Sie sind auf Seite 1von 43

Unified Model Documentation Paper 020

CLASSIC Aerosol Scheme

UM Version : 10.1
Last Updated : 2014-12-05 (for vn10.0)
Owner : Andrew Jones

Contributors:
C. E. Johnson, N. Bellouin, P. S. Davison, A. Jones, J. G. L. Rae,
D. L. Roberts, M. J. Woodage, S. Woodward, C. Ordóñez and
N. H. Savage

Met Office
FitzRoy Road
Exeter
Devon EX1 3PB
United Kingdom

© Crown Copyright 2015

This document has not been published; Permission to quote from it must be obtained from the Unified Model
system manager at the above address
UMDP: 020
CLASSIC Aerosol Scheme

Contents
1 Sulphur Cycle 3
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Model considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Dry or gas phase oxidation of SO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.2 Wet or aqueous phase oxidation of SO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.3 Oxidation of DMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.4 Exchange processes between the sulphate modes . . . . . . . . . . . . . . . . . . . . . 9
1.4 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Dry deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6 Wet scavenging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.6.1 Rainout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6.2 Washout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.7 Radiative effects in the radiation scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.8 Effects on Precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9 Coupling with UKCA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9.2 One-way coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9.3 Two-way coupling (recommended) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2 Soot (Black Carbon) Aerosol Scheme 21


2.1 Using soot in the Unified Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Transformation of soot species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Dry deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Wet deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5.1 Scavenging by large scale rain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.2 Scavenging by convective rain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6 Radiative effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3 Biomass Smoke Aerosol Scheme 23


3.1 Using the biomass scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Transformation of biomass smoke species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.4 Dry Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Wet deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.6 Radiative and Precipitation effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4 OCFF Scheme 25
4.1 Using the OCFF Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Transformation of OCFF Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.4 Dry Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5 Wet Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.6 Radiative and Precipitation Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

5 Nitrate Aerosol 25
5.1 Chemistry and inter-modal processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.2 Radiative and Precipitation Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3 Logical variables controlling the nitrate scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.4 STASH items available with the nitrate scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

6 Mineral Dust scheme 28


6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.2 Dust Emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.2.1 Dust Emission From Seasonal Vegetation . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.3 Dry Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.4 Wet Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

6.5 Radiative Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31


6.6 Coding Consideration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.7 Use of the Dust Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.7.1 The two-bin variant scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.7.2 Preferential sources for HadGEM2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

7 Sea-salt Aerosol Scheme 34


7.1 Using the sea-salt scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.2 Calculation of sea-salt number concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.3 Radiative and Precipitation effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

8 Particulate matter (PM10, PM2.5 and TSP) diagnostics 35


8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
8.2 Using PM diagnostics in the UM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
8.3 Calculation of PM diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

9 Aerosol climatologies 39
9.1 Biogenic aerosol climatology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

10 Lateral boundary conditions (LBCs) 39


10.1 Generation of LBCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
10.2 Using LBCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
10.3 Stash codes for CLASSIC LBCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

11 References 40

2 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Preface

This documentation paper includes details of the Coupled Large-scale Aerosol Simulator for Studies In Climate
(CLASSIC) aerosol scheme. It does not cover any of the aerosol processes in the UK Chemistry and aerosols
(UKCA) scheme.

1 Sulphur Cycle

1.1 Introduction

The Sulphur Cycle is in UM Section 17 (control routine AERO CTL), which is called towards the end of routine
ATM STEP, after the model physics and tracer advection have been done. The Sulphur Cycle can be initiated
from emissions of anthropogenic sulphur dioxide, natural dimethyl sulphide, and/or volcanic sulphur, which are
inserted into the appropriate model level via calls to routine TRSRCE from AERO CTL. In the SULPHR routine,
3 modes of sulphate aerosol (Aitken, accumulation and dissolved) are generated from the oxidation of sulphur
dioxide gas; also dimethyl sulphide gas is oxidised to sulphur dioxide, sulphate, and methyl sulphonic acid
(MSA). Required oxidants (listed below) are accessed from ancillary files or by coupling with UKCA (see section
1.9 for details of coupling the aerosol scheme with UKCA). The Sulphur Cycle tracer variables are advected
by the tracer transport scheme, mixed and dry deposited in the boundary layer, mixed convectively, and wet
scavenged by the large-scale and convective precipitation as the model proceeds. There are also options to
allow the aerosol to affect the model’s radiation scheme (so-called ‘direct’ and ‘first indirect’ effects), and also
affect the model’s precipitation efficiency in the large-scale precipitation scheme (the ‘second indirect’ effect).
Ammonia may also be included as a prognostic variable to buffer the wet oxidation reaction and allow further
oxidation of SO2 by ozone to occur. Any other ‘free’ tracers may be chosen in addition (although the number of
vertical levels for all tracers must be the same).

1.2 Model considerations

When the Sulphur Cycle is run in the UM, some additional prognostic variables are invoked; they are:
1. The following 3D mass mixing ratio fields, subject to tracer advection, boundary layer mixing and dry
deposition, convective transport, and wet scavenging from large-scale and convective precipitation:
(a) S in SO2 (sulphur dioxide gas)
(b) S in Aitken mode SO4 (sulphate particles with lognormal size distribution:
median radius rAit = 6.5 10−9 m, geometric standard deviation σAit = 1.30)
(c) S in accumulation mode SO4 (sulphate particles with lognormal size distribution:
median radius racc = 95 10−9 m, geometric standard deviation σacc = 1.40
(d) S in dissolved mode SO4 (sulphate dissolved in cloud water droplets)
(e) S in DMS (dimethyl sulphide gas) if selected
(f) N in NH3 (ammonia gas) if ozone oxidation is selected
2. The following 3D field as a passive non-advected variable subject only to depletion and replenishment in
the chemistry routine:
(a) H2 O2 (hydrogen peroxide) mass mixing ratio
3. The following 3D fields which are passive non-advected variables read from oxidant ancillary files (unless
coupling with UKCA is on):
(a) OH (hydroxyl radical) concentration in molecules/cc
(b) H2 O2 (hydrogen peroxide) mass mixing ratio ‘background’ limit
(c) HO2 (peroxide radical) concentration in molecules/cc
(d) O3 (ozone) mass mixing ratio if selected (required for DMS oxidation)

3 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

4. 3D natural emissions field for volcanic sulphur in kg m−2 s−1 (assumed to be SO2 )
5. The following 2D emissions fields in kg m−2 s−1 :
(a) anthropogenic SO2 surface emissions
(b) anthropogenic SO2 higher level ‘chimney stack’ emissions (level can be specified in the UMUI)
(c) natural DMS (dimethyl sulphide) surface emissions (interactive or from ancillary file)
(d) ammonia surface emissions if selected with ozone oxidation

1.3 Chemistry

In routine SULPHR, dry or gas-phase oxidation of SO2 by OH occurs, in daylight only, to produce Aitken and
accumulation mode SO4 ; wet or aqueous phase oxidation of SO2 by H2 O2 and/or O3 occurs in cloud droplets
to produce dissolved SO4 . Exchanges between the different sulphate modes occur within the chemistry routine
by parametrizing evaporation, nucleation, diffusion, coagulation, and mode-merging. If DMS is included, gas
phase oxidation in daylight by OH and O3 occurs to produce SO2 , Aitken and accumulation mode SO4 and
methyl sulphonic acid (MSA), the latter being a sink of sulphur. Increments are calculated and stored through
the routine based on the amounts of each variable available at the start; updating is done at the end, after
adjusting, if necessary, to ensure that negative values are not produced.

1.3.1 Dry or gas phase oxidation of SO2

Production of sulphuric acid vapour (H2 SO4 ) In general, it is assumed that the rate of oxidation of SO2 is
proportional to the amount of SO2 available, i.e. a simple exponential decay equation :

δ{SO2 }/δt = −RAT E ∗ {SO2 } (1.1)

Integrating this gives an expression for the incremental change in SO2 over timestep ∆t:

∆{SO2 } = −[1 − exp(−RAT E ∗ ∆t)] ∗ {SO2 } (1.2)

Following the STOCHEM model (Collins et al., 1997, Stevenson et al., 1997), the equation for the rate of dry
oxidation in daylight only (in darkness it is assumed to be zero) is:

DRY RAT E = K SO2 OH ∗ [OH] (1.3)

where [OH] is the average daytime concentration of OH (in molecules per cc, obtained from the ancillary files or
the UKCA scheme) and

K SO2 OH = Klo/(1 + Krat ) ∗ 0.45 ∗ ∗{1/[1 + (log10 Krat /1.1904)2]} (1.4)

where
Krat = Klo /Khi is the ratio of reaction rate limits at low and high pressures
Klo = Nair * 4.0E-31 * (300/T)3.3 where T is temp in K
Nair = number of air molecules per cc
Khi = 2.0E-12 cc/mcl/s
Typical values of K SO2 OH are of order 10−12 cc/molecule.sec, and OH concentrations up to 106 molecules/cc.
This gives DRYRATE values of around 10−6 s−1 , which is very small, so it is acceptable to use an approximate
form of the series expansion of the exponential term in equation (1.2) to obtain:

∆{SO2 }DRY = −K SO2 OH ∗ [OH] ∗ {SO2 } ∗ ∆t (1.5)

If there is no sulphate aerosol already present in a gridbox, then the sulphate produced by gas-phase oxidation
is assumed to form Aitken mode sulphate. However, if there is pre-existing sulphate aerosol, then the sulphate

4 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

produced in this reaction is partitioned in the model between the Aitken and accumulation modes in a manner
described in the following paragraphs.

Partitioning of H2 SO4 between Aitken and accumulation modes The parametrization is based on the
condensation theory used in the GLOMAP model (Spracklen et al., 2005). The fraction of H2 SO4 condensing
onto the Aitken mode is given by
Pn
i=1 CAit,i NAit,i
fAit = Pn Pn (1.6)
i=1 CAit,i NAit,i + i=1 Cacc,i Nacc,i

where the summations are over all particle radius bins. NAit,i and Nacc,i are the number of particles in the ith
radius bins of, respectively, the Aitken mode and accumulation mode. CAit,i and Cacc,i are given by

C = 4πRwet DF (Kn)A(Kn) (1.7)

where Rwet is the wet radius of the particle, D is the diffusivity, and F(Kn) and A(Kn) are functions of the Knudsen
number, Kn. The diffusivity is given by:
   1/2
µair

3
 RT µair 1 + µH2 SO4
D=   (1.8)
8ρair NA d2H2 SO4 2π

where ρair is the density of air in kg m−3 , NA is Avogadro’s number, dH2 SO4 is the diameter of a molecule
of H2 SO4 , R is the gas constant, T is the temperature, µair is the molecular weight of air, and µH2 SO4 is the
molecular weight of H2 SO4 . F(Kn) and A(Kn) are given by

1 + Kn
F (Kn) = (1.9)
1 + 1.71Kn + 1.33Kn2

and

1
A(Kn) = 1.33KnF (Kn)
(1.10)
1+ Se −1

where Se is the sticking efficiency. The Knudsen number Kn is

λ
Kn = (1.11)
Rwet

where λ is the mean free path of H2 SO4 molecules in air, given by


  1/2
µH2 SO4 µair
λ= (1.12)
8NA2 d2H2 SO4 M mair µair + µH2 SO4

where M is the air density in mol m−3 and mair is the mass of an air molecule.
Because Rwet depends on relative humidity, and because of the dependence of D on temperature and air
density, and the dependence of λ on air density, CAit and Cacc will vary with pressure, temperature and relative
humidity. However, it is found that the dependences on pressure and temperature are not significant, and
only the dependence on relative humidity is important. The partitioning of newly-formed H2 SO4 between the
Aitken and accumulation modes is therefore parametrized in SULPHR by using look-up tables for ΣCAit NAit and
ΣCacc Nacc per unit mass mixing ratio, for different values of relative humidity at intervals of 0.01 between 0 and
1. The model uses the values of these sums corresponding to the relative humidity which is closest to the actual
relative humidity.

5 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Let A and B respectively denote the values taken from the look-up tables for ΣCAit NAit per unit mass mixing ratio
and ΣCacc Nacc per unit mass mixing ratio. Then, the fraction, Ψ, of the dry-oxidised SO2 becoming Aitken mode
sulphate is calculated in the model as
A[Ait]
Ψ= (1.13)
A[Ait] + B[acc]

where [Ait] and [acc] are the mass-mixing ratios of Aitken and accumulation mode sulphate respectively.

1.3.2 Wet or aqueous phase oxidation of SO2

Exponential decay equations similar to (1.1) and (1.2) are used for the wet oxidation of SO2 , but a suitable
value for the wet oxidation rate must be derived. In cloud, SO2 is oxidised by H2 O2 on an assumed average
timescale τchem =15 mins. This is shorter than the model timestep (30 mins), and shorter than the average time
for a circulating air parcel to remain in a cloud, which is taken to be τcloud = 3 hours (a reasonable value for
layer cloud, although it varies widely for different cloud types). The problem then is to determine the overall
rate of oxidation of SO2 molecules moving randomly around a model grid box with cloud fraction f. A statistical
approach is taken, beginning with the premise that

E(T ) = E(T |I)P (I) + E(T |O)P (O) (1.14)

Where:
E(T) is the expected lifetime of an SO2 molecule (the inverse of the oxidation rate we require)
E(T|I) is the expected lifetime for SO2 starting the timestep in cloud
E(T|O) is the expected lifetime for SO2 starting the timestep outside cloud
P(I) is the probability that an SO2 molecule starts the timestep in cloud
P(O) is the probability that an SO2 molecule starts the timestep outside cloud
If p is the probability that a molecule is oxidised during its passage through 1 cloud, then

p = 1 − exp(−τcloud /τchem ) (1.15)

Then assuming that the number of clouds a molecule of SO2 has to pass through before it is oxidised has a
geometric distribution, it can be shown that:

E(T |I) = τchem + τclear exp(−τcloud /2τchem )/p (1.16)

E(T |O) = τchem + τclear (1/p − 0.5) (1.17)


where:
τclear = τcloud (1 − f )/f (1.18)
is the average time spent in clear air.
Assuming that the probabilities P(O) and P(I) are proportional to the number of SO2 molecules in the clear and
cloudy fractions of the grid box, it follows that

P (O) = (1 − f )/(1 − f + f λ) (1.19)

P (I) = f λ/(1 − f + f λ) (1.20)

where λ is the ratio of the SO2 concentrations in the cloudy and clear parts of the grid box. Consideration of the
fluxes between the clear and cloudy air enables an expression for λ to be obtained:

2
2f λ = −1 + 2f − (1 − f )τratio + sqrt{1 + 2τratio (1 − f )(1 − 2f ) + τratio (1 − f )2 } (1.21)

where τratio = τcloud /τchem .

6 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

A value of the rate of wet oxidation can now be found, since

W ET RAT E = 1/E(T ) (1.22)

Typical values of WETRATE are not very small (of order 10−2 s−1 ), so the approximation to the exponential
series expansion cannot be made in (1.2) Allowance must also be made for the depletion of H2 O2 , so that the
equation for the change in c is:

∆{SO2}WET = −[1 − exp(−WETRATE ∗ ∆t)] ∗ {S} (1.23)

where {S} is either the amount of SO2 in the gridbox, or the molar equivalent amount of H2 O2 , whichever is the
smaller.

