circumstance allows to conclude that the ratio Pn/I de- (1978) hypothesis.
pends on the intrinsic characteristics of rain. Finally, Carollo et al. (2018) presented the comparison between the
Carollo and Ferro (2015), using approximately 24,000 DSDs de- pairs (I, Pn/I and I, D0) corresponding to Palermo dataset with the ones
tected by an optical disdrometer at Palermo (Sicily) in the period June obtained by different measurement techniques (drop camera, piezo-
2006 – February 2012, showed that: (i) Eq. (63) reproduces adequately electric force transducer, blotting paper) in other sites of the world
the kinetic power measurements specially using μ and Λ values obtained (Marshall Islands, New Jersey, Alaska, Indonesia, Oregon, Franklin,
by momentum method; (ii) Eq. (60) is fully applicable to rainfall re- Hong Kong, Ethiopia, Spain). They concluded that the relationships Pn/I
corded in Sicily and has to be preferred to a power relationship cali- - I and D0-I are site specific and therefore a single, like Eq. (60) or Eq.
brated on collected data that tends to overestimate the rainfall kinetic (61), is not reliable for estimating rainfall kinetic power at any site
power for high values of I; (iii) Eq. (61) yields appreciable under- (Fig. 6). On the contrary, the experimental pairs (D0, Pn/I) relative to all
estimates of Pn for low values of rainfall intensity. available datasets resulted near to a single increasing curve that can be
Carollo et al. (2016), taking into account that the Marshall and described by Eq. (64) (Fig. 6). This theoretically derived relationship
Palmer (1948) distribution per unit time and area can be assumed resulted fully applicable to all available datasets demonstrating that Pn/
formally identical to the Ulbrich (1983) one, by combining Eq. (57) and I – D0 relationship is free from at-site effects and useful to adequately
Eq. (63) with μ = 0.67, obtained the following theoretical relationship: characterize the rainfall erosivity (Serio, 2017).
In other words, for a given rainfall intensity, kinetic power depends
on the raindrop size distribution reaching soil surface, and the median
Pn 9. 52 ⎡ 2 1 ⎤
= 10−6 ρ ⎢1 − 4.67
+ 4.67
⎥ volume diameter is an useful characteristic diameter able to synthesize
I 7.2 ⎢

D
(
6 4.340 + 1 ) ( D
6 2.170 +1 ) ⎥
⎦ (64) whole DSD information.

220
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Fig. 7. View of the experimental installation and cross section of the optical disdrometer.

5. Seasonal effects on precipitation fall.

The predicted soil loss is affected by the interaction between the 5.1. Influence of seasonality on DSD in literature
erosive power of rainfall and the soil protection action of vegetative
canopy. Crop canopy, which is the aboveground part of vegetation that Alonge and Afullo (2012), carrying out a seasonal analysis using
intercepts rainfall, varies during the year with natural growth pro- DSD measured at Durban (South Africa), suggested that seasonality
cesses. Maximum erosion control is obtained using a crop having the influences raindrop formation. In particular this analysis highlighted
maximum canopy during periods of high rainfall erosivity (Toy et al., that for rainfall intensity, I, up to 10 mm/h the shape of the raindrop
2002). Soil loss is greatest when the peak period of rainfall erosivity size probability density function in Durban (South Africa) may be si-
corresponds to the period when the soil is most exposed to raindrop milar for summer, autumn and spring seasons. However, for
impact. I > 10 mm/h winter season presents a raindrop size probability den-
At date very few studies have been carried out in order to under- sity function different from the other seasons, having larger raindrop
stand how the seasonality could influence the raindrop formation size. Alonge and Afullo (2012) verified also the reliability of the LN, W
(Alonge and Afullo, 2012; Hachani et al., 2017) and evolution during its and U statistical models establishing that a different probabilistic model

221
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Table 4
Characteristic data of the erosive events in each season.
Season N. of erosive events Total number of disaggregated DSD Total number of aggregated DSD Total number of raindrop I range [mm/h]

Spring 72 3609 27 826,918 0.76–53.19


Summer 38 2689 62 803,952 0.80–113.56
Autumn 101 6672 79 2,096,357 0.78–203.33
Winter 313 29,293 74 8,320,775 0.78–166.65

