2. “The rain patters, the leaf quivers”
Rabindranath Tagore
Saturday, September 11, 2010
3. Precipitations
Riccardo Rigon
Objectives:
3
•To Give an introduction to general circulation phenomena and a description
of the atmospheric phenomena that are correlated to precipitation
•To introduce a minimum of atmospheric thermodynamics and some clues
regarding cloud formation
•To speak of precipitations, their formation in the atmosphere, and their
characterisations on the ground
Saturday, September 11, 2010
5. Some Atmospheric Physics
Riccardo Rigon
!"#$#"%"#& '%($()"*#$(+%,-./'/#(./#./$(#
('$",-%'(#0'%+#(./#/12$(%'#(%#(./#-%3/,#
4%235#6/#7$'')/5#%2(#68#$#5)'/7(#(./'+$3#7/33
Foufula-Georgiou,2008
5
Saturday, September 11, 2010
6. Some Atmospheric Physics
Riccardo Rigon
D = 2 ω V sin φ
6
But the Earth rotates on its own axis
And this means that all bodies are subject to the Coriolis force
In the northern hemisphere, a body moving at non-null velocity is deviated
to the right. In the southern hemisphere, to the left.
Saturday, September 11, 2010
7. Some Atmospheric Physics
Riccardo Rigon
D = 2 ω V sin φ
7
But the Earth rotates on its own axis
And this means that all bodies are subject to the Coriolis force
Saturday, September 11, 2010
8. Some Atmospheric Physics
Riccardo Rigon
D = 2 ω V sin φ
7
Coriolis Force
But the Earth rotates on its own axis
And this means that all bodies are subject to the Coriolis force
Saturday, September 11, 2010
9. Some Atmospheric Physics
Riccardo Rigon
D = 2 ω V sin φ
7
Coriolis Force
Rotational velocity of
the Earth
But the Earth rotates on its own axis
And this means that all bodies are subject to the Coriolis force
Saturday, September 11, 2010
10. Some Atmospheric Physics
Riccardo Rigon
D = 2 ω V sin φ
7
Coriolis Force
Rotational velocity of
the Earth
Relative velocity of
the object considered
But the Earth rotates on its own axis
And this means that all bodies are subject to the Coriolis force
Saturday, September 11, 2010
11. Some Atmospheric Physics
Riccardo Rigon
D = 2 ω V sin φ
7
Coriolis Force
Rotational velocity of
the Earth
Relative velocity of
the object considered
Latitude of the object
considered
But the Earth rotates on its own axis
And this means that all bodies are subject to the Coriolis force
Saturday, September 11, 2010
12. Some Atmospheric Physics
Riccardo Rigon
8
Thus, the air masses rotate around the
centres of low and high pressure
High pressure
polar, cold
Easterlies
cold
Westerlies, warm
High pressure
subtropical
warm
Polar
front
Low pressure
zone
Saturday, September 11, 2010
14. Some Atmospheric Physics
Riccardo Rigon
Foufula-Georgiou,2008
10
!"#$%#&#'()$*+'*,)(-+.&
+&$($'.-(-+&%$(-/.01"#'#
Forming a complex global
circulation system
Saturday, September 11, 2010
16. Some Atmospheric Physics
Riccardo Rigon
12
The forces of the pressure gradient...
Pressure, mb
Isobaric surfaces
surface of the ground
surface of the ground
Pressure, mb
pressure gradienthigher
pressure
lower
pressure
map at 1,000m altitude
isobar
Saturday, September 11, 2010
17. Some Atmospheric Physics
Riccardo Rigon
13
...generate winds
The sea breeze
Sea Land
Day
Night
Sea Land
Plane
Valley
Plane
Valley
WarmWarm
ColdCold
Pressure
gradient
Pressure
gradient
Saturday, September 11, 2010
19. Some Atmospheric Physics
Riccardo Rigon
15
The hydrostatic equilibrium of the atmosphere
Column with
section of unit area
Ground
Pressure = p + dp
Pressure = p
Saturday, September 11, 2010
20. Some Atmospheric Physics
Riccardo Rigon
16
dp = −g(z) ρ(z)dz
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
21. Some Atmospheric Physics
Riccardo Rigon
16
dp = −g(z) ρ(z)dz
V a r i a t i o n i n
pressure
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
22. Some Atmospheric Physics
Riccardo Rigon
16
dp = −g(z) ρ(z)dz
V a r i a t i o n i n
pressure
Acceleration
due to gravity
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
23. Some Atmospheric Physics
Riccardo Rigon
16
dp = −g(z) ρ(z)dz
V a r i a t i o n i n
pressure
Acceleration
due to gravity
Air density
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
24. Some Atmospheric Physics
Riccardo Rigon
16
dp = −g(z) ρ(z)dz
V a r i a t i o n i n
pressure
Acceleration
due to gravity
Air density
Thickness of the
air layer
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
25. Some Atmospheric Physics
Riccardo Rigon
17
dp = −g(z) ρ(z)dz
Ideal Gas Law
ρ(z) =
p(z)
R T(z)
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
26. Some Atmospheric Physics
Riccardo Rigon
18
dp = −g(z) ρ(z)dz
Temperature
Pressure
ρ(z) =
p(z)
R T(z)
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
27. Some Atmospheric Physics
Riccardo Rigon
18
dp = −g(z) ρ(z)dz
Air constant
Temperature
Pressure
ρ(z) =
p(z)
R T(z)
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
28. Some Atmospheric Physics
Riccardo Rigon
18
dp = −g(z) ρ(z)dz
Air constant
Temperature
Air density
Pressure
ρ(z) =
p(z)
R T(z)
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
29. Some Atmospheric Physics
Riccardo Rigon
19
dp(z) = −g(z)
p(z)
R T(z)
dz
dp
p
= −g(z)
p(z)
R T(z)
dz
p(z)
p(0)
dp
p
= −
z
0
g(z)
p(z)
R T(z)
dz
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
30. Some Atmospheric Physics
Riccardo Rigon
20
log
p(z)
p(0)
= −
z
0
g(z)
R T(z)
dz
log
p(z)
p(0)
≈
g
R
z
0
1
T(z)
dz
The hydrostatic equilibrium of the atmosphere
Saturday, September 11, 2010
31. Some Atmospheric Physics
Riccardo Rigon
The first law of thermodynamics
with the help of the second
U = U(S, V )
Equilibrium thermodynamics states that the internal energy of a system is a
function of Entropy and Volume:
As a consequence, every variation in internal energy is given by:
∂U()
∂S
:= T(S, V )
dU() = T()dS − pU ()dV
∂U()
∂V
:= −pU (S, V )
21
Saturday, September 11, 2010
32. Some Atmospheric Physics
Riccardo Rigon
The first law of thermodynamics
with the help of the second
U = U(S, V )
Equilibrium thermodynamics states that the internal energy of a system is a
function of Entropy and Volume:
As a consequence, every variation in internal energy is given by:
∂U()
∂S
:= T(S, V )
Temperature
dU() = T()dS − pU ()dV
∂U()
∂V
:= −pU (S, V )
21
Saturday, September 11, 2010
33. Some Atmospheric Physics
Riccardo Rigon
The first law of thermodynamics
with the help of the second
U = U(S, V )
Equilibrium thermodynamics states that the internal energy of a system is a
function of Entropy and Volume:
As a consequence, every variation in internal energy is given by:
∂U()
∂S
:= T(S, V )
Temperature pressure
dU() = T()dS − pU ()dV
∂U()
∂V
:= −pU (S, V )
21
Saturday, September 11, 2010
34. Some Atmospheric Physics
Riccardo Rigon
U = U(S, V )
Variation of
internal
energy
heat
exchanged by
the system
work done by the
system
dU() = T()dS − pU ()dV
The first law of thermodynamics
with the help of the second
As a consequence, every variation in internal energy is given by:
22
Equilibrium thermodynamics states that the internal energy of a system is a
function of Entropy and Volume:
Saturday, September 11, 2010
35. Some Atmospheric Physics
Riccardo Rigon
UT := U(S(T, V ), V )
However, while temperature is directly measurable, entropy is not - a
consequence of the second law of thermodynamics. For this reason it is
preferred to express entropy as a function of temperature, by means of a
change of variables. In this case, it should be observed that entropy is not
solely a function of temperature, but also of volume:
pS() :=
∂U()
∂S
∂S()
∂V
dUT = CV ()dT + (pS() − pU ())dV
The first law of thermodynamics
with the help of the second
23
Saturday, September 11, 2010
36. Some Atmospheric Physics
Riccardo Rigon
UT := U(S(T, V ), V )
However, while temperature is directly measurable, entropy is not - a
consequence of the second law of thermodynamics. For this reason it is
preferred to express entropy as a function of temperature, by means of a
change of variables. In this case, it should be observed that entropy is not
solely a function of temperature, but also of volume:
Entropic
PressurepS() :=
∂U()
∂S
∂S()
∂V
dUT = CV ()dT + (pS() − pU ())dV
The first law of thermodynamics
with the help of the second
23
Saturday, September 11, 2010
37. Some Atmospheric Physics
Riccardo Rigon
The sum of the two pressures, entropic ed energetic, if so they can be defined,
is the normal pressure:
p() := pS() − pU ()
The first law of thermodynamics
with the help of the second
24
Saturday, September 11, 2010
38. Some Atmospheric Physics
Riccardo Rigon
By definition (!) the internal energy of an ideal gas does NOT explicitly
depend on the volume. Therefore:
Variation of
internal
energy
heat
exchanged by
the system
U = U(S)
dU() = T()dS !!!!!!! =⇒ dQ() = dU()
The first law of thermodynamics
with the help of the second
As a consequence, every variation in internal energy is given by:
25
Saturday, September 11, 2010
39. Some Atmospheric Physics
Riccardo Rigon
Therefore, for an ideal gas:
CV () :=
∂UT
∂T
or:
dividing the expression by the mass of air present in the volume:
dUT = dQ() = CV ()dT + ps()dV
dUT = CV ()dT + d(ps() V ) − V dps()
The first law of thermodynamics
with the help of the second
26
Saturday, September 11, 2010
40. Some Atmospheric Physics
Riccardo Rigon
Therefore, for an ideal gas:
CV () :=
∂UT
∂T
or:
dividing the expression by the mass of air present in the volume:
dUT = dQ() = CV ()dT + ps()dV
dUT = CV ()dT + d(ps() V ) − V dps()
The first law of thermodynamics
with the help of the second
26
specific heat at
constant volume
Saturday, September 11, 2010
41. Some Atmospheric Physics
Riccardo Rigon
v :=
1
ρ
duT = cV ()dT + d(ps() v) − v dps()
dividing the expression by the mass of air present in the volume:
The first law of thermodynamics
with the help of the second
27
Saturday, September 11, 2010
42. Some Atmospheric Physics
Riccardo Rigon
v :=
1
ρ
specific
volume
duT = cV ()dT + d(ps() v) − v dps()
dividing the expression by the mass of air present in the volume:
The first law of thermodynamics
with the help of the second
27
Saturday, September 11, 2010
43. Some Atmospheric Physics
Riccardo Rigon
And using the ideal gas law per unit of mass:
ps() v = R T
The following results:
duT = cV ()dT + d(R T) − v dps()
duT = cV ()dT − d(ps() v) + v dps()
The first law of thermodynamics
with the help of the second
28
Saturday, September 11, 2010
44. Some Atmospheric Physics
Riccardo Rigon
Which can be rewritten as (in this case being du = dq):
During isobaric transformations, by definition, dp() = 0, and
dq|p = (cV () + R) dT = cpdT
cp() := cv() + R
cp is known as specific heat at constant pressure
dq = (cV () + R) dT − v dp()
The first law of thermodynamics
with the help of the second
29
Saturday, September 11, 2010
45. Some Atmospheric Physics
Riccardo Rigon
Adiabatic lapse rate
The information given in the first law of thermodynamics can be
combined with that obtained from the law of hydrostatics. In fact,
assuming that a rising parcel of air is subject to an adiabatic
process, then:
v dps() = −g dz
dq() = cp() dT + v dps()
dq() = 0
30
Saturday, September 11, 2010
46. Some Atmospheric Physics
Riccardo Rigon
Resolving the previous system results in:
dT
dz
= −Γd
Γd :=
g
cp
≈ 9.8◦
K Km−1
Adiabatic lapse rate
31
Saturday, September 11, 2010
48. Some Atmospheric Physics
Riccardo Rigon
33
The conditions of atmospheric stability
Temperature
STABLE AIR