Depletion and replenishment of H2 O2 At the end of the SULPHR routine, after all increments have been
adjusted, H2 O2 is depleted by the molar equivalent of ∆{SO2 }WET . H2 O2 is naturally replaced in the atmo-
sphere, in daylight, from sources of the peroxide radical (HO2 ). Because the depletion rate is quite fast and the
replenishment rate relatively slow (and zero in darkness), it is not appropriate to simply reset the H2 O2 values
back to their reference values. The reaction rate for the replenishment (from the STOCHEM model) is:

d{H2O2}/dt = K HO2 HO2 ∗ [HO2 ] ∗ [HO2 ] (1.24)

where

K HO2 HO2 = [2.2E−13∗exp(600/T )+1.9E−33∗Nair∗exp(890/T )]∗[1+1.4E−21∗Nwater∗exp(2200/T )] (1.25)

T is the temperature in K
Nwater is number of water vapour molecules per cc
[HO2 ] is the concentration of HO2 (in molecules per cc, provided by ancil or coupling with UKCA)
The local value of H2 O2 is replenished using this equation in the cloud-free part of the grid box only, up to a
maximum value equal to the reference H2 O2 field.

Oxidation by O3 (optional) If H2 O2 is limiting and there is SO2 left unoxidised, then it is possible to continue
oxidation by ozone, if there is sufficient ammonia present to neutralise the sulphuric acid produced (to form
ammonium sulphate). This reaction is highly pH dependent, and a precise calculation of the pH (an iterative
scheme) is beyond the scope of the simple parametrization used here; instead, it is assumed that if sufficient
ammonia is present to neutralise all the SO2 in the gridbox, then O3 oxidation proceeds at the same WETRATE
as for H2 O2 . The amount of SO2 removed by this process is given by an equation similar to (1.23) where {S}
is now either the amount of SO2 in the gridbox, or the molar equivalent amount of NH3 , whichever is the smaller
(remembering that the ratio of S:N atoms in ammonium sulphate is 1:2). (It will be noted that oxidation by both
H2 O2 and O3 can occur on all the available SO2 , leading to an unrealistically large decrement to the SO2 field;
however, this is consistent with the philosophy of the whole SULPHR routine, where all increments are summed
and scaled down if necessary before updating at the end of the routine.)
A reference field of O3 is provided by ancillary files or the UKCA scheme, and a threshold value of 10ppbv is
required for oxidation to occur (which is found almost everywhere except around Antarctica). The NH3 field is
generated within the model from surface emissions in a similar manner to the SO2 field, undergoing advection,
boundary layer and convective mixing, wet scavenging and dry deposition. Neutralising oxidation products is
also assumed to be a sink of NH3 , as it is no longer available for buffering; this amount can be diagnosed for
mass balance calculations.

1.3.3 Oxidation of DMS

There are two versions of the DMS oxidation scheme available.

7 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Simple scheme A very simple scheme may be used if an ozone field is not available, but it is not the recom-
mended version. In the gas phase, and in daylight only, DMS is assumed to be oxidised by OH to produce SO2
and MSA in a constant proportion of 9 parts SO2 to 1 part MSA. This is an extreme simplification of a series of
chemical reactions which is poorly understood and very complex. An exponential decay rate equation for the
depletion of DMS is assumed, similar to (1.1), with a solution for the DMS increment:

∆{DMS} = −[1 − exp(−DMSRATE ∗ ∆T )] ∗ {DMS} (1.26)

where
DMSRATE = K DMS OH ∗ [OH] (1.27)

and a constant value of 9.1 E-12 cc/molecule.sec is assumed for K DMS OH. Because this is small, the first
order approximation to the exponential term in 3.3.1 can be made, to give:

∆{DM S} = −K DM S OH ∗ [OH] ∗ {DM S} ∗ ∆t (1.28)

In the update step at the end of the routine, SO2 is incremented by 0.9∆{DMS}, while the MSA produced (=
0.1∆{DMS} ) is a sink of sulphur; although not stored, this can be diagnosed for mass balance calculations.

Recommended scheme The recommended version is more complex and requires a 3D ozone field (as well
as OH and HO2) to parametrize DMS oxidation. Two independent reactions of DMS with OH are modelled
(representing the addition and abstraction pathways); CH3 SO2 and CH3 SO3 are assumed to be intermediate
species produced by the reactions, which then break down to produce SO2 , SO4 and MSA (so do not need to
be stored as model variables between timesteps).
The following daylight-dependent reactions are represented, with their respective reaction rates:

(i)
K1 DMS OH
DM S + OH −−−−−−−→ CH3 SO2 (1.29)

where K1 DM S OH = 1.13E − 11 ∗ exp(−254/T ), and T is temperature in K

(ii)
K2 DMS OH
DM S + OH −−−−−−−→ CH3 SO2 + M SA (1.30)

where K2 DM S OH = 1.7E − 42 ∗ [O2 ] ∗ exp(7810/T ))/{1.0 + 5.5E − 31 ∗ [O2 ] ∗ exp(7460/T )} and [O2 ] is the
concentration of oxygen (calculated as 20.95% air concentration).
The products of this reaction are assumed to be formed in the ratio 9 parts CH3 SO2 to 1 part MSA.
The total rate of oxidation of DMS is therefore :

DMSRATE = (K1 DM S OH + K2 DM S OH) ∗ [OH] (1.31)

And the total amount oxidised (by analogy with (1.28)) is given by:

∆{DM S} = −(K1 DM S OH + K2 DM S OH) ∗ [OH] ∗ {DM S} ∗ ∆t (1.32)

The fraction of this becoming CH3 SO2 is:

F DM S T O CH3SO2 = (K1 DM S OH + 0.9 ∗ K2 DM S OH)/(K1 DM S OH + K2 DM S OH) (1.33)

(iii)
K3 CH3SO2
CH3 SO2 −−−−−−−→ CH3 + SO2 (thermal decomposition) (1.34)

where K3 CH3SO2 = 2.6E11*exp(-9056/T)

8 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

(iv)
K4 CH3SO2 O3
CH3 SO2 + O3 −−−−−−−−−→ CH3 SO3 + O2 (1.35)

where K4 CH3SO2 O3 = 1.0E-14


So the fraction of CH3 SO2 producing SO2 is given by:

F CH3SO2 T O SO2 = K3 CH3SO2/(K3 CH3SO2 + K4 CH3SO2 O3 ∗ [O3 ]) (1.36)

where [O3 ] is the concentration of ozone (calculated from the ancillary mass mixing ratio field).
The fraction of oxidised DMS converted to SO2 is then:

F DM S T O SO2 = F DM S T OCH3SO2 ∗ F CH3SO2 T O SO2 (1.37)

(v)
K5 CH3SO3 HO2
CH3 SO3 + HO2 −−−−−−−−−−→ M SA + O2 (1.38)
where K5 CH3SO3 HO2 = 4.0E-11.

(vi)
K6 CH3SO3
CH3 SO3 −−−−−−−→ CH3 + SO3 (thermal decomposition) (1.39)
where K6 CH3SO3 = 1.1E17 * exp(-12057/T).
The SO3 thus produced is assumed to be converted instantaneously to sulphate aerosol, divided between the
Aitken and accumulation modes in proportion to their surface areas (using Ψ calculated in Equation (1.13) for
SULPHR).
The fraction of CH3 SO3 becoming SO4 is then given by:

F CH3SO3 T O SO4 = K6 CH3SO3/(K6 CH3SO3 + K5 CH3SO3 HO2 ∗ [HO2 ]) (1.40)

Where [HO2 ] is the concentration of HO2 (obtained from the ancillary file).
The fraction of oxidised DMS becoming SO4 directly (i.e. not via SO2 ) is then:

F DM S T O SO4 = F DM S T O CH3SO2 ∗ (1.0 − F CH3SO2 T O SO2) ∗ F CH3SO3 T O SO4 (1.41)

Finally, the fraction of oxidised DMS becoming MSA is the remaining portion:

F DM S T O M SA = 1.0 − F DM S T O SO2 − F DM S T O SO4 (1.42)

The MSA thus produced is treated as a sink of sulphur, and may be diagnosed for mass balance calculations.

1.3.4 Exchange processes between the sulphate modes

The following processes are parametrized:


1. Evaporation of dissolved sulphate in cloud-free grid boxes or parts thereof, to form accumulation mode
SO4 ;
2. Nucleation of cloud droplets by accumulation mode SO4 to form dissolved SO4 ;
3. Diffusion of Aitken mode sulphate into cloud droplets to form dissolved SO4 ;
4. Coagulation of Aitken mode with accumulation mode SO4 to form more of the accumulation mode SO4 ;
5. Mode-merging, which parametrizes the growth of Aitken mode SO4 to accumulation mode SO4 by con-
densation of H2 SO4 .

9 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

The same strategy is used to formulate equations for processes 1-3 as was used for the wet oxidation of SO2 ,
i.e. a statistical approach considering the overall circulation of aerosol around a grid box with a given cloud
fraction. However, some simplifications can be made if the timescale τ for the process under consideration is
very short.

Release of dissolved aerosol from evaporating cloud droplets First, any dissolved sulphate found in a
gridbox with negligible cloud water or ice is converted to the accumulation mode. Then, if the cloud fraction is
less than 0.95, it is assumed that some evaporation of dissolved sulphate will occur in the non-cloudy part of
the grid box. From (1.18), we find the minimum value for the time spent in clear air is:
τclear > τcloud ∗ 0.05/0.95 = 568s
which is larger than the typical timescale for a cloud droplet to evaporate, τevap = 300s
We need to find an overall evaporation rate for all dissolved sulphate in the grid box. Using an equation similar
to (1.2), the probability of evaporation occurring once dissolved sulphate enters clear air is then given by:

pevap = 1 − exp(τclear /τevap ) ≈ 1 (1.43)

Using equations analogous to (1.16) and (1.17) we then have:

E(Tevap |O) ≈ τevap (1.44)

E(Tevap |I) ≈ τevap + 0.5 ∗ τcloud (1.45)

and P(O) ≈ 0 , P(I) ≈ 1 are the probabilities of dissolved sulphate starting the timestep in clear air and in cloud,
respectively.
The overall timescale for evaporation is then given by:

E(Tevap ) ≈ τevap + 0.5 ∗ τcloud (1.46)

and the amount of sulphate ∆Sevap exchanged from dissolved to accumulation mode in this process is given by
:

∆Sevap = [1 − exp(−∆t/E(Tevap ))] ∗ {SO4 diss } (1.47)

where {SO4diss } is the mass mixing ratio of dissolved sulphate in the grid box.

Nucleation of accumulation mode aerosol Accumulation mode sulphate particles nucleate very rapidly on
entering cloud, with a typical timescale of τnuc = 30 secs. As this is very much shorter than the average time
spent in cloud ( τcloud ≈ 3 hours ), the working can be simplified as in section 1.3.4 above, so that:

E(Tnuc ) = τnuc + 0.5 ∗ τcloud (1 − f )/f (1.48)

and the amount of sulphate ∆Snuc exchanged from accumulation to dissolved mode in this process is given by:

∆Snuc = [1 − exp(−∆t/E(Tnuc ))] ∗ {SO4 acc } (1.49)

where {SO4 acc } is the mass mixing ratio of accumulation mode sulphate in the grid box.

10 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Diffusion of Aitken mode sulphate into cloud water This is a much slower process than the evaporation
and nucleation described above (typically several hours), and similar approximations cannot be made; instead,
the full probability arguments used in section 1.3.2 are used. Equations 1.14 to 1.21 are therefore applied with
τdif f , the diffusive lifetime of Aitken mode particles in cloud, replacing τchem . The method used to calculate τdif f
takes into account the relative motion of cloud droplets and aerosol particles, so that the diffusional scavenging
is enhanced due to the cloud droplets falling slowly relative to the aerosol (see Pruppacher and Klett). The
relative strength of the advective and diffusive terms is measured by the Peclet number Pe where:

P e = advective term/diffusive term = 2U ∗ R/D (1.50)

where
U is the fall speed of the cloud droplet
R is the droplet radius
D is the particle diffusion coefficient
Using the Stokes’ Law approximation for U (see Pruppacher and Klett p. 416, Batchelor p. 234),

U = 2g ∗ R2 ∗ ρ/9µ (1.51)

Where
g is the gravitational acceleration,
ρ is the density of water,
µ is the viscosity of air.

So
P e = 4g ∗ R3 ∗ ρ/(9µ ∗ D) (1.52)

A rough calculation using a typical value of R for clouds of 10−5 m, and D for Aitken mode sulphate particles
= 1.7E-9 m2 /s , indicates that Pe ≈ 150, so the advective contribution to the scavenging will be significant.
Equation [17-12] of Pruppacher and Klett gives an expression for the combined advective-diffusive flux, F, of
particles to a single droplet of radius R :

F = −δna /δt = 4π ∗ R ∗ D ∗ (1 + 0.5P e1/3 ) ∗ na (1.53)

Where na is the ambient aerosol number concentration.


This can be integrated over the whole range of cloud droplet sizes assuming R has the Khrgian-Mazin distribu-
tion (see Pruppacher and Klett p. 26), and remembering that Pe is also a function of R, to give an expression
for the average diffusional scavenging coefficient for Aitken mode particles as:

SCAV CD = 6π ∗ D ∗ N ∗ WRAD ∗ (1 + KPEC ∗ WRAD ) (1.54)

Where
WRAD = {W / (10 π*ρ*N) }1/3
KPEC = { 4 g *ρ / (9 µ*D) }1/3
The cloud droplet number N is calculated using the function NUMBER DROPLET (called in routine AERO CTL)
which either relates N to ambient aerosol concentrations or sets fixed values. The diffusivity D of particles of
radius r can be calculated using the formula (see Pruppacher and Klett):

D = kT ∗ (1 + α ∗ Kn )/6πµ ∗ r (1.55)

where
k is Boltzmann’s constant ( = 1.38*10−23 ),
T is temperature (K) and
α = 1.257 + 0.4 exp( -1.1 / Kn )
where Kn = λ / r is the Knudsen number, and
λ = 6.6E-8 *p0 T / (pT0 ) is the mean free path of the air molecules,

11 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

p, T being the pressure and temperature at the required level,


p0 , T0 the reference pressure (101325 Pa) and temperature (293.15K)
µ is the viscosity of air , which is given by: [1.718 + 0.0049 ∗ (T − 273.15)] ∗ 10−5
To obtain an effective value of D over all Aitken mode particles, we use the assumed lognormal size distribution
and also approximate α by its value α when r = r Ait (the median radius). This enables a value of the mass-
weighted average diffusivity D∗ to be calculated as:

2 2
D∗ = (kT /6πµ) ∗ {r−1 −2
Ait ∗ exp[−2.5(logeσAit ) + α ∗ λ ∗ rAit ∗ exp[−4.0 ∗ (loge σAit ) ]} (1.56)

Using this value of diffusivity in 1.54, a value of SCAVCD may be obtained; this then gives the diffusion rate for
Aitken mode particles into cloud droplets as τdiff = SCAVCD−1 . Equations similar to 1.14 to 1.21 can then be
used, with τdiff replacing τchem , to obtain a value for the overall diffusion rate DIFFRATE. Then the amount of
sulphate ∆Sdiff exchanged from Aitken to dissolved mode in this process is given by:

∆Sdiff = [1 − exp(−DIF F RAT E ∗ ∆t)] ∗ {SO4 Ait } (1.57)

where {SO4 Ait } is the mass mixing ratio of Aitken mode sulphate in the grid box.

Coagulation of Aitken mode particles onto accumulation mode sulphate Only Aitken-accumulation mode
coagulation is modelled; the Aitken-Aitken mode form is not parametrized. Assuming that each mode has a log-
normal size distribution and that all particles are spherical (even after coagulation), then the volume distribution
n(v) for each mode i can be shown to be given by (see Park et al., 1999):

ni (v) = Ni ∗ exp{−(loge v − loge vi )2 /18(logeσi )2 }/[3(2π)1/2 (loge σi ) ∗ v] (1.58)

where
Ni is the total number concentration for mode i, per unit volume
σi the geometric standard deviation of mode i
v is the volume of the particle
vi is the median volume of the distribution of particle mode i
The rate of transfer of volume from the Aitken to accumulation mode, ζ, is found by integrating:
Z ∞ Z ∞
ζ= vAit ∗ nAit (vAit ) ∗ nacc (vacc ) ∗ K(vAit , vacc )dvAit dvacc (1.59)
0 0

where K(vAit , vacc ) is the coagulation kernel.