should be used for each season. In particular the Lognormal distribution 1 min.
best fits the summer and autumn; the Ulbrich distribution fits winter In order to better focus the influence of seasonality and rainfall
and Weibull distribution fits spring season. The analysis also shows that intensity on both DSD and rainfall energetic characteristics (Laws and
the highest number of the rain drops per unit area and time occurs in Parsons, 1943; Carter et al., 1974; Sauvageot and Lacaux, 1995;
summer, while the winter is characterized by the lowest number of Jayawardena and Rezaur, 2000), the 42,264 DSDs (named in-
raindrops. The Authors also highlighted the existence of seasonality stantaneous DSDs) were firstly distinguished in spring (April, May,
respect to Dm, and σ(D). June), summer (July, August, September), autumn (October, No-
For the Weibull model, referred to raindrop size distribution per vember) and winter (December, January, February, March) and then
unit area and time, the values of the shape parameter, δ, resulted si- aggregated in intensity classes having width equal to 1 mm/h.
milar for the seasons of summer, autumn and spring. In addition, the The rainfall distribution during the year is characterized by the
scale parameter, ω, for the seasons of autumn and spring are also found highest values of monthly rainfall in autumn and winter (wet period)
to be close, instead winter season has ω values higher than summer and the lowest values in spring and summer (dry period). In particular,
season. the mean annual value of the rainfall depth at Palermo site is equal to
For the Ulbrich distribution all seasons are observed to take dif- 410 mm in the winter period while in autumn, spring and summer,
ferent values of the scaling parameter, N0. However, the values of the Λ 220 mm, 120 mm and 100 mm occur, respectively.
parameter estimated for summer and autumn are the closest (Alonge For each class, the rainfall intensity was calculated as average of the
and Afullo, 2012). intensities of the instantaneous DSDs falling into the class. This proce-
Hachani et al. (2017) used DSDs detected at the Cévennes-Vivarais dure allows to have number of aggregated DSDs (named DSDs) equal to
region (France) by six OTT Parsivel optical disdrometers to investigate 27 DSDs for spring, 62 for summer, 79 for autumn and 74 for winter.
the seasonal influence on rain drop size distributions. At first the Ul- Table 4 reports some characteristic data of the erosive events referred to
brich raindrop size distribution was fitted to measured DSDs. The each season.
analysis highlighted that the shape parameter is almost identical in Fig. 8 shows the number of raindrops falling in each intensity class
spring and autumn. The number of drops was most similar in autumn versus mean rainfall intensity referred to each class. In general, for each
and winter. The spring and autumn seasons present hybrid character- season the number of the raindrops increases for very low values of I
istics between those of winter and summer. (I < 5 mm/h) (Fig. 8a), while for 5 < I < 40 mm/h (Fig. 8a) the
The intra-annual variability of rainfall erosivity can be high number of the raindrops decreases as rainfall intensity increases. For
(Panagos et al., 2016a) and this information can be useful for identi- winter and autumn seasons the number of the raindrops becomes
fying the high risk periods in which soil exposure to rainfall coincides constant for I > 100 mm/h (Fig. 8b).
with occurring high erosive events. The Ulbrich’s distribution was fitted to each DSDs, using the max-
To date, no study has been carried out to know if the seasonality of imum likelihood method to estimate μ and Λ parameters. The frequency
drop size distribution could have any influence on rainfall erosivity, i.e. distribution of μ and Λ parameters referred to each season datasets are
on the energetic characteristics of rainfall that reaches soil surface. calculated and reported in Fig. 9. In particular the frequency distribu-
tions of the shape parameter (μ) are found to be closer for summer,
spring and autumn rains than the winter one. In fact in agreement to
5.2. New findings: Seasonality on DSDs detected at Palermo experimental Hachani et al. (2017), the values of statistical indices (mean, median,
area standard deviation and coefficient of variation) (Table 5) are closest
and suggest that a similarity in DSD shape in these seasons exists.
In this section an analysis of the seasonality effects on character- Fig. 9b shows the frequency distributions of the DSD slope parameter,
istics of DSDs that reach an unit area in an unit time and on erosive Λ, referred to each season. In winter the DSDs are characterized by the
power of the precipitation at Palermo (Italy) is presented. highest values of Λ parameter. In addiction from spring to winter the
Measurements of drop size distributions were carried out using an raindrop size distributions present an increasing slope parameter value.
optical disdrometer placed at the experimental area equipped near the Each DSD is characterized by a pair (μ, Λ), thus the possible influ-
Department of Agricultural, Food and Forestry Sciences of the ence of the seasonality on the μ and Λ relationship has to be in-
University of Palermo (Fig. 7). vestigated. Fig. 10 shows the pairs (μ, Λ) relative to precipitations oc-
This disdrometer (model ODM 470 made by Eigenbrodt) measures curred in each season. A single relationship between μ and Λ
drop diameters in the range 0.05–0.6 cm. Each drop is separately parameters can be observed for spring, summer and autumn seasons,
measured and registered into classes of about 0.005 cm width. Drop which are also characterized by the nearest values of the total rainfall
diameter is measured by registering light damping due to the passage of depth. While in winter season, the precipitations characterized by
the drop in the control volume between two diodes (Fig. 7). More de- I > 40 mm/h present a μ − Λ relationship different from those with
tails about construction and functioning features of disdrometer are I < 40 mm/h (Fig. 10). In addition, the pairs (μ, Λ) relative to winter
reported in Grossklaus et al. (1998) and Carollo and Ferro (2015). precipitations with I < 40 mm/h are quasi overlapped to spring,
The disdrometer registered 524 rainfall events in the period June 2, summer and autumn ones, suggesting that a DSD similarity exists.
2006 – March 10, 2014. For each rainfall event, only the DSDs for Therefore, these results suggest that for I > 40 mm/h winter DSDs are
which the rainfall intensity was > 0.5 mm/h and measured diameter different from the other season ones, and within winter season DSDs
classes were at least 20 were considered. This choice excluded both characterized by I < 40 mm/h are different from those with
rainfalls having low erosive power and DSDs having a small sample I > 40 mm/h.
size. This procedure provided 42,264 DSDs with a sampling time of

222
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Fig. 8. Number of the raindrops falling in each intensity class versus mean rainfall intensity referred to each class for I < 40 mm/h (a) and for I > 40 mm/h (b).

For each DSDs the median volume diameter value was also de- For each season the kinetic power per unit volume of rainfall, Pn/I,
termined. Fig. 11 shows the pairs (I; D0) referred to each season. The by DSDs was also determined by associating each diameter with the
measurements suggest that for all rainfall intensity classes, the pairs (I, terminal velocity estimated by Eq. (15) with Vmax and an, respectively
D0) are overlapped and thus the relationship between D0 and I is not equal to 9.5 m s−1 and 6 cm. Fig. 12 shows the comparison between Pn/
seasonally dependent (Fig. 11). For I < 40 mm/h the median volume I values versus rainfall intensity of aggregated DSDs for I < 40 mm/h
diameter shows an increasing trend with I, and a single power law can (a) and for I > 40 mm/h (b) in different seasons. The pairs (I, Pn/I) are
be fitted (Fig. 11a). Instead for I > 40 mm/h D0 does not depend on overlapped highlighting that no season dependence of the relationship
rainfall intensity and can be considered quasi constant and equal to between Pn/I and I exists for the precipitation registered at Palermo.
0.25 cm (Fig. 11b). The Pn/I values measured at Palermo show an increasing trend with

Fig. 9. Frequency distributions of μ and Λ parameters in different seasons.

223
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Table 5
Statistic indices of frequency distributions of μ and Λ parameters of Ulbrich distribution in different seasons.
μ Λ

Season Spring Summer Autumn Winter Spring Summer Autumn Winter

Mean 0.99 1.08 1.35 2.32 24.59 24.82 27.71 30.48


Median 0.79 0.97 1.11 2.09 20.01 21.51 23.49 25.88
Standard deviation 1.12 0.99 1.00 1.73 14.87 11.81 13.02 14.17
Coefficient of Variation 1.13 0.92 0.74 0.74 0.60 0.48 0.47 0.46