Altitude Temperature
GROUND LEVEL
1. The wind pushes
the parcels of air at
21°C up the hill
2. The moving air
cools to 18.3°C
3. The air is cooler
than the surrounding
air and therefore it
drops
Altitude
Saturday, September 11, 2010
49. Some Atmospheric Physics
Riccardo Rigon
34
The conditions of atmospheric stability
Temperature
STABLE AIR
Altitude Temperature
GROUND LEVEL
1. The wind pushes
the parcels of air at
21°C up the hill
2. The moving air
cools to 18.3°C
3. The air is cooler
than the surrounding
air and therefore it
drops
Altitude
Saturday, September 11, 2010
50. Some Atmospheric Physics
Riccardo Rigon
35
The conditions of atmospheric stability
Temperature
STABLE AIR
Altitude Temperature
GROUND LEVEL
1. The wind pushes
the parcels of air at
21°C up the hill
2. The moving air
cools to 18.3°C
3. The air is cooler
than the surrounding
air and therefore it
drops
Altitude
Saturday, September 11, 2010
51. Some Atmospheric Physics
Riccardo Rigon
36
The conditions of atmospheric instability
Temperature
UNSTABLE AIR
Altitude Temperature
GROUND LEVEL
1. The wind pushes
the parcels of air at
21°C up the hill
2. The
moving air
cools to
18.1°C
3. The air is warmer
than the surrounding
air and therefore
continues to rise
4. The air at 15.1°C
continues to rise
5. The air at
12.1°C continues
to rise
6. The air at
9.1°C continues
to rise
Altitude
At altitude the air is relatively cool
Saturday, September 11, 2010
52. Some Atmospheric Physics
Riccardo Rigon
37
The conditions of atmospheric instability
Temperature
UNSTABLE AIR
Altitude Temperature
GROUND LEVEL
1. The wind pushes
the parcels of air at
21°C up the hill
2. The
moving air
cools to
18.1°C
3. The air is warmer
than the surrounding
air and therefore
continues to rise
4. The air at 15.1°C
continues to rise
5. The air at
12.1°C continues
to rise
6. The air at
9.1°C continues
to rise
Altitude
At altitude the air is relatively cool
Saturday, September 11, 2010
53. Some Atmospheric Physics
Riccardo Rigon
38
What happens when water vapour is added?
The water content of the atmosphere is specified by the mixing
ratio w :
w =
Mv
Md
=
ρv
ρd
where Mv is the mass of vapour and Md is the mass of dry air.
Alternatively, one can refer to the specific humidity, q:
q =
Mv
Md + Mv
=
ρv
ρd + ρv
≈ w
where the last equality is valid for MvMd, which is generally true.
Given that humid air can be considered, in good approximation, an
ideal gas its degrees of freedom are restricted once more by the
ideal gas law:
p = ρRT
where the value of the constant depends on the humidity. At the
extremes the values are Rd=287J.K-1kg-1 for dry air and
Rv=461J.K-1kg-1 for vapour.
Saturday, September 11, 2010
54. Some Atmospheric Physics
Riccardo Rigon
39
What happens when water vapour is added?
Let us now introduce a thermodynamic parameter, the potential
temperature θ, that takes account of this phenomenon. It is
defined as the temperature of a parcel of air that has moved
adiabatically from a starting point with temperature T and
pressure p to a reference altitude (and therefore reference
pressure), conventionally set at p0=1,000hPa (sea level). In other
words it describes an adiabatic transformation from (p,T) to
(p0, θ). Qualitatively, the potential temperature represents a
temperature correction based on the altitude.
θv = Tv
p0
p
Rd/co
p
Saturday, September 11, 2010
69. Some Atmospheric Physics
Riccardo Rigon
54
High pressure
polar, cold
Easterlies
cold
Westerlies,
warm
High pressure
subtropical
warm
Polar
front
Low pressure
zone
DejaVu
Saturday, September 11, 2010
79. Precipitations
Riccardo Rigon
Why it rains
•Large-scale atmospheric movements are caused by the variability of solar
radiation at the Earth’s surface, due to the spherical shape of the Earth.
Saturday, September 11, 2010
80. Precipitations
Riccardo Rigon
Why it rains
•Large-scale atmospheric movements are caused by the variability of solar
radiation at the Earth’s surface, due to the spherical shape of the Earth.
•Also, given the rotation of the Earth about its own axis, every air mass in
movement is deflected because of the (apparent) Coriolis force.
Saturday, September 11, 2010
81. Precipitations
Riccardo Rigon
Why it rains
•Large-scale atmospheric movements are caused by the variability of solar
radiation at the Earth’s surface, due to the spherical shape of the Earth.
•This situation:
•generates movements between “quasi-stable” positions of high and low
pressures
•causes large-scale discontinuities in the air’s flow field and discontinuities
of the thermodynamic properties of the air masses in contact with one
another
•generates, therefore, the situation where the lighter masses of air “slide”
over heavier ones, being lifted upwards in the process.
•Also, given the rotation of the Earth about its own axis, every air mass in
movement is deflected because of the (apparent) Coriolis force.
Saturday, September 11, 2010
83. Precipitations
Riccardo Rigon
•The surface of the Earth is composed of various material masses (air, water,
soil) that are oriented differently. They each respond to solar radiation in
different ways causing further movements of the air masses (at the scale of the
variability that presents itself) in order to redistribute the incoming radiant
energy.
Why it rains
Saturday, September 11, 2010
84. Precipitations
Riccardo Rigon
•The surface of the Earth is composed of various material masses (air, water,
soil) that are oriented differently. They each respond to solar radiation in
different ways causing further movements of the air masses (at the scale of the
variability that presents itself) in order to redistribute the incoming radiant
energy.