Assuming that the particles are sufficiently small that Brownian coagulation is the dominant process, then where
rAit , racc are the radii corresponding to vAit , vacc respectively, we have (see Pruppacher and Klett p. 450, p. 460):

−1 −1 −2 −2
K(rAit , racc ) = (2kT /3µ) ∗ (rAit + racc ){rAit + racc + λ ∗ ACunn ∗ (rAit + racc )} (1.60)

where λ , k ,T and µ are as defined in (1.55) above, and ACunn is the Cunningham slip-flow correction, which we
assume= 1.591 for both particle modes (following Park et al.,1999).
Substituting r = c *v 1/3 where c = (3/4π)1/3 , and for K into (1.59) we obtain:

Z ∞ Z ∞
1/3 1/3
ζ =(2kT /3µ) ∗ nAit (vAit ) ∗ nacc (vacc ) ∗ vAit ∗ (vAit + vacc )∗
0 0 (1.61)
−1/3 −1/3 −2/3 −2/3
{vAit + vacc + λ ∗ ACunn ∗ (3/4π)−1/3 (vAit + vacc )}dvAit dvacc

Hence ζ is the sum of terms which are all of the form:


Z ∞ Z ∞
a b
Ia,b = { vAit ∗ nAit (vAit )dvAit } ∗ { vacc ∗ nacc (vacc )dvacc } (1.62)
0 0

12 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Using 1.58, we can evaluate these integrals, e.g.:

Z ∞
b
Jb = vacc ∗ nacc (vacc )dvacc
0
Z ∞ (1.63)
= Nacc /[3(2π)1/2 (loge σacc )] ∗ b−1
vacc ∗ exp{−(loge vacc − loge vacc )2 /18(logeσacc )2 }dvacc
0

Substituting x = loge vacc so that v = ex , dvacc = ex dx , and writing x = loge vacc

Z ∞
Jb = Nacc /[3(2π)1/2 (loge σacc )] ∗ ebx ∗ exp{−(x − x)2 /18(loge σacc )2 }dx
−∞
Z ∞ (1.64)
= Nacc ∗ ebx /[3(2π)1/2 (loge σacc )] ∗ exp{−y 2 /18(loge σacc )2 + by}dy
−∞

where y = x - x . This can be integrated to give:

b
Jb = Nacc ∗ vacc ∗ exp{4.5b2(loge σacc )2 } (1.65)

(which agrees with Park et al., 1999).


Hence
a b
Ia,b = NAit ∗ Nacc ∗ vAit ∗ vacc ∗ exp{4.5[a2 (loge σAit )2 + b2 (loge σacc )2 ]} (1.66)

And
ζ =(2kT /3µ) ∗ {2I1,0 + I4/3,−1/3 + I2/3,1/3 +
(1.67)
λ ∗ ACunn ∗ (3/4π)−1/3 ∗ [I2/3,0 + I4/3,−2/3 + I1/3,1/3 + I1,−1/3 ]}

is the rate of volume transfer to the accumulation mode. To obtain the rate of mass transfer, we multiply the
RHS of (1.67) by the density of the particles, ρ , and also rewrite the expression for Ia,b in terms of the mass
mixing ratios of the aerosol modes, mAit and macc . Then mi = Mi /ρair where Mi is the mass of aerosol in mode
i per m3 of air, and
Z ∞
Mi = ρ ∗ v ∗ ni (v)dv = ρ ∗ Ni ∗ vi ∗ exp{4.5(loge σi )2 } (1.68)
0

Rearranging,

N i = [ρair mi /(ρvi )] ∗ exp−4.5(logeσi )2 (1.69)

And substituting into (1.66) gives:

3(a−1) 3(b−1)
Ia,b =(4π/3)a+b−2 ∗ (ρair /ρ)2 ∗ mAit ∗ macc ∗ rAit ∗ racc ∗
2 2 2 2 2
(1.70)
exp{4.5 ∗ β ∗ [(a − 1)(loge σAit ) + (b − 1)(loge σacc ) ]}

So far the equations derived have been for dry aerosol particles; however, we can allow for hygroscopic growth
of the particles using the same parametrization (Fitzgerald, 1975 ) that is used in the radiation scheme. This
assumes that if a dry particle has radius rd and grows to radius r(S) at relative humidity S, then:

r(S) = α ∗ rdβ (for 0.81 ≤ S ≤ 0.995) (1.71)

where α and β are functions of S (see below, and note that α here is not the same as that in (1.55) and (1.56)
in the previous section).

13 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Then : loge r = loge α ∗ +βloge rd


So that if rd is lognormally distributed then so also is r. This means that we can use the same formula for ζ as in
the dry case (1.67), except that when evaluating Ia,b (from (1.70)) the parameters from the dry size distribution
should be replaced by those from the corresponding moist size distribution, so ri , loge σi are replaced by αrβi ,
βloge σi respectively. However, there is an extra complication in that we want to compute the rate of transfer of
dry aerosol mass, excluding any water associated with the particles. To take account of this, let vd Ait be the dry
volume of an Aitken mode particle, and vAit , vacc now represent the volumes of Aitken and accumulation mode
particles after hygroscopic growth. Then if ρ is the density of a dry particle, then the rate of (dry) mass transfer,
ξ , is given by:
Z ∞ Z ∞
ξ=ρ vd Ait ∗ nAit (vAit ) ∗ nacc (vacc ) ∗ K(vAit , vacc )dvAit dvacc (1.72)
0 0

where nAit , and nacc are still lognormal distributions but with their parameters adjusted for hygroscopic growth.
Now
vd Ait /vAit = (rd Ait /rAit )3 , and rd Ait = (rAit /α)1/β
So
3(1/β−1)
vd Ait = vAit ∗ α−3/β ∗ rAit = α−3/β ∗ vAit (3vAit /4π)(1/β−1) (1.73)

Substituting for vd Ait into (1.72), and writing γ = 1/β, we obtain an expression similar to (1.59) but with vAit
γ
replaced by vAit . We can therefore deduce an expressions for ζ in terms of I functions similar to (1.67), but with
the first index increased by (γ - 1):

ξ =(2kT ρ/3µ)α−3γ (3/4π)γ−1 ∗ {2Iγ,0 + I(γ+1/3),−1/3 + I(γ−1/3),1/3 +


(1.74)
λ ∗ ACunn ∗ (3/4π)−1/3 [I(γ−1/3),0 + I(γ+1/3),−2/3 + I(γ−2/3),1/3 + Iγ,−1/3 ]}

where, since we now need to use the adjusted size distribution parameters, (1.70) becomes:

3β(a−1) 3β(b−1)
Ia,b =(4π/3)a+b−2 ∗ (ρair /ρ)2 ∗ mAit ∗ macc ∗ α3(a+b−2) ∗ rAit ∗ racc ∗
2 2 2 2
(1.75)
exp{4.5 ∗ [(a − 1)(loge σAit ) + (b − 1)(loge σacc ) ]}

It remains to find the hygroscopic growth parameters α and β; the code can be simplified considerably without
much loss of accuracy by making the following approximations according to the value of the relative humidity S:
For S < 0.3 assume there is no hygroscopic growth, i.e. α = β= 1, and r = rd
For 0.3 ≤ S < 0.81 assume that all particles grow at the same rate, so β = 1, and obtain α(S) by linear
interpolation between its values at S=0.3 and S=0.81, the latter being derived from Fitzgerald’s formula ((1.77)
below), so:

α(S) = 1 + {(S − 0.3) ∗ [α(0.81) − 1]/0.51} (1.76)

For 0.81 ≤ S < 0.97, Fitzgerald’s formulae give:

α(S) = 1.2exp{0.066 ∗ S/(1.058 − S)}


(1.77)
β(S) = exp{0.00077 ∗ S/(1.009 − S)}

in which we can again make the approximation β = 1.


For S > 0.97 coagulation is neglected as other processes are likely to be dominant. Note that the hygroscopic
growth tends to reduce the coagulation rate sharply, at least for Brownian coagulation as assumed here.

14 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Mode-merging of Aitken to accumulation mode particles As sulphuric acid is deposited onto Aitken mode
particles, their radii will increase until they become accumulation mode particles. The Aitken and accumulation
mode size distributions adopted in the model overlap at a radius of approximately 7x10−8 m, so any increase in
the mass mixing ratio of Aitken particles which are above this size can be transferred to the accumulation mode.
The rate of deposition of sulphuric acid vapour onto a uniform size aerosol is given by:

F = −CN [H2 SO4 ] (1.78)

where N is the number of particles, [H2 SO4 ] is the sulphuric acid vapour concentration, and C is a constant
depending on the particle wet radius, the diffusivity of H2 SO4 in air and the Knudsen number. This theory is not
applied directly to the UM, because of the wide range of particle sizes in the Aitken mode. The UM employs a
parametrization of Equation (1.78) developed using a model representing the range of sizes in the distribution
with a large number of bins, and assuming that the deposition to aerosol is the same as the rate of oxidation of
SO2 and DMS to form H2 SO4 , which is not modelled. For each size in the model, the H2 SO4 concentration, and
hence the deposition flux to each size class of aerosol, was modelled using:

k1 [OH][SO2 ] = [H2 SO4 ]CN (1.79)


where k1 is the rate coefficient, and C and N depend on the size class.
The parametrization used in the UM proceeds by: 1) prediction of the split of the deposition between Aitken
and accumulation mode aerosol using a look-up table using the masses of the two modes; and 2) predicting
the mass transferred from the Aitken mode to the accumulation mode assuming that the effects of pressure,
temperature and humidity variations are small, and that this mass depends linearly on the deposition flux to the
Aitken mode. These parametrizations have been developed for the Aitken and accumulation modes described
above (Section 1.2), and will need to be modified if the size parameters are changed. The fraction of the Aitken
mode transferred to the accumulation mode at each timestep is therefore parametrized as a constant times the
deposition, p(Ait), to the Aitken mode so the change in the Aitken mode MMR, {Ait}, is:

∆{Ait} = 3.068 × 109 p(Ait){Ait} (1.80)


In practise this process leads to the loss of Aitken mode to accumulation mode with a timescale of around 1.7
days, much faster than the competing coagulation process.

1.4 Emissions

Emissions (derived from ancillary files) are added into the appropriate model levels by calling the routine
TRSRCE from AERO CTL before the chemistry routine SULPHR. TRSRCE converts emissions in kg m−2
sec−1 to mass mixing ratio (kg / kg dry air). Surface emissions (anthropogenic SO2 , land and ocean DMS,
and NH3 if selected) are added into the lowest layer of the model. Oceanic DMS emissions can be specified
directly from a surface-emissions ancillary file. Alternatively, one of two interactive approaches for oceanic DMS
emissions may be chosen. One calculates oceanic DMS emissions from the concentration of DMS in seawater,
surface windspeed and sea-surface temperature; see HCTN No. 47 (Jones and Roberts, 2003) for details. If
this approach is chosen, a further ancillary file of seawater DMS concentration is required. The other interactive
approach calculates oceanic DMS emissions from the ocean model, requiring that the model is run in coupled
atmosphere-ocean mode. One of three different parametrizations for sea-air exchange of DMS may be chosen
if an interactive approach to DMS emissions is taken. High level anthropogenic SO2 emissions are taken from
another 2-D ancillary file and are inserted at a user-specified model level which has, historically, corresponded
to an altitude approximately 500 m above the surface. The level specified will naturally depend on the model’s
vertical resolution. Finally, volcanic SO2 emissions are specified in a 3-D ancillary file.

1.5 Dry deposition

Dry deposition of ammonia gas (if included) and of all the sulphur species except DMS (for which deposition is
neglected) is parametrized using an approach analogous to electrical resistances (Seinfeld and Pandis, 1998).

15 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

In general, the total resistance to deposition Rtot is then given by:

Rtot = Ra + Rb + Rc (1.81)

Where
Ra is the local aerodynamic resistance (neglecting orographic drag),
Rb is the resistance to passing through the quasi-laminar layer next to the surface and
Rc is the resistance to uptake at the surface.
Values of Rb and Rc are obtained by scaling the values used for water vapour in the boundary layer scheme.
According to Monteith and Unsworth (1990),

Rb ∝ D−2/3 (1.82)
and
Rc ∝ D−1 (1.83)

where D is the molecular diffusivity in the case of gases, and

D ∝ M −1/2 (1.84)

where M is the molecular weight of the gas, or D is the particle diffusivity for the free sulphate particle modes.
From equations (1.82) and (1.84) we can find the following scaling factors for SO2 and NH3 :

RbSO2 /RbH2 O = [MSO2 /MH2 O ]1/3 = 1.53


RbN H3 /RbH2 O = [MN H3 /MH2 O ]1/3 = 0.981

The model does not explicitly compute a value of RbH2 O , so this has to be inferred from quantities that are avail-
able. Rb is related to the difference between the drag coefficients for momentum and heat; equation P243.61 of
UMDP-024 gives an expression for the ‘aerodynamic resistance’ for moisture, in which the resistance subsumes
both Ra and Rb in the sense in which they are used here. Hence we can rewrite P243.61 as:

Ra + RbH2 O = 1/(CH ∗ vshr )


where CH is the drag coefficient for heat,
and vshr is the magnitude of the surface wind shear.
Since Ra = 1/(CD ∗vshr ), where here CD is the drag coefficient for momentum neglecting orographic roughness,
RbH2 O can be calculated in the boundary layer routine from the following equation:

RbH2 O = (1/CH − 1/CD )/vshr (1.85)

In stable conditions CD > CH , so RbH2 O > 0. However, in unstable conditions CD < CH , so RbH2 O < 0 (see
UMPD No. 24); if this occurs, RbH2 O is set to zero for these calculations.
From equations (1.83) and (1.84) we can find:

RcSO2 /RcH2 O = [MSO2 /MH2 O ]1/2 = 1.89


RcN H3 /RcH2 O = [MN H3 /MH2 O ]1/2 = 0.972

RcH2 O can be found from the model on land points, as it is the inverse of the canopy conductance which is
available as a prognostic variable in the MOSES I and II boundary layer schemes. The canopy water variable
is also used to reduce the resistance to deposition for gases such as SO2 and NH3 which dissolve readily in
water, by a factor of:
1 − fcan ∗ 2/3

16 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

where fcan is the ratio of the canopy water to the surface capacity (where the latter is > 0.01). Similarly, because
of the high solubility of SO2 and NH3 , RcSO2 and RcNH3 are set to zero over oceans and lakes. For grid boxes
containing snow, the snowy and clear parts represent parallel resistances, so

1/Rc = fsnow /Rcsnow + (1 − fsnow )/Rcclear (1.86)


where
fsnow is the snow fraction (calculated from the snow depth in the model),
Rcsnow is the resistance to uptake over snow, taken as 103 s/m (Padro et al., 1993), and
Rcclear is the resistance to uptake over the snow-free surface (as calculated above).
Over sea ice, we ignore the leads and simply take Rc = Rcsnow .
For particles, the diffusivity depends on their size; for processes like dry deposition, which alter the mass
concentration, it is appropriate to characterise the particle size by the mass-weighted radius. For the Aitken and
accumulation mode particles

Rmass = r exp[1.5 ∗ (loge σ)2 ] (1.87)

so Rmass = 7.21 nm for Aitken mode particles,


and Rmass = 112.6 nm for accumulation mode particles.
To calculate the particle diffusivities, the Stokes-Cunningham formula is used (see equation (1.55) above),
assuming standard near-surface values of T and P.
This gives D = 2.53*10−4 cm2 s−1 for Aitken mode,
and D = 1.8389*10−6 cm2 s−1 for accumulation mode particles.
If the value of D for water vapour is taken as 0.23406 cm2 /s (Pruppacher and Klett p.413), the scaling factors
required for the particle modes are:

RbAit /RbH2 O = [DH2 O /DAit ]2/3 = 94.9


Rbacc /RbH2 O = [DH2 O /Dacc ]2/3 = 2530.0

For sulphate dissolved in cloud droplets, it is known that ‘dry’ (or ‘occult’) deposition to the surface is relatively
efficient (Monteith and Unsworth 1990), so Rb is set to zero for this mode.
For all particle modes, bouncing-off and resuspension from all surface types are neglected, and Rc is set to zero
universally. (This is a reasonable assumption, since sulphate particles are small enough to attach themselves
easily to dry surfaces, and will also dissolve readily in water.)
Note that in the MOSES II tiled land-surface scheme, the model arrays of Ra, RbH2 O and RcH2 O contain
values for each tile on land points. Therefore the total resistance to deposition (Rtot) for each grid-box must be
calculated by treating the tiled values as parallel resistances, i.e.:

1/Rtot = Σtile-fraction/(Ratile + Rbtile + Rctile ) (1.88)

where Σ represents the sum over all tiles in the grid-box (this is done in the subroutine SRESFACT).