velocity and drop diameter. This paper presented the most commonly
used relationship for raindrop terminal velocity estimate, highlighting
that the rainfall terminal velocity depends only on rainfall diameter and
fall height.
The most widely used theoretical DSDs were also reviewed, high-
lighting the need to distinguish the raindrop size distribution per unit
volume of air, from that per unit area and time. The raindrop size
distribution per unit volume of air yields to know the static properties
of the raindrop population, i.e. the spatial size distribution of the
raindrops in a volume of air (which governs the raindrop concentra-
tion) and the probability distribution of their sizes. In addition to these
static properties, the raindrop size distribution per unit area and time
involves the dynamic properties of the raindrop population as well its
velocity distribution. However, if the effects of wind, turbulence and
raindrop interaction are neglected, the two size distributions are re-
lated. To date, at the best of our knowledge, no relationship to rainfall
kinetic power estimate has been deduced for the Weibull and
Lognormal distribution.
The reliability of the most commonly used relationships to estimate
Fig. 10. Comparison between μ and Λ parameter relationship in different sea-
rainfall kinetic power in several part of the world, was also reviewed,
sons. highlighting that a single relationship Pn/I - I is not applicable at any
site. Furthermore, a relationship relating Pn/I and I as those proposed
by Wischmeier and Smith and by Kinnell, needs to be calibrated on
rainfall intensity for I ≤ 40 mm/h (Fig. 12a) after which becomes
data. The theoretically derived relationship between Pn/I – D0, from the
constant (Pn/I = 26 J/m2 mm) (Fig. 12b). This performance is similar
Marshall and Palmer distribution per unit time and area, suggested by
to D0-I one. These results represent a further confirmation of the ap-
Carollo et al. resulted fully applicable to all available datasets. This
plicability of the Wischmeier and Smith (1978) approach, even if the
result demonstrates that Pn/I – D0 relationship is free from at-site effects
threshold value of rainfall intensity (It = 40 mm/h) resulted less than
and useful to adequately characterize the rainfall erosivity. This cir-
the one proposed by Wischmeier and Smith (1978) (It = 76 mm/h)
cumstance represents a theoretical confirmation of Wischmeier and
(Carollo et al., 2017).
Smith hypothesis, which established that the trend of Pn/I depends
strictly on the D0 one; in other words, the global parameter D0 is suf-
6. Conclusive remarks and research needs ficient to determine the ratio Pn/I and it is a variable free from at site
effects and useful to characterized rainfall erosivity.
The rainfall kinetic power can be calculated by Drop Size Finally, an analysis on influence of seasonality in DSD and rainfall
Distribution (DSD) measurements combined with fall velocity values energetic characteristics at Palermo experimental area was also carried
measured or estimated by empirical relationship linking terminal fall

Fig. 11. Comparison between D0 values versus rainfall intensity of aggregated DSDs in different seasons for I < 40 mm/h (a) and for I > 40 mm/h (b).

224
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Fig. 12. Comparison between Pn/I values versus rainfall intensity of aggregated DSDs for I < 40 mm/h (a) and for I > 40 mm/h (b) in different seasons.

out. This investigation yields to conclude that the seasonality influences erosion under simulated rainfall in Mediterranean vineyards. Soil Tillage Res. 93,
the raindrop formation only in winter season. However, the relation- 324–334.
Aronica, G., Ferro, V., 1997. Rainfall erosivity over Calabrian region. J. Hydrol. Sci. 42,
ship between Pn/I and I is not season dependent as D0 and I too. In other 35–48.
words, at the best of our knowledge, the erosive power of the pre- Asante, E.A., 2011. Effect of Mulch Type, Mulch Rate and Slope on Soil Loss, Runoff and
cipitation does not depend on the season when the rainfall occurs. Infiltration Under Simulated Rainfall for Two Agricultural Soils in Ghana. Ph.D.
Thesis. University of Science and Technology, Kumasi, Ghana.
Therefore, this paper highlights that future works on rainfall ero- Asseline, J., Valentine, C., 1978. Construction et mise au point d’ un infiltrometre a as-
sivity should be addressed: persion. Chaiers ORSTOM. Serie Hidrologie 15, 321–349.
Assouline, S., 2009. Drop size distributions and kinetic energy rates in variable intensity

• to establish techniques able to give measurements of the raindrop rainfall. Water Resour. Res. 45, 1–7.
Assouline, S., Mualem, Y., 1989. Similarity of regional rainfall. A dimensionless model of
terminal velocity; drop size distribution. Trans. Am. Soc. Agric. Eng. 32, 1216–1222.
• to carry out contemporaneous measurements of drop size and Atlas, D., 1953. Optical extinction by rainfall. J. Meteorol. 10, 486–488.
Atlas, D., Ulbrich, C.W., 1977. Path- and area-integrated rainfall measurement by mi-
terminal velocity by laser disdrometers;
• to deduce relationships to estimate rainfall kinetic power by Weibull
crowave attenuation in the 1–3 cm band. J. Appl. Meteorol. 16, 1322–1331.
Austin, P.M., 1987. Relation between measured radar reflectivity and surface rainfall.
and Lognormal distributions; Mon. Weather Rev. 115, 1053–1070.

• to verify the reliability of the probability distribution function, F Baruah, P.C., 1973. An investigation of drop size distribution of rainfall in Thailand. PhD
Dissertation. Asian Institute of Technology, Bangkok, Thailand.
(D*), proposed by Assouline and Mualem; Beard, K.V., 1976. Terminal velocity of cloud and precipitation drops aloft. J. Atmos. Sci.
• to verify the theoretically derived relationship between P /I – D n 0 33, 851–864.
Beard, K.V., Pruppacher, H.R., 1969. A determination of the terminal velocity and drag of
suggested by Carollo et al. in different site of the world;
• to find new models or techniques able to provide good estimates or
small water drops by means of a wind tunnel. J. Atmos. Sci. 26, 1066–1072.
Bentley, W.A., 1904. Studies of raindrops and raindrop phenomena. Mon. Weather Rev.
measurements of the median volume diameter of the distribution; 32, 450–456.
• to investigate the influence of the seasonality on both rain drop size Best, A.C., 1950. The size distribution of raindrops. Q. J. R. Meteorol. Soc. 76, 1636.
Bezdek, J., Solomon, K., 1983. upper limit lognormal distribution for drop size data. J.
distribution per unit area and time, and rainfall erosivity in other Irrig. Drain. Eng. 1 (72), 72–88.
site of the world; Blanchard, D.C., 1950. Behavior of water drops at terminal velocity. Trans. Am. Geoph.
• to relate the seasonality of drop size distribution with the rain Union. 31, 836–842.
Blanchard, D.C., 1952. Raindrop Size Distribution and Associated Phenomena in
forming systems;