•Because of these movements, localised lifting of air masses can occur.
Why it rains
Saturday, September 11, 2010
85. Precipitations
Riccardo Rigon
•The surface of the Earth is composed of various material masses (air, water,
soil) that are oriented differently. They each respond to solar radiation in
different ways causing further movements of the air masses (at the scale of the
variability that presents itself) in order to redistribute the incoming radiant
energy.
•Because of these movements, localised lifting of air masses can occur.
•Moving masses of air are lifted by the presence of orography.
Why it rains
Saturday, September 11, 2010
86. Precipitations
Riccardo Rigon
•The surface of the Earth is composed of various material masses (air, water,
soil) that are oriented differently. They each respond to solar radiation in
different ways causing further movements of the air masses (at the scale of the
variability that presents itself) in order to redistribute the incoming radiant
energy.
•Because of these movements, localised lifting of air masses can occur.
•Moving masses of air are lifted by the presence of orography.
• Heating of the Earth’s surface also causes air to be lifted, causing conditions
of atmospheric instability.
Why it rains
Saturday, September 11, 2010
88. Precipitations
Riccardo Rigon
•As air rises it cools, due to adiabatic (isentropic) expansion, and the
equilibrium vapour pressure is reduced. Hence, the condensation of water
vapour becomes possible (though not always probable).
Why it rains
Saturday, September 11, 2010
89. Precipitations
Riccardo Rigon
•As air rises it cools, due to adiabatic (isentropic) expansion, and the
equilibrium vapour pressure is reduced. Hence, the condensation of water
vapour becomes possible (though not always probable).
•In this way, at a suitable altitude above the ground, clouds are formed: particles
of liquid or solid water suspended in the air.
Why it rains
Saturday, September 11, 2010
90. Precipitations
Riccardo Rigon
•As air rises it cools, due to adiabatic (isentropic) expansion, and the
equilibrium vapour pressure is reduced. Hence, the condensation of water
vapour becomes possible (though not always probable).
•In this way, at a suitable altitude above the ground, clouds are formed: particles
of liquid or solid water suspended in the air.
Why it rains
Saturday, September 11, 2010
100. Precipitations
Riccardo Rigon
Factors that influence the nature and quantity of
precipitation at the ground
•Latitude: precipitations are distributed over the surface of the Earth in
function of the general circulation systems.
•Altitude: precipitation (mean annual) tends to grow with altitude - up to a
limit (the highest altitudes are arid, on average).
•Position with respect to the oceanic masses, the prevalent winds, and the
general orographic position.
Saturday, September 11, 2010
103. Precipitations
Riccardo Rigon
Precipitation exhibits spatial variability at
a large range of scales
(mm/hr)
512km
pixel = 4 km
0 4 9 13 17 21 26 30
R (mm/hr)
2
km
4
km
pixel = 125 m
Foufula-Georgiou,2008
77
Spatialdistribution
Saturday, September 11, 2010
106. Precipitations
Riccardo Rigon
Characteristics of precipitation at the ground
•The physical state (rain, snow, hail, dew)
•Depth: the quantity of precipitation per unit area (projection),
often expressed in mm or cm.
•Duration: the time interval during which continuous precipitation is
registered, or, depending on the context, the duration to register a
certain amount of precipitation (independently of its continuity)
•Cumulative depth, the depth of precipitation measured in a pre-fixed
time interval, even if due to more than one event.
Saturday, September 11, 2010
107. Precipitations
Riccardo Rigon
•Storm inter-arrival time
•The spatial distribution of the rain volumes
•The frequency or return period of a certain precipitation event with
assigned depth and duration
•The quality, that is to say the chemical composition of the
precipitation
Characteristics of precipitation at the ground
Saturday, September 11, 2010
117. Extreme precipitations
Riccardo Rigon
Let is consider the maximum annual precipitations
These can be found in hydrological records, registered by characteristic durations:
1h, 3h, 6h,12h 24 h and they represent the maximum cumulative rainfall over the
pre-fixed time.
91
year 1h 3h 6h 12h 24h
1 1925 50.0 NA NA NA NA
2 1928 35.0 47.0 50.0 50.4 67.6
......................................
......................................
46 1979 38.6 52.8 54.8 70.2 84.2
47 1980 28.2 42.4 71.4 97.4 107.4
51 1987 32.6 40.6 64.6 77.2 81.2
52 1988 89.2 102.0 102.0 102.0 104.2
Saturday, September 11, 2010
118. Extreme precipitations
Riccardo Rigon
92
Let is consider the maximum annual precipitations
for each duration there is a precipitation distribution
Precipitazioni Massime a Paperopoli
durata
Precipitazione(mm)
1 3 6 12 24
5010015050100150
Precipitation(mm)
Duration
Maximum Precipitations at Toontown
Saturday, September 11, 2010
119. Extreme precipitations
Riccardo Rigon
1 3 6 12 24
50100150
Precipitazioni Massime a Paperopoli
durata
Precipitazione(mm)
Median
boxplot(hh ~ h,xlab=duration,ylab=Precipitation
(mm),main=Maximum Precipitations at Toontown) 93
Let is consider the maximum annual precipitations
Precipitation(mm)
Duration
Maximum Precipitations at Toontown
Saturday, September 11, 2010
120. Extreme precipitations
Riccardo Rigon
1 3 6 12 24
50100150
Precipitazioni Massime a Paperopoli
durata
Precipitazione(mm)
upper quantile
94
Let is consider the maximum annual precipitations
Precipitation(mm)
Duration
Maximum Precipitations at Toontown
Saturday, September 11, 2010
121. Extreme precipitations
Riccardo Rigon
1 3 6 12 24
50100150
Precipitazioni Massime a Paperopoli
durata
Precipitazione(mm)
lower quantile
95
Let is consider the maximum annual precipitations
Precipitation(mm)
Duration
Maximum Precipitations at Toontown
Saturday, September 11, 2010
122. Extreme precipitations
Riccardo Rigon
1 ora
Precipitazion in mm
Frequenza
20 40 60 80
0510152025
3 ore
Precipitazion in mm
Frequenza
20 40 60 80 100
051015
6 ore
Precipitazion in mm
Frequenza
40 60 80 100
051015
96
Frequency
Precipitation (mm)
Frequency
Frequency
Precipitation (mm) Precipitation (mm)
6 hours3 hour1 hour
Saturday, September 11, 2010
123. Extreme precipitations
Riccardo Rigon
12 ore
Precipitazion in mm
Frequenza
40 60 80 100 120
02468
24 ore
Precipitazion in mm
Frequenza
40 80 120 160
024681012
97
Frequency
Precipitation (mm)
12 hours
Frequency
Precipitation (mm)
24 hours
Saturday, September 11, 2010
124. Extreme precipitations
Riccardo Rigon
Return period
It is the average time interval in which a certain precipitation intensity is
repeated (or exceeded).