1.6 Wet scavenging

There are two main mechanisms for the removal of species from the atmosphere by wet scavenging: ‘rainout’ or
‘in-cloud’ scavenging, in which dissolved species are removed as precipitation forms, and ‘washout’ or ‘below-
cloud’ scavenging, in which sulphur species are captured and removed by falling precipitation. Since the model
treats large-scale and convective cloud separately and differently, the parametrization of wet scavenging also
differs for the two types.

17 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

1.6.1 Rainout

The subroutine AEROSOL RAINOUT does the rain-out of dissolved aerosols by large-scale precipitation, amongst
them dissolved sulphate. The rain-out is done layer by layer, from the top of the atmosphere down to the surface.
For each atmospheric layer, the rained-out mixing ratio of dissolved aerosols (sulphate, here) depends on the
conversion rate of condensed water into precipitated water, β in s−1 , given by:

β = ∆Ls/[ρdry ∗ Qprevious ] (1.89)

where ∆ Ls is the net flux of large-scale precipitation going through the layer, ρdry is the density of dry air and
Qprevious is the amount of liquid and frozen water in the layer at the previous time-step. The rained-out mixing
ratio of dissolved sulphate, ∆{SO4 }, is then given by:

∆{SO4 } = {SO4 } ∗ [exp(−fdiss ∗ β ∗ t) − 1] (1.90)

where {SO4 } is the mixing ratio of dissolved sulphate before the rain-out, fdiss is the fraction of dissolved
sulphate that is dissolved in the cloud droplet and t is the length of a time-step, in seconds. Obviously, fdiss
is 1, but this parameter may be used to adjust the strength of the rain-out in the future. ∆{SO4 } is negative
(sink). Accumulated rained-out mixing ratio, ∆{SO4 }cumul (variable RNOUT AERO in the routine) is computed
by adding −∆{SO4 } from each layer, including the current one. ∆{SO4 }cumul is positive, by convention.
All the precipitated water does not reach the surface: some of it may re-evaporate in the layer, translating
into a negative ∆Ls. If it is so, the positive fraction of precipitated water that re-evaporates is given by fre
= -∆Ls / Ls(layer), where Ls(layer) is the flux of large-scale precipitation that falls out of the layer. fre ranges
between 0 (no re-evaporation) and 1 (total re-evaporation). The re-evaporated mixing ratio of dissolved sulphate,
∆{SO4 }re , is given by:

∆{SO4 }re = fre ∗ ∆{SO4 }cumul /ρdry (1.91)


∆{SO4 }re is positive and is added to the accumulation-sulphate mixing ratio. The re-evaporation process then
moves mass from the dissolved to the accumulation mode. The accumulated rained-out mixing ratio is corrected
for the re-evaporation. Note that in the code, the variable BETA is re-used to hold the value of fre . If the
re-evaporation is not total, fre is (arbitrarily) divided by two to account for those droplets that shrink without
completely evaporating.
For convective precipitation there is no explicit link with the dissolved sulphate (as this is associated only with
stratiform cloud), so all sulphate particle modes are removed using a simple linear scheme (in routine SCN-
SCV2, called from NI CONV CTL) based on the theory of Scott (1982):

δ{SO4 }/δt = −λC ∗ Ψ ∗ {SO4 } (1.92)


where {SO4 } is the mass mixing ratio of S for each SO4 mode, Ψ is the total convective precipitation rate (rain
and snow, in mm/hr), and λC = 0.3E-4 s−1 is a parameter dependent on the horizontal grid size.
Integrating this gives an expression for the incremental change in SO4 over timestep ∆t (in seconds) at all levels
between cloud base and cloud top:

∆{SO4 } = −[1 − exp(−λC ∗ Ψ ∗ ∆t)] ∗ {SO4 } (1.93)


We also incorporate the assumption that the convective precipitation occurs over just 5% of the horizontal grid-
box area by dividing Ψ by 0.05 and multiplying ∆{SO4 } by 0.05.

1.6.2 Washout

For all the sulphate particle modes, ‘washout’ by all types of precipitation is neglected, as it is generally consid-
ered less important than ‘rainout’. For the gaseous species (SO2 and NH3 ), scavenging by rain is parametrized

18 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

in a more complex way to reflect their different solubility properties and the possibility of raindrops becoming
saturated. For SO2 we assume the rate of removal is given by:

δS/δt = −Λ(R, S) ∗ S (1.94)


where
S is the concentration by volume (in ppbv) of SO2 ,
R is the rain rate in mm/hr (here, snow is neglected),
and the scavenging function Λ(R,S) is defined as:

Λ(R, S) = λ0 R2/3 f orS ≤ S0


2/3
(1.95)
= λ1 (R/S) f orS > S0

where the threshold concentration value S0 and the scavenging parameters λ0 and λ1 must be chosen so such
that Λ(R,S) is continuous. Thus,

−2/3
λ0 = λ1 S0 (1.96)

Assuming that for very small values of S the scavenging process is controlled by the gas-phase transport, and
comparing with results for HNO3 when R = 1 mm/hr (Levine and Schwartz, 1982), which has similar diffusivity,
we obtain
λ0 = 6.5E-5 s−1
To find λ1 , we use data from Adamowicz (1979) which gives the e-folding times for the reduction in SO2 for a
range of concentrations, at a rain rate of 1 mm/hr. The scavenging function Λ is then the inverse of this e-folding
time. Curves in Fig. 17 of Adamowicz (1979) give
For 1 < S < 10 ppbv, e-folding time t = 80 hours, Λ= 1/t = 3.4782E-6 s−1
For 10 < S < 100 ppbv, e-folding time t= 30 hours, Λ = 1/t = 9.259E-6 s−1
So the average scavenging rate for S = 10 ppbv is 6.3655 E-6 s−1 ; substituting this into (1.90) we obtain:
λ1 = 102/3 6.3655 E-6 ≈ 2.955 E-5 s−1 .
Then from (1.91), we obtain S0 = (λ1 /λ0 )3/2 = 0.3065 ppbv
For ammonia, the calculation is simplified by the assumption that it is sufficiently soluble that the raindrops
never become saturated, so the rate of washout is independent of its concentration. Then, where A is the
concentration of ammonia:

δA/δt = −λa R2/3 A (1.97)

Using Eq. (2) of Levine and Schwartz (1982), we assume that the ratio of the scavenging functions for NH3 and
HNO3 scale as the ratio of their diffusivities to the 2/3 power , to obtain a value of λa = 1.0E-4 s−1 .
For large-scale precipitation, we scavenge SO2 and NH3 at all levels where the rain is non-zero, provided it
reaches the ground. For convective precipitation, we scavenge at all levels from the ground to the cloud top,
again making the assumption that cloud only occupies 5% of the grid-box area.
Finally, wet scavenging of DMS is not represented in the model.

1.7 Radiative effects in the radiation scheme.

There are two ways in which sulphate aerosol from the Sulphur Cycle can be made to alter radiative fluxes in
the model and thus alter the meteorology of an experiment:

19 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Direct effect The so-called ‘direct’ effect of sulphate aerosol produced by the Sulphur Cycle can be repre-
sented in the radiation scheme.
For this purpose, the particles are assumed to be composed of a mixture of ammonium sulphate and water,
the refractive indices of ammonium sulphate at selected wavelengths being taken from the paper by Toon et
al. (1976). The particles are also assumed to be spherical, so that Mie theory can be used to calculate the
scattering and absorption coefficients. This is done separately for the Aitken and accumulation modes, which
are assigned the lognormal size distributions stated previously (section 1.2). However, these are the distributions
of the dry particles; for radiative purposes the hygroscopic growth of the particles is taken into account so that
their effective sizes are often larger, by an amount that depends on the relative humidity in the model grid
box. The growth function is based partly on the parametrization of Fitzgerald (1975), which has the convenient
property that the size distribution of the moist aerosol continues to be lognormal.
It should be noted that the Mie theory calculations are not performed during the model integration, but be-
forehand; the results are supplied to the model in the form of look-up tables of the aerosol single scattering
parameters for both the Aitken and accumulation modes, which are included in the spectral files that control the
configuration of the radiation scheme. For computing the specific extinction and scattering coefficients stored in
spectral files, ammonium sulphate density is taken at 1769 kg m−3 .

First indirect effect The ‘first indirect’ effect can also be represented, in which the sulphate aerosols from
the Sulphur Cycle are used to determine cloud droplet number concentrations, thereby affecting cloud albedo.
The modelling of this effect is described in Jones et al. (2001), although this describes the parametrization as
used in an earlier version of the Aerosol Modelling scheme. The current version uses the same procedure to
calculate droplet number with the exceptions (1) that the sulphate size parameters used are consistent with
those in section 1.2, and (2) that Aitken mode sulphate is not used in the calculation, as such particles are too
small to act as cloud condensation nuclei (CCN).
Further details may be found in the radiation documentation paper ( UMDP-023 ).

1.8 Effects on Precipitation

Sulphate aerosol can also affect the model’s meteorology via the ‘second indirect’ effect, this time by determining
cloud droplet number concentration in the large-scale precipitation scheme. This affects the rate at which cloud
water is converted to rain via the autoconversion process, thereby affecting cloud water contents and lifetime.
Cloud droplet number is calculated as for the first indirect effect; see Jones et al. (2001) and the large-scale
precipitation documentation paper ( UMDP-026 ) for details.

1.9 Coupling with UKCA

1.9.1 Introduction

There is an option to couple the UKCA tropospheric chemistry scheme and the CLASSIC sulphate aerosol
scheme. This allows the concentrations of sulphur-cycle oxidants calculated by UKCA to be used in oxidation
reactions in the sulphur-cycle scheme, where prescribed monthly-mean concentrations from ancillary files were
previously used. The coupling may be one-way, i.e. the oxidant concentrations from UKCA are passed to the
sulphur scheme but are not depleted by the reactions there and passed back to UKCA, or two-way, in which the
depleted oxidant concentrations are passed back, thereby completing the feedback loop.

1.9.2 One-way coupling

With one-way coupling, the sulphur-cycle scheme uses oxidant concentrations and MMRs from UKCA instead
of the prescribed monthly-mean values, but there is no feedback to UKCA. The H2 O2 MMR is still an upper limit,
and H2 O2 is still depleted and replenished in the sulphur-cycle scheme in exactly the same way as described
above. However, this has no impact on H2 O2 in the UKCA chemistry scheme. The reading of data from the
UKCA arrays and input into the arrays used by the sulphur cycle is performed in subroutine get sulpc oxidants

20 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

which also converts the concentrations from mass mixing ratios to molecules cm−3 as used in the sulphur cycle
routines.

1.9.3 Two-way coupling (recommended)

With the two-way coupling, the depletion of the oxidants is carried out in routine SULPHR. HO2 is removed only
in the DMS oxidation scheme. H2 O2 is removed only by reaction with SO2 . OH and O3 are removed through
oxidation of both DMS and SO2 . In the case of OH and O3 , the rates of removal through oxidation of DMS and
SO2 are calculated separately. It is possible for the total amount removed calculated in this way to be greater
than the amount available in that gridbox. If this is the case, then the removal rates in that gridbox are reduced
by multiplying by an appropriate scaling factor. This is analogous to what is already done for the sulphur-cycle
species in routine SULPHR.
The depleted concentrations and MMRs are then passed back up to AERO CTL and then to ATM STEP, from
where they are passed to a new routine called WRITE SULPC OXIDANTS which converts the concentrations
from molecules cm−3 to mass mixing ratio and updates the concentration arrays used in UKCA.
The replenishment of H2 O2 through the reaction of HO2 with itself is not included in the sulphur-cycle scheme
when two-way coupling is used, as this reaction is already included in the UKCA chemistry scheme.

2 Soot (Black Carbon) Aerosol Scheme

2.1 Using soot in the Unified Model

UMUI specifications Soot modelling has been included in Section 17 of the Unified Model, alongside the
Sulphur Cycle. The two aerosol schemes act independently, i.e. the user can choose either, both or neither,
and the model’s free tracer variables can be used in addition to any of these combinations. As with the sul-
phur cycle, the user must specify an ancillary file containing the emissions data. The routines AGESOOT and
SOOTDIFFSCAV are called from the Section 17 aerosol control routine AERO CTL, described in section 1 of
this paper.

Model variables If soot modelling is chosen the following additional prognostic variables are invoked:
• Fresh soot surface emissions, if selected (2D field)
• Fresh soot elevated emissions, if selected (2D field)
• Fresh soot mass mixing ratio (3D field)
• Aged soot mass mixing ratio (3D field)
• Soot in cloud water mass mixing ratio (3D field)
Both fresh and aged soot are assumed to have lognormal size distributions, with a median radius of 40 nm and
a geometric standard deviation of 2.0 in each case. This is an idealisation of what is a very complex situation in
the real atmosphere. These assumptions impact parameters used in the calculation of rates of dry deposition
and diffusional scavenging as well as the parametrized optical properties used in radiation calculations. It is
important, therefore, that the parameters used in these sections are consistent. See sections 2.3,2.4 and 2.6
for further details.

2.2 Emissions

The soot may be used as a surface-level source within the model, an elevated source, or both. Methods of
introducing the emissions into the model are the same as for the Sulphur Cycle (see section 1.4). Note that
surface level emissions are in fact introduced at model level 1. Elevated emissions may be injected at any model
level, but at no more than one. There is an argument in favour of injecting some fossil fuel emissions at a level
representative of tall chimneys.

21 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

2.3 Transformation of soot species

Three varieties of the aerosol are transported using the model’s tracer advection scheme. It is emitted as fresh
soot, considered to be hydrophobic and therefore unaffected by wet deposition processes. It decays with an
exponential conversion rate to aged soot, according to the equation

δS/δt = −kS (2.1)

where S is the mass mixing ratio of fresh soot, and k is a rate constant, set as k = 1.157E-5 s−1 (giving an
e-folding timescale for the conversion process of 1.0 days). Note that this is a change from version 4.5 of the
model, where the e-folding timescale was 1.6 days.
This can be solved to give the amount of soot ∆S converted per timestep ∆t as:

∆S = [1 − exp(−k∆t)] ∗ S (2.2)

Aged soot represents an internal mixture of the soot particles with hygroscopic material, such as sulphate
droplets. (Note that, despite the theoretical interaction with sulphate aerosol, no aerosol is removed from the
Sulphur Cycle to form these composite particles.)
Finally, some aged soot becomes internally mixed with cloud water, forming soot in cloud water. This transfor-
mation uses a parametrization of diffusional scavenging of the same form as that used for Aitken mode sulphate.
See 1.3.4 for details.
The calculated rate of diffusional scavenging depends on cloud droplet concentration, which is influenced by
the choice of whether other aerosols are used in the calculation of droplet number. Soot itself is assumed not
to nucleate cloud droplets, and thus does not influence droplet number.