Hawaiian Rains. Woods Hole Oceanographic Institution, Woods Hole, MA, USA No,
to investigate the effect of high-resolution precipitation data on the pp. R52.
calculation of the rinfall erosivity factor. Blanchard, D.C., 1953. Raindrop size-distribution in Hawaiian rains. J. Meteorol. 10,
457–473.
Bollinne, A., Florins, P., Hecq, P., Homerin, D., Renard, V., Wolfs, J.L., 1984. Etude de
Declaration of Competing Interest l’énergie des pluies en climat tempéré océanique d’Europe Atlantique. Z.
Geomorphol. 27–35.
None. Bonilla, C.A., Vidal, K.L., 2011. Rainfall erosivity in Central Chile. J. Hydrol. 410,
126–133.
Bowen, E.G., Davidson, K.A., 1951. A raindrop spectrograph. Quart. J. R. Meteorol Soc.
Acknowledgements 77, 445–449.
Bradley, S.G., Stow, C.D., 1974. The measurement of charge and size of raindrops, Part II.
Results and analysis at ground level. J. Appl. Meteorol. 13, 131–147.
All authors set up the research, analyzed and interpreted the results Brandt, C.J., 1988. The transformation of the rainfall energy by a tropical rain forest
and contributed to write the paper. canopy in relation to soil erosion. J. Biogeogr. 15, 41–48.
Brandt, C.J., 1990. Simulation of the size distribution and erosivity of raindrops and
through fall drops. Earth Surf. Process. 15, 687–698.
References Brawn, D., Upton, G., 2008. On the measurement of atmospheric gamma drop-size dis-
tributions. Atmos. Sci. Lett. 9, 245–247.
Abd Elbasit, M.A.M., Yasuda, H., Salmi, A., Anyoji, H., 2010. Characterization of rainfall Brodie, I., Rosewell, C., 2007. Theoretical relationships between rainfall intensity and
generated by dripper-type rainfall simulator using piezoelectric transducers and its kinetic energy variants associated with stormwater particle washoff. J. Hydrol. 340,
impact on splash soil erosion. Earth Surf. Process. Landforms. 35, 466–475. 40–47.
Alonge, A.A., Afullo, T.J., 2012. Seasonal analysis and prediction of rainfall effects in Brown, L.C., Foster, G.R., 1987. Storm erosivity using idealized intensity distribution.
eastern South Africa at microwave frequencies. Prog. Electromagn. Res. 40, 279–303. Trans. Am. Soc. Agric. Eng. 30, 379–386.
Angulo-Martínez, M., Beguería, S., Kyselý, J., 2016. Use of disdrometer data to evaluate Campos, E.F., 1999. On measurements of drop size distributions. Top. Meteorol.
the relationship of rainfall kinetic energy and intensity (KE-I). Sci. Total Environ. Oceanogr. 6, 24–30.
568, 83–94. Carbone, R., Nelson, L.D., 1978. The evolution of raindrop spectra in warm-based con-
Arnaez, J., Lasanta, T., Ruiz-Flaño, P., Ortigosa, L., 2007. Factors affecting runoff and vective storms as observed and numerically modeled. J. Atmos. Sci. 35, 2302–2314.