Let:
T
be the time interval for which a certain measure is available.
Let:
n
be the measurements made in T.
And let:
m=T/n
be the sampling interval of a single measurement (the duration of the event in
consideration).
98
Saturday, September 11, 2010
125. Extreme precipitations
Riccardo Rigon
Then, the return period for the depth h* is:
99
where Fr= l/n is the success frequency (depths greater or equal to h*).
If the sampling interval is unitary (m=1), then the return period is the
inverse of the exceedance frequency for the value h*.
Tr :=
T
l
= n
m
l
=
m
ECDF(h∗)
=
m
1 − Fr(H h∗)
N.B. On the basis of the above, there is a bijective relation between
quantiles and return period
Return period
Saturday, September 11, 2010
126. Extreme precipitations
Riccardo Rigon
1 3 6 12 24
50100150
Precipitazioni Massime a Paperopoli
durata
Precipitazione(mm)
Median - q(0.5) - Tr = 2 years
q(0.25) - Tr = 1.33 years
100
Precipitation(mm)
Duration
Maximum Precipitations at Toontown
q(0.75) - Tr = 4 years
Saturday, September 11, 2010
128. Extreme precipitations
Riccardo Rigon
h(tp, Tr) = a(Tr) tn
p
102
depth of
precipitation
power law
Rainfall Depth-Duration-Frequency (DDF) curves
Saturday, September 11, 2010
129. Extreme precipitations
Riccardo Rigon
h(tp, Tr) = a(Tr) tn
p
103
coefficient
dependent on
the return
period
depth of
precipitation
Rainfall Depth-Duration-Frequency (DDF) curves
Saturday, September 11, 2010
130. Extreme precipitations
Riccardo Rigon
h(tp, Tr) = a(Tr) tn
p
104
duration
considered
depth of
precipitation
Rainfall Depth-Duration-Frequency (DDF) curves
Saturday, September 11, 2010
131. Extreme precipitations
Riccardo Rigon
h(tp, Tr) = a(Tr) tn
p
105
exponent (not
dependent on
t h e r e t u r n
period)
depth of
precipitation
Rainfall Depth-Duration-Frequency (DDF) curves
Saturday, September 11, 2010
132. Extreme precipitations
Riccardo Rigon
h(tp, Tr) = a(Tr) tn
p
Given that the depth of cumulated precipitation is a non-decreasing function
of duration, it therefore stands that n 0
Also, it is known that average intensity of precipitation:
J(tp, Tr) :=
h(tp, Tr)
tp
= a(Tr) tn−1
p
decreases as the duration increases. Therefore, we also have n 1
Rainfall Depth-Duration-Frequency (DDF) curves
Saturday, September 11, 2010
133. Extreme precipitations
Riccardo Rigon
Tr = 50 years a = 36.46 n = 0.472
Tr = 100 years a = 40.31
Tr = 200 years a = 44.14
curve di possibilità pluviometrica
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
2.3
2.4
1 10 100tp[h]
log(prec) [mm]
tr=50 anni
tr=100 anni
tr=200 anni
a 50
a 100
a 200
107
Rainfall Depth-Duration-Frequency (DDF) curves
Tr=50 years
Tr=100 years
Tr=200 years
DDF Curve
Saturday, September 11, 2010
134. Extreme precipitations
Riccardo Rigon
curve di possibilità pluviometrica
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
2.3
2.4
1 10 100tp[h]
log(prec) [mm]
tr=50 anni
tr=100 anni
tr=200 anni
a 50
a 100
a 200
DDF curves are parallel to each other in the
bilogarithmic plane
108
Tr=50 years
Tr=100 years
Tr=200 years
DDF Curve
Saturday, September 11, 2010
135. Extreme precipitations
Riccardo Rigon
curve di possibilità pluviometrica
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
2.3
2.4
1 10 100tp[h]
log(prec) [mm]
tr=50 anni
tr=100 anni
tr=200 anni
a 50
a 100
a 200
tr = 500 years
tr = 200 years
h(,500) h(200)
109
DDF curves are parallel to each other in the
bilogarithmic plane
Tr=50 years
Tr=100 years
Tr=200 years
DDF Curve
Saturday, September 11, 2010
136. Extreme precipitations
Riccardo Rigon
curve di possibilità pluviometrica
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
2.3
2.4
1 10 100tp[h]
log(prec) [mm]
tr=50 anni
tr=100 anni
tr=200 anni
a 50
a 100
a 200
tr = 500 years
tr = 200 years
Invece h(,500) h(200) !!!!
110
DDF curves are parallel to each other in the
bilogarithmic plane
Tr=50 years
Tr=100 years
Tr=200 years
DDF Curve
Saturday, September 11, 2010
137. Extreme precipitations
Riccardo Rigon
The problem to solve using
probability theory and statistical analysis...