2.4 Dry deposition

The soot in all three of its forms is affected by dry deposition. This is parametrized, as in other studies (e.g.
Seinfeld and Pandis 1998), by analogy to electrical resistance. The aerodynamic resistance Ra, the quasi-
laminar resistance Rb and the surface resistance Rc act in series. Ra represents the resistance to motion
through the atmosphere’s surface layer and is a function of wind speed and surface drag. Rb, the resistance
to penetration of the laminar sub-layer, is proportional to D−2/3 , where D is the particle diffusivity coefficient.
These coefficients depend upon molecular mass and upon particle radius. More details on the calculation of the
diffusivity coefficients are available in the Sulphur Cycle documentation (see section 1.5). Rc is zero for particles,
i.e. particles are not considered to bounce back, having impacted upon the surface. (For gases, Rc would adopt
a value representative of intake by the stomata of the surface vegetation.) As with sulphate particles, the mass-
weighted radius is used as a characteristic size; with the assumed (number) size distribution, we obtain:
Rmass = 82.23 nm for both modes of soot.
The associated diffusion coefficients are:
D = 2.96366*10−6 cm2 /s
This gives Rb/RbH2 O = 1840.77 for both modes of soot.
These parameters are held in include file C ST BDY, and would need to be changed should the assumed soot
size distributions change in the future.

2.5 Wet deposition

Because the model treats large-scale cloud and convective cloud separately and quite differently, the parametriza-
tion of wet deposition also differs in the two cases.

22 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

2.5.1 Scavenging by large scale rain

Washout Code has been written to perform washout (scavenging of particles by falling precipitation) of aged
(hydrophilic) soot. Whereas this is undoubtedly a process taking place, the functionality is turned off since its
importance is smaller than that of rainout. This has been implemented by setting scavenging coefficients to
zero.

Rainout See section 1.6 for details of the parametrization of rainout .

2.5.2 Scavenging by convective rain

The soot in cloud water variable is considered to be associated with stratiform cloud only; there is no explicit link
with convective cloud, so convective scavenging cannot be parametrized in the same way. Instead we adopt
the method used in the Sulphur Cycle (see section 1.6), neglect ‘washout’, and model ‘rainout’ for aged soot by
applying the equation

δ{Saged }/δt = −0.108 ∗ P ∗ {Saged } (2.3)

where P is the total surface convective precipitation rate (rain and snow) in units of kg m−2 s−1 , between cloud
base and cloud top.

2.6 Radiative effects

Soot in all of its modes is an efficient absorber in the short-wave region of the spectrum. As such, it plays
an almost unique role among aerosols; others, such as sulphate mainly scatter incoming radiation. Soot’s
absorption properties allow it to provide a net positive forcing for the climate system, opposite in sign to the
forcing supplied by sulphate aerosol.
The direct effects of the fresh and aged modes of soot may be selected. Soot optical properties (specified in
the radiation scheme’s spectral files) are computed for a lognormal size distribution with mean radius of 0.04
µm and standard deviation of 2. The set of refractive indices for soot aerosols is taken from WCRP (1986).
Soot aerosols are assumed not take up water vapour and hygroscopic growth is not included in their optical
properties. Soot aerosol density is 1900 kg m−3 .
As noted above, soot is assumed not to nucleate cloud droplets, and so does not have ‘indirect’ effects on
clouds.

3 Biomass Smoke Aerosol Scheme

3.1 Using the biomass scheme

In many ways the biomass aerosol scheme is very similar to the soot scheme described in Section 2. The
scheme may be used alone or in addition to sulphur and/or soot. Routines AGEBMASS and BMASSNUCLSCAV
are called from the main aerosol control routine AERO CTL in Section 17.
If the scheme is selected, additional prognostic variables are used exactly as for the soot scheme. The assumed
sizes of the particles are different, although still assumed to obey log-normal distributions:
Fresh smoke - median radius 0.1 µm, geometric standard deviation 1.30;
Aged smoke - median radius 0.12 mum; geometric standard deviation 1.30.

Should these assumptions change it is important that consistent changes are made to the deposition coefficients
detailed in section 3.4 and the optical properties used in radiation calculations (section 3.6).

23 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

3.2 Emissions

The user will be asked to specify an ancillary file containing emissions data. As for soot, emissions sources may
be at the model surface, on higher model levels, or both. Surface emissions are in fact introduced on model
level 1.
Unlike soot, elevated emissions of biomass aerosol may be introduced on more than one model level. The total
rate of emission (in kg m−2 s−1 ) is specified by data in the emissions ancillary file. The user is asked to specify
a range of model levels on which elevated emissions will be introduced and this total rate is then divided equally
across all model levels in this range. The rate of emission in kg m−2 s−1 on each of these model levels is thus
equal, but because grid boxes on higher model levels have larger horizontal area, the mass of tracer emitted per
unit time will in fact be greater on higher model levels than lower levels; however, this is only a marginal effect.

3.3 Transformation of biomass smoke species

As for the soot scheme, fresh biomass smoke is converted to aged smoke at an exponential rate with an e-
folding time of 6 hours.
The biomass-aerosol mass is increased upon ageing by a factor of ∼ 1.62 (=8.75/5.4); this of course means that
the scheme does not conserve mass. The mass increase is to account consistently for the fact that the mass-
fraction of black (elemental) carbon assumed for aged biomass smoke (5.4%) is less than that of fresh biomass
smoke (8.75%). This reduction in the mass-fraction of black carbon is assumed to be caused by the deposition
of volatile organic compounds (VOCs), which are assumed always to be present; as VOCs are released during
biomass-burning, this is a plausible assumption.
Aged biomass smoke is assumed to act as CCN, i.e. it is converted to smoke-in-cloud-water by nucleation
scavenging, in a manner analogous to accumulation mode sulphate aerosol; see section 1.3.4 for details.

3.4 Dry Deposition

All three modes of smoke are subject to dry deposition. Again, the parametrization used is similar to that in the
soot scheme, although the deposition coefficients differ from those for soot because the assumed particle size
distributions are different. For biomass smoke we use:
Fresh smoke: Rb/RbH2O = 2492.4
Aged smoke: Rb/RbH2O = 2970.7
These are held in include file C BM BDY. Changing the size distribution of biomass-burning aerosols requires
an update to the above values.

3.5 Wet deposition

Once again, we use a similar treatment to that of soot. Large-scale washout of biomass smoke is neglected,
while large-scale rainout and convective scavenging are calculated as described in section 2.5.

3.6 Radiative and Precipitation effects

Biomass smoke both absorbs and scatters solar radiation and will thus exhibit a direct radiative effect, which may
be selected. The size distribution of the two non-dissolved modes of biomass-burning aerosols are assumed
lognormal. Mean radius for fresh biomass-burning aerosols is 0.1 µm with standard deviation 1.3. Numbers
for the aged mode are 0.12 µm and 1.3, respectively. Size distributions, hygroscopic growth curves, and the
set of refractive index for fresh and aged biomass aerosols come from aircraft measurements made during the
SAFARI-2000 field campaign (Haywood et al., 2003). Biomass-burning aerosol density is 1350 kg m−3 .
Additionally, since biomass smoke may act as CCN, there may be indirect effects associated with it. Such
effects may be selected, although biomass smoke is assumed to be supplementary to sulphate aerosol, i.e. the
relevant effect must be active for sulphate aerosol for it to be available for biomass smoke.

24 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

4 OCFF Scheme

4.1 Using the OCFF Scheme

The scheme to model organic carbon aerosols derived from fossil-fuel combustion (OCFF aerosols) is based
closely on the biomass-burning and black carbon (soot) aerosol schemes, but naturally having different emission
sources.
If selected, three new prognostic variables are introduced, namely the mass mixing ratios of fresh OCFF, aged
OCFF and OCFF in cloud droplets. The free particle modes are assumed to follow log-normal distributions with
size parameters the same as those assumed for biomass-burning aerosols.

4.2 Emissions

Ancillary files must be specified for OCFF emissions, in units of kg of carbon m−2 s−1 . The treatment of
emissions follows that of the black carbon scheme: emissions may be specified at the surface and/or at a single
elevated model layer.

4.3 Transformation of OCFF Species

Fresh OCFF aerosol is converted to aged OCFF at an exponential rate with an e-folding time of 1 day (as with
black carbon). Aged OCFF is assumed to act as a CCN, i.e. it is converted to OCFF in cloud droplets by
nucleation scavenging (as with biomass-burning aerosol).

4.4 Dry Deposition

All three modes are subject to dry deposition which is handled exactly as for biomass-burning aerosols.

4.5 Wet Deposition

This is performed in the same manner as for biomass-burning aerosol.

4.6 Radiative and Precipitation Effects

Again, the treatment of these effects is the same as those for biomass-burning aerosol, and size distributions of
fresh and aged OCFF aerosols are the same as the equivalent mode of biomass-burning aerosols. Hygroscopic
growth rates are also taken from biomass-burning aerosols. The real part of the refractive index of OCFF
aerosols is assumed to be that of aged biomass-burning aerosols, with an imaginary part set at 0.006i across
all wavelengths, leading to less absorption than aged biomass-burning aerosols. OCFF aerosol density is 1350
kg m−3 .

5 Nitrate Aerosol

In this scheme there are two modes of ammonium nitrate: accumulation mode (with the same size distribution
as accumulation-mode ammonium sulphate) and dissolved mode. There are no direct emissions of ammonium
nitrate in the model. It is formed from nitric acid (HNO3 ) and ammonia (NH3 ). HNO3 concentrations are obtained
online from the UKCA chemistry scheme. Ammonia is already present in the HadGEM sulphur cycle. At the end
of the sulphur cycle, the ammonia concentration is depleted to take account of ammonium sulphate production;
the depleted ammonia concentration is then passed to the nitrate scheme for formation of ammonium nitrate.
The chemistry involved is described in the Chemistry and inter-modal processes section below. Both modes

25 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

of ammonium nitrate are subject to scavenging by convective precipitation and to dry deposition. Dissolved
nitrate is also scavenged by large-scale precipitation (rainout only, not washout). As tracers, both modes are
also transported by advection and convection. The wet and dry deposition processes are treated in the same
way as for accumulation-mode and dissolved sulphate.
Because the size distribution is the same as for sulphate, the resistances used in the dry deposition scheme
are also the same. After the chemistry routine (NITRATE) is called, the updated HNO3 concentration is passed
back to UKCA. This is done in the same way as for the sulphur-cycle oxidants.

5.1 Chemistry and inter-modal processes

Routine NITRATE is called from AERO CTL immediately after the sulphur scheme. The mass-mixing ratios of
HNO3 (kg[HNO3 ] kg[air]−1 ) and NH3 (kg[N] kg[air]−1 ) are passed to NITRATE. The ammonium nitrate chemistry
scheme in this routine is based on the STOCHEM scheme, and is similar to the scheme used in the EMEP
model. In NITRATE, the deliquescence relative humidity is calculated for each gridbox as
 
618.3
DRH = exp − 2.551 (5.1)
T
where T is the temperature.
Then, the equilibrium coefficient, kp2 , is calculated for the reaction:

HN O3 + N H3 ⇋ N H4 N O3 (5.2)

The calculation of kp2 is different for relative humidities above and below the deliquescence relative humidity.
For relative humidities lower than DRH, kp2 depends only on temperature following
 
−6.025 24084
kp2 = T exp 118.87 − (5.3)
T
For relative humidities higher than DRH, kp2 depends on both temperature and relative humidity RH, following

kp2 = p1 − p2 (1 − RH) + p3 (1 − RH)2 ) (1 − RH)1.75 kp2


1

(5.4)
1
where kp2 is computed using equation 5.3, and p1 , p2 , and p3 are temperature-dependent parameters defined
as  
19.12 8763
p1 = T exp −135.94 + (5.5)
T
 
9969
p2 = T 16.22 exp −122.65 + (5.6)
T
 
13875
p3 = T 24.46 exp −182.61 + (5.7)
T

Total HNO3 and NH3 are calculated:

[Total HN O3 ] = [HN O3 ] + [N H4 N O3 ] (5.8)

[Total N H3 ] = [N H3 ] + [N H4 N O3 ]. (5.9)

The equilibrium concentration of NH4 NO3 is then calculated as:


1
[N H4 N O3 ] = {[Total N H3 ] + [Total HN O3 ] − ({[Total N H3 ] + [Total HN O3 ]}2
2 (5.10)
− 4{[Total N H3 ] ∗ [Total HN O3 ] − kp2 })1/2 }

To avoid negative concentrations, this calculation is only done if [Total NH3 ] * [Total HNO3 ] > kp2 Otherwise, it is
assumed that the NH4 NO3 concentration is zero in that gridbox.

26 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

It is assumed that all NH4 NO3 formed in this way goes into the accumulation mode. The concentrations of HNO3
and NH3 are then updated:

[HN O3 ] = [Total HN O3 ] − [N H4 N O3 ] (5.11)

[N H3 ] = [Total N H3 ] − [N H4 N O3 ] (5.12)

Note that conversion factors are included in the calculation to take account of the fact that volumetric con-
centrations of HNO3 , NH3 and NH4 NO3 (in molecules cm−3 ) are required for the equilibrium calculation but
mass-mixing ratios are used by UKCA and by the CLASSIC aerosol scheme (so the species are passed in/out
of the NITRATE routine as mass-mixing ratios).
The following inter-modal processes are included: nucleation of accumulation-mode nitrate to form dissolved
nitrate, and evaporation of cloud droplets to re-form accumulation-mode nitrate from dissolved nitrate. These
processes are represented in the same way as for sulphate.

5.2 Radiative and Precipitation Effects

The treatment of these effects is similar to other aerosol types; the optical properties of the aerosol are read
from the spectral file. Note that optical properties are those of ammonium nitrate aerosols. Accumulation-
mode nitrate is given the same size distribution as accumulation-mode sulphate. Hygroscopic growth follows
the parametrization by Fitzgerald (1975). The set of refractive index for ammonium nitrate is a composite of
different datasets. For wavelengths between 2 and 20 µm, data is taken from Jarzembski et al. (2003). For
wavelengths larger than 20 µm, the refractive index remains constant at the value for 20 µm. For wavelengths
between 0.1 and 2 µm, the real part of the refractive index is derived from the CRC Handbook of Chemistry and
Physics, 58th edition, 1971 by using the value they give at 0.5876 µm. For wavelengths between 0.7 and 2 µm,
the imaginary part of the refractive index is taken from Gosse et al. (1997) for an ammonium nitrate solution at
25%. Finally, imaginary parts for wavelengths between 0.1 and 0.7 µm are set to 10−8 to represent a scattering
aerosol. Ammonium nitrate density is 1725 kg m−3 .

5.3 Logical variables controlling the nitrate scheme

• The nitrate scheme is controlled by the logical L NITRATE


• Logical L SULPC 2 WAY COUPLING turns on the two-way coupling of the sulphur cycle and nitrate
scheme to the UKCA chemistry and should be set to true
• Logicals L SULPC NH3 and L NH3 EM to turn on the ammonia tracer
• Cloud and radiative effects of nitrate are controlled by the logicals L USE NITRATE DIRECT, L USE -
NITRATE INDIRECT and L USE NITRATE AUTOCONV

5.4 STASH items available with the nitrate scheme

• Mass-mixing ratio of accumulation-mode nitrate (kg[N] kg[air]−1 ) (STASH code 0-117).