225
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

Carollo, F., Ferro, V., 2015. Modeling rainfall erosivity by measured drop-size distribu- Herngren, L., 2005. Build-up and wash-off process kinetics of PAHs and heavy metals on
tions. J. Hydrol. Eng. 20 C4014006-1-7. paved surfaces using simulated rainfall. Ph.D. Thesis. Queensland University of
Carollo, F., Ferro, V., Serio, M.A., 2016. Estimating rainfall erosivity by aggregated drop Technology, Brisbane, Australia.
size distributions. Hydrol. Process. 30, 2119–2128. Hu, Z., Srivastava, R., 1995. Evolution of the raindrop size distribution by coalescence,
Carollo, F., Ferro, V., Serio, M.A., 2017. Reliability of rainfall kinetic power- intensity breakup, and evaporation: theory and observations. J. Atmos. Sci. 52, 1761–1783.
relationships. Hydrol. Process. 31, 1293–1300. Hudson, N.W., 1963. Raindrop size distribution in high intensity rain. Rhod. J. Agric. Res.
Carollo, F., Ferro, V., Serio, M.A., 2018. Predicting rainfall erosivity by momentum and 1, 6–11.
kinetic energy in Mediterranean environment. J. Hydrol. 560, 173–183. Hudson, N.W., 1971. In: Soil Conservation. Batsford Ltd,, London, pp. 338.
Carter, C.E., Greer, J.D., Braud, H.J., Floyd, J.M., 1974. Raindrop characteristics in South Hudson, N.W., 1965. The influence of rainfall mechanics on soil erosion. MSc Thesi, Cape
Central United States. Trans. Am. Soc. Agric. Eng. 17, 1033–1037. Town.
Caton, P.G.F., 1966. A study of raindrop size distributions in the free atmosphere. Quart. Illingworth, A.J., Stevens, C.J., 1987. An optical disdrometer for the measurement of
J. Roy. Meteorol. Soc. 92, 15–30. raindrop size spectra in windy conditions. J. Atmos. Ocean. Technol. 4, 411–421.
Cerro, C., Bech, J., Codina, B., Lorente, J., 1998. Modeling rain erosivity using disdro- Jau-yau, Lu., Chih-Chiang, Su., Tain-Fang, Lu., Maa, Ming-Ming, 2008. Number and vo-
metric techniques. Soil Sci. Soc. Am. J. 62, 731–735. lume raindrop size distributions in Taiwan. Hydrol. Process. 22, 2148–2158.
Chandrasekar, V., Bringi, V.N., 1987. Simulation of radar re-flectivity and surface mea- Jayawardena, A.W., Rezaur, R.B., 2000. Drop size distribution and kinetic energy load of
surements of rainfall. J. Atmos. Oceanic Technol. 4, 464–478. rainstorms in Hong Kong. Hydrol. Process. 14, 1069–1082.
Ciaccioni, A., Bezak, N., Rusjan, S., 2016. Analysis of rainfall erosivity using disdrometer Jiang, H., Sano, M., Sekine, M., 1997. Weibull raindrop-size distribution and its appli-
data at two stations in central Slovenia. Acta Hydrotech. 29, 89–102. cation to rain attenuation. IEE Proc. Microw. Antennas Propag. 144 (3), 197–200.
Cohen, A.C., 1965. Maximum likelihood estimation in the Weibull distribution based on Joss, J., Gori, E.G., 1978. Shapes of raindrop size distributions. J. Appl. Meteorol. 17,
complete and on censored sample. Technometrics 7, 579–588. 1054–1061.
Courshee, R.J., Byass, J.B., 1953. Study of the Methods of Measuring Small Spray Drops. Joss, J., Waldvogel, A., 1967. A spectrograph for raindrops with automatic interpretation.
National Institute of Agricultural Engineering, Bedford, UK. Pure Appl. Geophys. 68, 240–246.
Coutinho, M.A., Tomás, P.P., 1995. Characterization of raindrop size distributions ay the Kathiravelu, G., Lucke, T., Nichols, P., 2016. rain drop measurement techniques, a review.
vale formoso experimental erosion Center. Catena 25, 187–197. Water 8, 1–20.
Davies, C.N., 1942. Unpublished Ministry of Supply reports quoted by Sutton in Air Kelkar, V.N., 1959. Size distribution of raindrops- Part II. Indian J. Meteorol. Geoph. 4,
Ministry report, M.R.P. 40. 323–330.
Defant, A., 1905. Gesetzmässigkeiten in der Verteilung der Verschiedenen Kessler, E., Wilks, K.E., 1968. Radar measurements of precipitation for hydrologic pur-
Tropfenggrössen bei Regenfällen Wiener Sitzungsber. poses. Reports on WMO/IHD projects. Report No 5, 46.
MathematischWissenschaftliche Klasse 114, 585–646. Khrgian, A.K.A., Mazin, I.P., Cao, V., 1952. Distribution of drops according to size in
Dingle, A.N., Hardy, K.R., 1962. The description of rain by means of sequential rain drop cloud. Tr. Tsent. Aerol. Observ. 7, 56.
size distributions. Quar. J. R. Meteorol. Soc. 88, 301–314. Kincaid, D.C., Solomon, K.H., Oliphant, J.C., 1996. Drop size distributions for irrigation
Donnadieu, G., 1980. Comparison of results obtained with the VIDIAZ spectro-pluvi- sprinklers. Trans. Am. Soc. Agric. Eng. 39, 839–845.
ometer and the Joss-Waldvogel rainfall disdrometer in a “rain of a thundery type. J. Kinnell, P.I.A., 1981. Rainfall intensity-kinetic energy relationship for soil loss prediction.
Appl. Meteorol. 19, 593–597. Soil Sci. Soc. Am. Proc. 45, 153–155.
Egodawatta, P.K., 2007. Translation of Small-plot Scale Pollutant Build-up and Wash-off Kinnell, P.I.A., 1987. Rainfall energy in Eastern Australia, intensity-kinetic energy re-
Measurements to Urban Catchment Scale. Ph.D. Thesis. Queensland University of lationships for Canberra, A.C.T. Aust. J. Soil Res. 25, 547–553.
Technology, Brisbane, Australia. Klik, A., Konecny, F., 2013. Rainfall erosivity in northeastern Austria. Trans. ASABE 56,
Eigel, J.D., Moore, I.D., 1983. A simplified technique for measuring raindrop size and 719–725.
distribution. Trans. Am. Soc. Agric. Eng. 24 (4), 1079–1083. Kohl, R.A., 1974. Drop size distributions from medium-sized agricultural sprinklers. Am.
Epema, G.F., Riezebos, H.Th., 1983. Fall velocity of water drops at different heights as a Soc. Agr. Eng. Trans. 17, 690–693.
factor influencing erosivity of simulated rain. Rainfall simulation, runoff and soil Langmuir, I., 1948. The production of rain by a chain reaction in cumulus clouds at
erosion. In: DePloy, J. (Ed.), Catena Supplement. pp. 1–18. temperatures above freezing. J. Meteor. 5, 175–192.
Feingold, G., Levin, Z., 1986. The lognormal fit to raindrop spectra from convective Lanzinger, E., Theel, M., Windolph, H., 2006. Rainfall Amount and Intensity Measured by
clouds in Israel. J. Clim. Appl. Meteorol. 25, 1346–1363. the Thies Laser Precipitation Monitor. TECO-2006, Geneva, Switzerland.
Ferro, V., Giordano, G., Iovino, M., 1991. Isoerosivity and erosion risk map for Sicily. J. Laws, J.O., 1941. Measurements of the fall-velocity of water-drops and raindrops. Am.
Hydrol. Sci. 36, 549–564. Geophys. Union Trans. 22, 709–721.
Ferro, V., Porto, P., Yu, B., 1999. A comparative study of rainfall erosivity estimation for Laws, J.O., Parsons, D.A., 1943. The relation of raindrop size to intensity. Trans. Am.
southern Italy and southeastern Australia. J. Hydrol. Sci. 44, 3–24. Geophys. Union 24, 452–460.
Ferro, V., 2001. Tecniche di misura e monitoraggio dei processi erosivi. Quaderni di Lenard, P., 1904. Uber regen. Meteor. Z. 21, 248–262 [for English translation see. Quart.
Idronomia Montana 21/2, 63-128 (in Italian). J. R. Meteorol. Soc. 31, 62–73.
Fornis, R.I., Vermeulen, H.R., Nieuwenhuis, J.D., 2005. Kinetic energy-rainfall intensity Leone, A., Pica, M., 1993. Caratteristiche dinamiche e simulazione delle piogge. Parte
relationship for Central Cebu, Philippines for soil erosion studies. J. Hydrol. 300, prima, Fondamenti teorici. Rivista di Ingegneria Agraria 3, 167–175.
20–32. Levin, L.M., 1961. Studies in the physics of coarsely dispersed aerosols. Izv. Akad. Nauk.
Foster, G.R., 2004. User’s reference guide, Revised Universal Soil Loss Equation SSSR (in Russian).
(RUSLE2). Report USDA. Levin, L.M., 1971. Charge separation by slashing of naturally falling raindrops. J. Atmos.
Friedrich, K., Higgins, S., Masters, F.J., Lopez, C.R., 2013. Articulating and stationary Sci. 28, 543–548.
PARSIVEL disdrometer measurements in conditions with strong winds and heavy Lim, Y.S., Kim, J.K., Kim, J.W., Park, B.I., Kim, M.S., 2015. Analysis of the relationship
rainfall. J. Atmos. Ocean. Technol. 30, 2063–2080. between the kinetic energy and intensity of rainfall in Daejeon, Korea. Quat. Int. 384,
Fuchs, N., Petrjanoff, I., 1937. Microscopic examination of fog, cloud and rain droplets. 107–117.
Nature 139, 111–112. List, R., Low, T.B., Donaldson, N., Freire, E., Gillespie, J.R., 1987. Temporal evolution of
Gagin, A., 1980. The relationship between the depth of cumuliform clouds and their drop spectra to collisional equilibrium in steady and pulsating rain. J. Atmos. Sci. 44,
raindrop characteristics. J. Rech. Atmos. 14, 409–422. 362–372.
Gori, E.G., Geotis, S.G., 1981. Comparison of the raindrop size distributions observed in Lobo, G.P., Bonilla, C.A., 2015. Sensitivity analysis of kinetic energy-intensity relation-
New England and Switzerland. Preprint 20th Conf. on radar meteorology. Boston. ships and maximum rainfall intensities on rainfall erosivity using a long-term pre-
Am. Meteorol. Soc, 282–286. cipitation dataset. J. Hydrol. 527, 788–793.
Grosh, R.C., 1996. Weighted fall speed parameters. In: 26th Conference on Radar Löffler-Mang, M., Joss, J., 2000. An optical disdrometer for measuring size and velocity of
Meteorology. Ma. Meteor. Soc. Boston, pp. 607–610. hydrometeors. J. Atmos. Ocean. Technol. 17, 130–139.
Grossklaus, M., Uhlig, K., Hasse, L., 1998. An optical disdrometer for use in high wind Lowe, E.J., 1982. Rain drops. Q. J. R. Meteorol. Soc. 18, 242–245.
speeds. J. Atmos. Ocean. Technol. 15, 1051–1059. Lui, J.Y., Orville, H.D., 1969. Numerical modelling of precipitation and shadow effects on
Gucker Jr., F.T., 1949. Determination of concentration and size of particulate matter by mountain-induced cumuli. J. Atmos. Sci. 26, 1286.
light scattering and sonic techniques. In: Proceedings of First National Air Pollution Maahn, M., Kollias, P., 2012. Improved Micro Rain Radar snow measurements using
Symposium, 10–11 November. Stanford Research Institute, Los Angeles, CA, USA, pp. Doppler spectra post-processing. Atmos. Meas. Tech. 5 (11), 2661–2673.
14–25. Mache, H., 1904. Ueber die Geschwindigkeit un Grosse der Regentrpfen. Meteorol. Z. 39,
Gunn, R., Kinzer, G.D., 1949. The terminal velocity of fall for water droplets in stagnant 278.
air. J. Meteorol. 6, 243–248. Mannaerts, C.M., Gabriels, D., 2000. Rainfall erosivity in Cape Verde. Soil Tillage Res. 55,
Gunn, K.L.S., Marshall, J.S., 1955. The effect of wind shear on falling precipitation. J. 207–212.
Meteorol. 12, 339–349. Markowitz, A.H., 1976. Raindrop size distribution expressions. J. Appl. Meteorol. 15,
Hachani, S., Boudevillain, B., Delrieu, G., Bargaoui, Z., 2017. Drop size distribution cli- 1029–1031.
matology in Cévennes-Vivarais Region, France. Atmosphere 8. Marshall, J.S., Palmer, W.M., 1948. The distribution of raindrops with size. J. Meteorol. 5,
Hall, M.J., 1970. Use of the stain method in determining the drop-size distributions of 165–166.
coarse liquid sprays. Am. Soc. Agric. Eng. Trans. 13, 33–37. Mason, B.J., Ramanadham, R., 1953. A photoelectric raindrop spectrometer. Q. J. R.
Hall, R.L., Calder, I.R., 1993. Drop size modification by forest canopies, measurements Meteorol. Soc. 79, 490–495.
using a disdrometer. J. Geophys. Res. 90, 465–470. McCooll, D.K., 1982. Personal Communication. USDA-ARS, Agricultural Engineering
Hauser, D., Amayenc, P., Nutten, B., Waldteufe, l.P., 1984. A new optical instrument for Department, Washington State Univ, Pullman.
simultaneous measurement of raindrop diameter and fall speed distributions. J. McGregor, K.C., Mutchler, C.K., 1976. Status of the R factor in northern Mississippi, soil
Atmos. Ocean. Technol. 1, 256–269. erosion, prediction and control. S.C. S.A. pp. 135-142.

