Similarity theory

1. Buckingham Pi Theorem and examples

Turbulent closure problem requires empirical expressions for
determining turbulent eddy diffusion coefficients. The development
of turbulence closure is based on observations not theory.
We need to find an intelligent way of organizing observational data.
Similarity theory is a method to find relationships among variables
based on observations.

U
We want to find the relationship between
the cruising speed and the weight of airplane.
2
3
velocity ( / )
dimension ( )
mass ( )
acceleration of gravity ( / )
air density ( / )
U m s
m
W kg
g m s
kg m
a. Define relevant variables and
their dimensions.
b. Count number of fundamental
dimensions.
kg s m , ,
c. Form n dimensionless groups
where n is the number of variables
minus the number of fundamental
dimensions.
2 3 5 = = n
n
H H ,..
1
2 2
1
U
Wg

= H
gravitational force
lift force
3
2

W
= H
mass of airplane
mass of
displaced air
d. Measure as a function of
1
H
2
H
e. Further simplification; assume:
3
2 1
~ constant constant W H = H =
i.e.
6 3 / 2 2 2 2
~ ~ ~ U W W U U W
) ( f
2 1
t t =
Weight as a function of
cruising speed ( The
simple science of flight by
Tennekes, 1997, MIT press)
Flying objects range
from small insects to
Boeing 747
W~U
6

Speed (m/s)
The great flight diagram
W
e
i
g
h
t

(
N
e
w
t
o
n
s
)

Procedure of Buckingham Pi Analysis
Step 1, Hypothesize which variables could be important to the flow.
e.g., stress, density, viscosity, velocity, ..
Step 2, Find the dimensions of each of the variables in terms of the
fundamental dimensions. Fundamental dimensions are:
L=length
M=mass
T=time
K=temperature
Dimensions of any other variables can be represented by these
fundamental dimensions.
1 - 1 -
0
2 - 1 -
1 -
3 -
T ML t Coefficien Viscosity
L height Layer Boundary H
L roughness z
T ML stress wind
LT ty veloci U
ML density

Example
H; , , ; z U, , H; U, ,
0
t
Step 3, Count the number of fundamental dimensions in the problem
there are 3 dimensions in this example: L, M, T
Step 4, Pick up a subset of original variables to become key variables,
subject to the following restrictions:
The number of key variables must equal the number of fundamental dimensions.
All fundamental dimensions must be represented in terms of key variables.
No dimensionless group is allowed from any combination of key variables.
e.g. Pick up 3 variables:
Invalid set: U; , , ; z H, ,
0
t
Step 5, Form dimensionless equations of the remaining variables in terms of
the key variables.
e.g.
i h g
0
f e d
c b a
U H z
U H
U H

t
=
=
=
Step 6, Solve for the unknowns a, b, c, d, e, f, g, h, i
e.g.
2 c 0, b 1, a
) LT ( L) ( ) ML ( T ML
U H
c 1 - b a 3 - 2 - 1 -
c b a
= = =
=
= t
Step 7, Form dimensionless (PI) groups.
e.g.
, , ,
H
z
3
UH
2
U
1
0
2
= = = t t t

t
Step 8, Form other PI groups if you want as long as the total number is the same.
e.g.
, ,
3 3
2 1
5 4
t t
t
t t = =
, , ,
0 0
2
z
H
5
Uz
4
U
1
= = = t t t

t
Which PI groups are right?

They are all right, but some groups are more commonly used and follow
Conventions.
Next, find relations between
PIs through experiments.
roughness relative , number; Reynolds ,
H
z
3
UH
1 0
2
= = t

t
Surface layer similarity (Monin Obukhov similarity)
Surface layer: turbulent fluxes are nearly constant. 20-30 m
Relevant parameters:
( )
) s m ( ) ' ' w (
velocity frictional , ) s m ( ) w v ( ) w u ( / | | u
) m ( z
3 2
g
2 2
2 / 1
2 2 2
*

