Parameterization

Representation of any small-scale or complex physical processes using a

simplified approach in terms of the resolved variables.

Most atmospheric physical processes are either too small in

scales (e.g., cloud microphysics, aerosol processes, cumulus
convection), or too complex (e.g., radiative transfer) to be Grid-box size is limited by
explicitly simulated using the resolved fields; instead, they are computational capability
often represented using simplified models or schemes – physical
Processes that act on scales
parameterization. smaller than the grid box will
be excluded from the model-
resolved solutions.

We need to include them by

means of parameterization
(a largely statistical
description of what goes on
“inside” the box).

Similar idea to molecules

being summarized
statistically by temperature
and pressure, but much
more complex!
Why parameterization...?

The basic physical eqns describe the behavior of the atm on small scales.
We derive eqns that describe the behavior of the system on larger scales.

Large-scale eqns contain terms representing effects of smaller-scale processes.

Since cumulus parameterization is an attempt to formulate the statistical effects of cumulus

convection without predicting individual clouds, it is a ‘closure’ problem in which we seek a
limited number of eqns that govern the statistics of a system with huge dimensions. Therefore,
the core of the cumulus parameterization problem as distinguished from the dynamics and
thermodynamics of individual clouds is in the choice of appropriate closure assumptions.
(Arakawa, Met. Monograph, 1993)

Parameterizations are statistical theories describing the interactions of small scales

with larger scales. They involve idealizations as well as “closure assumptions” that are
at best, only approximately valid.
 Major parameterized processes in AGCMs

Model-resolved fields: u, v, w, T, Z, q, RH.

Parameterized physical processes:
 Atmospheric radiative transfer – parameterized using simplified
radiative transfer models – relatively accurate
 Atmospheric convection/convective precipitation – parameterized using
dry convection and cumulus schemes (Bechtold, 2013)
 Cloud fraction/amount: parameterized using RH, stability, and other
resolved fields (Forbes 2015; Jakob and Miller 2003)
 Large-scale precipitation: parameterized using RH, q, T, convergence,
etc.
 Cloud microphysics/aerosols: parameterized using simple models
 Turbulence mixing: parameterized using simplified turbulence models
 Surface SH and LH fluxes: bulk formulas
3
 Convective parameterization

Technique used in NWP to predict the collective effects of (many) convective clouds
that may exist within a single grid element… As a function of larger-scale processes
and/or conditions.

Why do NWP models need to worry about it..?

 Direct Concern: To Predict convective precipitation

 Feedback to larger Scales: Deep convection “overturns” the atmosphere, strongly
affecting mesoscale dynamics

- changes vertical stability

- generates and redistributes heat

- removes and redistributes moisture

- makes clouds, strongly affecting surface heating and atmospheric radiation

 Flow chart on convective parameterization

 What comes to the mind when we talk of moist convection?

 Why is it important and what are the different types of moist convection?
 Moist process- A multi-scale problem

 What is convective parameterization and why is it necessary?

 Point of uncertainties in convective parameterization
 Few well known schemes: KUO scheme, Arakawa-Schubert, Betts-Miller-Janjic and
Kain-Fritsch
Convection

EmanuelConvection([…]) → Emanuel convection scheme from `[Emanuel and

Zivkovic-Rothman]`_

EmanuelConvection.__call__(state, timestep) → Gets tendencies and diagnostics

from the passed model state.

DryConvectiveAdjustment([…]) → A conservative scheme to keep the temperature

profile close to the dry adiabat if it is super-adiabatic.

DryConvectiveAdjustment.__call__(state, timestep) → Gets diagnostics from the

current model state and steps the state forward in time according to the timestep.
What comes to our mind.....? (about moist convection)

 It could be severe thunderstorms with strong & gusty wind, heavy rain, lightning etc.

 Sometimes these storms can merge to form lines of organized deep convective
storms with trailing stratiform rainfall regions.

 It could be stratocumulus seen near coastline or stratus cloud producing light rainfall
over a large area.