226
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

McGregor, K.C., Bingner, R.L., Bowie, A.J., Foster, G.R., 1995. Erosivity index values for Climatol. 25, 1695–1701.
northern Mississippi. Trans. Am. Soc. Agric. Eng. 38, 1039–1047. Salles, C., Poesen, J., 1999. Performance of an optical spectro-pluviometer in measuring
Meshesha, D.T., Tsunekawa, A., Tsubo, M., Haregeweyn, N., Adgo, E., 2014. Drop size basic rain erosivity characteristics. J. Hydrol. 218, 142–156.
distribution and kinetic energy load of rainfall events in the highlands of the Central Salles, C., Poesen, J., 2000. Rain properties controlling soil splash detachment. Hydrol.
Rift Valley, Ethiopia. Hydrol. Sci. J. 59 (12), 2203–2215. https://doi.org/10.1080/ Process. 14, 271–282.
02626667.2013.865030. Salles, C., Poesen, J., Sempere-Torres, D., 2002. Kinetic energy of rain and its functional
Mikirov, A.E., 1957. A photoelectric method of investigating the distribution of particle relationship with intensity. J. Hydrol. 257, 256–270.
size precipitation. Izv. Akad. Nauk. SSSR 1, 104. Salvador, R., Bautista-Capetillo, C., Burguete, J., Zapata, N., Serreta, A., Playán, E., 2009.
Mueller, E.A., 1966. Rader cross sections from drop size spectra. PhD dissertation. A photographic method for drop characterization in agricultural sprinklers. Irrig. Sci.
University of Illinois, pp. 89. 27, 307–317.
Mueller, EA., Sims, A.L., 1966. Radar cross sections from drop size spectra. Tech. rep. Sanchez-Moreno, J.F., Mannaerts, C.M., Jetten, V., Löffler-Mang, M., 2012. Rainfall ki-
ECOM-00032-F, Contract DA-28-043 AMC-00032(E), Illinois State Water Survey, netic energy-intensity and rainfall momentum-intensity relationships for Cape Verde.
Urbana. 110. [AD-645218]. J. Hydrol. 454–455, 131–140.
Mueller, EA., Sims, A.L., 1967. Raindrop Distributions at Majuro Atoll, Marshall Islands. Sander, J., 1975. Rain attenuation of millimeter waves at L=5.77, 3.3 and 2 mm. IEEE
Research and development technical report (United States. Army Electronics Trans. 23, 213–220.
Command), ECOM-02071-RR1. Fort Monmouth, N.J. Sauvageot, H., 1994. Rainfall measurement by radar, a review. Atmos. Res. 35, 27–54.
Muller, E., 1965. Radar rainfall studies. Ph.D. dissertation. University of Illinois. Sauvageot, H., Lacaux, J.P., 1995. The shape of averaged drop size distributions. J.
Munchak, S.J., Kummerow, C.D., Elsaesser, G., 2012. Relationships between the raindrop Atmos. Sci. 52, 1070–1083.
size distribution and properties of the environment and clouds inferred from TRMM. Schmidt, W., 1909. Eine unmittelbare Bestimmung der Fallgeschwindigkeit von
J. Clim. 25, 2963–2977. Regenropfen. Sitz. Akas. Wiss. Wien. Mathem.-naturw. Klasse. 118 (2A), 71–84.
Mutchler, C.K., 1967. Parameters for describing raindrop splash. J. Soil Water Conserv. Schönhuber, M., Lammer, G., Randeu, W.L., 2007. One decade of imaging precipitation
22, 91–94. measurement by2D-video-distrometer. Adv. Geosci. 10, 85–90.
Mutchler, C.K., 1971. Splash droplet production by waterdrop impact. Water Resour. Res. Sekhon, R.S., Srivastava, R.C., 1971. Doppler radar observations of drop-size distributions
7 (4), 1024–1030. in a thunderstorm. J. Atmos. Sci. 28, 983–994.
Mutchler, C.K., Hansen, L.M., 1970. Splash of a waterdrop at terminal velocity. Science Sekine, M., Lind, G., 1982. Rain attenuation of centimeter, millimeter and submillimeter
169 (3952), 1311–1312. radio waves. In: Proceedings of the 12th European Microwave conference, Helsinki,
Mutchler, C.K., Larson, C.L., 1971. Splash amounts from water drop impact on a smooth Finland, pp. 586–589.
surface. Water Resour. Res. 7, 195–200. Sempere Torres, D., Porra‘, J.M., Creutin, J.D., 1994. A general formulation for raindrop
Navas, A., Alberto, F., Machín, J., Galán, A., 1990. Design and operation of a rainfall size distribution. J. Appl. Meteorol. 33, 1494–1502.
simulator for field studies of runoff and soil erosion. Soil Technol. 3, 385–397. Sempere-Torres, D., Salle, S.C., Creutin, J.D., Delrieu, G., 1992. Quantification of soil
Nawaby, A.S., 1970. A method of direct measurement of spray droplets in an oil bath. J. detachment by raindrop impact, performances of classical formulae of kinetic energy
Agric. Res. 15, 182–184. in Mediterranean storms. In: Bogen, J., Walling, D.E., Day, T. (Eds.), Erosion and
Nearing, M.A., Yin, S., Borrelli, P., Polyakov, V.O., 2017. Rainfall erosivity: a historical Sediment Transport Monitoring Programs in River Basins. International Association
review. Catena 157, 357–362. of Hydrological Sciences, Oslo, pp. 115–124.
Nešpor, V., Krajewski, W.F., Kruger, A., 2000. Wind-induced error of raindrop size dis- Serio, M.A., 2017. Estimating Rainfall Erosivity by Drop Size Distribution. University of
tribution measurement using a two-dimensional video disdrometer. J. Atmos. Ocean. Palermo Ph.D dissertation.
Technol. 17, 1483–1492. Smith, J.A., De Veaux, R.D., 1992. The temporal and spatial variability of rainfall power.
Nystuen, J.A., 2001. Listening to raindrops from underwater, an acoustic disdrometer. J. Environmetrics 3, 29–53.
Atmos. Oceanic Technol. 18, 1640–1657. Smith, J.A., Hui, E., Steiner, M., Baeck, M.L., Krajewski, W.F., Ntelekos, A.A., 2009.
Ochou, A.D., Nzeukou, A., Sauvageot, H., 2007. Parameterization of drop size distribution Variability of rainfall rate and raindrop size distributions in heavy rain. Water
with rain rate. Atmos. Res. 84, 58–66. Resour. Res. 45, 12.
Olver, F.W.J., 1997. Asymptotic and Special Functions. A. K. Peters, Wellesley, MA. Spilhaus, A., 1948. Raindrop size and falling speed. J. Meteorol. 5, 108–110.
Onaga, K., Shirai, K., Yoshinaga, A., 1988. Rainfall erosion and how to control its effects Steiner, M., Smith, J.A., 2000. Reflectivity, rain rate, and kinetic energy flux relationships
on farmland in Okinawa. Land Conservation for Future Generation. Department of based on raindrop spectra. J. Appl. Meteorol. 39, 1923–1940.