• Mass-mixing ratio of dissolved nitrate (kg[N] kg[air]−1 ) (STASH code 0-118).
• Dry deposition flux of accumulation-mode nitrate (kg[N] m−2 s−1 ) (STASH code 3-274).
• Dry deposition flux of dissolved nitrate (kg[N] m−2 s−1 ) (STASH code 3-275).
• Wet deposition flux of dissolved nitrate scavenged by large-scale precipitation (rainout) (kg[N] m−2 s−1 )
(STASH code 4-213).
• Wet deposition flux of dissolved nitrate scavenged by large-scale precipitation (washout; currently zero as
not included) (kg[N] m−2 s−1 ) (STASH code 4-214).

27 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

• Wet deposition flux of accumulation-mode nitrate scavenged by convective precipitation (kg[N] m−2 s−1 )
(STASH code 5-247).
• Wet deposition flux of dissolved nitrate scavenged by convective precipitation (kg[N] m−2 s−1 ) (STASH
code 5-248).
• Flux through equilibrium reaction HNO3 + NH3 ⇋ NH4 NO3 (kg[N] kg[air]−1 s−1 ) (STASH code 17-240).
As this is an equilibrium reaction, the flux may be positive or negative (if it is positive then NH4 NO3 is
produced; if negative then NH4 NO3 is destroyed to re-form HNO3 and NH3 ).
• Transfer of accumulation-mode nitrate to dissolved mode via nucleation (kg[N] kg[air]−1 s−1 ) (STASH code
17-241).
• Transfer of dissolved nitrate to accumulation mode via evaporation (kg[N] kg[air]−1 s−1 ) (STASH code
17-242).

6 Mineral Dust scheme

6.1 Introduction

The mineral dust scheme is unlike the other CLASSIC aerosol schemes in that emissions are calculated in-
teractively from atmospheric and surface fields each time-step. The large particle size range is divided into a
series of contiguous bins, and each bin is treated independently. Particles are assumed to be spherical. Dust is
transported through the atmosphere as a series of independent tracers and undergoes dry deposition through
turbulent mixing and gravitational settling and wet deposition through washout. When it is required, dust can
have a direct radiative effect on the atmosphere in both the shortwave and longwave spectral regions. The
direct radiative effect is usually enabled in weather and climate forecasts, although due to the interaction be-
tween model winds and the dust emissions the dust direct effect is often calculated using an aerosol climatology
when testing other model changes (see Section 9). In earth-system models the deposition of dust to the ocean
provides a source of iron to phytoplankton, and can thus affect the carbon cycle.

6.2 Dust Emissions

Dust emissions are calculated during each model time-step using prognostic model fields. The dust emission
scheme utilises the widely-used algorithm of Marticorena and Bergametti (1995) to calculate horizontal flux in
each of 9 bins with boundaries at 0.0316, 0.1, 0.316, 1.0, 3.16, 10.0, 31.6, 100., 316. and 1000. µm radius.
Within each bin dV /d(log(r)) is assumed constant, where V is particle volume and r is particle radius, giving a
”histogram-shaped” distribution of dV /d(log(r)). Horizontal flux in a bin is given by:

   2 !
∗3 1 + Uti∗ Uti∗ CD
Gi = ρBU 1− Mi (6.1)
U∗ U∗ g

where i refers to bin number, ρ is air density at the surface, B is bare soil fraction in the grid-box, Ut∗ is threshold
friction velocity, U ∗ is the bare soil friction velocity excluding orographic effects, M is mass fraction of soil
particles in the bin, C is a constant of proportionality, set to 2.61 from wind-tunnel experiments, D is a tunable
parameter (see section 6.7) and g is acceleration due to gravity. Dust is currently emitted only from the bare
soil fraction of the grid box. Dust emissions are inhibited if snow is present, if the ground is frozen, on steep
slopes, if soil moisture exceeds a threshold (see below) and at coastal points with fractional land-cover where
the lowest level windspeed over land may be anomalously high.
Equation 6.1 was derived from experimental measurements, essentially at point sources at a single time,
whereas the model calculates grid-box mean, time-step mean values. In order to correct for the effect of this

spatial and temporal averaging, U ∗ in equation 6.1 is calculated from the model value UM and a tunable constant
k1 :

U ∗ = k1 UM

(6.2)

28 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

k1 is obtained empirically, by performing simulations with a range of values to minimise errors in dust concen-
trations and optical depth compared with observations. Inevitably, the value chosen will be influenced by biases
in model fields as well as by the resolution of the model. See section 6.7 for information on setting k1 .
The mass fraction of particles (M ) is obtained from an ancillary file. This contains the soil clay, silt and sand
fraction fields, together with M for bins 1-6, calculated from those fractions according to the method described
in Woodward (2001). The data used for this ancillary should be consistent with that used for the soil hydrological
properties ancillary. M for bins 7-9 is calculated interactively from the sand fraction.

Dry threshold friction velocity (Utd ) for each bin is taken from Bagnold (1941). The effect of soil moisture is
treated according to the method of Fécan et al. (1999), which has been shown be in good agreement with
observations. This relates threshold friction velocity in moist and dry conditions by:

Ut∗ ′
∗ = 1 for w < w
Utd
Ut∗ 0.5 (6.3)
= 1 + 1.21(w − w′ )0.68 for w > w′
Utd

w′ = 0.14 ∗ FC2 + 17.0 ∗ FC

where FC is clay fraction and w represents volumetric soil moisture. In order to relate this point value to the
model’s grid-box mean soil moisture over the top soil level (w1 ), a tunable constant is applied (see section 6.7) ,
in a similar manner to that for friction velocity:

w = k2 ∗ w1 (6.4)

In order to apply the Fécan et al. scheme to moister regions than the arid and semi-arid areas for which it
was designed, a further restriction is put on dust emissions from moist soils. Emissions are inhibited when soil
moisture exceeds a threshold, depending on particle size.

FC + 0.12
wt = (6.5)
0.03

This limit corresponds approximately to the maximum soil moisture at which soil movement was observed in the
measurements used by Fécan et al.
The total vertical dust flux F is related to the total horizontal flux across the 9 bins discussed above following
the formulation of Marticorena and Bergametti (1995) based on the measurements of Gillette (1979):
9
X
(13.4FC −8.0)
F = 10 Gi (6.6)
i=1

Only dust in the first 6 bins with radius ≤ 31.6 microns can be be lifted into the model’s atmosphere and there
are two options as to how to distribute the total vertical dust flux F into the flux in each bin Fi . Either the size
distribution follows that of the horizontal flux in these bins such that:
Gi
Fi = P6 F
i=1 Gi
P9 (6.7)
Gi
= 10(13.4FC −8.0) Gi Pi=1
6
i=1 Gi

or it is possible to prescribe a fixed size distribution by setting the variable L fix size dist:
xi
Fi = P6 F
i=1 xi
P9 (6.8)
(13.4FC −8.0) Gi
= xi 10 Pi=1
6
i=1 xi

29 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

where xi is the array size dist discussed in section 6.7. The interactively calculated size distribution in (6.7)
is currently used in HadGEM climate model configurations, whilst the fixed size distribution (6.8) was developed
for use in Numerical Weather Prediction.
The highest clay fraction (FC ) reported in the measurements on which the algorithm was based was 0.2 (Gillette
1979). However it subsequently became clear that the single measurement with this high clay fraction was
contaminated by upstream dust (Gillette, pers. Comm., reported in Alfaro and Gomes, 2001). The next highest
clay fraction measured was 0.1. As high clay fractions result in unrealistically high emissions from the dust
scheme, a maximum is applied, with higher values of FC being reset to this. (See section 6.7.)
In most cases dust production areas can be simulated successfully using the algorithms above, based on the
model’s prognostic fields. However, under certain circumstances (e.g. if the interactive vegetation scheme in
unable to produce a sufficiently realistic simulation of bare soil areas) it has been found desirable to constrain
the source areas further by use of a preferential source term. This capability has not yet been built into the UM
trunk, but must be applied through an appropriate code modification.

6.2.1 Dust Emission From Seasonal Vegetation

The aim of this parametrization it to represent the emission of dust from regions of seasonal vegetation. There is
currently only 1 method in place to do this, (dust veg emiss=1), which calculates the bare soil radiative fraction
(f r(bs) ) from of each vegetated surface type using the leaf area index (LAI) in the same way that is used to
calculate the surface albedo of these region

f r(bs) = exp(−LAI/2). (6.9)

The dust fluxes are then calculated for each surface tile, using friction velocity of bare soil, and this is then
multiplied by a ’dust emission fraction’ which, for vegetated tiles, is the bare soil radiative fraction for that tile
multiplied by a tunable scaling parameter for that tile. The use of the scalable parameters by plant functional
type (PFT) means that dust emission from different vegetation types can be easily controlled. For example the
9 tile configuration has 5 PFTs (Broadleaf trees, Needleleaf trees, C3 grass, C4 grass, Shrubs) can be set to
emit dust only from grasses and shrubs using the scaling parameters (0.0, 0.0, 1.0, 1.0, 1.0).
In a 1 tile model the dust flux is still calculated using the bare soil U* and the dust emission fraction is aggregated
for all of the vegetated surface types and combined with the bare soil fraction. Note that this does imply that, over
vegetated tiles, the dust flux is calculated using a different roughness length than the fluxes of heat, moisture
and momentum, but it does mean that the dust emissions are consistent between the 1 and 9 tile configurations.
The dust emission fraction is available on tiles as STASH item 1,3,400. The impact of this change is still being
tested and this option should not be switched on in climate simulations until the long term impact has been more
clearly evaluated.

6.3 Dry Deposition

Dust particles undergo dry deposition through gravitational settling and turbulent mixing. These processes are
represented using a resistance analogue method, with the deposition velocities treated as inverse resistances
(Seinfeld and Pandis, 1998). Total deposition velocity VD is given by:

VD = (RA + RB + RA RB VS )−1 + VS (6.10)

RA is the aerodynamic resistance and RB the resistance of the quasi-laminar surface layer. VS is the Stokes
velocity, including the Cunningham correction factor for slip flow. Resuspension of particles due to bouncing of
the surface is assumed to be negligible, so no transfer resistance term is required.
RB is given by:

RB = (U ∗ (Sc−2.0/3.0 + 10.0−3.0/St))−1 (6.11)

where U ∗ is friction velocity, Sc is the Schmidt number, and St the Stokes number (Seinfeld, 1986)

30 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Vs is taken from Pruppacher and Klett (1996):

d2 gρp
  
1 + αC λ
VS = (6.12)
0.5d 18.0η

The term on the left of this equation is the Cunningham correction factor. d is particle diameter (for which a
representative value is assigned to each bin), g is acceleration due to gravity, ρp is particle density and η is the
viscosity of air. αC is given by:
 
CC (0.5d)
αC = AC + BC e λ
(6.13)

where AC = 1.257, BC = 0.4, CC = -1.1, and λ is the mean free path:


  
P0 T0
λ = λ0 (6.14)
P T

where λ0 =6.6x10−8 m, P is pressure, P0 is reference pressure (1.01325 × 105 Pa), T is temperature and T0 is
reference temperature (293.15 K).
Dust particles are heavier than most other aerosols and the effect of gravity on the particles is important not
only for deposition of dust to the ground, but also vertical transport of dust throughout the atmosphere. The
downward flux of dust is calculated using the Stokes velocity (equation 6.12). Dust may fall either one or two
levels in one time-step.

6.4 Wet Deposition

Wet deposition of dust by washout below cloud is calculated for convective and large-scale precipitation. Dust
particles are not hydrophilic, so rainout is relatively unimportant and is ignored. A first order removal rate is
assumed:

∂D
= −KRD (6.15)
∂t

where D is dust concentration, t is time, K is the scavenging coefficient and R is precipitation rate. The scaveng-
ing coefficients for each of the 6 bins is: 2.0 × 10−5, 2.0 × 10−5, 3.0 × 10−5, 6.0 × 10−5, 4.0 × 10−4 and 4.0 × 10−4.
These values are derived from the experimental data of Volken and Schumann (1993). Once particles are
subjected to washout they are assumed to be deposited directly to the ground; any effects of evaporation are
ignored.

6.5 Radiative Effects

Dust has a direct radiative effect through the absorption and scattering of both shortwave and longwave radia-
tion, with smaller particles tending to have a larger effect in the shortwave and larger particles tending to have a
larger effect in the longwave. If the model is run including this direct radiative effect, each dust division is treated
independently by the radiation scheme. The radiative properties in the spectral files are calculated from Mie
theory, assuming spherical particles.

6.6 Coding Consideration

The dust code is not collected together in one section but distributed in routines throughout the atmospheric
model. This is partly for historical reasons, but it also reflects the fact that dust processes are dependent on
many atmospheric and surface variables in several different sections. The order of dust processes within a
time-step is thus constrained, and is artificial, but was not chosen at random and changing it can change the
results significantly. Dust is particularly sensitive to such coding considerations because of the strong gradients

31 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

in dust concentrations in both space and time, which can be difficult to model numerically. These gradients are
caused by the episodic nature of the emissions from areas which can have very sharply defined edges, as well
as the rapid fall speeds of the larger particles.
The main dust processes are given below, by code section, in the order in which they are performed within a
time-step. Please note that this is by no means an exhaustive list of routines.
Microphysics: MICROPHYSICS CTL calculates washout of dust by large-scale precipitation.
Radiation: dust radiative effects are calculated in both shortwave and longwave sections.
Explicit Boundary Layer: DUST SRCE is the main routine for calculating dust emission flux, though emissions
are not actually mixed into the atmosphere in this section, but passed to the Implicit Solver. Some initialisation
and the calculation of gridbox mean fluxes is performed in BDY LAYR. Calculation of friction velocity, excluding
orographic effects, is performed in either FCDCH LAND in the MOSES-II land surface model or in FCDCH in
JULES. A term in the resistance across the sub-laminar boundary layer is calculated in DUSTRESB, which
obtains the Stokes velocity and other variables from VGRAV.
Convection: Washout by convective precipitation is calculated in NI CONV CTL.
Implicit Boundary Layer Solver: GRAVSET calculates the vertical transport of dust under the influence of gravity,
from the top of the atmosphere down to the lowest level, but not to the surface. BL TRMIX DD handles the
deposition from the lowest level to surface due to the interaction between the processes of sedimentation and
mixing, as well the emission of the dust flux (calculated in DUST SRCE) into the atmosphere, and turbulent
mixing through the boundary layer. Both these routines call VGRAV to calculate the Stokes velocity and related
variables.
The DUST PARAMETERS MOD module contains both fixed parameters used by the dust scheme and those
that are most likely to be changed for tuning purposes. These latter are set in the namelist RUN DUST (see
section 6.7).