Convection varies widely in shape and sizes & its manifestation is seen in the form of
clouds of different shapes and sizes.
 Why is it important and what are the types of moist convection?

 Moist convection is important to the prediction of atmospheric circulation

for many reasons.

 Large scale horizontal gradients of latent heating produced by deep moist

convection help to drive large scale vertical circulations (e. g. Hadley &
Walker cell).

 Deep convection: also is a major component in ENSO and can influence the
seasonal climate in the Northern Hemisphere. The SST in the tropical eastern
Pacific are warmer than normal during ENSO episodes.

Associated with this, deep convection develops, releasing latent heating in a

deep atmospheric column and producing upper level divergence. The upper
level divergence excites Rossby waves that alter the hemispheric flow
(Tribbia 1991).
 No single model can encompass all relevant processes

mm 10 m 100 m 1 km 10 km 100 km 1000 km 10000 km

Cloud
microphysics  turbulence Cumulus Cumulonimbus Mesoscale Extratropical Planetary
clouds clouds Convective systems Cyclones waves
DNS
Large Eddy Simulation (LES) Model
Cloud System Resolving Model (CSRM)
Numerical Weather Prediction (NWP) Model
Global Climate Model
 Length scales in the atmosphere

Earth 103 km Landsat 60 km 65km LES 10 km

~1mm-100mm ~mm ~100m

10
 Shallow convection

 In contrast to deep convection, shallow cumulus clouds are the most frequently
observed tropical cloud (Johnson et al. 1999).

 Shallow convection modifies the surface radiation budget, influences the structure
and turbulence of the PBL and thereby also affect the global climate (Randall et al
1985).

 Shallow convection also occurs in mid-latitude particularly when cold air moves over
warm water. Shallow cumulus cloud develop over water which commonly align
themselves in the form of bands or streaks (Houze 1993)
 Stratiform convection

Deep convection can be further sub divided into convective and stratiform
components (Houze, 1997, Chattopadhyay etal, 2009).

The convective components refer to convection associated with individual cells,

horizontally small regions of more intense updrafts and down drafts in association with
young and active convection.

The stratiform component refers to convection associated with older, less active
convection with vertical motion generally less than 1ms-1.
 How does the feedback occur in a model?

At every grid point, predictive variables change at each time step as a function
of a number of processes, including convection…

temp
d
= Prad + Pconv + Pcond / evap + Phdiff + Pvdiff + Psfc
dt
water vapour
dqv
= Pconv + Pcond / evap + Phdiff + Pvdiff + Psfc
dt
momentum
du 1 p
= − fv = (Pconv ) + Phdiff + Pvdiff + Psfc
dt  x

…when activated, a convective parameterization computes the changes in temp and moisture
(and possibly cloud water, momentum, etc.) that would occur at each vertical level if
convection developed in the given grid-point environment
- For example:   final −  initial
= = Pconv
t conv c

where τc is a convective timescale, typically 30 min to 1hr



How to parameterize..

 Relate unresolved effects (convection) to grid-scale properties using statistical or

empirical techniques

 What properties of convection do we need to predict?

 convective triggering (yes/no)
 convective intensity (how much rain?)
 vertical distribution of heating
 vertical distribution of drying
Governing equations

(1)

(2)

(3)
 When, where, and how much...

 Triggering (always requires positive area on a thermodynamic [e.g., Skew-T Log-P]

diagram)

- Different approaches

1) mass or moisture convergence exceeds certain threshold at a grid point,

2) positive destabilization rate,

3) perturbed parcels can reach their level of free convection,

4) sufficient cloud layer moisture

 Convective intensity (net heating)

- proportional to mass or moisture convergence
- sufficient to offset large-scale destabilization rate
- sufficient to eliminate CAPE (constrained by available moisture)
* Vertical distribution of heating and drying
- determined by nudging to empirical reference profiles
- estimated using a simple 1-D cloud model to satisfy the constraints on intensity

Evolution of cloud parameterization

a = cloud fraction I = condensate CP=convective precipitation

Prognostic = predicted Diagnostic = estimated

 Conceptualizing cumulus parameterization

 Since convective parameterization represents the effects of sub-grid scale processes

on the grid variables (implicit parameterization)

 Convective parameterization can be conceptualized in many ways and can be

separated into some basic types.