...is, therefore, to determine, for each duration, the correspondence between
quantiles (assigned return periods) and the depth of precipitation
For each duration, the data will need to be interpolated to a probability
distribution. The family of distribution curves suitable to this scope is the Type I
Extreme Value Distribution, or the Gumbel Distribution
b is a form parameter, a is a position parameter (it is, in effect, the mode)
P[H h; a, b] = e−e− h−a
b
− ∞ h ∞
Saturday, September 11, 2010
140. Extreme precipitations
Riccardo Rigon
The distribution mean is given by:
E[X] = bγ + a
where:
is the Euler-Mascheroni constant
γ ≈ 0.57721566490153228606
Gumbel Distribution
Saturday, September 11, 2010
142. Extreme precipitations
Riccardo Rigon
The standard form of the distribution (with respect to which there are tables
of the significant values) is
P[Y y] = ee−y
Gumbel Distribution
Saturday, September 11, 2010
144. Extreme precipitations
Riccardo Rigon
In order to adapt the family of Gumbel distributions to the data of interest
methods of adjusting the parameters are used.
We shall use three:
- The method of the least squares
- The method of moments
- The method of maximum likelihood
Let us consider, therefore, a series of n measures, h = {h1, ....., hn}
118
Methods of adjusting parameters
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
145. Extreme precipitations
Riccardo Rigon
The method of moments consists in equalising the moments of the sample
with the moments of the population. For example, let us consider
The mean and the variance and
the t-th moment of the SAMPLE
119
µH
σ2
H
M
(t)
H
Methods of adjusting parameters
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
146. Extreme precipitations
Riccardo Rigon
If the probabilistic model has t parameters, then the method of
moments consists in equalising the t sample moments with the t
population moments, which are defined by:
In order to obtain a sufficient number of equations one must consider as
many moments as there are parameters. Even though, in principle, the
resulting parameter function can be solved numerically by points, the
method becomes effective when the integral in the second member
admits an analytical solution.
120
MH[t; θ] =
∞
−∞
(h − EH[h])t
pdfH(h; θ) dh t 1
MH[1; θ] = EH[h] =
∞
−∞
h pdfH(h; θ) dh
Methods of adjusting parameters
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
147. Extreme precipitations
Riccardo Rigon
The application of the method of moments to the Gumbel distribution
consists, therefore, in imposing:
or:
bγ + a = µH
b2 π2
6 = σ2
H
MH[1; a, b] = µH
MH[2; a, b] = σ2
H
Methods of adjusting parameters
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
148. Extreme precipitations
Riccardo Rigon
The method is based on the evaluation of the (compound) probability of
obtaining the recorded temporal series:
P[{h1, · · ·, hN }; a, b]
In the hypothesis of independence of observations, the probability is:
P[{h1, · · ·, hN }; a, b] =
N
i=1
P[hi; a, b]
The method of maximum likelihood
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
149. Extreme precipitations
Riccardo Rigon
This probability is also called the likelihood function - it is evidently a
function of the parameters. In order to simplify calculation the log-
likelihood is also defined:
123
P[{h1, · · ·, hN }; a, b] =
N
i=1
P[hi; a, b]
log(P[{h1, · · ·, hN }; a, b]) =
N
i=1
log(P[hi; a, b])
The method of maximum likelihood
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
150. Extreme precipitations
Riccardo Rigon
124
If the observed series is sufficiently long, it is assumed that it must be such that
the probability of observing it is maximum. Then, the parameters of the curve
that describe the population can be obtained from:
∂ log(P [{h1,···,hN };a,b])
∂a = 0
∂ log(P [{h1,···,hN };a,b])
∂b = 0
Which gives a system of two non-linear equations with two unknowns.
The method of maximum likelihood
with respect to the Gumbel distribution but having general validity
Saturday, September 11, 2010
151. Extreme precipitations
Riccardo Rigon
125
e.g. Adjusting the Gumbel Distribution
The logarithm of the likelihood function, in this case, assumes the form:
Deriving with respect to u and α the following relations are obtained:
That is:
Saturday, September 11, 2010
152. Extreme precipitations
Riccardo Rigon
The method of least squares
It consists of defining the the standard deviation of the measures, the ECDF,
and the probability of non-exceedance:
δ2
(θ) =
n
i=1
(Fi − P[H hi; θ])
2
and then minimising it
126
Saturday, September 11, 2010
153. Extreme precipitations
Riccardo Rigon
Standard
deviation
The method of least squares
It consists of defining the the standard deviation of the measures, the ECDF,
and the probability of non-exceedance:
δ2
(θ) =
n
i=1
(Fi − P[H hi; θ])
2
and then minimising it
126
Saturday, September 11, 2010
154. Extreme precipitations
Riccardo Rigon
ECDF
Standard
deviation
The method of least squares
It consists of defining the the standard deviation of the measures, the ECDF,
and the probability of non-exceedance:
δ2
(θ) =
n
i=1
(Fi − P[H hi; θ])
2
and then minimising it
126
Saturday, September 11, 2010
155. Extreme precipitations
Riccardo Rigon
ProbabilityECDF
Standard
deviation
The method of least squares
It consists of defining the the standard deviation of the measures, the ECDF,
and the probability of non-exceedance:
δ2
(θ) =
n
i=1
(Fi − P[H hi; θ])
2
and then minimising it
126
Saturday, September 11, 2010
156. Extreme precipitations
Riccardo Rigon
∂δ2
(θj)
∂θj
= 0 j = 1 · · · m
The minimisation is obtained by deriving the standard deviation expression
with respect to the m parameters
so obtaining the m equations, with m unknowns, that are necessary.
127
The method of least squares
Saturday, September 11, 2010
157. Extreme precipitations
Riccardo Rigon
we have, as a result, three pairs of parameters which are all, to a certain extent,
optimal. In order to distinguish which of these sets of parameters is the best
we must use a confrontation criterion (a non-parametric test). We will use
Pearson’s Test.
128
After the application of the various adjusting methods...
Saturday, September 11, 2010
158. Extreme precipitations
Riccardo Rigon
Pearson’s test is NON-parametric and consists in:
1 - Sub-dividing the probability field into k parts. These can be, for example, of equal
size.
129
Pearson’s Test
Saturday, September 11, 2010
160. Extreme precipitations
Riccardo Rigon
131
Pearson’s Test
Pearson’s test is NON-parametric and consists in:
3 - Counting the number of data in each interval (of the five in the figure).