' '
+
' '
= =
o v
o o o
v
size eddy or height
u
t
u
Say we are interested in wind shear:
z
u
c
c
Four variables and two basic units result in two dimensionless numbers, e.g.:
3
*
0
* u
z ) w ( g
z
u
u
z
and
v
v
u
u
' '
c
c
The standard way of formulating this is by defining:
0
3
*
) w (
u
g
L
v
v
u
u
' '
=
Monin-Oubkhov length
constant Karman - Von
0.35(0.4),
), ( ) (
L
z
z
u
u
z
*
=
= =
c
c
k
| |
k
m m
PI relation
stable unstable
Empirical gradient functions to
describe these observations:
0 5 1
0 ) 16 1 (
4 / 1
> + =
< =

|
|
for
for
m
m
Note that eddy diffusion coefficients
and gradient functions are related:
z
u
w u
c
c
=
' '
m
k
, 0 w v Assuming =
' '
*
zu
1
k |
m
m
k
=
Now we are interested in the vertical gradient of virtual potential temperature.
z c
c
v
u
We can form a new variable
*
*
u
) ' w (
o
u
u

=
Again, four variables and two basic units result in two dimensionless numbers,
L
z
z
z
and
*
c
c
v
u
u
), ( ) (
L
z
z
z
*
| |
u
u
k
h h
v
= =
c
c
PI relation
Similarly, we have
), ( ) (
L
z
z
q
q
z
*
| |
k
q q
= =
c
c
Normally,
), ( ) ( | |
q m
=
Surface wind profile
1. Neutral condition
0
L
z
=
1 ) 0 ( =
m
|
1
z
u
u
z
*
=
c
c k
) ln( u
0
*
z
z
u
k
=
) exp(
z
z
*
u
u
0
k
=
disappear. winds re height whe the is z
0
) ln(
u
u
0
z
z
*
k
=
Aerodynamic roughness length
i
N
1 i
i
t
i
N
1 i
i
t
0
w h
L
25 . 0
s h
S
25 . 0
z
= =
~ =
element i of width : w length; total : L
element i of area : s element; i of height : h area; total : S
i t
i i t
Over land
Over water
Kondo and Yamazawa
(1986)
0.016 ,
g
u
z
2
*
0
= = o
o
If you have observations at three levels,
you may determine displacement as,
) ln(
u
) ln(
u
) ln(
u
u
0
3
0
2
0
1
z
d z
3
z
d z
2
z
d z
1
*

= = =
k k k
0
z
d
Displacement distance
) ln( u
0
*
z
d - z
u
k
=
) ln( ) ln(
u u
u u
d z
d z
d z
d z
1 3
1 2
1
2
1
3

2. Non-neutral condition 0
L
z
= =
0 5 1
0 ) 16 1 (
4 / 1
> + =
< =

|
|
for
for
m
m
0 5 1
0 ) 16 1 (
2 / 1
> + = =
< = =

| |
| |
for
for
h q
h q
)] ( ) [ln( u
0
*
z
z
u

k
m
+ =
Integral form of wind and temperature profiles in the surface layer
0 , 5 ) (
0 , ) 16 1 ( , tan 2 ) ln( ) ln( 2 ) (
4 / 1
2
1
2
1
2
1
2
> = +
< = + + = +

+ +


t
for
for x x
m
x x
m
*
/u u
) ln(
0
z
z
0 , L =
0 , 0 L > >
0 , 0 L < <
) ( ) ln(
t *
0
z
z
) (

u
u u k
h
v v
+ =

Integral form of wind and temperature profiles in the surface layer

0 , 5 ) (
0 , ) 16 1 ( ), ln( 2 ) (
2 / 1
2
1
> = +
< = = +
+


for
for y
h
y
h
t 0
z z at = =
v v
u u
0 t
z z Normally, =
Similarly,
) ( ) ln(
q *
0
z
z
q
) q q (

k
q
+ =

) ( ) (
h q
+ = +
q 0
z z at q q = =
0 t
z z Normally, =
Bulk transfer relations
How to estimate surface fluxes using conventional surface observations,
surface winds (10m), surface temperature (2m),?
). q q ( u ) q w (
), ( u ) w (
, u u
0 0
v0 v 0 v
2 2
*
=
' '
=
' '
=
Q
H
D
C
C
C
u u u
: , ,
Q H D
C C C
Drag coefficient of momentum, heat, and moisture.
, ) (
2
0
2
*
)] ( ) z/z [ln(
2
u
u