 Convective parameterization can be grouped as deep-layer control schemes and low

level control schemes.

 Deep layer control schemes relates the creation of CAPE by large scale processes to
the development of convection. These schemes could be termed “supply side”
approaches as it is assumed that convection consumes the CAPE that is created.

 Low level control schemes tie the development of convection to the initiation
processes by which CINE is removed.
 Point of uncertainties

•A number of uncertainties in modeling clouds and their associated processes exist.

•We do not adequately understand what determines the rate of entrainment of

“environmental” air into the updrafts, or how entrainment affects the evolution of a convective
cloud system.

• Cumulus entrainment entails the dilution of convective updraft by dry, cool environmental air.

•Current parameterizations incorporate the effects of entrainment through simple assumptions

• The environment of the hot towers is typically assumed to be uniform, but in reality its
properties vary on unresolved scales.

• Properties of the entrained air must therefore, depend on which part of the variable
environment in which an updraft happens to find itself. Also, representation of microphysical
processes is extremely crude.

• Cloud dynamics is highly simplified in large-scale models.

 Uncertainties in formulating clouds & associated processes

Examples for processes that need to be parameterized in the

atmosphere
Parametrization of cloud processes must be kept simple since the numerical
integrations must be carried out in a reasonable time.

Two simple parametrization schemes used in numerical models...

- First scheme: only condensation of water vapour is considered while clouds are not
parametrized. (scheme frequently used in numerical models).

- Second scheme: Cloud water is predicted and the release of precipitation is

parametrized in a simple way
 Scheme without cloud stage

Cloud stage is skipped and therefore, assumed that all the condensed water vapour is
immediately converted into precipitation (easily implemented in a numerical
model).

We consider the model eqns for large-scale values of temp & moisture which, in the
case of a grid-point model, prescribe the time changes of the temp and specif
humidity at the grid points.

r
= A(r) − C
t

where terms A(T) and A(r) accounts for all adiabatic & diabatic processes, except
condensation of water vapour. Condensation rate C is rate at which saturation
 mixing ratio r changes in saturated air:
sat

drsat
C =
dt
In numerical models the condensation rate is generally determined diagnostically.
Therefore, preliminary values T* and r* are first predicted by neglecting the effects
of condensation:

T * = T n −1 + 2tA(T)

If the air becomes supersaturated:

r *  rsat (T * )
T* and r* are adjusted to their saturation values, which are then the final values Tn+1
and rn+1 at time step n + 1

n +1 L r * − rsat (T * )
T =T +*

cp L drsat (T * )
1+
cp dT

r *
− r (T *
)
r n +1 =r +
* sat
L drsat (T * )
1+
cp dT
following from:
L
T = − r ;r = rsat (T n +1 ) − r *
cp
where rsat the saturation value is that for the corrected temp which is approximated
as:
n +1 drsat
rsat (T )  rsat (T ) +
*
(T * ) T
dT
and condensation rate C is given as:
1 r* − rsat (T * )
C =
2t 1 + L drsat
c p dT

Assumption:

condensated water vapour immediately falls out as rain, the precipitation rate
All the
is (units: mH2O sec-1):

1
P=
 water  Cdz
0
; ρwater is water density
Cloud parameterization scheme

Most numerical models consider the condensation processes as described before.

The cloud phase is skipped by assuming condensed water vapour falls immediately
out as precipitation.

Although the release of latent heat seems the most important process for the large-
scaleflow, formation of clouds & release of precipitation may also be important as
they area part of the hydrological cycle.

Consequently there have been several attempts to parameterize clouds &

precipitation processes (e.g. Kessler).

Kessler scheme designed for warm clouds, since it considers the release of
precipitation to be mainly due to coalescence; Ogura & Takahashi scheme designed
for cumulus clouds.
The scheme described here takes into account the two main processes important for
the release of precipitation —coalescence & the Bergeron process.