6.7 Use of the Dust Scheme

The dust scheme is switched on via the UMUI window Model → Atmosphere → Scientific Parameters and
Sections → Section by section → Aerosols. From there a button leads to the Dust window, where the dust
ancillary file should be specified (see below). In addition to running with the prognostic dust scheme, it is
also possible to diagnose what the dust emission would have been (ignoring the direct radiative effect) without
actually including dust in the model’s atmosphere. From the Dust window, buttons lead to the Aerosol Effects
window, from where the direct radiative effect of dust is switched on (if this is off, then the dust will simply be
transported as a set of passive tracers) and also to the LW-Gen2 window where the treatment of longwave
scattering should be set to ”Full” and a longwave spectral file which includes dust should be specified. A
shortwave spectral file including dust should also be specified on the equivalent shortwave window SW-Gen2.
A soil size distribution ancillary file is required as input to the dust scheme. It contains soil clay, silt and sand
fractions and the fractional mass of soil in each of the first 6 size bins. This ancillary should be consistent with
the soil hydrological properties ancillary, in that both sets of fields should be derived from the same data (e.g.
HWSD).
The dust scheme is unlike other aerosol schemes in that dust emissions are calculated interactively and are,
therefore, sensitive to model climate, particularly near-surface windspeed and soil moisture in arid areas. In
order to obtain realistic simulations the dust scheme has to be tuned to the model configuration being used, to
account for resolution-dependent effects and also, inevitably, model biases. The most readily tuned parameters
can be set via the UMUI panel, whilst other parameters can be tuned by creating a branch and editing the
DUST PARAMETERS MOD module.
The RUN DUST namelist contains the following parameters:

32 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

i dust main dust switch: 0 = dust scheme off, 1 = prognostic dust, 2 = diagnostic dust
us am friction velocity tuning term (multiplicative). k1 in (6.2)
sm corr top level soil moisture tuning term (multiplicative). k2 in (6.4)
horiz d global multiplier to horizontal dust flux. D in (6.1)
l fix size dist Switch to use the fixed size distribution in (6.8). The size distribution in
the code is (0.0005, 0.0049, 0.0299, 0.2329, 0.4839, 0.2479) for six-bin dust
and (0.1800, 0.8200) for two-bin dust.
dust veg emiss switch controlling emission of dust from seasonal vegetation (6.2.1).
0 is off (recommended), 1 uses LAI to calculate an emission fraction
As dust emissions are dependent on soil moisture, consideration must be given to its initialisation. In long
climate runs it can take several decades for soil moisture to equilibrate from the initial values supplied in an
ancillary or start dump, so it may be desirable to initialise soil moisture from the end of a long experiment as
close as possible to the one being set up. Other soil fields which should be kept consistent with soil moisture
are deep soil temperature, snow depth and all the soil hydrology fields. These can be set from the UMUI via
Model Selection → Atmosphere → Ancillary and input data files → Climatologies & potential climatologies →
Soil moisture and snow depth, and also → Deep soil temperatures, and → Soil : VSMC, hydrological/thermal
conductivity etc

6.7.1 The two-bin variant scheme

At UM 7.9 the ability to run a version of the dust scheme with two bins instead of the standard six was added.
This automatically uses a fixed size distribution of dust with the two bins having values equal to 0.1800 and
0.8200 (the two bins corresponding to 0-2 microns and 2-10 microns respectively). The horizontal fluxes in the
emissions subroutine are still performed over nine bins, however these are mapped onto two divisions rather
than six when the two-bin scheme is active. Conversion between the two dust schemes in lateral boundary
conditions is available using MakeBC. The two-bin scheme uses only the first two STASH items allocated to
dust. The scheme is currently only available using a handedit, and requires spectral files which include two-bin
dust data.

6.7.2 Preferential sources for HadGEM2

In rare cases model biases may render it impossible to obtain a realistic dust simulation through normal tuning.
In such situations it may be desirable to constrain source areas though the use of a preferential source multiplier.
This capability has not been built into the UM trunk but is currently available as branches at UM versions HG6.6.1
and HG6.6.3. An outline is included here because this option has been used in HadGEM2. (A preferential
source term was required in HadGEM2-ES because the interactive vegetation scheme was unable to produce a
sufficiently realistic simulation of bare soil source areas, and was also used in HadGEM2 AO and HadGEM2-A
for consistency.)
In general, a preferential source term represents the probability of the presence of accumulated erodible sed-
iment in a grid-box by relating this to a more readily available variable such as topography. The term used for
HadGEM2 was based on the formulation of Ginoux et al (2001):

 P
zmax − zi
S= (6.16)
zmax − zmin

where S represents the probability of the presence of accumulated sediment, zmax , zmin and zi are maximum,
minimum and local altitude in an area approximately 10°x 10°and P is an empirically chosen term, set to 3 here
to give the best simulation of dust concentrations and AODs compared with observations.

33 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

7 Sea-salt Aerosol Scheme

7.1 Using the sea-salt scheme

The sea-salt aerosol scheme is quite different from the other CLASSIC aerosol schemes in that it is a simple di-
agnostic scheme, i.e. instead of being emitted, transported, deposited &a , the concentration of sea-salt aerosol
is simply diagnosed as a function of other model variables. Hence sea-salt’s only purpose in the UM is for its di-
rect and indirect effects. There is only one version of the scheme available, in routine SET SEASALT. Depending
on which effects have been selected, this routine may be called from the radiation control routine NI RAD CTL,
the cloud microphysics control routine MICROPHYS CTL, and/or the aerosol control routine AERO CTL.
If sea-salt effects are selected, local variables are set up in the relevant control routine(s) to contain sea-salt
concentrations (number of particles per cubic metre) in two log-normal modes, representing the two main pro-
duction pathways of sea-salt aerosol:
Film-mode: median radius 0.1µm, geometric standard deviation 1.9;
Jet-mode: median radius 1.0µm, geometric standard deviation 2.0.

7.2 Calculation of sea-salt number concentration

When selected, sea-salt number in each mode is determined using local 10m windspeed (U10). This is used
to calculate number concentrations (m−3 ) of film- and jet-mode sea-salt aerosols (Af and Aj respectively),
following Jones et al. (2001):
For 0 ≤ U10 ≤ 2 ms−1 :
Af = 3.856 × 106 (1 − e−0.736U10 ) (7.1)

Aj = 0.671 × 106 (1 − e−1.351U10 ) (7.2)


−1
For 2 ≤ U10 ≤ 17.5 ms :
Af = 10(0.095U10+6.283) (7.3)

Aj = 10(0.0422U10+5.7122) (7.4)
−1
For U10 > 17.5 ms :
Af = 1.5 × 108 (1 − 97.87e−0.313U10) (7.5)

Aj = 3.6 × 106 (1 − 103.926e−0.353U10) (7.6)

Values are only calculated for points which have a non-zero fraction of open ocean, and for levels within the
boundary layer, concentrations being set to zero elsewhere. These initial values are then scaled exponentially
depending on height (h) above the surface, and linearly depending on the fraction of land (fl ) and of sea-ice (fi )
in the gridbox:

Af /j (scaled) = Af /j (initial) × exp(10/zf /j ) × exp(−h/zf /j ) × (1 − fl ) × (1 − fi ) (7.7)

where zf /j is the scale-height for film/jet modes (both currently set to 900m).

7.3 Radiative and Precipitation effects

These are the raison d’être of CLASSIC sea-salt aerosols. The direct radiative effect of sea-salt may be selected
independently of any other aerosols. Sea-salt complex refractive index is that of NaCl, following Toon et al.
(1976). Hygroscopic growth is parametrized following Fitzgerald (1975). Density is taken at 2165 kg m−3 .
For the indirect effects, sea-salt is assumed to be supplementary to sulphate, i.e. the relevant effect must be
active for sulphate aerosol for it to be available for sea-salt.

34 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

8 Particulate matter (PM10, PM2.5 and TSP) diagnostics

8.1 Introduction

Particulate matter (PM), also known as particulate pollution, is a complex mixture of small particles and liquid
droplets suspended in the air, many of which are hazardous. This mixture contains acids such as nitrates and
sulfates, soot, organic chemicals and crustal materials.
As particles have often irregular shapes there are many definitions of particulate size. The most widely used
definition is the aerodynamic diameter, which is the diameter of an idealised sphere of unit density (1g cm−3 )
with the same gravitational settling velocity as the particle in question. The aerodynamic diameter of the particles
determines how they are transported in air and how they can be removed from it. It also governs how far they
get into the air passages of the respiratory system. The sampling and description of particles is therefore often
based on this aerodynamic diameter.
The particles of concern in air pollution are usually those of 10 µm or smaller aerodynamic diameter (PM10),
because they can get into the lungs, potentially causing serious health effects. The fine fraction of these particles
contains the smaller ones with a size up to 2.5 µm (PM2.5). The coarse fraction contains the larger particles
with a size ranging from 2.5 to 10 µm (PM10 - PM2.5). Although most data and air quality standards relate to
PM10, it is believed that PM2.5 is responsible of much of the health effects attributable to PM10.

8.2 Using PM diagnostics in the UM

PM diagnostics have been included as items 220-235 in section 17 of the Unified Model. These diagnostics are
3D fields that represent mass concentrations (in µg/m3):
• mass concentration of PM10
• mass concentration of PM2.5
• contribution of sulphate to mass concentration of PM10
• contribution of sulphate to mass concentration of PM2.5
• contribution of black carbon to mass concentration of PM10
• contribution of black carbon to mass concentration of PM2.5
• contribution of biomass burning aerosol to mass concentration of PM10
• contribution of biomass burning aerosol to mass concentration of PM2.5
• contribution of OCFF to mass concentration of PM10
• contribution of OCFF to mass concentration of PM2.5
• contribution of SOA to mass concentration of PM10
• contribution of SOA to mass concentration of PM2.5
• contribution of sea salt aerosol to mass concentration of PM10
• contribution of sea salt aerosol to mass concentration of PM2.5
• contribution of dust to mass concentration of PM10
• contribution of dust to mass concentration of PM2.5
• contribution of nitrate aerosol to mass concentration of PM10
• contribution of nitrate aerosol to mass concentration of PM2.5
• total suspended particulate from dust aerosols
The last 16 diagnostics represent the individual contributions of the different aerosol species present in the
CLASSIC scheme sulphate, black carbon (also called soot), biomass-burning aerosol, organic carbon from fossil
fuel (OCFF), secondary organic aerosol or SOA (climatology of biogenic aerosol), sea salt aerosol, mineral dust

35 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

and nitrate aerosol to the mass concentrations of PM10 and PM2.5. The first two diagnostics are calculated as
sum of the individual contributions of the aerosol species considered to both particle sizes.

8.3 Calculation of PM diagnostics

The calculation is done by the routine calc pm diags, which is called by AERO CTL only if the STASH flag of
any of the items 220-235 in section 17 has been set to .TRUE., i.e. only if any of those diagnostic has been
selected. If that condition is met the mass mixing ratios of the insoluble modes of the above mentioned aerosol
types (or aerosol number in the case of sea salt) are passed as arguments from AERO CTL to calc pm diags.
In this routine each aerosol mode (except for the six divisions of dust) is assumed to have log-normal size
distribution, with known median radius and geometric standard deviation. The assumed radii are given in table
1. Given these parameters as well as the density of the aerosol species and the air density it is possible to use
the volume distribution to calculate the mass concentration for all particles below a certain cut-off diameter (10
µm in the case of PM10 and 2.5 µm for PM2.5). The formulation used here has been adapted from Seinfeld
and Pandis (2006).

Aerosol species and mode rmed σg ρi


Sulphate Aitken 0.0065 1.3 1769.0
Sulphate accumulation 0.0950 1.4 1769.0
Black carbon fresh 0.0400 2.0 1900.0
Black carbon aged 0.0400 2.0 1900.0
Biomass burning fresh 0.1000 1.3 1350.0
Biomass burning aged 0.1200 1.3 1350.0
OCFF fresh 0.1000 1.3 1350.0
OCFF aged 0.1200 1.3 1350.0
Biogenic SOA 0.0950 1.5 1300.0
Sea salt film 0.1000 1.9 2165.0
Sea salt jet 1.0000 2.0 2165.0

Table 1: Parameters of the log-normal size distributions of the aerosol modes used. rmed is the median radius
(µm) and σg is the geometric standard deviation. The density ρi (kg m−3 ) has also been included.

The calculation of the fraction of each aerosol with radii less than each diameter is done in a stand alone program
calc pm params.f90 as the code uses a non intrinsic Fortran function erf and also for reasons of efficiency. If
the size of any of the aerosols is changed the program will need to be modified.
The log-normal distribution has been proved to match well observed shapes of ambient distributions of aerosols
and is regularly used in atmospheric applications. If an aerosol species in the mode i is log-normally distributed
its aerosol number concentration (n) satisfies:

−(loge D − loge Di )2
 
dN Nt
n(loge D) = = exp (8.1)
dloge D (2π)1/2 loge σi 2(loge σi )2

where Nt is the total aerosol number concentration, D represents the diameter of the particles, and Di and ρi
are the parameters of the distribution. Their significance will be discussed later.
Using the distribution of n (D) instead of that of n (loge D) and combining it with

n(loge D) = D.n(D) (8.2)

we have that
−(loge D − loge Di )2
 
dN Nt
n(D) = = exp (8.3)
dloge D (2π)1/2 loge σi 2(loge σi )2

Let us define the cumulative distribution N (D∗ ) as the concentration of particles in the population with diameters
smaller or equal to D∗ :
Z D∗

N (D ) = n(D)dD (cm−3 ) (8.4)
0

36 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Since the aerosol distribution is log-normal, n (D) is given by (8.3) and therefore
D′
−(loge D − loge Di )2
 
Nt 1
Z

N (D ) = exp dD (8.5)
(2π)1/2 loge σi 0 D 2(loge σi )2

To evaluate this integral we let √


t = (loge D − loge Di )/ 2loge σi (8.6)

Differentiating
1 dD
dt = √ (8.7)
( 2loge σi ) D

and substituting (8.6) and (8.7) in (8.5) we obtain


loge (D′ −loge Di )

Nt
Z
2loge σi 2

N (D ) = √ e−t dt (8.8)
π −∞

This integral can be evaluated by using the error function erf (z) which is defined as
Z z
2 2
erf (z) = √ e−t dt (8.9)
π 0

with erf (-∞) = -1, erf (0) = 0, erf (∞) = 1



If we divide the integral into a
√first term from -∞ to 0 and a second
√ one from 0 to (loge D∗√− loge Di )/ 2σi , then
the first integral is equal to π/2 and the second one to ( π/2) · erf [loge D∗ − logDi / 2loge σi ]. Finally, the
cumulative distribution of the particle number will adopt this form:
 
∗ Nt Nt loge (D∗ /Di )
n(D ) = + erf √ (8.10)
2 2 2loge σi

For D∗ = Di
Nt Nt Nt
n(Di ) = + erf (0) = (8.11)
2 2 2

Di is therefore the median diameter (Dmed ), i.e. the diameter for which exactly one-half of the particles are
smaller and the other one-half are larger. It can be shown that σi is the ratio of the diameter below which 84.1%
of the particles lie to the median diameter and is termed the geometric standard deviation (σg ).
The concentration of particles in the population with diameters smaller or equal to D* can now be expressed as:
!
Nt Nt loge (D∗ /Dmed )

n(D ) = + erf √ (cm−3 ) (8.12)
2 2 2loge σg

Since we are looking for the mass concentration (µgcm-3) of particles with diameters up to D* it is necessary to
adopt the volume distribution v(D). Assuming spherical particles:
π
v(D) = n(D) D3 (8.13)
6

Substituting (8.3) in (8.13) and using the definitions of median diameter and geometric standard deviation:

πD3 Nt −(loge D − loge Dmed )2


 
v(D) = exp (8.14)
6(2π)1/2 Dloge σg 2(loge σg )2

37 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

By letting D3 = exp (3 loge D), expanding the exponential, and completing the square in the exponent (Seinfeld
and Pandis, 2006), (8.14) becomes

 
(π/6)Nt 9 2
v(D) = exp 3loge Dmed + (loge σg ) ×
(2π)1/2 loge σg 2
(8.15)
−[loge D − (loge Dmed + 3(loge σg )2 ]2
 
exp
2(loge σg )2

Comparing (8.15) with (8.3) , if the number distribution n(D) is log-normal then the volume distribution v(D) is
also log-normal with the same geometric standard deviation σg as the parent distribution and with the volume
median diameter VMD given by
loge VMD = loge Dmed + 3(loge σg )2 (8.16)

The volume concentration (µm3cm-3) of particles with diameter equal or lower to the cut-off diameter D* can be
obtained by integrating (8.15):
 
9
v(D∗ ) =exp 3loge Dmed + (loge σg )2 ×
2
Z D∗ (8.17)
−(loge D − loge VMD)2
 
(π/6)Nt 1
exp dD
(2π)1/2 loge σg 0 D 2(loge σg )2

Multiplying by the density of the particles (ρi ) and using the erf(z) function similarly as done for the number dis-
tribution, we can obtain the mass concentration of particles with diameter equal or lower to the cut-off diameter
D∗ :
" !#
∗ Nt Nt loge (D∗ − loge (VMD))
M (D ) = ρi A · + erf √ (8.18)
2 2 2loge σg

where
A = π6 exp 3loge Dmed + 29 (loge σg )2
 

loge VMD = loge Dmed + 3(loge σg )2


D∗ * = cut-off diameter
Dmed = median diameter of the aerosol mode i
σg = geometric standard deviation of the aerosol mode i
ρi = density of the aerosol
Nt = total aerosol number

By letting D* = 10 µm and D* = 2.5 µm, the routine calc pm diags uses (8.18) to calculate the individual contribu-
tions of the different aerosol modes to the mass concentration of PM10 and PM2.5, which are finally expressed
in µgm-3. The values of Dmed , σg and ρi for the aerosol modes present in the CLASSIC scheme are given in
Table X.3.1. Except in the case of sea-salt aerosol, AERO CTL does not pass Nt to calc pm diags but the mass
mixing ratios mi for the individual aerosol modes. Thus before the calculation of M(D∗ ) it is necessary to first
determine Nt as:

ρair 1
N t = mi (8.19)
ρi Vi

where Vi is the average volume of the particles.