Land development, Bangkok, pp. 627–639. Sulakvelidze, G. K., Dadali, Yu. A., 1971. Multiwave length radar measurements of pre-
Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K., cipitation intensity. Radar meteorology, Proc. Third All- Union Conf. (translated from
Montanarella, L., Alewell, C., 2015b. The new assessment of soil loss by water erosion Russian) Israel Program for Scientific Translations. 32-45. (available from NTIS).
in Europe. Environ. Sci. Policy 54, 438–447. Sulakvelidze, G. K., 1969. Rainstorms and Hail. Translated from Russian by Israel
Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., Perčec Tadić, Program for Scientific Translations, Jerusalem.
M., Michaelidis, S., Hrabalíková, M., Olsen, P., Aalto, J., Lakatos, M., Rymszewicz, A., Tokay, A., Short, D.A., 1996. Evidence from tropical raindrop spectra of the origin of rain
Dumitrescu, A., Beguería, S., Alewell, C., 2015a. Rainfall erosivity in Europe. Sci. from stratiform versus convective clouds. J. Appl. Meteorol. 35, 355–371.
Total Environ. 511, 801–814. Toy, T.J., Foster, G.R., Renard, K.G., 2002. Soil Erosion: Processes, Prediction,
Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., 2016a. Spatio-temporal analysis of Measurement and Control. John Wiley and Sons, New York, pp. 338.
rainfall erosivity and erosivity density in Greece. Catena 137, 161–172. Tracy, F.C., Renard, K.G., Fogel, M., 1984. Rainfall energy characteristics for southeastern
Park, S.W., Mitchell, J.K., Bubenzer, G.D., 1980. An analysis of slash erosion mechanics. Arizona. In: Water-Today and Tomorrow, Proc. of Am. Soc. Civil Engineers Irrig. and
In: ASAE Winter Meeting, Chicago, USA, pp. 2502. Drain. Div. Specialty Conference, July 24-26. Flagstaff, Arizona, pp. 559–566.
Park, S.W., Mitchell, J.K., Bubenzer, G.D., 1983. Rainfall characteristics and their relation Uijlenhoet, R., Stricker, J.N.M., 1999. Dependence of rainfall interception on drop size-a
to splash erosion. Trans. Am. Soc. Agric. Eng. 26 (3), 795–804. comment. J. Hydrol. 217, 157–163.
Parsakhoo, A., Lotfalian, M., Kavian, A., Hoseini, S.A., Demir, M., 2012. Calibration of a Ulbrich, C.W., 1983. Natural variations in the analytical form of the raindrop size dis-
portable single nozzle rainfall simulator for soil erodibility study in hyrcanian forests. tribution. J. Clim. Appl. Meteorol. 22, 1764–1775.
Afr. J. Agric. Res. 7, 3957–3963. Ulbrich, C.W., Atlas, D., 1998. Rainfall microphysics and radar properties, analysis
Parsons, D.A., Gadian, A.M., 2000. Uncertainty in modelling the detachment of soil by methods for drop size spectra. J. Appl. Meteorol. 37, 912–923.
rainfall. Earth Surf. Process. Landf. 25, 723–728. Uplinger, W.G., 1981. A new formula for raindrop terminal velocity. In: Proceedings of
Pasqualucci, F., 1982. The variation of drop size distribution in convective storms, a the 20th Conference on Radar Meteorology, 30 Nov–3 Dec 1981. Amer. Meteorol.
comparison between theory and measurement. Geophys. Res. Lett. 9, 839–841. Soc. Boston, Mass, USA, pp. 389–391.
Pérez-Latorre, F.J., Castro, L., Delgado, A., 2010. A comparison of two variable intensity Usòn, A., Ramos, M.C., 2001. An improved rainfall erosivity index obtained from ex-
rainfall simulators for runoff studies. Soil Tillage Res. 107, 11–16. perimental interrill soil losses in soils with Mediterranean climate. Catena 43,
Petan, S., Rusjan, S., Vidmar, A., Mikos, M., 2010. The rainfall kinetic energy– intensity 293–305.
relationship for rainfall erosivity estimation in the mediterranean part of Slovenia. J. Van Boxel, J.H., 1998. Numerical model for the fall speed of rain drops in a rain fall
Hydrol. 391, 314–321. simulator. In: Gabriels, D., Cornelis, W.M. (Eds.), Proceedings of the International
Pruppacher, H.R., Pitter, R.L., 1971. A semi-empirical determination of the shape of cloud Workshop on Technical Aspects and Use of Wind Tunnels for Wind-Erosion Control,
and rain drops. J. Atmos. Sci. 28, 86–94. Combined Effects of Wind and Water Erosion on Processes, November 17–18 1997,
Quinn, N.W.T., 1981. Properties of Transformed Simulated Rainfall under a Corn Canopy Ghent, Belgium ICE Spec. Rep. 1998/1. 77–85, Int. Cent. for Eremol. Univ. of Ghent,
at Different GrowthStages and RowWidths. Masters Thesis. Iowa State University, Ghent, Belgium.
Ames, IA, USA. Van Dijk, A.I.J.M., Bruijnzeel, L.A., Rosewell, C.J., 2002. Rainfall intensity-kinetic energy
Reddy, K.K., Kozu, T., 2003. Measurements of raindrop size distribution over Gadanki relationships, a critical literature appraisal-Review. J. Hydrol. 261, 1–23.
during south-west and north-east monsoon. Indian J. Radio Space 32, 286–295. Verstraeten, G., Poesen, J., 2006. Long-term (105 years) variability in rain erosivity as
Renard, K.G., Foste, r.G.R., Weesies, G.A., Porter, J.P., 1991. RUSLE: revised universal soil derived from 10-min rainfall depth data for Ukkel (Brussels, Belgium): implications
loss equation. J. Water Soil Conserv. 46, 30–33. for assessing soil erosion rates. J. Geophys. Res. 111.
Renard, K.G., Foste, r.G.R., Weesies, G.A., McCool, D.K., Yoder, D.C., 1997. Predicting Waldvogel, A., 1974. The N0 jump of raindrop spectra. J. Atmos. Sci. 31, 1067–1078.
Soil Erosion by Water, A Guide to Conservation Planning with the Revised Universal Weibull, W., 1951. A statistical distribution function of wide applicability. J. Appl. Mech.
Soil Loss Equation (RUSLE). USDA, pp. 404. 293–297.
Roels, J.M., 1981. Personal communication. Laboratory of Physical Geography, Wickerts, S., 1982. Drop size distribution in rain. FOA Report, C20438-E1(E2).
Geographical Institute, University of Utrecht, Netherlands. Wiesner, J., 1895. Beitragezur Kenntnis des troischenregens (About Contributions to the
Rose, C.W., 1960. Soil detachment caused by rainfall. Soil Sci. 89, 28–35. knowledge of the tropical rain). Atmos. Electr. 104, 1397–1434.
Rosewell, C.J., 1986. Rainfall kinetic energy in eastern Australia. J. Appl. Meteorol. Wilken, F., Baur, M., Sommer, M., Deumlich, D., Bens, O., 2018. Uncertainties in rainfall