As the scheme considers cloud water as a further variable, the prognostic eqns are
therefore:

T L
= A(T) + (C − E cld − E rain )
t cp

r
= A( r ) − (C − Ecld − Erain )
t
rcld
= A(rcld ) + (C − E cld − E rain )
t
P 

 (G
1
= − E rain )
t water 0
rain

where r and rcld are mixing ratios for water vapour & cloud water respectively. Ecld &
Erain are evaporation rates of cloud water & rain water; Grain is rate of release of
precipitation due to conversion from cloud water to rain water.
 Parametrization of condensation C is done the same way as described before, and
evaporation of cloud water Ecld due to increase of the saturation mixing ratio is
similarly calculated.

Release of precipitation is parametrized as follows. Two types of precipitating clouds

are considered

(a) Clouds with low temps at cloud top: Ttop < Tcrit with Tcrit ~ -15°K. For this cloud type
it is assumed that the total liquid cloud water is immediately released as
precipitation.

(b) Clouds with temps at top above Tcrit : Ttop > Tcrit but with a high cloud water content
Wcld > Wcrit, with Wcrit ~2mmH2O, where:

top

r
1
W cld = dz
water base
cld
It is further assumed that cloud water in excess of critical value is instantaneously
removed as precipitation.

Besides condensation processes & precipitation processes, the following processes

are also considered:

(a) Evaporation of rain

(b) Collection of cloud water in non precipitating cloud layers by precipitation from
above (Grain)coll1
 Task of convection parameterization
in practice this means:

Activation Trigger function

* Determine occurrence/localization of convection

Distribution Cloud model/specific profile

* Determine vertical distribution of heating, moistening and momentum
changes

Intensity Closure

* Determine the overall amount of the energy conversion, convective

precipitation = heat release
 Summary of convective parameterization

 There is no “best scheme

 Computational time is unlimited – choice of detail depends on the scientific

question to be answered..

 Convection parameterizations need to describe the collective effects of an

ensemble of convective clouds on a model column, (e.g. profiles of condensation,
evaporation, and transport).

 A number of approaches to convection parameterization exist (some benefits and

other drawbacks)

 Basic ingredients to present convective parameterizations are a method to trigger

convection, a cloud model and a closure assumption

 Small changes in the scheme formulation can have big impacts (e.g. triggering,
microphysics, closure...)
 The mass-flux approach has been successfully applied to both interpretation of
data and convection parameterization

 Other new approaches are also being developed in recent years, including the
super-parameterization.

 Feedbacks on the hydrological cycle, other meteorological parameters & transport

properties
Fluid dynamics (continuous flow)
Example..

Consider an fluid parcel of volume V moving through space; its position can be given
as (x (t), y (t), z (t)). Assume we have a field φ = φ(x,y,z,t) and wish to determine the
changes in φ within the volume as it moves through space. This can be expressed
as φ(x (t), y (t), z (t), t).

We can write the coordinates as (xi (t), t) and track the change from (xi (t), t) to (xi (t +
δt), t + δt), where i = 1, 2, 3.

i.e.  = (xi (t + t),t + t) − (xi (t),t)

Using Taylor expansion:
   2 (t) 2
(x i (t + t),t + t) = (x i (t),t) + (x i (t + t) − x i ((t)) + t + − − − + 2 +,
x i t t 2
Now, dividing by δt, and letting it go to zero:

(x i (t + t),t + t) − (x i (t),t)   (x i (t + t) − x i (t))   2 (t)2 

= + +−− −+ 2 +
t  x i t t t 2 
 Letting δt → 0, we notice that:
(x i (t + t), t + t) − (x i (t), t) d
→ for t → 0;
t dt

( xi (t + t ) − xi (t ))
→ui fort →0;
t
where ui is the velocity of our volume.