Saturday, September 11, 2010
161. Extreme precipitations
Riccardo Rigon
Pearson’s test is NON-parametric and consists in:
4 - Evaluating the function:
P[H h0] = P[H 0]
P[H hn+1] = P[H ∞]
where:
in the case of the figure of the previous slides we have:
(P[H hj+1] − P[H hj]) = 0.2
X2
=
1
n + 1
n+1
j=0
(Nj − n (P[H hj+1] − P[H hj])2
n (P[H hj+1] − P[H hj])
132
Pearson’s Test
Saturday, September 11, 2010
162. Extreme precipitations
Riccardo Rigon
0 50 100 150
0.00.20.40.60.81.0
Precipitazione [mm]
P[h]
1h
3h
6h
12h
24h
133
After having applied Pearson’s test...
Precipitation (mm)
Saturday, September 11, 2010
163. Extreme precipitations
Riccardo Rigon
0 50 100 150
0.00.20.40.60.81.0
Precipitazione [mm]
P[h]
1h
3h
6h
12h
24h
Tr = 10 anni
h1 h3 h6 h12 h24
134
After having applied Pearson’s test...
Precipitation (mm)
Tr = 10 years
Saturday, September 11, 2010
164. Extreme precipitations
Riccardo Rigon
0 5 10 15 20 25 30 35
406080100120140160180
Linee Segnalitrici di Possibilita' Pluviometrica
h [mm]
t[ore]
135
By interpolation one obtains...
DDF Curves
t(hours)
Saturday, September 11, 2010
165. Extreme precipitations
Riccardo Rigon
0.5 1.0 2.0 5.0 10.0 20.0
6080100120140160
Linee Segnalitrici di Possibilita' Pluviometrica
t [ore]
h[mm]
136
By interpolation one obtains...
DDF Curves
t (hours)
Saturday, September 11, 2010
166. Extreme precipitations - addendum
Riccardo Rigon
χ2
If a variable, X, is distributed normally with null mean and unit variance,
then the variable
is distributed according to the “Chi squared” distribution (as proved by Ernst
Abbe, 1840-1905) and it is indicated
which is a monoparametric distribution of the Gamma family of
distributions. The only parameter is called “degrees of freedom”.
137
Saturday, September 11, 2010
167. Extreme precipitations - addendum
Riccardo Rigon
In fact, the distribution is:
And its cumulated probability is:
where is the incomplete “gamma” functionγ()
χ2
from Wikipedia
138
Saturday, September 11, 2010
168. Extreme precipitations - addendum
Riccardo Rigon
γ(s, z) :=
x
0
ts−1
e−t
dt
The incomplete gamma function
Saturday, September 11, 2010
169. Extreme precipitations - addendum
Riccardo Rigon
χ2
from Wikipedia
140
Saturday, September 11, 2010
170. Extreme precipitations - addendum
Riccardo Rigon
The expected value of the distribution is equal to the number of degrees of
freedom
χ2
The variance is equal to twice the number of degrees of freedom
E(χk) = k
V ar(χk) = 2k
from Wikipedia
141
Saturday, September 11, 2010
171. Extreme precipitations - addendum
Riccardo Rigon
Generally, the distribution is used in statistics to estimate the goodness of an
inference. Its general form is:
χ2
Assuming that the root of the variables represented in the summation has a
gaussian distribution, then it is expected that the sum of squares variable is
distributed according to with a number of degrees of freedom equal to
the number of addenda reduced by 1.
χ2
χ2
from Wikipedia
142
χ2
=
(Observed − Expected)2
Expected
Saturday, September 11, 2010
172. Extreme precipitations - addendum
Riccardo Rigon
The distribution is important because we can make two mutually
exclusive hypotheses. The null hypothesis:
χ2
It is conventionally assumed that the alternative hypothesis can be excluded
from being valid if X^2 is inferior to the 0.05 quantile of the
distribution with the appropriate number of degrees of freedom.
χ2
from Wikipedia
And its opposite, the alternative hypothesis:
that the sample and the population have the same distribution
that the sample and the population do NOT have the same
distribution
χ2
143
Saturday, September 11, 2010
174. Extreme Events - GEV
Riccardo Rigon
A little more formally
The choice of the Gumbel distribution is not a whim, it is due to a
Theorem which states that, under quite general hypotheses, the
distribution of maxima chosen from samples that are sufficiently
numerous can only belong to one of the following families of
distributions:
I) The Gumbel Distribution
G(z) = e−e− z−b
a
− ∞ z ∞
a 0
145
Saturday, September 11, 2010
175. Extreme Events - GEV
Riccardo Rigon
II) The Frechèt Distribution
G(z) =
0 z ≤ b
e−(z−b
a )
−α
z b
α 0a 0
146
A little more formally
The choice of the Gumbel distribution is not a whim, it is due to a
Theorem, which states that, under quite general hypotheses, the
distribution of maxima chosen from samples that are sufficiently
numerous can only belong to one of the following families of
distributions:
Saturday, September 11, 2010
176. Extreme Events - GEV
Riccardo Rigon
Mean
Mode
Median
Variance
P[X x] = e−x−α
II) The Frechèt Distribution
from Wikipedia
147
A little more formally
Saturday, September 11, 2010
178. Extreme Events - GEV
Riccardo Rigon
α 0
a 0
G(z) =
e−[−(z−b
a )]
−α
z b
1 z ≥ b
III) The Weibull Distribution
149
A little more formally
The choice of the Gumbel distribution is not a whim, it is due to a
Theorem, which states that, under quite general hypotheses, the
distribution of maxima chosen from samples that are sufficiently
numerous can only belong to one of the following families of
distributions:
Saturday, September 11, 2010
179. Extreme Events - GEV
Riccardo Rigon
from Wikipedia
III) The Weibull Distribution
(P. Rosin and E. Rammler, 1933)
150
A little more formally
Saturday, September 11, 2010
180. Extreme Events - GEV
Riccardo Rigon
When k = 1, the Weibull distribution
reduces to the exponential distribution.
When k = 3.4, the Weibull distribution
becomes very similar to the normal
distribution.