k
m
D
C
+
= = , ) (
2
0
2
*
)] z/z [ln(
2
u
u
k
= =
DN
C
,
)] ( ) z/z )][ln( ( ) z/z [ln(
t 0
2

k
h m
H
C
+ +
=
,
)] z/z )][ln( z/z [ln(
t 0
2
k
=
HN
C
,
)] ( ) z/z )][ln( ( ) z/z [ln(
q 0
2

k
q m
Q
C
+ +
=
,
)] z/z )][ln( z/z [ln(
q 0
2
k
=
QN
C
0 -0.5 0.5
1.0
1.5
DN
D
C
C
L
z
2
z
z
10
0
=
5
z
z
10
0
=
0 -0.5 0.5
1.0
1.5
HN
H
C
C
L
z
2
z
z
10
0
=
5
z
z
10
0
=
Flux footprint
General concept of the flux footprint. The darker the red color,
the more contribution that is coming from the surface area certain
distance away for the instrument.
Relative contribution of the land surface area to the flux for two
different measurement heights at near-neutral stability.
Relative contribution of the land surface area to the flux for two different surface
roughnesses at near-neutral stability.
Relative contribution of the land surface area to the flux for two different cases
of thermal stability.
Bulk transfer relations
How to estimate surface fluxes using conventional surface observations,
surface winds (10m), surface temperature (2m),?
). q q ( u ) q w (
), ( u ) w (
, u u
0 0
v0 v 0 v
2 2
*
=
' '
=
' '
=
Q
H
D
C
C
C
u u u
: , ,
Q H D
C C C
Drag coefficient of momentum, heat, and moisture.
number Stanton :
H
C number Dalton :
Q
C
, ) (
2
0
2
*
)] ( ) z/z [ln(
2
u
u

k
m
D
C
+
= = , ) (
2
0
2
*
)] z/z [ln(
2
u
u
k
= =
DN
C
,
)] ( ) z/z )][ln( ( ) z/z [ln(
t 0
2

k
h m
H
C
+ +
=
,
)] z/z )][ln( z/z [ln(
t 0
2
k
=
HN
C
,
)] ( ) z/z )][ln( ( ) z/z [ln(
q 0
2

k
q m
Q
C
+ +
=
,
)] z/z )][ln( z/z [ln(
q 0
2
k
=
QN
C
0 -0.5 0.5
1.0
1.5
DN
D
C
C
L
z
2
z
z
10
0
=
5
z
z
10
0
=
0 -0.5 0.5
1.0
1.5
HN
H
C
C
L
z
2
z
z
10
0
=
5
z
z
10
0
=
Problem: Assuming we have wind observations but no temperature
observations at two levels, say, 5 m and 10 m, in the surface layer,
can we estimate surface roughness and stability?
), ( ) ln(
L
z
z
z
u
u
10
0
10
*
10
+ =
k
), ( ) ln(
L
z
z
z
u
u
5
0
5
*
5
+ =
k
), ( ) ( ) ln(
L
z
L
z
z
z
u
) u u (
5 10
5
10
*
5 10
+ + + =
k
0 ) ln( : Neutral
5
10
*
5 10
z
z
u
) u u (
=
k
) ( ) ln(
, 0 ) ln( : Stable
L
z
L
z
z
z
u
) u u (
z
z
u
) u u (
5 10
5
10
*
5 10
5
10
*
5 10
=
>

|
k
k
4 / 1
L
z
5
1
10
1
x 1
x 1
) x 1 (
) x 1 (
z
z
u
) u u (
z
z
u
) u u (
) 16 (1 x
)} x tan x (tan 2 ] ln[ ] {ln[ ) ln(
, 0 ) ln( : Unstable
2
5
2
10
2
5
2
10
5
10
*
5 10
5
10
*
5 10
=
+ =
<

+
+
+
+

k
k