Thus, we arrive at Euler’s eqn relating total derivatives to partial derivatives:

d  
= +ui
dt t xi
  
=u +v +w
x y z

= +u • ;
t
 Land surface processes

 SimplePhysics([simulate_cyclone, …]) → Interface to the simple

physics package.
 SimplePhysics.__call__(state, timestep) → Gets diagnostics from
the current model state and steps the state forward in time
according to the timestep.
 SlabSurface([tendencies_in_diagnostics, name]) → Calculate the
surface energy balance.
 SlabSurface.__call__(state) → Gets tendencies and diagnostics
from the passed model state.
 BucketHydrology([soil_moisture_max, …]) → Manages surface
energy and moisture balance
 BucketHydrology.__call__(state, timestep) → Gets diagnostics
from the current model state and steps the state forward in time
according to the timestep.
 Land surface processes (Cont’d..)

 Albedo: Ratio of total shortwave (0.3 - 3.0 mm) radiation flux reflected by a given
surface in all directions to the total downwelling solar flux.

Fraction of incoming solar radiation absorbed and not reflected goes to:

- warm the earth's surface (sensible heat)

- power the evaporation process (latent heat)

- furnish the energy to sustain various biochemical processes.

Here, we focus on the earth’s surface albedo

Mean albedo = average of reflectance factors in all spectral bands weighted by the
downwelling solar flux in each band.

Reflected radiance in each band measured with a radiometer. (Measurements can

provide estimates within 3% accuracy).

Alternatively, albedo can be measured by remote sensing (offers the advantage of

averaging albedo over large areas).

Several factors influence surface albedo (see Table below):

(i) time of day (i.e., the solar elevation or zenith angle); (ii) season; (iii) relief; (iv)
vegetative cover; (v) surface roughness and texture; (vi) soil and rock type; (vii) soil
moisture; and (viii) snow cover.

Ostensibly, vegetative cover, surface roughness and texture & soil moisture are
subject to alteration by human action
-
Land use changes have the potential to change albedo. Water and vegetation
surfaces have lower albedos than dry, bare soil surfaces.

Soil moisture is a crucial parameter in landuse-albedo relation (albedo increases

markedly with a decrease in soil moisture).

Forests have lower albedos than rangelands; in turn, rangelands have lower albedos
than dry bare ground.

Albedo and climate change

Precipitation occurs when the following conditions are met:

 Presence of sufficient moisture in the atmospheric column, supplied either

externally through advection of moist air or internally by evaporation from the
ground.
 Presence of an uplift mechanism strong enough to cause rapid cooling (e.g.
thermal, frontal, or orographic lifting, or as a result of the horizontal convergence
of two adjacent air masses) & subsequent condensation (and cloud formation) of
the moisture present in the atmospheric column.
 Growth of cloud elements to sizes large enough to precipitate (e,g. ice crystal or
coalescence processes).

Albedo & global energy budget

Use the law of energy conservation and Stefan-Boltzmann law to evaluate average
global energy.
 Qo  (1 − albedo) 
T = 4  
 4 
where T = temp (K); Qo = solar constant (1361 Wm-2); σ = Stefan-Boltzmann constant
(5.67037 ✕ 10-8 Wm-2K-4)


 Global energy budget

Energy absorbed = energy radiated

The rate of emission by the earth is assumed to be of a black body and, given by the
Stefan–Boltzmann law. Assuming the earth’s temp to be same everywhere, the rate
of emission is given by Te and Te the radiative equilibrium temp or effective
4

radiating earth temp.

In radiative equilibrium, the rate at which radiation is absorbed equals the rate at
which radiation is emitted. The radiative equilibrium earth temp is given by:
1 1
 Qo (1 −  ) 
4  1361Wm (1 − 0.3) 
−2 4
Te =      = 255 K
 4  −8 −2 −4
 4 5.67  10 W m K 
The equilibrium temperature is equivalent to an atmospheric temperature at
analtitude of about 5 km.

Example..

Suppose there was no atmosphere and the earth is entirely covered by oceans (α =
0.06). Determine Te and comment on your answer.
 Exercise:

1. Determine the earth's average albedo in order to produce an average global temperature
above the freezing point of water.