Mean
Mode
Median
Variance
from Wikipedia
151
A little more formally
III) The Weibull Distribution
(P. Rosin and E. Rammler, 1933)
Saturday, September 11, 2010
182. Extreme Events - GEV
Riccardo Rigon
For the distribution reduces to the Gumbel distribution
For the distribution becomes a Frechèt distribution
For the distribution becomes a Weibull distribution
ξ = 0
ξ 0
ξ 0
The aforementioned theorem can be reformulated in terms of a three-parameter
distribution called the Generalised Extreme Values (GEV) Distribution.
G(z) = e−[1+ξ(z−µ
σ )]−1/ξ
z : 1 + ξ(z − µ)/σ 0
−∞ µ ∞ σ 0
−∞ ξ ∞
153
A little more formally
Saturday, September 11, 2010
183. Extreme Events - GEV
Riccardo Rigon
G(z) = e−[1+ξ(z−µ
σ )]−1/ξ
z : 1 + ξ(z − µ)/σ 0
−∞ µ ∞ σ 0
−∞ ξ ∞
154
A little more formally
The aforementioned theorem can be reformulated in terms of a three-parameter
distribution called the Generalised Extreme Values (GEV) Distribution.
Saturday, September 11, 2010
184. Extreme Events - GEV
Riccardo Rigon
gk = Γ(1 − kξ)
155
A little more formally
The aforementioned theorem can be reformulated in terms of a three-parameter
distribution called the Generalised Extreme Values (GEV) Distribution.
Saturday, September 11, 2010
185. Extreme Events - GEV
Riccardo Rigon
dgev(x, loc=0, scale=1, shape=0, log = FALSE)
pgev(q, loc=0, scale=1, shape=0, lower.tail = TRUE)
qgev(p, loc=0, scale=1, shape=0, lower.tail = TRUE)
rgev(n, loc=0, scale=1, shape=0)
R
156
A little more formally
Saturday, September 11, 2010
186. Bibliography and Further Reading
Riccardo Rigon
•Albertson, J., and M. Parlange, Surface Length Scales and Shear Stress: Implications
for Land-Atmosphere Interaction Over Complex Terrain, Water Resour. Res., vol. 35,
n. 7, p. 2121-2132, 1999
•Burlando, P. and R. Rosso, (1992) Extreme storm rainfall and climatic change,
Atmospheric Res., 27 (1-3), 169-189.
•Burlando, P. and R. Rosso, (1993) Stochastic Models of Temporal Rainfall:
Reproducibility, Estimation and Prediction of Extreme Events, in: Salas, J.D., R.
Harboe, e J. Marco-Segura (eds.), Stochastic Hydrology in its Use in Water Resources
Systems Simulation and Optimization, Proc. of NATO-ASI Workshop, Peniscola,
Spain, September 18-29, 1989, Kluwer, pp. 137-173.
Bibliography and Further Reading
Saturday, September 11, 2010
187. Bibliography and Further Reading
Riccardo Rigon
•Burlando, P. e R. Rosso, (1996) Scaling and multiscaling Depth-Duration-Frequency
curves of storm precipitation, J. Hydrol., vol. 187/1-2, pp. 45-64.
•Burlando, P. and R. Rosso, (2002) Effects of transient climate change on basin
hydrology. 1. Precipitation scenarios for the Arno River, central Italy, Hydrol.
Process., 16, 1151-1175.
•Burlando, P. and R. Rosso, (2002) Effects of transient climate change on basin
hydrology. 2. Impacts on runoff variability of the Arno River, central Italy, Hydrol.
Process., 16, 1177-1199.
• Coles S.,ʻʻAn Introduction to Statistical Modeling of Extreme Values, Springer,
2001
• Coles, S., and Davinson E., Statistical Modelling of Extreme Values, 2008
Saturday, September 11, 2010
188. Bibliography and Further Reading
Riccardo Rigon
•Foufula-Georgiou, Lectures at 2008 Summer School on Environmental Dynamics,
2008
•Fréchet M., Sur la loi de probabilité de l'écart maximum, Annales de la Société
Polonaise de Mathematique, Crocovie, vol. 6, p. 93-116, 1927
•Gumbel, On the criterion that a given system of deviations from the probable in
the case of a correlated system of variables is such that it can be reasonably
supposed to have arisen from random sampling, Phil. Mag. vol. 6, p. 157-175, 1900
• Houze, Clouds Dynamics, Academic Press, 1994
Saturday, September 11, 2010
189. Bibliography and Further Reading
Riccardo Rigon
•Kleissl J., V. Kumar, C. Meneveau, M. B. Parlange, Numerical study of dynamic
Smagorinsky models in large-eddy simulation of the atmospheric boundary layer:
Validation in stable and unstable conditions, Water Resour. Res., 42, W06D10, doi:
10.1029/2005WR004685, 2006
•Kottegoda and R. Rosso, Applied statistics for civil and environmental engineers,
Blackwell, 2008
•Kumar V., J. Kleissl, C. Meneveau, M. B. Parlange, Large-eddy simulation of a diurnal
cycle of the atmospheric boundary layer: Atmospheric stability and scaling issues,
Water Resour. Res., 42, W06D09, doi:10.1029/2005WR004651, 2006
•Lettenmaier D., Stochastic modeling of precipitation with applications to climate
model downscaling, in von Storch and, Navarra A., Analysis of Climate Variability:
Applications and Statistical Techniques,1995
Saturday, September 11, 2010
190. Bibliography and Further Reading
Riccardo Rigon
•Salzman, William R. (2001-08-21). Clapeyron and Clausius–Clapeyron
Equations (in English). Chemical Thermodynamics. University of Arizona. Archived
from the original on 2007-07-07. http://web.archive.org/web/20070607143600/
http://www.chem.arizona.edu/~salzmanr/480a/480ants/clapeyro/clapeyro.html.
Retrieved 2007-10-11.
•von Storch H, and Zwiers F. W, Statistical Analysis in climate Research, Cambridge
University Press, 2001
•Whiteman, Mountain Meteorology, Oxford University Press, p. 355, 2000
Saturday, September 11, 2010