For a log-normal distributed aerosol, different diameters D can be related to Dmed by the so-called Hatch-Choate
equation (Hatch and Choate, 1929):

D = Dmed exp[p(loge σg )2 ] (8.20)

38 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

where p is a parameter that defines the various possible diameters. In the case of the diameter of average mass
(Dm ), p = 3/2 (see e.g. Reist, 1984; Zhang, 2004). Therefore the average volume used in calc pm diags for the
calculation of Nt in (8.19) is given by

π π 3 9
Vi = (Dm )3 = Dmed exp( (loge σg )2 ) (8.21)
6 6 2

Since the size distribution of the different dust sizes is not log-normal, an alternative empirical formulation has
been used to fit the cumulative volume/mass distribution. In this formulae dust divisions 1-3 and a small fraction
(∼0.19) of div4 are considered to contribute to PM2.5, while divs1-4 and a fraction (∼0.40) of div5 contribute to
PM10. The final mass concentrations of PM10 and PM2.5 are finally calculated as the sum of all the individual
contributions considered. The PM diagnostics calculated in calc pm diags are passed to AERO CTL and finally
stored in the routine DIAGNOSTICS AERO.

9 Aerosol climatologies

The model can use climatologies for all of the currently held prognostic aerosols in order to include the direct
radiative effect of aerosols without the expense of running the prognostic aerosol scheme. It is possible to
run with some species as a prognostic, but others as a climatology. It is also possible to run with a particular
aerosol species with both prognostic and climatological aerosol - in this configuration the radiation scheme sees
the climatological aerosol, allowing changes to be tested in the model which could drastically change aerosol
distribution, while maintaining a realistic aerosol radiative forcing.
When these aerosol climatologies are turned on an additional ancillary file containing a 3D field of the aerosol is
read in and passed down to the radiation code. Any prognostic aerosols which are overridden by a climatology
are also passed into the radiation scheme although these do not effect the model evolution through changing
the radiative fluxes, it is possible to obtain the Aerosol Optical Depth diagnostics. AOD diagnostics with STASH
codes 2284-2299 refer to the AOD of the aerosols as seen by the radiation scheme, while STASH codes 2421-
2427 are derived from the prognostic aerosol scheme regardless of their radiative impact.
The current standard aerosol climatology ancillaries were generated by a twenty year integration of HadGEM2-
A (HadGAM1a for mineral dust), using year 2000 emissions, with the addition of (i) fossil-fuel organic carbon
scheme, and (ii) the inclusion of ozone oxidation of SO2 in producing sulphate aerosol.

9.1 Biogenic aerosol climatology

Biogenic aerosol, or secondary organic aerosol from terpene emissions from vegetation, is only available as
a climatology of three-dimensional mass-mixing ratios. The climatology currently in use was obtained from
STOCHEM simulations (Derwent al., 2003). Biogenic aerosols exert direct and, contrary to the other climatolo-
gies, indirect effects. The inclusion of both direct and indirect effects are controlled by the same model logical,
L USE BIOGENIC. Their size distribution is assumed log-normal with median radius 0.095 µm and standard
deviation 1.5. Hygroscopic growth follows Varutbangkul et al. (2006) The complex refractive index is assumed
wavelength-independent and follows Lund-Myhre and Nielsen (2004), with no absorption. Density is taken at
1050 kg m−3 (Bahreini et al., 2005).

10 Lateral boundary conditions (LBCs)

Code is available to write and read lateral boundary conditions (LBCs) which include CLASSIC aerosols in addi-
tion to the other prognostic variables normally used in limited area models. This section is simply a description
of how to enable this functionality for CLASSIC aerosols. For more details please see UMDP-C71 .

39 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

10.1 Generation of LBCs

There are currently two methods of generating LBCs from the MetUM: output directly from the driving model or
via the MakeBC small utility. The former method will shortly be retired from the model as it is inefficient. Aerosol
LBCs can be generated for any or none of the active transported CLASSIC aerosols or gases (including DMS
and NH3 ).
To include aerosol tracers in the LBCs generated by the model, set up LBC output as normal using the panel
Atmosphere → Control → Output data files (including LBCs). Go to Atmosphere → Section by section choices
→ aerosol panel and then to the LBCs follow on window and choose which aerosols you wish to include in the
output LBCs using the checkboxes on the panel.
To use MakeBC to generate LBCs from the MetUM output after a model run, ensure that you have all the
required model output going to a fields file and then run MakeBC. See the instructions in UMDP-F54 .

10.2 Using LBCs

Set up the input of the LBCs as usual using the panel Atmosphere → Ancillary and input data files → Other
ancillary files and lateral boundary files → Lateral Boundary conditions. The aerosol tracers to use are set by
the checkboxes at the bottom of the panel. Ensure that the CLASSIC aerosol LBC inputs which you choose
match the fields present in the LBC file you are reading in or the model will fail on the first timestep. If a LBC for
a tracer is not present, this means it will be treated as if there is no source of the tracer outside your model i.e.
the LBC is effectively zero.

10.3 Stash codes for CLASSIC LBCs

As with standard LBC fields for input all the LBCs for CLASSIC tracers are in section 31. For reference the
STASH codes for the CLASSIC LBCs are given in table 2.

11 References

Adamowicz, R. F., 1979. Atmos. Environ., 13, 105–121


Alfaro, S. C. and L. Gomes, 2001, J. Geophys. Res., D16.
Bagnold, R. A., 1941, The Physics of Blown Sand and Desert Dunes., Methuen, New York.
Balkanski, Y., Schulz, M., Claquin, T., and Guibert, S. (2007), Reevaluation of mineral aerosol radiative forcing
suggests a better agreement with satellite and AERONET data. Atmos. Chem. Phys., 7, 81–95.
Bahreini, R., et al. Measurements of secondary organic aerosol from oxidation of cycloalkenes, terpenes, and
m-xylene using an Aerodyne aerosol mass spectrometer. Environ. Sci. Technol., 39, 5674–5688, 2005.
Collins, W. J., Stevenson, D. S., Johnson, C. E. and Derwent, R. G., 1997. J. Atmos. Chem., 26, 223–274
Cooke, W. F. and Wilson, J. J. N., 1996. J. Geophys. Res., 101, 19395–19410
Derwent, R.G., Collins, W.J., Jenkin, M.E., Johnson C.E., and Stevenson D.S., 2003, The global distribution of
secondary particulate matter in a 3D Lagrangian chemistry transport model. J. Atmos. Chem., 44, 57–95.
Fécan, F. et al, 1999, Parametrization of the increase of the aeolian erosion threshold wind friction velocity due
to soil moisture for arid and semi-arid areas. Ann. Geophys. 17, 149–157.
Fitzgerald, J. W., 1975. J. Applied Met., 14, 1044–1049
Gillette, D.A., 1979, Environmental factors affecting dust emissions by wind erosion., Saharan Dust, ed. C
Morales, John Wiley, New York.
Ginoux, P. et al, 2001, Sources and distributions of dust aerosols simulated with the GOCART model, J. Geo-
phys. Res., 106, D17, 20255–20273.

40 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Gosse, S.F., Wang, M., Labrie, D., and Chylek P (1997), Imaginary part of the refractive index of sulfates and
nitrates in the 0.7-2.6-µm spectral region. Appl. Opt, 36, 16, 3622–3634.
Hatch, T., and Choate, S.P., 1929. Statistical description of the size properties of non-uniform particulate sub-
stances. J. Franklin. Inst.
Haywood JM, Osborne SR, Francis PN, Keil A, Formenti P, Andreae MO, Kaye PH, 2003. The mean physical
and optical properties of regional haze dominated by biomass burning aerosol measured from the C-130 aircraft
during SAFARI 2000. J. Geophys. Res., 108, D13, 8473.
Jarzembski, M.A, Norman, M.L., Fuller, K.A., Srivastava, V., and Cutten, D.R. (2003). Complex refractive index
of ammonium nitrate in the 2-20-µm spectral range. Appl. Opt., 42, 6, 922–930.
Jones, A. and Roberts, D. L., 2004. Hadley Centre Tech. Note No. 47
Jones, A., Roberts, D. L., Woodage, M. J. and Johnson, C. E., 2001. J. Geophys. Res., 106, 20293–20310
Levine, S. Z. and Schwartz, S. E., 1982. Atmos. Environ., 16, 1725–1734
Lund-Myhre, C., and Nielsen, C.J. Optical properties in the UV and visible spectral region of organic acids
relevant to tropospheric aerosols. Atmos. Chem. Phys., 4, 1759–1769, 2004.
Marticorena, B. and Bergametti, G., 1995, Modeling the atmospheric dust cycle. 1. Design of a soil-derived
dust emission scheme, J. Geophys. Res., 100, D8, 16415–16430.
Monteith, J. L. and Unsworth, M. H., 1990. Principles of Environmental Physics (2nd Edn), Edward Arnold,
London
Padro, J., Neumann, H. H., and DenHartog, G., 1993. Water, Air and Soil Pollution, 68, 325–339
Park et al., 1999. J. Aerosol Sci., 30, 3–16
Pruppacher, H. R. and Klett, J. D., 1980. Microphysics of Clouds and Precipitation (1st Edn), Kluwer, Dordrecht
Reist, P. C., 1984, Introduction to aerosol science, MacMillan, New York.
Scott, B. C., 1982. Atmos. Environ., 16, 1753–1762
Seinfeld, J. H. and Pandis, S. N., 1998. Atmospheric Chemistry and Physics: From Air Pollution To Climate
Change, John Wiley, New York
Seinfeld, J. H., and Pandis, S. N., 2006. Atmospheric Chemistry and Physics: From Air Pollution to Climate
Change. John Wiley and Sons, New Jersey.
Spracklen, D. V., Pringle, K. J., Carslaw, K. S., Chipperfield, M. P., and Mann, G. W., 2005. Atmos. Chem. Phys.
Discuss., 5, 179–215
Stevenson, D. S., Collins, W. J., Johnson, C. E. and Derwent, R. G., 1997. Atmos. Environ., 31, 1837–1850
Toon, O. B., Pollack, J. B., and Khare, B. N., 1976. J. Geophys Res., 81, 5733–5748
Twomey, S., 1977. Atmospheric Aerosols (Developments in Atmospheric Science, Vol. 7), Elsevier, London
Varutbangkul, V., et al. Hygroscopicity of secondary organic aerosols formed by oxidation of cycloalkenes,
monoterpenes, sesquiterpenes, and related compounds. Atmos. Chem. Phys., 6, 2367–2388, 2006.
Volken, M. and T. Schuman, 1993, A critical-review of below-cloud aerosol scavenging results on Mt. Rigi, Water
Air Soil Pollut., 68, 15–28.
WCRP, 1986. IAMAP Radiation Commission. WCP-112 WMO/TD No. 24.
Woodward, S., 2001. J. Geophys. Res., 106, 18,155–18,166
Zhang, Y., 2004. Indoor Air Quality Engineering, CRC Press.

41 © Crown Copyright 2015


UMDP: 020
CLASSIC Aerosol Scheme

Section Item no. Lateral boundary value/tendency


31 23 DUST DIV 1 BOUNDARY VALUE
31 24 DUST DIV 2 BOUNDARY VALUE
31 25 DUST DIV 3 BOUNDARY VALUE
31 26 DUST DIV 4 BOUNDARY VALUE
31 27 DUST DIV 5 BOUNDARY VALUE
31 28 DUST DIV 6 BOUNDARY VALUE
31 29 SO2 BOUNDARY VALUE
31 30 DMS BOUNDARY VALUE
31 31 SO4 AITKEN BOUNDARY VALUE
31 32 SO4 ACCU BOUNDARY VALUE
31 33 SO4 DISS BOUNDARY VALUE
31 35 NH3 BOUNDARY VALUE
31 36 SOOT NEW BOUNDARY VALUE
31 37 SOOT AGD BOUNDARY VALUE
31 38 SOOT CLD BOUNDARY VALUE
31 39 BMASS NEW BOUNDARY VALUE
31 40 BMASS AGD BOUNDARY VALUE
31 41 BMASS CLD BOUNDARY VALUE
31 42 OCFF NEW BOUNDARY VALUE
31 43 OCFF AGD BOUNDARY VALUE
31 44 OCFF CLD BOUNDARY VALUE
31 45 NITR ACC BOUNDARY VALUE
31 46 NITR DISS BOUNDARY VALUE
31 276 DUST DIV 1 BOUNDARY TEND.
31 277 DUST DIV 2 BOUNDARY TEND.
31 278 DUST DIV 3 BOUNDARY TEND.
31 279 DUST DIV 4 BOUNDARY TEND.
31 280 DUST DIV 5 BOUNDARY TEND.
31 281 DUST DIV 6 BOUNDARY TEND.
31 282 SO2 BOUNDARY TEND.
31 283 DMS BOUNDARY TEND.
31 284 SO4 AITKEN BOUNDARY TEND.
31 285 SO4 ACCU BOUNDARY TEND.
31 286 SO4 DISS BOUNDARY TEND.
31 288 NH3 BOUNDARY TEND.
31 289 SOOT NEW BOUNDARY TEND.
31 290 SOOT AGD BOUNDARY TEND.
31 291 SOOT CLD BOUNDARY TEND.
31 292 BMASS NEW BOUNDARY TEND.
31 293 BMASS AGD BOUNDARY TEND.
31 294 BMASS CLD BOUNDARY TEND.
31 295 OCFF NEW BOUNDARY TEND.
31 296 OCFF AGD BOUNDARY TEND.
31 297 OCFF CLD BOUNDARY TEND.
31 298 NITR ACC BOUNDARY TEND.
31 299 NITR DISS BOUNDARY TEND.

Table 2: Stash codes for CLASSIC LBCs

42 © Crown Copyright 2015

Das könnte Ihnen auch gefallen