Soln: From the same equation:  Qo  (1 − albedo) 

T =4  
 4 

Rearranging, we solve for albedo using temp as an input:

4T 4
albedo = 1 −
 Qo
Freezing point of water is 32°F or 0°C (273.15 K).

Now: 4 5.67  10−8  (273.15) 4

albedo = 1 − = 0.072
1361
Thus, overall average albedo would need to be about 7% or less for the average global temp.


2. If the earth was completely black, what would its average surface temperature be?

Soltn: Assume “completely black” means albedo is equal to zero.

Now:  Qo  (1 − albedo)   1361 (1 − 0) 

T=  4 = 
4
−8  = 278.3 K
 4   4 5.67 10 
i.e. TC = 5.2° C or TF = 41.3° F
Role of land surface
Role of land surface (Cont’d)...
Land surface (cont’d..)
 Global budgets

• Land surface
- Net radiative flux at the surface (RS + RT) is downward. Small storage at the surface
(G) implies upward sensible & latent heat fluxes
• Bowen ratio: Land vs Sea
- Different physical mechanism control the exchanges at surface
*Continents: fast responsive surface; surface temp adjusts quickly to maintain zero
ground heat flux
*Oceans: large thermal inertia; small variations of surface temp allow imbalances on
a much longer time scale
*Surface fluxes & atmosphere
- Sensible heat (H) at the bottom = energy immediately available close to surface
 - Latent heat (LE) = delayed availability through condensation processes for the
whole tropospheric column

- Net radiative cooling of the whole atm is balanced by condensation & sensible heat
flux at the surface. Land surface processes directly affect H or indirectly affect
condensation, radiative cooling, ...)
Surface time scales

 Diurnal time scale

- Forcing time scale determined by quasi-sinusoidal radiation modulated by clouds

 Diurnal/weekly time scales

- Forcing time scale determined by “quasi-random” random precipitation

 Weekly/monthly time scale

- Internal time scale determined by the physics of soil water exchanges/transfer

- Evaporation time scale determined by the ratio (net radiative forcing)/available soil
water
 Hydrological cycle (HC)
Hydrological cycle traces the largest movement of any substance on earth.
Water has always had and will continue to have a controlling influence on
the earth's evolution.

HC influences climate in a variety of ways:

- Exchanges of moisture and heat btwn the atm & earth's surface
fundamentally affect the dynamics & thermodynamics of the climate
system.

- In terms of vapour, douds, liquid, snow and ice, as well as phase transitions;
water plays opposing roles in heating & cooling the environment.

NB: Evaporation & precipitation are major components of the global HC.

* Energy: needed to convert soil water to vapour. Mostly comes from

radiation absorbed by the surface (depends on surface albedo & other
factors).
Land surface affect the hydrological cycle in various ways:

- Extratropics = experience large seasonal fluctuations, soil stores some of the

precipitation it received in winter and returns it to the atm in summer.

Snow & land ice = must be inc!uded in land surface parameterization.

Not surprisingly, oceans are the dominant reservoir in the global water cycle (hold
over 97% of the world's water). Contrastingly, atm holds only 0.001%; rest is locked
up in ice caps, snow & underground storage.

* Surface albedo: determined by vegetation and soil conditions. Changes in vegetation

and soil moisture alter the partition btwn evaporation and runoff all which in turn
changes surface conditions
Hydrological cycle is indeed global –continents & oceans exchange water.

- Over the oceans, evaporation exceeds precipitation & difference contributes to

precipitation over land.

- Over land, 35% of the rainfall comes from marine evaporation driven by winds; 65%
comes from evaporation from the land. As precipitation exceeds evaporation over
land, excess water returns to the oceans as runoff. Mean residence time of water in
the atm and oceans is an important climate parameter

Realistic modelling of land surface processes is essential for successful simulation of

climate, but there are many obstacles.

To improve land surface biological models for GCMs we must learn how to combine
effects of processes from local to GCM grid scales rigorously.
Governing laws and equations
Governing laws and
equations