227
M.A. Serio, et al. Journal of Hydrology 576 (2019) 210–228

kinetic energy-intensity relations for soil erosion. Catena 171, 234–244. revised Universal Soil Loss Equation. Trans.ASAE 42, 1615–1620.
Wischmeier, W.H., 1959. A rainfall erosion index for a universal soil loss equation. Soil Yu, B., 1998a. Rainfall erosivity and its estimation for Australia’s tropics. Aust. J. Soil Res.
Sci. Soc. Am. Proc. 23, 246–249. 36, 143–165.
Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses-a guide to con- Zanchi, C., Torri, D., 1980. Evaluation of rainfall energy in central Italy. In: De Boodt, M.,
servation planning. In: Agriculture Handbook. United States Department of Gabriels, D. (Eds.), Assessment of Erosion. John Wiley & Sons, Chichester, pp.
Agricolture, Washingtin, DC, pp. 537. 133–142.
Yin, S., Nearing, M.A., Borrelli, P., Xue, X., 2017. Rainfall erosivity: an overview of Zhang, G., Vivekanandan, J., Brandes, E., Meneghini, R., Kozu, T., 2003. The shape-slope
methodologies and applications. Vadose Zone J. 16, 12. relation in observed gamma raindrop size distribution, statistical error or useful in-
Yu, B., 1998b. A comparison of the R-factor in the Universal Soil Loss Equation and formation? J. Atmos. Ocean. Tech. 20, 1106–1119.

228

You might also like